Fgf10 Gene Dosage from a Single Allele Is Insufficient for Forming Multilayered Epithelial Cells in the Murine Lacrimal Gland

TL;DR

This study shows that one copy of the Fgf10 gene is not enough to form proper lacrimal glands in mice, explaining why some human patients lack these glands.

Contribution

The study demonstrates Fgf10 haploinsufficiency in lacrimal gland development in mice, providing a model for human ALSG.

Findings

Fgf10+/− mice lacked bilateral lacrimal gland primordia at E16.5.

A single Fgf10 allele was insufficient to support budding during lacrimal gland morphogenesis.

FGF10 and FGF7 are mesenchymal factors critical for initial lacrimal gland formation.

Abstract

Mutations in the fibroblast growth factor 10 (FGF10) gene in humans cause aplasia of the lacrimal and salivary glands (ALSG). In patients with ALSG, heterozygous loss-of-function mutations are found, and FGF10 haploinsufficiency results in the absence of these secretory organs. Lacrimal glands (LGs) are formed through epithelial thickening, budding, and branching morphogenesis. To compare the variable phenotypes of the Fgf10+/− Harderian glands (HGs) previously reported, we examined the development of LGs in wild-type (WT), Fgf10+/−, and Fgf10-null mice. Pax6 immunostaining was performed to visualize the LG primordia from embryonic day 15.5 (E15.5) onwards. In situ hybridization of the genes encoding the epithelial receptor of FGF10, FGFR2b, and its other ligands was performed to determine their potential involvement in LG development. LG primordia were not observed in Fgf10+/− mice bilaterally at E16.5 or later stages. At E15.5, budding from the developing conjunctival epithelium (CE) was observed in a small fraction of the Fgf10+/− LG primordia. In contrast, the Fgf10-null CE failed to promote budding. Among Fgf1, Fgf3, Fgf7, Fgf10, and Fgf22, Fgf10 was expressed in the mesenchyme surrounding developing LG epithelial cells, whereas Fgf1 was expressed in the LG epithelium of WT mice. Fgf7 was initially expressed in the mesenchyme surrounding the nascent LG epithelium, but its expression subsequently became diffused. Thus, we conclude that among the FGFR2b ligands, initial LG formation is dependent on the mesenchymal factors FGF10 and FGF7, and FGF1 is likely to function as an epithelial factor in the LG primordia. A single allele of Fgf10 was found to be insufficient to support the budding process during LG morphogenesis.

Figures

Full text

1. Introduction

The lacrimal gland (LG) contributes to the production of the aqueous portion of tears, consisting of water, electrolytes, and proteins, which protect the ocular surface of the cornea and conjunctiva. In humans, the main LG is located within the orbit of the eye, whereas in rodents, the main LG lies exorbitally just below the auricle, with its long axis perpendicular to the zygomatic arch [1]. The exorbital LG is connected to the ocular surface by a single, long secretory duct, which then runs forward and is joined by the duct of the intraorbital LG and finally, to the lower eyelid. The developmental processes of the mouse LG and their molecular mechanisms have been well-studied [2].

Numerous factors are known to be involved in the development, cellular differentiation, tissue repair, and regeneration of the LG [2,3]. One such crucial factor is fibroblast growth factor 10 (FGF10). Since Fgf10-null mice lack lacrimal, Harderian, and salivary glands, Fgf10 is required for the development of these glands [4,5,6,7,8]. In contrast, transgenic mouse lines in which exogenous Fgf10 is overexpressed under a lens-specific promoter have extra glands within the cornea [5]. These results indicate that FGF10 induces LGs under competence factors such as Pax6 [4]. Thus, FGF10 is used to produce LG organoids from adult stem cells and induced pluripotent stem cells for clinical applications [9,10].

