The functions of the vertebrate inner ear are to maintain balance and detect sound. The vestibule of the inner ear, responsible for maintaining balance, comprises three sensory cristae and their associated semicircular canals, as well as maculae of the utricle and saccule. The maculae detect linear acceleration, whereas the three cristae and their associated semicircular canals sense angular head movements. Fluid displacements within the semicircular canals as a result of angular movements can lead to stimulation of sensory hair cells within the associated cristae. Thus, malformation of any of the three cristae or canals could result in balance problems.
Increasing evidence supports the hypothesis that the crista induces the formation of its associated non-sensory structure, the semicircular canal (Cantos et al., 2000). First, transplantation experiments in chicken suggest that specification of sensory cristae develops before the semicircular canals (Wu et al., 1998). Second, genes such as Bone morphogenetic protein-4 (Bmp4) and Fibroblast growth factor 10 (Fgf10) are primarily expressed in the presumptive cristae, yet engineered deletions of these genes affect cristae as well as canal formation (Pauley et al., 2003; Chang et al., 2008). Third, most existing inner ear mutants of mouse and zebrafish show malformations that involve both cristae and canals, or canals alone (Kiernan et al., 2005; Omata et al., 2007; Hammond et al., 2009). While all of the aforementioned lines of evidence are consistent with the hypothesis of a prospective crista inducing formation of an associated canal, there are a few exceptions in which mouse mutants are reported to have a normal canal in the absence of its crista/ampulla. For example, both Foxg1−/− and one of the reported Hmx3 mutant strains have a lateral semicircular canal in the absence of the lateral ampulla (Wang et al., 1998; Pauley et al., 2006). Whether the presumptive lateral crista is initially present in these mutants long enough to induce lateral canal formation is not clear.
Foxg1 is a winged helix transcription factor that is expressed in the telencephalon, eye, foregut, and otic placode of mouse embryos (Xuan et al., 1995; Hebert and McConnell, 2000). Foxg1 knockout mice show hypoplasia of the telencephalon, abnormal eye and ear development, and they die shortly after birth (Xuan et al., 1995; Huh et al., 1999; Pauley et al., 2006). In humans, mutations of FOXG1 are associated with a congenital variant form of Rett syndrome, characterized by severe neural developmental problems, cognitive impairment, and growth retardation (Ariani et al., 2008).
Inner ears from Foxg1−/− mice show a shortened cochlear duct with multiple rows of sensory hair cells (Pauley et al., 2006). There is aberrant axonal growth and targeting of vestibular neurons. In addition, the lateral ampulla is missing and both anterior and lateral canals are connected to the anterior ampulla. In this study, we focus on the crista and canal phenotypes of Foxg1−/− mutants. Using gene expression analyses, we investigated whether the presumptive lateral crista is ever present in Foxg1−/− inner ears. Our data show that a presumptive lateral crista is initially present in Foxg1−/− mutants at a developmental time that is early enough for lateral canal induction. Thus, the phenotype of Foxg1−/− mutants does not negate the canal genesis hypothesis (Chang et al., 2004). Furthermore, we provide genetic fate mapping evidence that Foxg1 is required for proper separation of the prospective anterior and lateral cristae.
Crista Phenotypes in Foxg1−/− Mouse Embryos
Analyses of paint-filled Foxg1−/− inner ears show a consistent phenotype of a single, anterior-positioned ampulla connecting to the anterior and lateral canals as previously reported (Fig. 1, arrow, n=36/36 ears; Pauley et al., 2006). The cochlear duct is shortened with only a single coil and an incomplete separation from the saccule. In addition to the reported phenotypes, some mutant embryos show a failure of posterior semicircular canal resorption (Fig. 1C, asterisk, n=6/36 ears, 16.7%, Table 1). None of these phenotypes are detected in the Foxg1 heterozygotes. Within the anterior ampulla of Foxg1−/− inner ear, a single crista is present in most cases (Fig. 2B, B', CA; n=22/24 ears, 91.7%), but occasionally two separate cristae can be found (Fig. 2C, C', CA1, CA2; n=2/24 ears, 8.3%). Normal anterior and posterior cristae have a non-sensory region, septum cruciatum, which divides the organ into two equal parts (Fig. 2D, G, Cr). However, the cruciatum within the mutant anterior crista is eccentrically located and does not bisect the sensory crista completely (Fig. 2E, F). The posterior crista shows similar malformed cruciatum (Fig. 2H, I). Both anterior and posterior cristae are misshapen and smaller in size, compared to wildtype (Fig. 2J). The average length of mutant anterior cristae is 31% less than wildtype (wildtype: 435.5 ± 9.6 μm (mean ± SEM), n=7; mutant: 302.0 ± 15.0 μm, n=7; P < 0.001; see Supp. Table S1, which is available online). In addition, the average length of mutant posterior cristae is reduced by 46% (wildtype: 413.0 ± 9.6 μm, n=6; mutant: 223.1 ± 17.3 μm, n=7; P < 0.001; Table S1).
