Sculpting the skull through neurosensory epithelial-mesenchymal signaling

Summary

The vertebrate skull is a complex structure housing the brain and specialized sensory organs, including the eye, the inner ear, and the olfactory system. The close association between bones of the skull and the sensory organs they encase has posed interesting developmental questions about how the tissues scale with one another. Mechanisms that regulate morphogenesis of the skull are hypothesized to originate in part from the encased neurosensory organs. Conversely, the developing skull is hypothesized to regulate the growth of neurosensory organs, through mechanical forces or molecular signaling. Here, we review studies of epithelial-mesenchymal interactions during inner ear and olfactory system development that may coordinate the growth of the two sensory organs with their surrounding bone. We highlight recent progress in the field and provide evidence that mechanical forces arising from bone growth may affect olfactory epithelium development.

Introduction

The vertebrate skull is an intricate and protective housing for the brain and specialized sensory organs, including the eye, the inner ear, and the olfactory system. The size and shape of the vertebrate head differs dramatically among species. Such morphological diversity has fueled hypotheses about the adaptive importance of the size and shape of the skull and the structures it supports (see Fish, 2017). While bone protects neurosensory organs, it also limits their size and shape. Therefore, selective pressures driving evolution of the size and shape of neurosensory structures must also act on the skull. For example, the sizes of the orbit and the eye have both increased substantially in the evolution of digited tetrapods from finned fish, providing an evolutionary advantage for terrestrial (above water) vision (MacIver et al., 2017). In a more striking example, the structural complexity of the posterior nasal turbinates, which is linked to the surface area of the overlying olfactory epithelium, varies highly among mammals (Van Valkenburgh et al., 2014). This variation has been hypothesized to correlate with the adaptive importance of olfaction during evolution. For instance, terrestrial carnivorous caniforms have increased olfactory surface area relative to body size compared to omnivorous caniforms, correlating with their need to detect far-away prey due to their more specialized diet (Green et al., 2012). Furthermore, a recent study found that across mammals, the number of olfactory receptor genes encoded in the genome correlates with the size of the cribriform plate, a bony structure that olfactory axons cross to reach the brain (Bird et al., 2018).

The close association of parts of the skull and neuro- and neurosensory epithelia, such as that between the cranium and the brain, poses interesting developmental questions. It has been hypothesized that crosstalk between neurosensory structures and the surrounding bone during development must exist to ensure that the tissues scale proportionally (reviewed in Adameyko and Fried, 2016). Such signaling interactions are potential targets for adaptive forces during evolution, accounting for the diversity of skull and neurosensory organ morphologies. Indeed, existence of epithelial-mesenchymal interactions between the ectoderm and underlying facial primordia during early craniofacial development has been known for a long time (see Francis-West et al., 1998). However, the identification of molecular signals mediating such interactions during neurosensory development is lacking.

In this review, we discuss the current understanding of how neurosensory epithelia shape the surrounding skull, and vice versa, by reviewing studies of epithelial-mesenchymal interactions during development. Specifically, we focus on the development of the inner ear otic epithelium and its surrounding otic capsule, and of the main olfactory epithelium and the posterior nasal capsule and turbinates.

Inner Ear

Structure and Function

The inner ear, divided into the dorsal vestibule and the ventral cochlea, contains six structures required for balance and hearing: three semicircular canals, utricle, saccule, and cochlea (Fig. 1D). These structures contain neurosensory epithelia filled with mechanosensory hair cells and supporting cells, and are uniquely shaped. The three semicircular canals (horizontal, superior, and posterior), which detect head movement in three-dimensional space, are oriented to three different planes at right angles to one another. The size of the semicircular canals is believed to be functionally important, as larger canals are associated with higher sensitivity to rotations (Ekdale, 2015; Alsina and Whitfield, 2017). In mammals, the cochlea, which contains the sound-detecting organ of Corti, is a coiled duct. Although the functional significance of the coiling is not entirely known (Ekdale, 2015), the length of the cochlea likely has a role in hearing. Movement of the basilar membrane in response to different frequencies of detected sound elicit activation of hair cells at specific positions along the length of the cochlea (Mann and Kelley, 2011). This is referred to as the tonotopic organization of the cochlea. Another epithelial structure of the inner ear is the endolymphatic duct and sac (Fig. 1D), which contains endolymph, a fluid essential for the functions of the inner ear. The entire adult inner ear is encased in the bony otic capsule, which is a part of the temporal bone.

Figure 1. Schematic of inner ear and periotic mesenchyme development.

