The mammalian inner ear is a complex structure including specialized vestibular sensory organs to detect head position and movements and a coiled cochlea for hearing. Inner ear is derived from the cranial ectoderm, which thickens and forms the otic placode adjacent to the prospective hindbrain. Subsequently, otic placode invaginates into the otic cup, which then deepens until it closes and pinches off from the ectoderm giving rise to an epithelial otic vesicle. Series of morphological changes convert the simple otic vesicle into a three-dimensional (3D) structure harboring three semicircular ducts, utricle, saccule, cochlear duct, endolymphatic duct and sac, and neurons of the eighth cranial nerve ganglion. The head mesenchyme surrounding the developing epithelial membranous labyrinth condenses to form the bony capsule that closely follows the forms of the epithelium (Torres and Giraldez, 1998; Noramly and Grainger, 2002). The inner ear contains two closed fluid compartments that play important roles in the process of sound perception and they also respond to the mechanical changes associated with body position and acceleration (Beitz et al., 2003). The lumen of the labyrinth is filled with endolymph produced by the marginal cells of the stria vascularis in the cochlea and by the secretory dark cells that lay adjacent to the utricular and semicircular duct sensory epithelia (Wangemann, 1995; Torres and Giraldez, 1998; Ciuman, 2009). Between the bony capsule and the membranous labyrinth is a space containing perilymph, which origin has remained unclear although it resembles cerebrospinal fluid and appears to be maintained locally by transport or diffusion of solutes from blood plasma (Salt, 2005).
Gata family of zinc finger domain transcription factors in vertebrates includes six members, Gata1–6, which all bind to a consensus “A/T-GATA-A/G” DNA motif. Gata factors are important for the development of several organs controlling proliferation, differentiation and movement as well as cell-fate specification (Patient and McGhee, 2002). Here we have focused on Gata2, a factor known to be particularly important for the control of hematopoietic precursor cell fate (Tsai and Orkin, 1997; Persons et al., 1999). Targeted inactivation of Gata2 results in lethality due to severe anemia and the mouse embryos die around embryonic day (E) 10.5 (Tsai et al., 1994).
Gata2 is expressed in the developing inner ear of mouse and chicken together with another family member, Gata3 (Lilleväli et al., 2004, 2006). The expression of Gata2 starts in mouse otic vesicle at E9.5, while Gata3 expression is activated already in the otic placode (Lilleväli et al., 2004, 2006). The expression of these two family members is highly overlapping in the early otic vesicle, but becomes more diverse at later developmental stages. While Gata2 expression becomes more prominent in the vestibular nonsensory epithelium, Gata3 is restricted to the sensory areas and with a particularly strong expression in the cochlea (Lilleväli et al., 2004).
There is a regulatory relationship between the two factors both in the nervous system (Nardelli et al., 1999; Pata et al., 1999) and in the inner ear (Lilleväli et al., 2004). While in brain, Gata2 is required to activate Gata3 expression (Nardelli et al., 1999; Pata et al., 1999), in inner ear Gata3 is upstream of Gata2 at least at early stages (Lilleväli et al., 2004). In addition, the two factors have high affinity to similar genomic DNA binding sites (Ko and Engel, 1993; Merika and Orkin, 1993) and, therefore, may have redundant functions and common target genes. This hypothesis is in line with the finding that the loss of Gata2 gene does not cause any inner ear defects before lethality at E10.5 most likely because Gata3 is expressed in overlapping regions of the otic vesicle at the early stages (Lilleväli et al., 2004). Instead, inactivation of Gata3 leads to a very early inner ear defects that are evident before Gata2 is expressed in the otic region (Lilleväli et al., 2006).
To investigate the roles of Gata2 later in inner ear development, we used a conditional mutagenesis approach to inactivate mouse Gata2 in the otic region using the Cre-loxP system. No defects in the early inner ear morphogenesis, in the otic sensory development or in the cochlear duct could be observed. Instead, a severe growth defect in all three semicircular ducts was detected from E14.5 onward. In addition to the epithelial defect, the mesenchymal cell removal from the perilymphatic space was inefficient.
Generation of the Gata2 Conditional Allele
Our previous studies have shown that Gata2−/− null mutant mice do not exhibit any inner ear defects before death at E10.5 (Lilleväli et al., 2004). To analyze otic development at later stages, we created a conditional allele of Gata2 that could be inactivated by Cre-mediated recombination. The Gata2 gene consists of two noncoding exons (IS and IG in Fig. 1A) and five coding exons (II–VI in Fig. 1A) and spans roughly 13 kb in mouse chromosome 6. To build a targeting vector we inserted a loxP site upstream of exon IG and two loxP sites downstream of exon III flanking the Neomycin resistance gene (Neo) for positive selection of homologous recombination events. For negative selection, a PGK-TK cassette was added downstream of exon IV (Fig. 1A).
The linearized Gata2 targeting construct was introduced into embryonic stem (ES) cells and selected clones were tested for correct targeting by Southern blot analysis (Fig. 1B). One targeted ES cell clone gave rise to chimeric animals, which transmitted the conditional allele (Gata2c) through the germline (Fig. 1C). Further breeding showed that the targeted allele was transmitted at the expected Mendelian frequency and that the homozygous Gata2c/c mice were viable and fertile, and did not exhibit any obvious abnormalities (data not shown).