Table 1. Presence of the Presumptive Lateral Crista in Foxg1−/− Inner Eara
Presence of LC (n)
Separated entity of LC (n)
Presence of LC (n)
Separated entity of LC (n)
Lateral cristae at E11.5 and E12.0 were confirmed with overlapping expression of Bmp4 and Otx1 expressions. n = number of ears analyzed.
Foxg1 Is Expressed in the Three Presumptive Cristae
Presumptive anterior and lateral cristae are thought to arise from a Bmp4-positive stripe located in the anterior otocyst (Morsli et al., 1998). Starting at E11.5, this stripe begins to separate to form the two cristae (Fig. 3A–C, compare the weak Bmp4 region with arrowheads in B to A and C). Comparisons of Bmp4 and lacZ expression patterns in Foxg1 heterozygous embryos indicate that Foxg1 is clearly expressed in the presumptive anterior and posterior cristae (Fig. 3A', B'; data not shown). Its expression in the lateral crista, on the other hand, is much weaker (Fig. 3C', bracket). Nevertheless, the region between the two cristae that are separating, which shows weaker expression of Bmp4, is clearly positive for Foxg1 (Fig. 3B, B', arrowheads). Similar results are observed by probing for Foxg1 transcripts in wildtype specimens (data not shown).
Presumptive Lateral Crista Is Initially Present in Foxg1−/− Inner Ears
At E11.5, there is no obvious abnormality in the presumptive anterior and lateral cristae of Foxg1−/− mutants, based on the expression pattern of Bmp4 (Fig. 3D, F), and the two presumptive cristae appear to be separating (Fig. 3D–F, weaker Bmp4 expression in E than D and F). To further verify the presence of the lateral crista component within the Bmp4 domain, we used Otx1. Otx1 is expressed in the prospective lateral crista and its canal between E10.5 to E13.5 (Morsli et al., 1999). Thus, we define the region of overlap between Bmp4 and Otx1 as the prospective lateral crista, where the epithelium is also thicker (Fig. 4C, C', bracket). Such a region of overlap between Bmp4 and Otx1 is also evident in Foxg1−/− mutants, suggesting that the prospective lateral crista is present at this stage (Fig. 4F, F', bracket, Table 1).
Disappearance of the Presumptive Lateral Crista in Foxg1−/− Inner Ears at E12
By E12, more than half of the wildtype specimens show the anterior and lateral cristae as separate entities (Fig. 5A–C, n=12/20 ears, 60%; Table 1). Otx1 expression remained overlapped with Bmp4 in the prospective lateral crista (Fig. 5C', short bracket) and Otx1 is also expressed in the prospective lateral canal (Fig. 5C', long bracket). Most of the Foxg1 mutant specimens no longer show an overlap in expression domain between Bmp4 and Otx1 (Fig. 5D–F'; asterisk, n=15/20 ears; Table 1), even though Otx1 expression associated with the lateral canal is present. In 25% of the specimens, a reduced presumptive lateral crista is present expressing both Bmp4 and Otx1 (Fig. 5G–I', bracket, n=5/20 ears; Table 1).
Lateral Canal Formation in Foxg1−/− Mutants
The precise timing of canal specification in the mouse inner ear is not known, but the two canal pouches that give rise to the three canals are anatomically distinct by E11.5. Most of the genes that are known to be required for canal formation are already expressed by this age, and some such as Dlx5 and Lmo4 are initiated as early as the otic placode stage (Acampora et al., 1999; Merlo et al., 2002; Deng et al., 2006; Lin Gan, unpublished results). On the other hand, the expression of Bmp2 is only detectable in the canal pouch starting around E11.5. Previous studies in chicken have proposed that Bmp2 expression in the canal pouch is induced by signals in the presumptive cristae, possibly Fgfs (Chang et al., 2004). In support of the hypothesis, lack of Bmp2 affects formation of all three canals in both zebrafish and mice (Hammond et al., 2009; Hwang and Wu, unpublished results). In Foxg1−/− inner ears, the expression of Dlx5, Lmo4, and Bmp2 in the lateral canal pouch are all initiated by E11.5, similar to the wildtype (Fig. 6). Thus, these results are consistent with the idea that the program for the lateral canal formation is well underway before the disappearance of the prospective lateral crista at E12.