(A-D) Whole mount (top) and cross-sectional (bottom) view of the inner ear at E10.5 (A), E12.5 (B), E14.5 (C), and E17.5 (D). The dorsal (vestibular; blue) and ventral (cochlear; purple) parts of the inner ear develop from the otocyst (ot), surrounded by periotic mesenchyme. The outer layer of the periotic mesenchyme develops into the otic capsule (red), which becomes cartilaginous at around E14, while the inner part differentiates into otic fibrocytes (pink). Condensing mesenchyme is represented by reddish-pink. D, dorsal; M, medial; amp, cristae ampullaris of the semicircular canal; co, cochlea; ed, endolymphatic duct and sac; fp, fusion plate; hb, hindbrain; sa, saccule; sc, semicircular canals; ut, utricle. (E) Cross-sectional view of the developed cochlear duct, showing the organ of Corti (purple) containing sensory hair and supporting cells in the scala media (scm), as well as the fluid-filled scala vestibuli (scv) and scala tympani (sct). The spiral ganglion (sg) contains neurons that relay electrical signals from the sensory hair cells to the brain. bm, basilar membrane; oc, otic capsule; slb, spiral limbus; slg, spiral ligament.

Development

Development of the inner ear begins at around embryonic day 8.5 (E8.5) with the thickening of the otic placode, which invaginates and pinches off to form the otic vesicle (otocyst; reviewed in Basch et al., 2015; Sai and Ladher, 2015). The otocyst then undergoes outgrowth and morphogenesis to form the vestibule and the cochlea (see Figs. 1A-D). Throughout its development, the otic capsule tightly associates with the inner ear, with the contours of the capsule matching the intricate shape of the inner ear structures (Figs. 1A-D). Spatial restriction placed on the inner ear due to the size of the otic capsule is thought to influence the development of the inner ear, in particular the coiling of the cochlea (Li and McPhee, 1978, 1979). However, this has not been shown definitively or mechanistically.

The otic capsule develops from the periotic mesenchyme (POM), a tissue of mixed neural crest and cranial paraxial mesoderm origin that surrounds the inner ear (otic) epithelia (Trainor and Tam, 1995; Ladher, 2017). Specifically, the otic capsule develops from the outer POM layer via endochondral ossification (Sher, 1971; McPhee and Van De Water, 1985). In the developing mouse, the POM begins to condense at E10 (Frenz and Van De Water, 1991), differentiates into chondrocytes shortly after, and ossifies postnatally. The inner POM layer forms other mesenchyme-derived structures, such as the spiral ligament and the spiral limbus (refer to Fig. 1E), which are made up of differentiated otic fibrocytes (Sher, 1971; Cohen-Salmon et al., 2000; Trowe et al., 2010). The fluid-filled chambers of the cochlea, scala vestibuli and scala tympani (refer to Fig. 1E), are formed by cavitation of the inner POM layer (Phippard et al., 1999).

Epithelial-to-mesenchymal signaling in the developing inner ear

Early studies in mouse embryos using cultures of dissociated POM, dissected free of otic epithelium, showed that the presence of otic epithelial tissue affects chondrocyte differentiation (see McPhee and Water, 1986; Van De Water and Ruben, 1974). These experiments found that the otic epithelium promotes mesenchymal condensation in POM from E10.5 to E12, but inhibits it at E14 (Frenz and Van De Water, 1991). They also showed that TGFβ1 (Frenz et al., 1992), FGF2 (Frenz et al., 1994), SHH (Liu et al., 2002), retinoic acid (Frenz et al., 1996; Frenz and Liu, 1997, 2000) and BMP4 (Chang et al., 2002; Liu et al., 2003) could be signals from the otic epithelium regulating POM chondrogenesis. However, these in vitro experiments do not rule out the possibility of other tissues serving as sources of these signals, including the POM itself. Furthermore, no in vivo genetic data currently exists to back up the epithelial-to-mesenchymal (E→M) function of any of these particular signals.

The first genetic evidence of E→M cross talk in the inner ear came from experiments with Fgf9 null mice. FGF9, like FGF2, is a member of the Fibroblast Growth Factor (FGF) family of signaling molecules (Ornitz and Itoh, 2015). FGF9 is a member of the FGF9 subfamily of FGFs, which also includes FGF16 and FGF20. This subfamily is often expressed in epithelia and has a preference for binding the mesenchyme-expressed IIIc splice forms of FGF receptors 1 and 2 (FGFR1 and FGFR2; Ornitz and Itoh, 2015; Zhang et al., 2006), making them ideal candidates for mediating E→M signaling.