Analysis of Foxg1-Cre Mediated Recombination
To verify Cre recombination activity in inner ear, the Foxg1-Cre mice (Hébert and McConnell, 2000) were crossed to the Rosa26 reporter (R26R) line that expresses β-galactosidase (β-gal) in response to Cre recombination (Soriano, 1999). The Foxg1-Cre mice have previously been reported to drive recombination in otic vesicle (Hébert and McConnell, 2000; Pirvola et al., 2002). The resulting Foxg1-Cre;R26R embryos and extraembryonic membranes were analyzed for β-gal activity with X-gal staining. Unexpectedly, rather widespread recombination of the R26R allele was observed in our Foxg1-Cre;R26R embryos. At E9.5, the Foxg1-Cre-driven recombination had occurred in most cells of the embryo including the otic vesicle epithelium and the surrounding mesenchyme (Fig. 2A,B). In the yolk sac, the Cre recombination efficiency was lower and approximately 50% of the cells were stained with X-gal (Fig. 2C). In control R26R embryos, no β-gal activity was detected (Fig. 2D–F).
Inactivation of Gata2 With Foxg1-Cre Mediated Recombination
To gain insight into the requirement of Gata2 in otic epithelium and surrounding mesenchyme, we mated Gata2c/c mice with Foxg1-Cre mouse line. The expected genomic deletion after this mating removes the IG noncoding exon and the two first coding exons (II and III) including the translation initiation codon. The resulting floxed allele (Gata2fl) was distinguished in genomic DNA by polymerase chain reaction (PCR) reactions with primers c and d (Fig. 1A), which gave rise to a 1.6-kb fragment (Fig. 1D). The Cre transgene was detected with a Cre-specific PCR (Fig. 1D).
Despite the broad pattern of recombination occurring with the Foxg1-Cre mice, the Foxg1-Cre;Gata2fl/fl embryos (hereafter referred to as Gata2cko) survived up to birth most likely due to only a partial recombination in the yolk sac, thus overcoming the early hematopoietic lethal phenotype (Tsai et al., 1994). At E18.5, Gata2cko embryos had, however, widespread edema (Fig. 3C,D), while at E14.5–E16.5 the embryos had often local hemorrhages in the body that were not detected in littermate controls (Fig. 3A,B, and data not shown). These defects could be related to the lack of Gata2 in endothelial cells, where it is normally expressed during embryogenesis (Tsai et al., 1994).
A weak Gata2 expression is first detected in mouse inner ear at E9.5 (Lilleväli et al., 2004). At E10.5, Gata2 was expressed in the dorsolateral epithelium of the wild-type otic vesicle (Fig. 3E, arrowhead in Fig. 3F) as well as in a small compartment of the adjacent ventral hindbrain (arrow in Fig. 3F). No Gata2 expression was observed in the Gata2cko embryos at E10.5 (Fig. 3I,J). In wild-type E14.5 embryos, Gata2 expression was confined to the nonsensory epithelium of the vestibule (Fig. 3G) and the cochlea (arrow in Fig. 3H) as well as to the mesenchymal cells next to the developing vestibular structures (Fig. 3G), whereas no Gata2 mRNA could be detected in corresponding areas of the Gata2cko ears (Fig. 3K,L). To verify the absence of Gata2 protein in Gata2cko ears we stained sections from E12.5–E18.5 embryos (n = 10) with anti-Gata2 antibody. While a strong Gata2 staining could be observed in the nuclei of the vestibular epithelium and the surrounding mesenchyme of the control embryos, no Gata2 protein could be detected in the corresponding areas of the mutant embryos (Fig. 3M–P, and data not shown). These results confirmed that Foxg1-driven Cre recombination had inactivated Gata2 from the inner ear epithelium and the surrounding mesenchyme by E10.5, thus, prior any phenotype could be observed in the Gata2−/− null mutant embryos.
Defects in the Otic Epithelium in Gata2cko Embryos
Analysis of inner ear morphology in Gata2cko embryos did not reveal any major defects at E10.5, E12.5, or E14.5 (n = 3–4 embryos/stage/genotype, Figs. 3E–L, 4A–D, and data not shown). However, at E15.5, we noticed that all three semicircular ducts were slightly smaller in diameter in the mutant embryos than in control littermates (n = 2 embryos/genotype, Figs. 6E–G,I–K, 7A–D, and data not shown). At E16.5 and E18.5 the diametrical size difference in the semicircular ducts was more pronounced between controls and mutants (n = 5 embryos/stage/genotype, Fig. 4E–H,M,N, dashed line indicates cross diameter in Fig. 4G,H) while no size difference was observed in other vestibular compartments or in the cochlear duct (Fig. 4I–L,O,P). Also no defects in the vestibular sensory organs could be detected in Gata2cko embryos at E14.5–E18.5 (Figs. 3G,H,K,L, 4I–L,O,P, 6H,L and data not shown).
To analyze the semicircular duct phenotype in more detail, we counted the number of epithelial cells in the three ducts per diametrical cross-section and compared the numbers between control and Gata2cko ears at stages E14.5, E15.5, and E16.5 (n = 2 embryos [(or n = 4 ears] /stage/genotype, 20 sections/ear. This analysis revealed a small but significant reduction in cell numbers in Gata2cko ducts already at E14.5 (Fig. 4R). At E15.5 and E16.5, the number of cells in semicircular duct cross-sections prepared from Gata2cko embryos did not increase significantly while in controls, the cell numbers continued to increase (Fig. 4R). These results suggested that the diametrical growth of the Gata2cko semicircular ducts had arrested while the control ducts continued to grow as the development proceeded. At E16.5, the number of cells/cross-section of the superior, lateral and posterior ducts in Gata2cko ears was only 68.6%, 61.0% and 58.9% of the corresponding number in controls, respectively (Fig. 4R).