FGF10 belongs to the FGF family of extracellular secreted molecules that bind to transmembrane FGF receptors (FGFRs) for signal transduction and to extracellular matrix heparan sulfates to modulate their own functions [11]. The binding affinity of each FGF to FGFRs has been studied through a cell proliferation assay by comparison with FGF1 as the control [12,13]. FGF10 almost exclusively binds to FGFR2 isoform b (FGFR2b), although it can bind to FGFR1b with a much lower affinity. FGFR2b mRNA was found to be expressed in the developing epithelium of mouse embryos [14], whereas Fgf10 mRNA was detected in the mesenchyme underneath or surrounding the developing epithelia, such as in limb buds, lung buds, and eyelid primordia [15,16,17,18]. Thus, FGF10–FGFR2b signaling mediates a vital mesenchymal-to-epithelial cue. As FGFs with structural similarities to FGF10 are also ligands for FGFR2b [19], the roles of FGFR2b signaling cascade in LG development have been extensively clarified [2,20,21]. Most recently, it was reported that the phospholipase C gamma (PLCγ) pathway regulates LG development, competing with phospho-inositol 3 kinase (PI3K) [22]. Intracellularly, Shp2, Sprouty 2, Kras, and Pea3 function as FGFR2b downstream effectors during mouse LG development [23,24,25,26].

The specific and indispensable involvement of FGF10 in LG development or maintenance has been deduced from a human genetic disease, aplasia of the lacrimal and salivary glands (ALSG), caused by mutations in the FGF10 gene [27] (OMIM#180920). One allele of loss-of-function mutations in the FGF10 gene was detected; thus, haploinsufficiency for FGF10 underlies ALSG. However, since patients with a missense FGF10 mutation display a different phenotype as LADD syndrome [28] (OMIM#620193), exhibiting broader anomalies with affected ears, teeth, and digits, the existence of modifier genes or a dominant negative effect of the FGF10 mutation has been postulated as well.

As a disease model for ALSG with autosomal dominant inheritance, the phenotypes of *Fgf10-*heterozygous knockout (KO) (Fgf10^+/−^) mice have been reported [18,27]. In the Fgf10^+/−^ adult mouse, the LGs were absent, whereas residual salivary glands were observed, corresponding well to the clinical manifestations of ALSG patients [27]. To date, mouse Fgf10^+/^^−^ phenotypes have been described in the LGs, salivary glands, Harderian glands (HGs), ears, lungs, tracheal submucosal glands, and craniofacial bones [29,30,31,32,33,34,35]. Although Puk et al. (2009) reported that a new heterozygous loss-of-function mutation in the mouse Fgf10 gene (Aey17) causes atrophy of the HG [31], we recently re-examined another mouse line of Fgf10^+/−^ and found that more than half of the mutant mice lost their unilateral HGs in the early postnatal period [36].

The HG was originally described as a second LG, but the products and histology of the glandular secretory unit are different: HGs exhibit tubuloalveolar glands with a wide lumen secreting lipid-rich products, whereas LGs exhibit tubuloacinar glands connected to intercalated ducts secreting serous products [3,37]. In haploinsufficient conditions, phenotypic variability is often observed with a range of possible underlying molecular mechanisms [38]. In this study, we clarified when and how LGs deteriorated in our Fgf10^+/−^ mouse line and compared them with HGs to obtain a phenotypic view of this disease model animal. A considerable number of Fgf10 KO mice are available (Mouse Genome Informatics, https://www.informatics.jax.org) (MGI:1099809). Here, we describe the LG phenotype in one such strain, Fgf10^tm1Ska^, where a neomycin resistance cassette replaces the exon containing the ATG [18], for the foundation of consecutive studies to elucidate the mechanisms underlying phenotypic variance in the haploinsufficiency state.