Fgf9 is expressed in the otic epithelium as early as E10.5 (Pirvola et al., 2004). In Fgf9-null mice, the otic capsule, especially the vestibular portion, appeared hypoplastic at E14.5 and was thin and incompletely formed at E18.5. In addition, while Fgf9-null mice had normal vestibular and cochlear sensory epithelia, other epithelial structures, particularly the semicircular canals, were malformed (see next section). These defects were attributed to decreased proliferation of the POM at E12.5. Other signs of mesenchymal malformation included an enlarged scala vestibuli, which may be due to mesenchymal remodeling (Pirvola et al., 2004). It is currently not known if loss of FGF9 leads to all of these defects directly.

More recently, FGF9 and FGF20 from the developing cochlear epithelium were shown in vivo to signal to FGFR1 and FGFR2 in the POM at E11.5-E12.5 (Huh et al., 2015). Deletion of both Fgf9 and Fgf20 (here referred to as Fgf9/20-null) led to a severely shortened cochlear duct with decreased coiling attributable to decreased epithelial progenitor proliferation. This phenotype was reproduced with mesenchyme-specific deletion of both Fgfr1 and Fgfr2 (here referred to as Fgfr1/2^mes:null), but not with epithelium-specific deletion of these Fgfrs. Interestingly, this implies that FGF signaling in the mesenchyme activates a mesenchymal-to-epithelial (M→E) signal that, in turn, regulates epithelial proliferation and cochlear duct growth (see Fig. 4A; Huh et al., 2015). The identity of this M→E signal(s) is currently unknown.

Figure 4. Model of established epithelial-mesenchymal interactions during inner ear and olfactory system development.

(A) In the inner ear at E12.5, FGF9 (and Netrin 1, not shown) from the dorsal otocyst regulates the proliferation of adjacent mesenchymal cells, some of which differentiate into chondrocytes. Mechanical force from mesenchymal proliferation, in turn, is required to push together epithelial walls for fusion plate formation to initiate semicircular duct development. FGF9 from the roof of the cochlear duct (cd) is required for remodeling of the adjacent mesenchyme during scala vestibuli formation (Pirvola et al., 2004). FGF9 plus FGF20 from the floor of the cochlear duct signals to the adjacent mesenchyme, which sends a yet unidentified signal back to the cochlear epithelium to regulate progenitor proliferation (Huh et al., 2015). (B) In the olfactory system at E17.5, FGF20 from progenitors located at negatively-curved olfactory epithelium regulates the proliferation of adjacent mesenchymal cells, which differentiate into chondrocytes that form the protruding turbinate (Yang et al., 2018). Mechanical force from turbinate growth, in turn, generates negative curvature in the epithelium, which may be required to maintain Fgf20-expressing olfactory progenitors (FEP cells).

Canonical Wnt/βCatenin (βCat, also Ctnnb1) signaling has also been proposed, based on in vivo genetic evidence, to be an E→M signal during inner ear development. Axin2, a negative regulator and widely used marker of Wnt/βCat activity, is highly expressed in the POM throughout development, including the inner, but not outer, POM layer surrounding the cochlea, (Bohnenpoll et al., 2014). Wnt ligands 4, 5a, 7a, and 7b, meanwhile, are expressed by the cochlear duct at E12.5. Mesenchyme-specific deletion of βCat (here referred to as βCat^mes:null) led to a slight thickening of the otic capsule and loss of parts of the capsule medial to the cochlea. There was a further defect in otic fibrocyte proliferation and differentiation, leading to a loss of the spiral ligament and spiral limbus. The pericochlear cavities, scala vestibuli and scala tympani, were missing as well (Bohnenpoll et al., 2014). These findings are suggestive of epithelia-expressed Wnt ligands regulating fibrocyte and otic capsule development. A complementary experiment in which Wnt ligands are specifically deleted from the epithelium will help make this conclusion more definitive.

Mesenchymal deletions of Fgfr1 and Fgfr2 (Huh et al., 2015) and βCat (Bohnenpoll et al., 2014) had similar effects on the cochlea. In both studies, hair cell differentiation in the organ of Corti was not disrupted, but the cochlear duct was much shorter. However, while the short cochlea phenotype in Fgfr1/2^mes:null mice was attributed to decreased proliferation at E11.5 and E12.5, the similar phenotype in βCat^mes:null mice was attributed to increased apoptosis in the cochlear epithelium at E13.5 and E14.5. Despite these differences, the two studies both indicated that there are M→E signals regulating otic epithelium development. As Bohnenpoll et al. (2014) hypothesized, these signals could be a result of spatial restrictions on the growing cochlea due to the thickened capsule wall in βCat^mes:null inner ears.