To reveal the 3D organization of the epithelial semicircular ducts in controls and Gata2cko embryos we used the paintfill method (Kiernan, 2006). Our analysis confirmed that at E16.5 all three ducts were considerably thinner in Gata2cko embryos compared with control embryos (n = 10 ears/genotype; Fig. 5A,B). However, no change in the duct length was observed. Also in newborn mice (P0), severe thinning was observed in all three ducts (n = 3; Fig. 5C,D). In most severe cases, the lumen of the lateral duct was so narrow that no paint penetrated the area (arrowhead in Fig. 5D).
Normal Expression of EphB2 in Endolymph Producing Vestibular Dark Cells of Gata2cko Embryos
The ionic composition and the proper amount of endolymph inside the epithelial labyrinth are particularly important for inner ear sensory functions (Beitz et al., 2003), but very little is known of the molecules controlling the production of endolymph. Eph/ephrin bidirectional signaling pathways control events such as axon extension, cell migration, and adhesion during central nervous system development (Palmer and Klein, 2003). Some family members are also expressed in inner ear and participate in the production and ionic homeostasis of endolymph fluid. EphB2 is expressed in K(+) -secreting dark cells near the vestibular sensory epithelia, utricular macula and the cristae, while its ligand, ephrin-B2, is expressed in the adjacent nonsensory transitional cells separating the dark cells from the hair cells (Cowan et al., 2000; Dravis et al., 2007). Mutations in these genes in mice result in a hyperactive circling behavior associated with a decreased amount of endolymph fluid and reduced size of the semicircular ducts (Cowan et al., 2000; Dravis et al., 2007).
In Gata2cko embryos, the decrease in semicircular duct diameter appeared similar to that observed in EphB2 mutants (Cowan et al., 2000). Therefore, we decided to verify whether Gata2 protein was present in dark or transitional cells in the vestibule and to check whether the expression of EphB2 had altered in Gata2cko embryos using an anti-EphB2 antibody. Our immunohistological analyses suggested that Gata2 factor was not present in the nuclei of the transitional or dark cells (arrowheads in Fig. 5G) but was instead detectable in the adjacent nonsensory epithelium (arrow in Fig. 5G). Furthermore, no change in the expression of EphB2 was observed between control and Gata2cko embryos (arrowhead in Fig. 5E,F). Thus, the reduced size of the semicircular ducts in Gata2 conditional mutant embryos is not likely to be a result from the perturbation of EphB2 signaling in endolymph producing cells.
The endolymphatic duct and sac are enlarged in ephrin-B2−/− mice while no enlargement was reported in EphB2 mutants (Cowan et al., 2000; Dravis et al., 2007). Gata2 is not expressed in the endolymphatic duct or sac (Lilleväli et al., 2004), but we verified the morphology of these structures on serial sections at E16.5 and E18.5 (n = 6 ears/genotype/stage) to detect any indications of possible secondary problems in endolymphatic fluid balance. As shown in Figure 5H–K, no obvious enlargement of the endolymphatic structures was observed in Gata2cko ears.
Otic Cell Fate Determination Occurs Normally in Gata2cko Embryos
One reason behind the semicircular duct growth arrest could lie in changes in cell differentiation and/or identity in mutant embryos. To characterize the inner ear phenotype in more detail, we verified the expression of several otic marker genes in Gata2cko embryos at E12.5–E18.5. At these stages, Gata2 was strongly expressed all through the nonsensory epithelium of the three semicircular ducts (Fig. 6A–D,O–P and Lilleväli et al., 2004).
Netrin1, Nor1, and Dlx5 are expressed especially in the nonsensory epithelium of the developing semicircular ducts and the inactivation of these genes results in defective duct formation and/or growth (Acampora et al., 1999; Salminen et al., 2000; Ponnio et al., 2002). While Netrin1 and Nor1 marked the thinner inner wall of the semicircular ducts (arrowheads in Fig. 6E,F,I,J), Dlx5 was expressed in the thicker outer wall in both control and Gata2cko ears (arrowheads in Fig. 6G,K).
The expression of both early and late otic sensory markers such as Fgf10 (Pirvola et al., 2000), Atoh1 (Math1) (Chen et al., 2002; Woods et al., 2004), Jag1 (Morrison et al., 1999), Cdkn1b (p27Kip1) (Chen and Segil, 1999), and Pou4f3 (Erkman et al., 1996) was unaltered in Gata2cko embryos at E12.5–E18.5 (Fig. 6H,L and data not shown). Similarly, no change was observed in the expression of Gata3 at E12.5 or E14.5 (data not shown), which is at these stages expressed in restricted areas of vestibular nonsensory epithelium and periotic mesenchyme in addition to the sensory areas (Lilleväli et al., 2004, 2006). In summary, these observations suggested that the inactivation of Gata2 did not cause any major defects in cell differentiation and fate determination of the otic epithelium.
Changes in Proliferation and Programmed Cell Death in Gata2cko Otic Epithelium
Gata2 has been implicated in the control of proliferation and programmed cell death in various tissues (Patient and McGhee, 2002). We investigated whether the size-reduction of semicircular ducts could be related to defects in cell proliferation and/or programmed death.