Anterior and Lateral Cristae Fail to Separate in Foxg1−/− Mutants
Thus far, our results suggest that the lateral crista is present initially in Foxg1−/− inner ears but defects appear during the time of separation from the anterior crista. While some of the specimens showed a separated lateral crista at E12, many of them failed to be maintained by E18.5 (Table 1). These phenotypes could all be attributed to an abnormal crista separation. However, programmed cell death analyses of the region between the presumptive anterior and lateral crista did not reveal anything abnormal (data not shown). This is a region with high numbers of apoptotic cells normally, which may obscure detection of additional cell death, if it exists, due to abnormal crista separation. In addition, in the course of our study, we identified several genes such as Dlx5, Lmo4, Hey1, and Emx2, which are expressed in this region of crista separation, similar to what is shown for Foxg1 (Fig. 3B'). None of these genes show any obvious difference between wildtype and Foxg1−/− inner ears during these stages (data not shown).
To further investigate the possibility of abnormal crista separation, we genetically fate-mapped cells of the lateral crista in the Foxg1−/− inner ears. We utilized an Otx1-cre+/− mouse strain, in which cre expression is under the control of Otx1 promoter, and we analyzed YFP expression in Foxg1−/−; Otx1-cre+/−; RosaEYFP/+ compound mutants at E18.5. Similar shortened cristae in these compound mutants are observed as in Foxg1−/− ears on the C57BL6 background (Table S1). In Foxg1+/−; Otx1-cre+/−;RosaEYFP/+ embryos, robust YFP signals are located in the lateral crista and canal but not the anterior crista (Fig. 7A). This finding is consistent with the expression pattern of Otx1 between E11.5 to E13.5 (Morsli et al., 1999). In Foxg1−/− inner ears, three types of mutant cristae were found (Fig. 7B–D). The first type is a small single crista, which does not show any YFP signal suggesting the lateral crista has completely disappeared leaving a dysmorphic anterior crista (Fig. 7B, n=2/6 ears; Table S1). The second type is two separated malformed cristae showing YFP signal only in the lateral but not the anterior crista (Fig. 7C, n=1/6 ear; Table S1). The third type is most common showing a shortened crista with YFP expression in the lateral region, indicating that this type of crista represents a fused anterior and lateral crista (Fig. 7D, n=3/6 ears, Table S1). Magnified views of the YFP-positive region reveal YFP expression is located within the sensory region and includes sensory hair cells (Fig. 7E,F). The smallest of the three cristae containing lateral crista components (277.2 μm; Table S1) is shorter than the mean length of Foxg1−/− cristae (302 μm). Extrapolating from this limited number of specimens, our results suggest the likelihood that more than 50% of the single crista observed in Foxg1−/− inner ears represent fused anterior and lateral cristae.
Requirement of Foxg1 in Sensory Crista Formation
Our results indicate that while a small percentage of Foxg1−/− inner ears show two small separated anterior and lateral cristae, a majority of them have a single crista that is likely fused (Fig. 8). We attribute most of the crista phenotype in Foxg1−/− mutants to a failure of crista to separate. In vertebrate inner ears, splitting a common sensory domain to form separate sensory organs could be an important mechanism to generate diversity and accommodate functional requirements of a given species. This notion is based on the variable number of sensory organs found among vertebrate inner ears, ranging from a total number of five sensory organs in fish, to six in mice, and eight in chicken. Very little is known about the molecular mechanisms underlying this cellular process. In principle, proper splitting of a sensory patch could involve intrinsic signaling within the prosensory tissue converting a portion of it to a non-sensory fate, as well as extrinsic signaling, which pulls a sensory patch apart. Evidence for the latter mechanism is provided by a number of knockout mice such as Otx1 and Hmx3, in which the two maculae are either fused or sharing a single chamber (Acampora et al., 1996; Wang et al., 1998; Morsli et al., 1999). Both Otx1 and Hmx3 encode transcription factors that are expressed exclusively in non-sensory tissues adjacent to the macular primordium. Here, we show for the first time that a transcription factor, Foxg1, expressed within a crista primordium is required for the proper separation of the anterior and lateral cristae. In the absence of Foxg1, anterior and lateral cristae are often fused.
Furthermore, the abnormal formation of the septum cruciatum in the anterior and posterior cristae of Foxg1 mutants could also be a result of improper separation between sensory and non-sensory components. During septum cruciatum formation in chicken, sensory genes that appear to be uniformly expressed in the prospective sensory crista such as Bmp4 and Fgf10 are observed to segregate from the center region of the sensory patch (Chang et al., 2008).