Mesenchymal-to-epithelial signaling in the developing inner ear

Early studies of inner ear development involving grafting and explant cultures of mesenchyme-free otocysts showed that the POM is necessary for proper inner ear development (see Anniko and Schacht, 1984; Doetzlhofer et al., 2004; Miura et al., 2004; Montcouquiol and Kelley, 2003; Swanson et al., 1990). However, the extent to which the POM affects epithelial differentiation and morphogenesis was unclear, as results from different studies were not always consistent with one another. For example, Anniko and Schacht (1984) found that E12.5 mesenchyme-free otocyst explants failed to undergo much vestibular and cochlear morphogenesis, while Miura et al. (2004) observed both semicircular canal formation and cochlear spiraling. This difference could be explained by differences in culturing method; Miura et al. (2004) used Matrigel for their experiments. Further confounding these grafting and explant experiments, other nearby structures such as the notochord, hindbrain, and ganglion cells are sources of developmental signals, in particular Sonic Hedgehog (SHH) and Wnt, for the developing inner ear (Riccomagno et al., 2002, 2005; Miura et al., 2004; Bok et al., 2007b, 2007a, 2013; Brown and Epstein, 2011; Tateya et al., 2013).

In terms of in vivo experiments, no M→E signal has been identified. However, studies have suggested that during semicircular canal morphogenesis, mechanical forces from the proliferation of mesenchymal cells are required to push epithelial walls together to form the fusion plate (Figs. 1B, 4B). This process, potentially regulated by Netrin 1 and FGF9 from the epithelia, is required for the initiation of duct formation (See Fig. 4A, Netrin 1 not shown; Pirvola et al., 2004; Salminen et al., 2000). Additionally, a few mesenchyme-expressed transcription factors have been found to affect otic epithelial development. Two of these, Prx1 and Prx2, are expressed in the POM surrounding the developing semicircular canals and are required for their outgrowth (ten Berge et al., 1998). Others include Tbx1, Pou3f4, Sox9, and Zic2, as detailed below. Notably, while deletion of each of these genes led to defects in growth, morphogenesis, or differentiation of the otic epithelia, as detailed below, the mechanism by which they affect epithelial development has not been elucidated.

Tbx1 is expressed in both the otic epithelium and the POM during development (Vitelli et al., 2003; Raft et al., 2004). Tbx1 constitutive KO mice exhibited a smaller otocyst and failure of sensory organ development (Jerome and Papaioannou, 2001; Vitelli et al., 2003), a phenotype recapitulated by otic epithelium-specific deletion of Tbx1 (Arnold et al., 2006; Xu et al., 2007b). Mesenchyme-specific deletion of Tbx1 resulted in a severely shortened cochlear duct with a lack of organ of Corti hair cells, except at the very base (Xu et al., 2007a; Braunstein et al., 2009). This cochlear phenotype was attributed to a decrease in progenitor proliferation in the epithelium at E11.5 and E12.5, similar to what was observed in mice lacking mesenchymal FGFR1/2 (Huh et al., 2015), as well as to increased cell death. Furthermore, the otic capsule medial to the cochlea was missing (Xu et al., 2007a; Braunstein et al., 2009), similar to what was observed in mice lacking mesenchymal βCat (Bohnenpoll et al., 2014).

Pou3f4 (also Brn4) is expressed in the condensing POM during development, and some knockout mice had a shortened cochlea with reduced coiling (Phippard et al., 1999). Interestingly, the penetrance of this cochlear phenotype was enhanced when a copy of Tbx1 was also lost (Pou3f4^−/−; Tbx1^+/−), which could be partially accounted for by increased apoptosis and decreased proliferation in the cochlear duct at E12.5 (Braunstein et al., 2008, 2009). This indicates a genetic interaction between Pou3f4 and Tbx1. In fact, mesenchymal Tbx1 has been shown to regulate Pou3f4 expression (Arnold et al., 2006; Braunstein et al., 2009). The M→E signals regulated by Pou3f4 and Tbx1 that, in turn, regulates cochlear growth have not been identified, although retinoic acid is a candidate (Braunstein et al., 2009; Monks and Morrow, 2012).