To analyze cell proliferation in semicircular ducts at E14.5–E16.5, we injected time-mated pregnant females with BrdU and detected dividing cells on tissue sections with anti-BrdU antibody. The proportion of proliferating cells in otic epithelium from all epithelial cells in semicircular ducts per section was highest at E14.5 (around 15–16%) and decreased during the next stages (to around 5–7%) in both controls and Gata2cko embryos (Table 1). The proportion of proliferating cells had decreased significantly in superior and lateral ducts being 22.18% and 23.02% lower in Gata2 mutants than in controls, respectively. However, in the posterior duct, the proliferation was higher in mutants than in control littermates (Table 1). At E15.5, a clear reduction in the proportion of proliferating cells was observed in all ducts of the Gata2 mutants being 30% to 50% lower than that of the control littermates (Fig. 7A–D; Table 1). At E16.5, a statistically significant reduction in proliferation was detected only in the superior duct epithelium (Fig. 7E–H; Table 1). Thus, a general reduction in proliferation was observed in the Gata2 mutant semicircular ducts. However, the developmental time point at which the decrease in proliferation was observed differed slightly between the ducts.
Table 1. Proportion of BrdU+ Cells in Semicircular Duct Epithelium of Gata2cko and Control Embryos at E14.5, E15.5, and E16.5a
% BrdU+ cells/section sd
% BrdU+ cells/section ld
% BrdU+ cells/section pd
The average proportion of BrdU-positive nuclei/section from all nuclei in the superior (sd), lateral (ld), and posterior (pd) ducts of control and Gata2cko embryos is shown with standard deviation. The Statistical analysis was performed with Student's t-test. Significant differences are indicated (P < 0.05*, P < 0.01** and P < 0.001***) and the % change in Gata2cko ears is given (−) decrease, (+) increase.
15.60 ± 5.60
16.11 ± 4.70
12.81 ± 4.64
12.14 ± 4.09*
12.40 ± 5.09*
15.99 ± 4.13*
decrease (−)/increase (+) in Gata2cko
% BrdU+ cells/section sd
% BrdU+ cells/section ld
% BrdU+ cells/section pd
9.43 ± 3.78
9.68 ± 3.25
10.18 ± 5.58
4.78 ± 2.76***
6.87 ± 2.70*
6.61 ± 3.79*
decrease (−)/increase (+) in Gata2cko
% BrdU+ cells/section sd
% BrdU+ cells/section ld
% BrdU+ cells/section pd
7.32 ± 2.45
5.62 ± 0.88
5.72 ± 2.05
4.66 ± 2.65**
4.65 ± 2.28
4.80 ± 2.35
decrease (−)/increase (+) in Gata2cko
Programmed cell death has been detected in restricted areas of the otic epithelium during early inner ear morphogenesis in both rodents and chicken (Martin and Swanson, 1993; Fekete et al., 1997; Cecconi et al., 2004; León et al., 2004). Developmental cell death in mouse inner ear occurs mainly through the Apa f1/caspase 9 apoptosome pathway and when components of this pathway are inactivated, severe morphological defects appear in the “sculpting” of the epithelial structures and in the growth of the semicircular ducts (Cecconi et al., 2004).
We calculated the proportion of TUNEL (terminal deoxynucleotidyl transferase–mediated deoxyuridinetriphosphate nick end-labeling) -labeled apoptotic nuclei from all nuclei counterstained with Hoechst in sections through the semicircular ducts at E14.5–E16.5. In most analyses, we did not detect any differences between the control and Gata2cko duct epithelia (Fig. 8A–D; Table 2). However, an increase in the proportion of apoptotic nuclei could be observed in the lateral duct at E14.5 (Fig. 8E–H; Table 2) and a decrease in the posterior duct at E16.5 (Table 2).
Table 2. Number of Apoptotic TUNEL-Positive Cells in the Semicircular Duct Epithelia at E14.5–16.5a
% of TUNEL+ cells
Lower confidence limit
Upper confidence limit
The average proportion of TUNEL+ cells/section in the superior (sd), lateral (ld) and posterior (pd) ducts of control and Gata2cko embryos is shown with lower and upper 95% confidence limits. The data was generated with the SAS system using the statistical procedure GENMOD. Significance indicated with P < 0.01**; P < 0.001***.
Of interest, expression of the anti-apoptotic Bcl-XL isoform coded by the Bcl-X (also called Bcl2l1) gene is decreased in adult Gata2+/− bone marrow stem cells and this is associated with a greater frequency of apoptotic cells suggesting that Gata2 is required for cell survival in bone marrow stem cell pool (Rodrigues et al., 2005). In addition, Gata2 has been shown to bind to Bcl-X promoter in leukemic cells (Koga et al., 2007). Bcl-X is strongly expressed in the inner ear epithelium during development at E10.5–E12.5, and its inactivation leads to developmental defects especially in the posterior duct (Cecconi et al., 2004). Here, we verified whether Bcl-X gene products were present also at E14.5 in inner ear epithelium. As seen in Figure 8I,J, Bcl-X derived protein products were abundantly expressed in ear epithelium and no obvious difference was detected between control and Gata2cko embryos, indicating that Gata2 is not necessary to maintain Bcl-X expression in inner ear and that the increase in apoptosis observed in the lateral duct is not linked to a specific loss of Bcl-X derived proteins.
Inefficient Clearance of the Perilymphatic Space in Gata2cko Embryos
Normally, most of the inner mesenchymal cells (also called fibrocytes) surrounding the developing semicircular ducts, utricle, and saccule are cleared to form a cell-free perilymphatic space between the membranous labyrinth and the outer mesenchyme condensing to form the bony otic capsule (Torres and Giraldez, 1998). Only a thin layer of inner mesenchyme remain lining the inner wall of the otic capsule (Henson and Henson, 1988). However, knowledge on the generation of the perilymphatic space is scarce.