Simultaneously, genes that are eventually associated with the non-sensory region such as p75Ngfr and Gata3 show a reciprocal relationship and start to increase their levels of expression in the center. These expression patterns eventually lead to the formation of a ridge of non-sensory tissue, which bisects the sensory crista. Thus, the formation of the septum cruciatum may be mechanistically similar to the separation process between anterior and lateral cristae: a patch of non-sensory tissue arising within a prosensory domain. Interestingly, both of these processes are affected in Foxg1−/− inner ears, which support a role of Foxg1 in defining sensory and non-sensory domains in this organ. In addition to the abnormal cruciatum formation, the size of the anterior and posterior cristae is smaller. Whether the smaller size of the crista is related to the malformed cruciatum is not clear.
Molecular Interactions of Foxg1 in the Inner Ear
Foxg1 is required for the proper formation of several organs and tissues such as the telencephalon, eye, and olfactory epithelium (Duggan et al., 2008; Kawauchi et al., 2009). In the telencephalon and olfactory epithelium, Foxg1 is thought to be required cell autonomously and has been implicated in the regulation of Fgfs and Tgf-β signaling pathways. In the telencephalon, Foxg1 and Fgf8 are thought to positively regulate each other (Hebert and Fishell, 2008). In the absence of Foxg1, Fgf8 expression is down-regulated in the ventral telencephalon (Martynoga et al., 2005). Other studies show that Fgf8 may be upstream of Foxg1 (Shimamura and Rubenstein, 1997; Storm et al., 2006). In contrast, Foxg1 is thought to negatively regulate the Tgf-β pathway in both the dorsal telencephalon and olfactory epithelium. In the cerebral cortex of Foxg1−/− mutants, Bmp signaling is increased based on the expression pattern of Bmps and phospho-smads (Dou et al., 2000; Martynoga et al., 2005). In addition, the loss of olfactory epithelium formation in Foxg1−/− mutants is rescued in Foxg1 and Gdf11 (Growth differentiation factor 11) double mutants (Kawauchi et al., 2009). While these results are interesting, it is not clear if Foxg1 regulates any of these pathways directly. In the olfactory epithelium, it has been suggested that Foxg1 inhibits Gdf11 activities by upregulating the expression of one of the Gdf11 antagonists, Follistatin, in both the olfactory epithelium and surrounding mesenchyme (Kawauchi et al., 2009). Even though both Fgfs and Bmps are important for crista formation in the inner ear (Pauley et al., 2003; Chang et al., 2008), there is no direct evidence that Foxg1 regulates anterior and lateral crista separation via the regulation of either Fgfs or Bmps. Inner ears of Fgf10−/− lack all three canals as well as the posterior crista (Pauley et al., 2003). There is no rescue of the lateral crista phenotype in Foxg1−/−; Fgf10+/− inner ears (Pauley et al., 2006), suggesting that Foxg1 does not mediate anterior and lateral crista separation by negatively regulating Fgf10. In contrast, the posterior ampulla and canal are missing in these compound mutants. Here, we show that a small percentage of Foxg1−/− mutants show non-resorption of the posterior canal as well. Thus, the more severe posterior vestibular phenotype observed in Foxg1−/−; Foxg1+/− inner ears suggests that Fgf10 could function in parallel or upstream of Foxg1 in mediating posterior canal and/or crista formation.
An emerging function of Foxg1 from studies in brain and olfactory epithelium appears to be maintaining cells in a progenitor pool and repressing specific differentiated fates. Experimental evidence suggests that the anti-differentiation role of Foxg1 is dependent on its DNA-binding domain, whereas its proliferation function is not (Dou et al., 2000; Hanashima et al., 2002). In the dorsal telencephalon, Foxg1 functions to suppress early cortical cell fate. In Foxg1−/− brain, an excess number of early cortical neurons, Cajal-Retzius, are produced (Hanashima et al., 2004, 2007). In the olfactory epithelium, Foxg1 promotes neurogenesis and inhibits Gdf11's activity in neuronal differentiation (Kawauchi et al., 2009). Furthermore, in the retina, Foxg1 functions to suppress temporal ganglion traits in nasal retinal ganglion neurons and promote contralateral projections at the optic chiasm (Pratt et al., 2004; Tian et al., 2008). Thus, the lack of Foxg1 causes an increase in ipsilateral projections of nasal ganglion neurons at the optic chiasm. Extrapolating from these results, Foxg1 could be maintaining a pool of progenitor cells in the inner ear. A reduction in progenitor cells could affect the overall size of the inner ear and/or the size of the sensory organs. Indeed, both the cristae and the membranous labyrinth are smaller in size in Foxg1−/− mutants. Simultaneously, Foxg1 could also function to suppress the sensory fate in the inner ear. In the absence of Foxg1, there is a loss of normal restriction on sensory fate causing improper sensory patch separation and malformed septum cruciata in the cristae, which could also be attributing to the abnormal size and morphogenesis.