2. Results

2.1. Absence of LGs in the Fgf10+/− Mouse at 5 Weeks and Perinatal Days

In the adult Fgf10^+/−^ mouse, adipose tissue occupies the area where the LG should normally exist [27]. The authors of that study examined two Fgf10^+/−^ female mice at 8–9 weeks old, but whether the LG was not formed at all during development or if the once-formed LG degenerated later by the stage of sexual maturity was not clarified. In this study, we first confirmed the absence of LG in Fgf10^+/−^ male mice at 5 weeks of age (n = 5, Figure 1, Supplementary Table S1). The LG was already replaced by adipose tissue (Figure 1D–F).

As we found that the Fgf10^+/−^ HG, a large post-bulbar gland characteristic of tetrapods, is often unilaterally degenerated by the early postnatal stage [36], we examined LG phenotypes at postnatal day 10 (P10) and earlier stages. We performed immunostaining with the anti-Pax6 antibody on P10 head sections because Pax6 identifies the LG primordia in mice [4]. None of Pax6-positive LG primordia were present in the Fgf10^+/−^ mouse, whereas Pax6-negative developing salivary glands were observed, corresponding to submandibular, sublingual, and parotid gland primordia (Supplementary Figure S1). At P6, P0.5, and embryonic day 19 (E19), histology showed only hollow connective tissues lacking LG primordia in the Fgf10^+/−^ mouse (Supplementary Figure S2). Whole-mount acetocarmine staining also showed the absence of LG primordia in Fgf10^+/−^ mice at P0.5 (n = 2 for WT, n = 4 for Fgf10^+/−^) and E19.5 (n = 2 for WT, n = 3 for Fgf10^+/−^) (Supplementary Figure S2 and Table S1) [22]. Intraorbital LGs were also absent in Fgf10^+/−^ mice at E19.5, as confirmed by hematoxylin–eosin (HE) staining (n = 2 for WT, n = 2 for Fgf10^+/−^) (Supplementary Figure S3). At E18.5, we performed 3D morphological analyses on a whole head of WT or Fgf10^+/−^ mice by correlative light microscopy and block-face imaging [39] and revealed that the LGs were absent bilaterally, while the nearby parotid glands were present in the Fgf10^+/−^ head, as reported previously in the adult mutant [27] (Supplementary Figure S4 and Movie S1).

2.2. Absence of LGs in the Fgf10+/− Mouse in Late Embryonic Stages

We then sought to examine the embryonic LG primordia before E18.5. In developing mice, the LG emerges from ingrowing at the temporal nascent conjunctival epithelium (CE), whereas the HG develops from ingrowing at the nasal CE. Mouse LG development involves three distinct processes: thickening of the developing CE, budding from the thickened epithelium, and subsequent branching morphogenesis [2]. LG branching morphogenesis has been reported not to begin before E16 in C57BL/6J mice, being delayed compared to ICR mice [40]. Since the origin of our Fgf10 KO mouse strain was (C57BL/6NCrlj × CBA/JNCrlj)F1 [18], we observed LG development in WT litters at E15,5, E16.5, E17, E18, and E18.5 by Pax6-immunostaining (Figure 2). At E15.5, budding from the developing CE was observed. Some sections showed isolated nascent acinar-like cells (Figure 2A), whereas other serial sections showed a bifurcation of the CE (Figure 3A) according to their horizontal levels. By E16.5, the terminal portion of the bifurcated epithelium had become multilayered and exhibited a club-like appearance (Figure 2C). At E17 and onwards, multiple isolated glandular cells were observed in the mesenchymal capsule-like sheath (Figure 2E,G,I), indicating cross-sections of branched glandular cells. In contrast, none of the Fgf10^+/−^ embryos had Pax6-positive LG primordia from E16.5 onwards (Figure 2D,F,H,J).