The otic epithelium has been hypothesized to regulate Pou3f4 and Tbx1 expression in the POM. The shortened cochlea phenotypes of Pou3f4 and Tbx1 mutants resemble that of Fgfr1/2^mes:null and βCat^mes:null mice. However, the expression of the two transcription factors were not affected in Fgf9/20-null mice (Huh et al., 2015). Interestingly, Pou3f4 expression in the POM was lost in βCat^mes:null mice at E18.5, which could partially account for the cochlear and otic capsule phenotype observed in these mice. Contrarily, Pou3f4 expression was not changed in βCat^mes:null mice at E12.5, which is when the proliferation and apoptosis phenotypes were observed in Pou3f4-null mice (Bohnenpoll et al., 2014). Experiments with Shh-null mice suggested that non-otic tissues regulate Tbx1 and Pou3f4 expression. In Shh-null mice, the expression of both Tbx1 and Pou3f4 was lost from the POM at E10.5, indicating that SHH signaling is upstream of the two genes (Riccomagno et al., 2002). At this stage, Shh is expressed in the notochord and hindbrain floor plate, but not in the otic epithelium. The entire inner ear was severely hypoplastic in Shh-null mice, however, so the loss of Tbx1 and Pou3f4 expression may have been secondary to otic epithelial defects (Riccomagno et al., 2002).

Sox9 is expressed in the ventral POM starting at E10.5, and is mostly restricted to the outer POM layer at E12.5 (Trowe et al., 2010). After E15.5, Sox9 was found in the spiral ligament. Mesenchyme-specific deletion of Sox9 resulted in decreased POM proliferation, a lack of otic capsule from the cochlear region, underdeveloped/lack of the spiral ligament, scala vestibuli, and scala tympani, and a shortened cochlea with normal differentiation (Trowe et al., 2010). These phenotypes are very similar to those of βCat^mes:null mice, except the spiral limbus appeared to be intact in Sox9-null mice, but not in βCat^mes:null mice. However, Sox9 expression was unchanged in βCat^mes:null mice at E12.5, and only decreased in the spiral ligament at E18.5, potentially accounting for the loss of the spiral ligament in βCat^mes:null mice (Bohnenpoll et al., 2014). It is not known if Sox9 may be acting upstream of or in parallel with Wnt/βCat signaling in other parts of the POM.

All five members of the Zic family of transcription factors are expressed in the POM starting as early as E9.5, but not in the otic epithelium (Chervenak et al., 2013). Zic2 is additionally highly expressed in the mesenchyme surrounding the growing cochlear duct at E12.5 and E13.5. Thus far, only Zic2-null mice have been found to have an inner ear development phenotype, although redundancy between Zic family members has not been thoroughly investigated. In Zic2-null mice, the inner ear was severely hypoplastic, with shortened semicircular canals and cochlea, and loss of the endolymphatic duct and sac (Chervenak et al., 2014). Currently, signaling pathways acting upstream and downstream of Zic2 during inner ear development have not been identified. Chervenak et al. (2014) hypothesized that the epithelial phenotypes in Zic2 null mutants could be attributable to displacement of the developing otocyst relative to the hindbrain, a source of signals that pattern the otocyst. This idea highlights the potential complexity of mesenchymal regulation of epithelial development, which may involve direct signaling via diffusible molecules, as well as mechanical forces and indirect regulation.

Main Olfactory System

Structure and Function

The olfactory epithelium (OE) is required for the sense of smell. The OE contains three main cell types: olfactory receptor neurons (ORNs), the nuclei of which are located in the middle layers of the OE and project ciliated membrane structures to the apical surface and axons to the olfactory bulb; sustentacular cells, a supporting cell population with nuclei located apically; and basal cells, a progenitor population that gives rise to both sustentacular cells and ORNs (reviewed in Schwob et al., 2017). The OE lines the interior of the posterior nasal cavity, called the olfactory recess, protected by the walls of the nasal capsule. Bony scrolls called turbinates project inward from the walls of the nasal capsule and provide increased surface area for the OE (Fig. 2D). Here, we use the term turbinate to refer exclusively to ethmoturbinates or “olfactory” turbinates, which are mainly lined by OE in the mouse. Turbinate size and complexity vary greatly among mammalian species. As mentioned in the introduction, OE surface area is hypothesized to be important to olfactory ability (Van Valkenburgh et al., 2014).

Figure 2. Schematic of main olfactory epithelium, posterior nasal capsule, septum, and turbinate development.