Induction of Canal Formation by Presumptive Cristae
We hypothesized that presumptive cristae induce the formation of the associated non-sensory structure, the semicircular canals (Cantos et al., 2000). This hypothesis is supported by a number of mutant phenotypes (Pauley et al., 2003; Kiernan et al., 2005; Chang et al., 2008). In addition, based on studies using gain- and loss-of-function approaches as well as DiI fate mapping in ovo, it was proposed that Fgfs expressed in the presumptive crista induce a canal genesis zone in the canal pouch, possibly by up-regulating Bmp2 (Chang et al., 2004). Despite the accumulating evidence, the canal genesis hypothesis required further validation as existing inner ear phenotypes including those in Foxg1−/− mutants directly challenge this hypothesis. Here, we show that at the time when canal pouches are morphologically evident and Bmp2 is expressed in Foxg1−/− inner ears, the presumptive lateral crista is present as well. Therefore, the lateral canal could be induced before the subsequent demise of the lateral crista. A similar scenario could be taking place in one of the Hmx3 knockout strains, reported to have a lateral canal without its ampulla (Wang et al., 1998). Interestingly, in another Hmx3−/− strain generated by Hadrys et al. (1998), both the lateral canal and crista are affected suggesting that the presence of lateral canal without its ampulla could be a milder consequence of lack of Hmx3 functions. Until further gene expression analyses of the Hmx3 mutants are performed, there are no other existing data that directly argue against the hypothesis of presumptive cristae inducing semicircular canal formation.
Mice and Genotyping
The Foxg1+/tm1M (Fog1+/−) mice have a lacZ cassette knocked into the Foxg1 locus on a 129S1/Sv * C57BL/6J (B6.129S1/SV) background (Xuan et al., 1995). Foxg1−/− embryos were generated by timed mating of Foxg1+/− mice on a C57BL/6J background. The Foxg1+/− and Otx1tm4(cre)Asim/+(Otx1-cre+/−) mice were bred to generate Foxg1+/−; Otx1-cre+/− compound mutants and they were maintained through sibling matings (Puelles et al., 2003). Gt(ROSA)26Sortm1(EYFP)Cos (RosaEYFP/+) mice were purchased from Jackson Lab. Foxg1−/−; Otx1-cre+/−; RosaEYFP/+ embryos were generated by timed matings of Foxg1+/−; Otx1-cre+/− and Foxg1+/−; RosaEYFP /EYFP mice.
Immunohistochemistry and Length Measurement of Cristae
Foxg1−/− and Foxg1−/−; Otx1-cre+/−; RosaEYFP/+ embryos at embryonic day 18.5 (E18.5) were used for the whole mount immunohistochemistry staining. Specimens were fixed with 4% paraformaldehyde overnight at 4°C before further dissection and labeling. Tissues were incubated with FITC-conjugated anti-GFP (1:200; GeneTex®) overnight at 4°C, followed by incubation with rhodamine conjugated phalloidin for 1 hr at room temperature (1:100; Chemicon®). Specimens were mounted with DAPI containing anti-fade agent (Invitrogen®) before examination under a confocal microscope (LSM710, Carl Zeiss®). The sizes of cristae were measured from 100× confocal images using Zen 2008 software (Carl Zeiss®).
In Situ Hybridization
Foxg1−/− embryos between E11.25 and E12 were harvested for in situ hybridization analyses. RNA probes for Bmp4 (Morsli et al., 1998), Otx1 (Simeone et al., 1993), Dlx5 (Depew et al., 1999), Lmo4 (Deng et al., 2006), Bmp2 (McMahon et al., 1998), and lacZ (Bok et al., 2007), were prepared as previously described (Morsli et al., 1998).
We thank Drs. Thomas Friedman and Matthew Kelley for critical reading of the manuscript, and Ms. Lydia Lui and Michael Mulheisen for technical assistance. We also thank Drs. Richard Harland and John Rubenstein for providing plasmids for generating RNA probes. This work was supported by the NIDCD intramural program (D.K.W.); the Italian Association for Cancer Research (AIRC) and the “Stem Cell Project” of Fondazione Roma (A.S.), as well as NIH grant HD29584 (E.L.).