2.3. LG Primordia of the Fgf10+/− Mouse at E15.5: A Small Fraction Has Stunted Buds

In the first Fgf10^+/−^ embryo at E15.5 examined, we could not find budding from the developing CE (Figure 2B). Since we previously found variable phenotypes in Fgf10^+/−^ HGs, we examined multiple Fgf10^+/−^ embryos at E15.5 and compared their phenotypes among the three genotypes. We found that most of the WT embryos at E15.5 had budding from the developing CE (n = 12 WT embryos; Figure 3A,E; Supplementary Movie S2). In contrast, three out of 22 Fgf10^+/−^ LG primordia examined (n = 11 Fgf10^+/−^ embryos) budded from the CE, but their subsequent growth was stunted (Figure 3B,E; Supplementary Movie S3). The remaining Fgf10^+/−^ LG primordia had no budding, and the LG primordial epithelium was flat or thickened only (Supplementary Movie S4), similar to that found in the Fgf10-null embryo (Figure 3C,E). As for Fgf10-null embryos, LG primordia have been shown to be absent at P0 [4]. We observed a thickened epithelium at the tip of the ingrowing CE (Figure 3D); however, none of the Fgf10-null embryos had budding from the CE (n = 7; Figure 3E).

2.4. Histomorphometric Analysis of HG Primordia of Fgf10+/− Mice at E15.5

As we previously reported on histology [36], Pax6-immunostaining showed that Fgf10^+/−^ HG primordia at E15.5 had multilayered, club-like glandular cells similar to those of WT (Supplementary Figure S5A–C). To quantitatively evaluate the size of the HG primordia, we compared their length, area, and number of epithelial layers between WT and Fgf10^+/−^ (Supplementary Figure S5E) (n = 12 for WT, n = 11 for Fgf10^+/−^). All these parameters were significantly reduced in Fgf10^+/−^ HGs (Supplementary Figure S5F). We also examined whether the difference in the size of the Fgf10^+/−^ HG primordia between the right and left sides could be detected at this stage, as we found that more than half of Fgf10^+/−^ mice exhibited unilateral HG degeneration [36]. At E15.5, none of the three parameters were significantly reduced on either side in Fgf10^+/−^ and WT mice (Supplementary Figure S5G,H). Thus, differential developmental delay or degeneration did not occur on either side of the developing HGs at E15.5. In the Fgf10-null embryos, a small process for the nictitating membrane, the third eyelid, was observed, and the flanking epithelia were thickened; however, none developed multilayered, club-like glandular cells, as seen in the WT and Fgf10^+/−^ embryos at E15.5 (n = 7, Supplementary Figure S5D).

2.5. mRNA Localization of Fgf10 and Other FGFR2b-Ligand Fgf Members During LG Development

FGF receptor type 2 isoform b (FGFR2b), whose mRNA is present in the developing epithelium [14], mediates crucial mesenchymal–epithelial signaling required for organogenesis, including that in the LGs [2,41]. FGFR2b ligands include FGF1, FGF3, FGF7, and FGF22, besides FGF10 [12,13], which may be synergistically involved in LG development. Therefore, we focused on these FGF members and examined their mRNA localization along with that of Pax6 and Fgfr2b in WT mice. In situ hybridization (ISH) revealed that Pax6 and Fgfr2b were expressed in developing LG epithelial cells at E15.5 and E18.5 (Figure 4A,A′,G,G′). As expected, distinct expression of Fgf10 was observed in the mesenchymal cells surrounding the LG epithelial cells (Figure 4B,B′). In contrast, we found that Fgf1 was clearly expressed in developing LG epithelial cells, similar to Pax6 and Fgfr2b (Figure 4C,C′). Fgf3 expression was not observed at E15.5 but was diffusely observed in E18.5 LG epithelial cells and the surrounding mesenchymal tissue (Figure 4D,D′). Fgf7 was diffusely expressed in the mesenchymal tissue, but its expression was not restricted to the developing LG region (Figure 4E,E′) compared to the data obtained with sense control probes (Supplementary Figure S6). None of the Fgf22 transcripts were detected in association with the LG primordia at E15.5 or E18.5 (Figure 4F,F′), while its mRNA was distinctly detected in developing hair follicles such as vibrissae (Supplementary Figure S7) as reported [42].