(A-D) Frontal sections through the posterior nasal cavity at E10.5 (A), E12.5 (B), E14.5 (C), and E17.5 (D). The olfactory epithelium (OE; purple) develops from the nasal pit (np). The nasal pit rim (light blue), where Fgf8 is expressed, has been shown to give rise to epidermal and respiratory epithelia, but not OE (Forni et al., 2013). The OE is encased within the posterior nasal capsule, which is separated into left and right halves by the nasal septum (S), and have interior projections called ethmoturbinates. A layer of mesenchyme (pink) separates the OE from the underlying nasal capsule, septum, and turbinates (red), which become cartilaginous at around E14. Condensing mesenchyme is represented by reddish-pink. There are six ethmoturbinates in mice, which are mainly lined by OE, but also respiratory epithelium (RE; light blue). From most anterodorsal to posteroventral, the turbinates are endoturbinate I (n1), ectoturbinate i (c1), endoturbinate II (n2), ectoturbinate ii (c2), endoturbinate III (n3), and endoturbinate IV (n4). D, dorsal; M, medial; fb, forebrain.

Development

Development of the olfactory system begins at around E9.5 with the thickening of the olfactory placode, which invaginates to form the olfactory (nasal) pit (reviewed in Sokpor et al., 2018). The epithelia lining the nasal pit develops into the OE, while the skeletogenic mesenchyme surrounding the pit develops into the nasal capsule and the nasal septum, which divides the left and right sides of the nasal cavity (see Fig. 2). This mesenchyme is derived from the neural crest (McBratney-Owen et al., 2008). Development of the olfactory turbinates has long been described as budding of the OE followed by proliferation and development of the underlying mesenchyme, adjacent to the developing nasal capsule, into turbinates (Dieulafé, 1906; Martineau-Doizé et al., 1992). Like the otic capsule, the posterior nasal capsule, septum, and turbinates (here referred to collectively as olfactory structures) develop by endochondral ossification.

The close association and correlated growth of the OE and underlying olfactory structures implies, even more enticingly than in the inner ear, the existence of intricate epithelial-mesenchymal crosstalk (Adameyko and Fried, 2016; Kaucka et al., 2018). Consistent with this idea, experiments in which the OE was blocked from forming or arrested at very early stages also led to malformed nasal cavity and surrounding craniofacial structures. This was observed in Fgf8-conditional null (Kawauchi et al., 2005), Foxg1-null (Duggan et al., 2008; Kawauchi et al., 2009), Six1/4-double null (Laclef et al., 2003; Kaucka et al., 2018), Dlx5/6-double null (Gitton et al., 2011), and Dicer1-conditional null (Kersigo et al., 2011) mice. Up until recently, however, an epithelial-mesenchymal signal regulating OE and olfactory structure development has remained elusive (Van Valkenburgh et al., 2014; Adameyko and Fried, 2016). Interestingly, it has been pointed out that while early disruptions to OE development affects nasal cavity development, later disruptions (i.e. to more differentiated progenitors) do not (Kaucka et al., 2018). For instance, deletion of Ascl1, a transcription factor required for neurogenesis, resulted in an OE devoid of immediate neuronal progenitors and olfactory neurons, but did not affect nasal capsule or turbinate morphology (Guillemot et al., 1993; Cau et al., 1997; Kaucka et al., 2018). This suggests that an early, less differentiated OE progenitor cell is the source of signals regulating the development of the olfactory structures.

Epithelial-to-mesenchymal signaling in the developing main olfactory system

Most studies of E→M interactions in the OE have focused on early development at the nasal pit stage. During these stages, FGF8 from the rim of the nasal pit is important for nasal pit invagination, and conditional inactivation of FGF8 in the OE resulted in failed development of the OE and olfactory structures (Kawauchi et al., 2005; Maier et al., 2010; Griffin et al., 2013). In the chick, explant experiments showed that FGF signaling is similarly required for early craniofacial morphogenesis (Abzhanov and Tabin, 2004; Szabo-Rogers et al., 2008), potentially via direct signaling to the early nasal mesenchyme (Firnberg and Neubüser, 2002). However, many factors could contribute to the disruption of olfactory structure development in Fgf8-null mice besides the loss of direct FGF8 signaling. These include malformation of other craniofacial structures affecting the development of olfactory structures, loss of the OE and any OE-derived signal required for the development of olfactory structures, and a combination of direct and such indirect effects. The early-onset and severity of the Fgf8-null craniofacial phenotype has made it difficult to tease out the exact mechanism leading to olfactory developmental defects. For instance, the OE phenotype in these mice has been attributed to apoptosis of OE progenitors (Kawauchi et al., 2005), inhibition of neuronal fate in OE progenitors (Tucker et al., 2010), loss of inhibition of respiratory epithelium fate (Maier et al., 2010), and loss of mesenchymal BMP4/Noggin signals required for OE development (Forni et al., 2013). Moreover, whether Fgf8 is even expressed in the OE or by olfactory progenitor cells has been called into question. Forni et al. (2013) showed that Fgf8 expression and lineage is restricted to the respiratory epithelium and epidermis. Therefore, OE signals that regulate olfactory structure development remained elusive.