In addition to the semicircular duct growth defect, we observed also a clearance problem in the perilymphatic space surrounding the ducts in the Gata2cko embryos. At E18.5 in control embryos, there was an almost completely cleared area close to the ducts (asterisk in Fig. 4M). However, closer to the condensed otic capsule, there were still inner mesenchymal cells present in the perilymphatic space (area indicated with ** in Fig. 4M). On one side (lateral side in case of superior duct) of the perilymphatic space, the inner mesencymal cells were reduced to a thin layer (arrows in Fig. 4M). In the Gata2cko mutants (n = 5), much more inner mesenchymal cells remained in the area closest to the duct epithelium (asterisk in Fig. 4N) and the cells were spread all through the perilymphatic area and were not squeezed toward the outer borders of the area against the capsule as in the controls (** in Fig. 4M). These results indicate that there is a delay or block in maturation of the perilymphatic space in Gata2-deficient inner ears. These observations prompted us to have a closer look at the clearance process during semicircular duct development in control and Gata2cko embryos and to verify Gata2 expression in the area.
Our histological analyses suggested that in controls the clearance of the perilymphatic space started around E14.5 when the condensing outer mesenchyme started to become distinguished from the looser type of inner mesenchyme (Fig. 4A,C, and data not shown). The clearance advanced during E15.5–E16.5 (Figs. 4E,G,I, 6C,E–G,M), and at E18.5 the perilymphatic space surrounding the vestibule was well developed, although some inner mesenchymal cells were still present (Figs. 4M,O, 6D). Before and during the clearance (E12.5–E16.5), Gata2 was expressed in the innermost mesenchymal cells closest to the epithelium, but expression was not detected in regions close to or in the condensing outer mesenchyme of the prospective otic capsule (Fig. 6A–C and data not shown). In the inner mesenchyme the expression appeared strongest at E14.5 (Fig. 6A–D). At E18.5, Gata2 expression persisted in the semicircular duct epithelium while the Gata2-expressing inner mesenchymal cells were for the most part cleared (Fig. 6D). Antibody staining at E15.5–E18.5 confirmed the presence of Gata2 protein in the semicircular duct epithelium and in the adjacent inner mesenchymal cells (Fig. 6O,Q,S and data not shown). At E15.5 Gata2 staining was particularly strong in the innermost mesenchymal cells closest to the duct epithelium, while it became gradually weaker toward the condensed mesenchyme (Fig. 6O). At E18.5 Gata2-expressing cells were for the most part cleared but Gata2 protein was, however, detected in cells forming a thin layer (membrane-like structure) between the cleared space and the condensed capsule (arrows in Fig. 6Q,S). Gata2 protein was also detected in the utricle epithelium and in the surrounding mesenchyme (Fig. 5G) where a clearance defect was also observed E16.5–E18.5 (asterisks in Fig. 4I,J,O,P).
To study the clearance of the semicircular duct perilymphatic space in more detail, we calculated the total number of cells/section in the area located between the condensing mesenchyme and the duct epithelium and compared the proportion of proliferating and TUNEL-positive cells in the area between mutant and control embryos.
Concerning the perilymphatic area surrounding the superior duct, there were more cells in the Gata2-deficient ears than in the control ears at all stages (n = 4 ears/stage, 28 sections/genotype/stage), E15.5, E16.5, and E18.5 (Table 3) especially in the region indicated with an asterisk in Figures 4G,H,M,N and 6E–G,I–K,M,N. However, significant differences in cell numbers were not observed in the perilymphatic areas surrounding the posterior and lateral ducts (Table 3), indicating slight differences in the clearing mechanism, extent or time schedule between the ducts.
Table 3. Number of Mesenchymal Cells in the Perilymphatic Space at E15.5–18.5a
No. of cells/section sd
No. of cells/section ld
No. of cells/section pd
The average number of mesenchymal cells/section in the perilymphatic area surrounding the superior (sd), lateral (ld), and posterior (pd) ducts of control and Gata2cko embryos is shown with standard deviation. Statistical analysis was performed with Student's t-test (P < 0.001***). n.a., not analyzed.
272.54 ± 17.64
275.33 ± 28.64
271.88 ± 18.65
314.14 ± 20.64 ***
289.82 ± 21.62
279.0 ± 19.62
203.67 ± 19.89
215.50 ± 3.54
181.5 ± 19.28
245.17 ± 21.46 ***
240.36 ± 20.38
204.29 ± 37.93
122.23 ± 16.43
222.12 ± 13.4 ***
We performed a bromodeoxyuridine (BrdU) analysis at the most active clearance phase at E15.5–E16.5 (n = 4 ears/stage, 28 sections/genotype/stage). At E15.5, no difference in the proportion of proliferating cells was detected between mutant and control samples (Fig. 7A–D; Table 4). At E16.5, a decrease in the proportion of proliferating cells in the superior duct perilymphatic area was observed in Gata2cko embryos while no change in lateral or posterior duct was detected (Fig. 7E–H; Table 4).
Table 4. Proportion of BrdU-Positive Nuclei in the Perilymphatic Space at E15.5 and 16.5a
The average proportion of BrdU-positive nuclei/section from all nuclei in the perilymphatic area surrounding the superior (sd), lateral (ld), and posterior (pd) ducts of control and Gata2cko embryos is shown with standard deviation. Statistical analysis was performed with Student's t-test (P < 0.01**).