Since we found that a small fraction of Fgf10^+/−^ LG primordia proceeded with budding from the nascent conjunctival epithelium, we examined the expression patterns of Fgf10, Fgf7, and Fgf1 at the position of bifurcated LG primordia of Fgf10^+/−^ mice at E15.5 (Figure 5). We found that Fgf10 was expressed in the proximal and anterior mesenchyme of the LG bud (Figure 5A). In contrast, Fgf7 was expressed in the more posterior periocular mesenchyme, which was localized in the portion toward which the LG bud was elongated (Figure 5B). Two-color fluorescent ISH (FISH) verified the overlapping and differentially expressed domains of Fgf10 and Fgf7 (Figure 5E). As for Fgf1, it was expressed in both conjunctival and bifurcated LG bud epithelia (Figure 5C; compare with the sense control in Figure 5C′).

3. Discussion

In this study, we used young male, postnatal, and embryonic mice and identified the LG phenotype in a strain of Fgf10^+/−^ mice [18] (MGI: 1099809): Fgf10^+/−^ LG does not develop after budding at the latest. Although Qu et al. (2011) briefly showed a similar LG phenotype in another Fgf10-heterozygous KO mouse [29,43], we wanted to know at what stage the Fgf10^+/−^ LG deteriorates in the case of our strain. As Kuony et al. (2019) described [40], the developmental stage of the LG seemed to be slightly delayed in Fgf10 KO mice with the C57BL/6J background. Pan et al. (2008) showed ectopic LG budding was induced by FGF10-soaked bead implantation into the conjunctival mesenchyme near the authentic LG bud [24]. Thus, it is tempting to reverse lacrimal gland aplasia in the Fgf10-heterozygous mice as a preliminary test for ALSG patients.

We visualized the WT budding portion from the developing CE wall at E15.5 on histological sections, which morphologically resembles the lateral or side branching in the mammary gland [44]. Because this budding does not occur at all in *Fgf10-*null mice, the Fgf10 gene dosage from the two alleles is required for budding in the LG. The WT LG bud, which morphologically resembles the terminal end bud in the mammary gland, consists of a multilayered epithelium, and subsequent clefting, typically observed in the developing salivary gland, seems to advance branching morphogenesis of the LG [45]. We witnessed the LG buds emerge in a small fraction of Fgf10^+/−^ embryos, but the bifurcated domain exhibited merely a folding of one-cell layer cells. Thus, we suggest that the Fgf10 gene dosage from a single allele is insufficient for forming multilayered epithelial cells to advance further branching morphogenesis. Most Fgf10^+/−^ LG primordia (19 of 22 LGs from 11 embryos) exhibited no thickening or thickening only, implying that developmental delay [46] and phenotypic variations could obscure the definite phenotype.

Although Fgf10^+/−^ and Fgf10^−/−^ mice differ in the extent to which LG development proceeds, the Fgf10 gene dosage required in the initial phase of bilateral LG development are considered rather high. This differs from the dose required for the formation of the HGs, eyelids, and limbs. Why the degree of phenotypes for these organs is so different in the *Fgf10-*heterozygous state (haploinsufficiency) is not clear, whereas all these organs are lacking in the Fgf10-null mouse. This is most likely due to genetic redundancy among FGFR2b ligands. It was postulated that the duration of transcriptional bursting from only one allele could not compensate for the stochastic, transiently low expression levels below the threshold required for the formation and development of organ primordia [47], which might correlate with phenotypic variations.