To date, only two OE-expressed signaling factors have been proposed to regulate olfactory structure development through E→M signaling: SHH and FGF20 (see next paragraph). Shh expression was detected in regions of the developing OE as early as E11.5 and has been implicated in nasal capsule roof development (Kaucka et al., 2018). Despite the suggestive evidence from this study, OE-specific deletion of Shh is required before a definitive conclusion can be reached. Shh is also expressed by other tissues nearby olfactory structures, such as the brain; brain-specific deletion of Shh resulted in malformation of the posterior nasal capsule and the nasal septum (Kaucka et al., 2018).

Fgf20 is expressed in the embryonic and early postnatal OE, and its expression seems to correlate with turbinate development (Yang et al., 2018). At early stages, Fgf20 is expressed in the OE overlying sites of future turbinate formation; at later stages, Fgf20 is expressed in regions of negatively-curved OE formed by the growing turbinates (see Fig 4B). Deletion of Fgf20 led specifically to malformed turbinates, due to defects in formation of mesenchymal condensations at the initial stages of turbinate growth. Turbinates were not completely absent in Fgf20-null mice, however, implying that there are additional signals that may collaborate with FGF20. These signals could be other FGFs, many of which are expressed in and around the developing OE (Bachler and Neubüser, 2001; Kawauchi et al., 2004; Zhu et al., 2016).

Interestingly, lineage tracing suggested that Fgf20-expressing cells are multipotent OE progenitors that expand the OE, thereby tying OE expansion with turbinate growth (Yang et al., 2018). These cells are referred to as FEP (Fgf20-expressing, epithelium-spanning progenitor) cells. Furthermore, specific disruption of Wnt/βCat signaling in the OE showed that Wnt/βCat is required to maintain FEP cells in an undifferentiated state, as well as to regulate Fgf20 expression. Deletion of βCat in the OE, therefore, also resulted in malformed turbinates, but much more severely than deletion of Fgf20. This could be due to the loss of other Wnt/βCat directly-regulated E→M signals or to the loss of additional signals secreted by FEP cells, since these progenitors were not maintained in the absence of Wnt/βCat signaling (Yang et al., 2018).

Mesenchymal-to-epithelial signaling in the developing main olfactory system

Interestingly, the total OE surface area in Fgf20-null mice was decreased by roughly ~17% (Yang et al., 2018). This suggests that an M→E feedback signal regulates OE surface area, coordinating it with changes to turbinate size and shape. The identity this signal is currently not known, although candidates exist, such as Wnt ligands. Mesenchymal sources of Noggin, a BMP/TGF-β inhibitor, has been implicated in defining the boundary of the olfactory pit (Forni et al., 2013), and is another candidate. The role of Noggin in later stages of OE development (i.e. OE expansion) is not known. Other candidates include Follistatin and Activin, two epithelial signals that regulate OE differentiation but that are also expressed in the underlying mesenchyme (Kawauchi et al., 2009; Gokoffski et al., 2011). Mechanical forces from the mesenchyme and developing turbinates could also play a role in regulating OE surface area. Consistent with this, after E14.5, FEP cells reside in negatively-curved regions of the OE, which are formed by the developing turbinates (see Figs. 2D, 4D). This suggests that the growth of turbinates may play a role in maintaining the FEP cell niche.

To explore the potential for Noggin signaling and turbinate growth to affect OE development, we examined Fgf20^GFP-Cre expression as a marker of FEP cells in Noggin-null mice (Noggin^LacZ; McMahon et al., 1998). Using an antibody to β-galactosidase (LacZ), we detected Noggin expression exclusively in chondrocytes of the nasal capsule, septum, and turbinates at E17.5 (Figs. 3A, 3B). Compared to control mice (Noggin^LacZ/+; Fgf20^GFP-Cre/+), Noggin-null mice (Noggin^LacZ/LacZ; Fgf20^GFP-Cre/+) had slightly malformed turbinates with dramatic chondrocyte hyperplasia in the nasal capsule, but relatively normal appearing OE. Turbinate protrusion was not severely affected at E17.5, and therefore most of the negatively-curved OE regions were maintained, along with the corresponding FEP cell niche (Figs. 3A, 3B, arrowheads). This suggests that Noggin signaling is not directly required for OE development or the maintenance of the FEP cell niche, but rather acts tissue-intrinsically to regulate turbinate chondrogenesis. Interestingly, turbinate n2, which in control mice forms two branches with negatively-curved OE at the branch point, was not branched in Noggin-null mice. As a result, the branch-point negative curvature of the OE was absent, along with FEP cells (Figs. 3A, 3C, arrows). This result suggests that negative curvatures in the OE, and therefore compressive forces from turbinate development, are required to maintain the FEP cell niche. A more specific experiment altering turbinate shape and OE negative curvature formation will be required to support this conclusion. Nevertheless, this idea gives rise to an interesting model in which progenitor cells in the developing OE regulate turbinate development, the growth of which then exert mechanical forces to shape the OE and regulate OE progenitor cells (see Fig. 4B).