4.41 ± 1.25
6.31 ± 1.75
6.07 ± 1.89
4.98 ± 0.73
5.70 ± 1.12
6.19 ± 1.09
4.58 ± 1.47
5.11 ± 0.74
6.61 ± 2.66
2.38 ± 1.02 (t-test **)
3.81 ± 1.46
4.46 ± 2.41
It has been suggested that the mesenchymal cells surrounding the growing semicircular ducts would be destined to undergo apoptosis to create the perilymphatic space (Chang et al., 2002). To verify whether the inactivation of Gata2 impaired programmed cell death in the inner mesenchymal cells, we calculated TUNEL-positive nuclei during the clearance on sections (20 sections/embryo) of control and Gata2cko embryos (n = 2 embryos/stage/genotype). In general, very little or no apoptosis was detected in the perilymphatic space mesenchyme at E14.5–E18.5. At stages when the clearance was actively ongoing in the control embryos (E15.5–E16.5), only one TUNEL-positive cell was detected in the 20 sections analyzed (on the average 0.05 cells/section) in the perilymphatic space surrounding the superior, posterior or lateral ducts (data not shown). The same result was also obtained with the Gata2cko embryos (data not shown). At E14.5, however, we detected more apoptosis (0–4 cells/section) in the perilymphatic space both in control (on the average 1.4 cells/section) and Gata2-deficient embryos (on the average 1.5 cells/section) but no significant difference between control and mutant embryos was found (Fig. 8A–H, arrows in A, C, E, and G point to the rare TUNEL-positive cells detected in the inner mesenchyme). Thus, our analyses suggest that programmed cell death is not a major player in the clearance process to generate the perilymphatic space and that the extent of cell death is not altered in Gata2-deficient embryos to explain the observed defect.
One reason behind defective clearance could be problems in cell-fate determination. Prx2 is required together with Prx1 for inner ear semicircular duct development and it is expressed in both the semicircular duct epithelium and the surrounding mesenchyme during development (ten Berge et al., 1998). We detected Prx2 expression in the cells of the perilymphatic space surrounding the ducts at E15.5–E16.5 in both control and Gata2cko embryos (arrows in Fig. 6M,N and data not shown) suggesting that, although the maturation of the perilymphatic space was delayed in Gata2cko embryos, the mesenchymal cell identity was not altered.
Gata factors have essential roles in many tissues; however, their roles vary in a cell-type specific manner. In hematopoietic and adipose progenitors, Gata2 is involved in maintaining the cells in an undifferentiated state, and overexpression of Gata2 promotes cell cycle progression and inhibits differentiation (Tsai et al., 2005; Lugus et al., 2007). In the central nervous system, Gata2 is required to specify the progenitors to serotonergic fate in rhombomere 1 and to GABAergic fate in midbrain (Craven et al., 2004; Kala et al., 2009). In addition, loss of Gata2 in midbrain leads to a complete fate switch of GABAergic neural progenitors to glutamatergic fate, whereas it did not affect the proliferation of progenitors in this region (Kala et al., 2009). On the other hand, in caudal hindbrain and spinal cord, Gata2 expression inhibits proliferation and forces cycling neural progenitors to a quiescent and differentiating state by interfering with cell cycle regulators and Notch pathway (El Wakil et al., 2006).
Here, we have with conditional mutagenesis approach investigated the roles of Gata2 in inner ear development and show that in this tissue, Gata2 is required to maintain a normal level of proliferation in the semicircular duct epithelium and to assure efficient clearing of the mesenchymal cells to generate the vestibular perilymphatic space.
Gata2 Controls Semicircular Duct Growth
For inner ear function, its 3D structure must be properly formed. The vestibular system is required to maintain head and body posture and to control coordinated movements and any malformation of the vestibule may lead to imbalance, dizziness, or vertigo in humans. In mice, vestibular dysfunction is often manifested as circling behavior, head bobbing, and/or hyperactivity. Three endolymph-filled nearly orthogonally arranged semicircular ducts and their associated sensory epithelia, the cristae, detect angular acceleration (Riley and Phillips, 2003). Their morphogenesis from the hollow epithelial otic vesicle can be divided in four critical steps. First, the otic epithelium grows out as two-layered pouches at E10.5–E11.5, then, the epithelial layers in the middle of the pouches become closer to each other so that areas of epithelial fusion can form at E11.5–E12.5. Subsequently, the fused epithelia are removed giving rise to three epithelial duct structures by E12.5–E13.5, which then in the fourth step grow to their final form and size (Martin and Swanson, 1993). Each step in the semicircular duct morphogenesis is regulated by genes expressed either in the ducts or in the surrounding tissues (Torres and Giraldez, 1998; Noramly and Grainger, 2002; Fritzsch et al., 2006). Of the GATA-factors, Gata3 is required at a very early morphogenetic stage during otic vesicle formation and the inactivation of Gata3 leads to a dramatic early developmental arrest and to a complete lack of semicircular ducts (Karis et al., 2001; Lilleväli et al., 2006). Instead, inner ear morphogenesis is not disturbed in Gata2−/− mouse embryos before E10.5, most likely due to compensation by Gata3 (Lilleväli et al., 2004).
To uncover the role of Gata2 in inner ear development, we used conditional mutagenesis in mouse. Inactivation of the conditional Gata2 allele with Foxg1-Cre mice resulted in the loss of Gata2 expression in otic vesicle and the surrounding mesenchyme before E10.5. Although the expression of Gata2 is particularly strong in the nonsensory epithelium of the vestibule and in the surrounding mesenchyme at E11.5–E14.5 (this work and Lilleväli et al., 2004), the loss of Gata2 did not cause any clear vestibular defects before E14.5. At E11.5–E14.5, the vestibular expression of Gata3 is rather divergent (Lilleväli et al., 2004); thus, it is not likely to have a compensating role anymore. Interestingly, even if Gata2 is expressed during duct formation at E11.5–E12.5 all through the duct epithelium (Lilleväli et al. 2004), no defects in the three first duct formation stages could be detected in Gata2cko embryos. However, at the last growth stage (E14.5–P0), we observed a growth arrest in the Gata2cko ducts so that at E16.5 the Gata2cko duct diameter was only approximately 59–69% of the size of the control ducts.