Organogenesis requires epithelial–mesenchymal interactions, in which FGFR2b signaling is thought to be crucial [2]. Mice and humans have 22 FGF members, and they have distinct affinities for seven types of FGF receptors [11]. FGF10 belongs to the FGF7 subfamily, which comprises FGF3, FGF22, FGF7, and FGF10. As described, FGFR2b specifically binds to the members of the FGF7 family along with FGF1 [12,13]. This study showed that Fgf10 and Fgf7 are expressed in the overlapping domain of mesenchyme near the developing LGs, but their expression domains are not identical. Functional redundancy with Fgf7 could partially rescue the deficiency of Fgf10 during LG development.

We observed overlapping expression patterns of Pax6 and Fgfr2b in the developing LG epithelium. Because Fgfr2b conditional KO mice lose Pax6 expression in the CE [48], FGFR2 signaling is thought to be important for maintaining Pax6 expression in ocular epithelial cells. Conversely, the transcription factor Pax6 directly regulates cell cycle exit in lens cells by binding to the Fgfr2 gene [49]. Although these previous studies covered other ocular epithelial cell components, mutual interactions between Pax6 transcriptional control and FGFR2b signaling in the developing LG epithelium are conceivable.

Both Pax6 and Fgf1 are expressed in developing LG epithelial cells. FGF1 is known to bind all seven FGF receptor proteins and does not specifically activate FGFR2b [12,13]. Pan et al. (2008) reported that FGF1 binding to FGFR2b does not require heparan sulfate and that the site of the FGF1–FGFR2b binding complex is not confined to the LG bud (tip) [24]. Our ISH data also showed a rather broad expression domain of Fgf1 in the developing conjunctival and LG epithelia, implying a unique role for FGF1 via FGFR2b and other FGFRs. FGF1 is known to maintain metabolic homeostasis and insulin sensitivity by upregulating its transcription via PPARγ gamma when exposed to a high-fat diet [50]. Therefore, it is tempting to speculate that the transcription of Fgf1 might be likewise regulated by Pax6 in LG formation, regeneration, and homeostasis.

In the present study, we focused on male LG phenotypes in the case of adult mice to exclude the influence of sexual dimorphism on their severity because a higher prevalence in females is seen in LG disorders [51]. As in the salivary glands, sexual difference in morphology is observed in the LG of rodents and humans [52], in which acinar area in LGs of 8-week male mice is larger than female equivalents. Given that the effect of testosterone on gender-specific gene expression like lacrimal Mup (Major urinary protein) mRNA levels occur after 2 weeks [53], the influence of sex hormones on LG expansion after morphogenesis seems to begin around this stage. Therefore, after the LG morphogenesis is complete, the acinar growth is likely enhanced by sex hormones. As previously reported [27], female Fgf10^+/−^ mice also lack the LG. Although whether the precise timing of LG morphogenesis is identical between the sexes is not well determined, it is most likely that Fgf10^+/−^ female mice also primarily fails to develop LGs as in the male.

4. Materials and Methods

4.1. Fgf10 KO Mice and Tissue Collection at Different Stages

Fgf10 KO mice used in this study have been described previously [18]. Genotyping was performed using tail DNA as previously described [36]. Male and female Fgf10^+/−^ mice were crossed, and the morning when vaginal plugs were detected was considered day 0.5 post coitum. Embryos were collected via cesarean section at E15.5, E16.5, E17, E18, E18.5, E19, and E19.5. The numbers of embryos examined are listed in Supplementary Table S1. Postnatal (P0.5, P6, and P10) and young (5 weeks old, male) mice were processed for subsequent experimental procedures after appropriate euthanasia. Acclimatization periods for the 5-week mice to experimental locations were more than 24 h.