Figure 3. Turbinate protrusion shapes the OE to maintain OE progenitor cells.

Serial anterior (A) and posterior (B) frontal sections through the posterior nasal cavity of E17.5 control (Noggin^LacZ/+; Fgf20^GFP-Cre/+) and Noggin-null (Noggin^LacZ/LacZ; Fgf20^GFP-Cre/+) mice. (A, B) Noggin^LacZ expression (in chondrocytes) is identified by an antibody to β-galactosidase (LacZ; red); Fgf20^GFP-Cre expression in the olfactory epithelium (OE) is identified by an antibody to GFP (GFP-Cre; green); arrowheads indicate regions of negatively-curved OE formed by the turbinates, which act as niches for Fgf20-expressing OE progenitors (FEP cells). (C) Magnification of boxed region in (A), on an adjacent section; Sox2 expression in apical and basal OE nuclei and FEP cells is identified by an antibody (red); arrows indicate loss of negatively-curved OE and FEP cells (green) in Noggin-null mice. D, dorsal; M, medial; n2, endoturbinate II; n3, endoturbinate III; RE, respiratory epithelium; S, septum; Scale bars, 500 μm. Reference Fig. 2D for olfactory system schematic at E17.5. Method: serial, frontal, frozen sections of the E17.5 nasal cavity were immunostained with rabbit anti-GFP and chick anti-β-galactosidase antibodies, followed by appropriate secondary antibodies, and imaged with a Zeiss LSM 700 confocal microscope (see Yang et al., 2018).

Conclusions and Future Directions

The hypothesis that neurosensory epithelia and the surrounding skull communicate during development is widely held. The close juxtaposition of sensory epithelia and parts of the skull strongly hints at this. Moreover, evolutionary factors acting on such crosstalk could provide an elegant mechanism accounting for the diversity of neurosensory organ and skull morphologies across vertebrates. While many studies over the past few decades have proposed the existence of intricate developmental epithelial-mesenchymal interactions in both the inner ear and the olfactory system, the identities of these signals have just recently begun to be uncovered. Furthermore, the idea of mechanical forces contributing to epithelial-mesenchymal interactions has been suggested, but not addressed experimentally. Additionally, all studies discussed in this review have focused on stages of chondrogenesis in otic and nasal capsule development. It is not known whether the inner ear and the OE regulate osteogenesis.

In the inner ear, only FGF signaling has been identified as an E→M signal (see Fig. 4A), but it is unclear whether the otic capsule defect in Fgf9-null mice is due directly to the loss of FGF9 signaling. Wnt/βCat signaling has also been heavily implicated, but the otic epithelium has not been definitively shown to be a source of Wnt ligands that signal to the mesenchyme. No M→E signals have been identified in vivo. However, the importance of mechanical forces resulting from mesenchymal proliferation was demonstrated in semicircular canal development (Salminen et al., 2000; Pirvola et al., 2004). The POM, in addition, is critical for sensory hair cell differentiation in the cochlear epithelium (Xu et al., 2007a; Braunstein et al., 2009). Identifying M→E signals, therefore, may potentially be important for therapeutic regeneration of lost or injured sensory hair cells.

In the OE, FGF20 is the first identified E→M signal that regulates turbinate development (see Fig. 4B). However, other unidentified signals likely exist that function similarly or redundantly with FGF20. Additional Wnt/βCat-dependent or FEP-expressed signals have yet to be found, and may be a good starting point in the search for other E→M signals. The identification of such signals has partly been hindered by the lack of a specific genetic tool to target the OE. The identification of Fgf20^GFP-Cre expression in FEP cells should allow gene deletion widely and specifically in the OE (see Yang et al., 2018). Additionally, we present here, for the first time, in vivo evidence suggesting that turbinate growth affects OE development, potentially by regulating an OE progenitor niche. In future studies, this idea of mechanical forces from turbinate growth deserves more attention as a mechanism for regulating OE development.