One reason for the epithelial growth arrest in Gata2cko ears could be a defective endolymph production as shown for EphB2 and ephrin-B2-deficient mice (Cowan et al., 2000; Dravis et al., 2007). However, Gata2 was not expressed in the endolymph producing cells and accordingly, EphB2 expression was not lost in Gata2cko ears, thus, suggesting that loss of Gata2 does not directly affect endolymph production. Furthermore, Gata2 is not expressed in the endolymphatic duct or sac (Lilleväli et al., 2004) and, thus, is not likely to directly participate in the control of fluid homeostasis. However, an indirect effect on endolymph production or inner ear fluid homeostasis cannot be completely ruled out. Also no clear overall increase in programmed cell death could be observed in the three semicircular ducts to explain the growth arrest. Furthermore, no obvious change in cell identity in the duct epithelium was detected. Instead, we show that Gata2 is required for normal proliferation in the semicircular duct epithelium at E14.5–E16.5, thus, at stages when the size difference between control and Gata2cko ducts becomes evident. These results suggest that Gata2 can be regarded as a “vestibular gene” involved in the fine tuning of ear morphogenesis and specifically influencing the final size of the duct diameter together with Nor1, Ntn1, and EphB2 (Fritzsch et al., 2006).
Gata2 Is Required for the Formation of the Perilymphatic Space
During inner ear development the otic vesicle derived epithelial labyrinth and the outer mesenchyme derived bony labyrinth or otic capsule are thought to interact and influence each other's development so that in the end the size and form of the capsule closely follows the contours of the epithelium (Noramly and Grainger, 2002). The inner mesenchymal cells originally residing between the epithelium and the condensing capsule are removed to form a cell-free space filled with perilymph. Formation of acellular structures is called cavitation concerning the cochlea, where the scala vestibuli and scala tympani are cleared and filled with perilymph (Sher, 1971). The formation of the perilymphatic space is important for inner ear sensory functions (Carey and Amin, 2006); however, the cellular and molecular mechanisms of the clearance process are poorly understood.
Gata2 expression in the inner mesenchyme is detected at E11.5 next to the outgrowing semicircular ducts and after duct formation at E12.5–E18.5 expression is detected in the inner mesenchyme surrounding the ducts and the utricle, but not around the saccule or in the cochlear perilymphatic areas (Lilleväli et al., 2004; this work; and data not shown). Accordingly, a defect in the maturation of the perilymphatic space surrounding the semicircular ducts and the utricle was observed in Gata2cko ears, whereas no perturbation in the perilymphatic clearing was observed around the saccule or in the cochlear scala vestibuli and scala tympani by E18.5 (data not shown). These observations suggest that Gata2 may act in a direct cell-autonomous way to control inner mesenchymal cell clearance in areas where it is expressed. However, because Gata2 is also expressed in the semicircular duct and utricle epithelium next to the cells undergoing clearance, an indirect effect from the epithelium cannot be excluded. In fact, a general growth defect of the epithelial ducts could slow down the clearance process in the mesenchyme. However, no clearance defect has been reported linked to other genes whose inactivation causes similar semicircular duct growth defects, such as Nor1 (Ponnio et al., 2002), Ntn1 (Salminen et al., 2000 and data not shown), EphB2 (Cowan et al., 2000), and ephrinB2 (Dravis et al., 2007), suggesting that the defect observed in Gata2cko perilymphatic space is specific and related to the lack of Gata2.
Of interest, the most striking difference between the Gata2-deficient and control perilymphatic spaces around all three semicircular ducts was the distribution of inner mesenchymal cells. While the cells were distributed quite evenly in the mutant perilymphatic space, they were arranged in two areas separated by a membrane-like ring in controls during E16.5–E18.5. One population of cells located in the innermost mesenchyme expressed high levels of Gata2. During clearance, these cells appeared to participate in the formation of a membrane-like ring, which is unable to form without Gata2 in Gata2cko embryos. The second, “outermost” population (with no detectable levels of Gata2 expression) appeared as if pushed out toward the condensing otic capsule wall. These cells were overlaid by the Gata2-expressing ring of cells. Also this second population of densely packed cells could not be distinguished in Gata2cko ears, suggesting that the formation of the Gata2-expressing membrane-like structure is required to reposition the outermost cells that do not express Gata2 and to “push” them toward the capsule.
The division of the inner mesenchymal cells in two populations distinguished by Gata2 expression indicates that the cells may have two different fates. An attractive hypothesis for the fate of the cells that seem to be pushed toward the condensing capsule could be that they are incorporated into the forming temporal bone. In fact, we have noticed that in certain areas of the periotic mesenchyme, depending of the development stage, there is no clear boundary between the inner and outer mesenchyme. Instead, the cells appear to form a continuous tissue (data not shown). The fate of the innermost Gata2-expressing mesenchyme is even more intriguing. Although some Gata2-expressing cells located in the immediate vicinity of the epithelium at E15.5–E18.5 (Figs. 4G,M, 6O,Q,S, and data not shown), most Gata2-positive nuclei at E18.5 were detected forming a continuous membrane-like layer inside the condensed capsule (Fig. 6Q,S). Because this thin membrane-like structure is missing in Gata2cko embryos, it suggests that Gata2 is required to form it. In addition, the Gata2-expressing cells most likely form the thin layer of inner mesenchyme that remains lining the inner wall of the otic capsule even after birth (Henson and Henson, 1988). Thus, without Gata2, the clearance of the innermost mesenchyme and the concomitant formation of the membrane-like structure do not occur normally by E18.5 in Gata2cko embryos. This kind of problem could be related to defects in cell adhesion molecule expression. However, nothing is known of the regulation of adhesion in the perilymphatic area.