4.2. Fixation and Histology

The collected embryos were immediately washed in ice-cold phosphate-buffered saline (PBS), immersed in 4% paraformaldehyde (PFA)/PBS at 4 °C for 1 h, and their heads were dissected out. For P6, P10, and 5-week-old mice, transcardial perfusion fixation was performed before dissection. The heads were fixed in ice-cold 4% PFA/PBS overnight, washed with PBS, dehydrated in an ethanol series, cleared in xylene, and embedded in paraffin. Sections were cut at 5 μm, transferred onto CREST-coated glass slides (Cat. No. CRE-01; Matsunami Glass, Osaka, Japan), deparaffinized using G-NOX (GenoStaff, Tokyo, Japan), and stained with Mayer’s hematoxylin (Cat. No. 1.09249.0500; Merck Millipore, Sigma-Aldrich, St. Louis, MO, USA) and Eosin Y (Cat. No. HT110116; Sigma-Aldrich) according to the standard procedure. To identify the location of the developing LGs, paraffin sections were stained with 0.05% toluidine blue (pH 7.0, Cat. no. 40981, Muto Pure Chemicals, Tokyo, Japan).

For in situ hybridization, frozen samples were prepared according to standard procedures. Cryoprotection was achieved by immersing the samples in 30% sucrose/PBS. Frozen sections were cut at 7 μm for E15.5 embryos and at 15 μm for later stages using a cryostat (Tissue-Tek Polar^®^; Sakura Finetek Japan, Tokyo, Japan).

4.3. Whole-Mount Acetocarmine Staining of LGs

To visualize LGs as whole mounts, acetocarmine staining was performed as previously described [22]. Briefly, fixed and trimmed mouse heads at E19.5 and P0.5 were dissected to expose the exorbital LGs and dehydrated in 70% ethanol overnight. The samples were then stained with a 0.5% carmine solution (Cat. No. C-1022; Sigma-Aldrich) prepared in 45% boiling acetic acid for 5 min. Successive destaining was performed using 70% ethanol (3 min), 1% HCl in 70% ethanol (2 min), and 5% HCl in 70% ethanol (1 min). The exorbital LGs in 70% ethanol were observed using a Leica stereomicroscope. After carmine staining, the samples were washed vigorously in 70% ethanol and processed for paraffin embedding, sectioning, deparaffinization, and HE staining to determine the presence or absence of intraorbital LGs (Supplementary Figure S3).

4.4. Correlative Microscopy and Block-Face Imaging (CoMBI)

Fixed embryonic heads (E18.5) were immersed overnight in 1% tannic acid/PBS. After cryoprotection with 30% sucrose/PBS, the samples were embedded in OCT compound and sectioned at 15 μm with a cryostat. Three-dimensional imaging using CoMBI was performed as previously described [39,54]. Briefly, the block-faces were captured serially using a digital camera (Sony α7RIII, Tokyo, Japan) with a macro lens (Tamron SP AF 180 mm F/3.5 Di LD [IF] MACRO 1:1 (Model B01), Saitama, Japan) and a teleconverter (Kenko Digital Teleplus Pro 300 2× DGX, Tokyo, Japan). Segmentation of the LGs and parotid glands was performed manually. As landmarks, the eyeball, lens, optic nerve, mandibular ramus, and inner ear were manually segmented and reconstructed using a 3D slicer (https://www.slicer.org). To verify the interpretation of the structures in the block-face images, several sections were stained with 0.05% toluidine blue, and their histology was correlated with the block-face images, as shown in Supplementary Figure S4.

4.5. Immunohistochemistry

Immunohistochemistry was performed on deparaffinized sections according to standard procedures. Briefly, the sections were antigen-retrieved, immersed in 10 mM citrate buffer (pH 6.0), and heat-treated in a pressure cooker (124 °C, 10 min). After sections were cooled to room temperature, they were washed with H_2_O and quenched against endogenous peroxidase using BLOXALL (Cat. No. SP-6000; Vector Laboratories, Newark, CA, USA) for 10 min, washed in PBS-Tween 20 (0.05%), and blocked with 2.5% normal horse serum (NHS). Next, anti-Pax6 antibody (rabbit, diluted 1:2000 in 2.5% NHS, Cat. No. 901301; BioLegend, San Diego, CA, USA) was added to the sections and incubated at room temperature for 1 h. After washing, ImmPRESS Peroxidase Polymer anti-rabbit IgG reagent (Cat. No. MP-7401; Vector Laboratories) was added, incubated for 10 min, washed with PBS-Tween, and visualized by incubation with the ImmPACT DAB EqV substrate (Cat. No. SK-4103; Vector Laboratories) for 2 min. After vigorous washing in H_2_O and PBS, and 1 min counterstaining with Mayer’s hematoxylin, the sections were mounted with a mixture of polyvinyl alcohol/glycerol. For frozen sections, antigen retrieval was shortened to 5 min at 95 °C in a thermo pot for better preservation of sections.

4.6. ISH

Conventional ISH was performed as previously described [36,55]. Target cDNAs were amplified from mouse brain first-strand cDNA by PCR using the primers listed in Supplementary Table S2 and inserted into the pBluescript II KS(+) vector. To generate digoxigenin-labeled riboprobes, template cDNAs flanked with T3, T7, and/or SP6 RNA promoters were prepared by PCR. For Fgf22, Fgfr2b, and Pax6, the coding regions were used as probes. For Fgf10, Fgf7, Fgf1, and Fgf3, 5′-untranslated regions were included in addition to the coding regions for the probes. The 3′-untranslated regions were not included in probe sequences not to detect signals when hybridized with sense probes. Two-color FISH was performed using signal amplification by exchange reaction (SABER)-FISH combined with hybridization chain reaction (HCR) according to previous reports [56,57]. HCR hairpins were prepared according to a previous report [58] Briefly, DNA hairpins labeled at the 5′ end with a C12-amino linker (IDT, Coralville, IA, USA) dissolved in 0.1 M sodium borate buffer (pH 8.5) was mixed with the N-hydroxysuccinimide esters of ATTO488 for S9_H1 and S9_H2 and with ATTO565 for S25_H1 and S25 H2 dissolved in dimethylformamide for coupling reaction. Fluorophore-conjugated DNAs were purified by denaturing polyacrylamide gel electrophoresis with 20% polyacrylamide gels. The DNA oligos used in SABER-HCR-FISH and related information are listed in Supplementary Table S3.

4.7. Image Capture and Processing

For dissection and observation of the dissected samples, a Leica dissecting microscope (M165FC or S9D; Leica Microsystems, Wetzlar, Germany) equipped with a digital camera was used (DFC310 FX or FLEXACAM C1; Leica Microsystems). Sections were observed under a Leica microscope (DM5000B; Leica Microsystems) or a Carl Zeiss Axioplan microscope (Carl Zeiss, Oberkochen, Germany). Images were captured using a digital camera (DS-Fi1; Nikon, Tokyo, Japan) or an IDS UI-3290SE-C-HQ camera (Imaging Development Systems, Obersulm, Germany). To identify the horizontal level of the sections, images were captured using a Keyence microscope with a 4x objective lens (BZ-X700; Keyence, Osaka, Japan) (CRL_552; Core Facility Portal, Okayama University). Fluorescence images were acquired using an IX71 inverted microscope (Olympus, Tokyo, Japan) equipped with an sCMOS camera (pco.panda 4.2; Excelitas PCO GmbH, Kelheim, Germany).

4.8. Histomorphometric Analysis and Statistics

Pax6 immunostaining was performed on 10–24 horizontal sections per individual embryonic head at E15.5 (n = 12 for WT, n = 11 for Fgf10^+/−^, and n = 7 for Fgf10^−/−^), and sections containing the largest HGs were selected and digitally photographed. The length, area, and number of epithelial layers were measured from the images as shown in Supplementary Figure S5. The average values were tested for significance using Student’s t-test. The graph was drawn as an Excel scatter plot, and the error bars indicate the standard deviation (SD). The same Pax6-immunostained sections were examined for the LG phenotype and categorized into four morphologically distinct types: normal budding, stunted buds, thickening only, and no thickening (Figure 3).