Of interest, the mesenchymal defect in the semicircular ducts occurs simultaneously with the epithelial growth arrest; thus, the clearance and epithelial growth might be closely linked through Gata2. Of interest, the inner mesenchyme cells express Gata2 in a gradually decreasing gradient so that highest levels of Gata2 is expressed in the cells next to the epithelium. This could suggest that a diffusible signal from the epithelium controls Gata2 expression in the mesenchyme.
A few additional genes are known to be expressed in the inner mesenchymal cells during otic development. Prx1 and Prx2 are expressed both in the otic epithelium and mesenchyme and their co-inactivation causes reduction in the size of the otic epithelium and capsule, but no defects have been reported in mesenchyme clearance (ten Berge et al., 1998). Two additional transcription factors, Pou3f4 (also called Brn4) and Tbx18 are expressed in the inner mesenchyme surrounding the entire developing inner ear and in addition, Pou3f4 is also expressed in the condensing capsule (Phippard et al., 1998; Trowe et al., 2008). Nevertheless, inactivation of these genes causes defects only in the cochlear compartment and moreover, no problems in the mesenchymal clearance was reported (Phippard et al., 1999; Trowe et al., 2008). We also verified the state of perilymphatic space maturation in three lines of knock-out embryos, where either Neogenin (Neo1), Unc5b, or Unc5c genes, known to be expressed in periotic mesenchyme during ear morphogenesis, have been inactivated (Ackerman et al., 1997; Lu et al., 2004; Matilainen et al., 2007). No clearance defect could be observed in these embryos at E16.5–E18.5 (data not shown). Thus, to date, Gata2 is the only gene shown to be involved in the perilymphatic space clearing process. There is, however, a congenital human disease called Cornelia de Lange syndrome, where among many other developmental defects in various organs, the vestibular perilymphatic space is filled with mesenchymal cells, whereas the cochlear perilymphatic spaces are cleared (Sasaki et al., 1996). Three genes have been linked to the disease (Kline et al., 2007), but whether these could be targets for Gata2 regulation remains open.
Programmed Cell Death Does Not Play a Major Part in the Clearance of the Perilymphatic Space
We have previously shown that focal cell death driven by the Apaf1/caspase9 apoptosome in inner ear epithelium underlies ear morphogenesis and growth. When caspase9 or Apaf1 are inactivated, no programmed cell death is observed by TUNEL in the developing inner ear and severe defects in its epithelial morphogenesis occur (Cecconi et al., 2004). In contrast to otic epithelium, very little or no cell death can be observed in the surrounding inner mesenchyme throughout development (Cecconi et al., 2004 and this work), although apoptosis has been proposed to be the clearance mechanism to generate the vestibular (Chang et al., 2002) and cochlear perilymphatic spaces (Nikolic et al., 2000). We did not observe any clear decrease in inner mesenchyme cell death in Gata2cko embryos to explain the maturation defect observed. In addition, there were no evident clearance problems in the vestibule or the cochlea in the absence of programmed cell death in Apaf1- and caspase9-deficient ears (Cecconi et al., 2004 and data not shown) strongly suggesting that Apaf1/caspase9-driven apoptosis is not a major player in the perilymphatic clearance process.
Generation of Gata2 Conditional Targeting Construct
A 10,413 bp NotI-SalI Gata2 genomic DNA fragment containing exons IG to IV was subcloned into the multiple cloning site of pBluescriptKSII (Stratagene, La Jolla, CA). The sequence of the Gata2 fragment is available in Ensembl (ENSMUSG00000015053).
To generate a loxP site, the following oligonucleotides were used: 5′-ATA ACT TCG TAT AGC ATA CAT TAT ACG AAG TTA TGC ATA CGT-3′ and 5′-ATG CAT AAC TTC GTA TAA TGT ATG CTA TAC GAA GTT ATA CGT-3′. After annealing (in 50 mM NaCl, 10 mM Tris, pH 8; continuous temperature decrease from +100°C to +20°C), a fragment with AatII sticky ends and an additional NsiI site was created. The loxP sequence was inserted into the AatII site located upstream of exon II (Fig. 1A). A neoflox8 plasmid (Torres and Kuhn, 1997) -derived neomycin cassette flanked by two loxP sequences was inserted into a BglII site between exons III and IV (Fig. 1A). The orientation of the single loxP site and the neomycin cassette was verified by DNA sequencing. A PGK-thymidine kinase (TK) selection gene was added to the 3′ end of the Gata2 genomic fragment (Fig. 1A). For electroporation the targeting construct was linearized with NotI at the 5′ end of the insert (Fig. 1A).
ES Cell Transfection, Mouse Breeding, and Genotyping
The linearized Gata2 targeting vector (25 μg) was electroporated with Bio-Rad Gene Pulser II into 9 × 106 R1 ES cells (Nagy et al., 1993) and transformants were selected with Geneticin (150 μg/ml; Invitrogen) and Gancyclovir (0.5 μg/ml; Sigma) containing medium for 7–10 days. Genomic DNA from resistant clones was digested with SacI and analyzed by Southern blot analysis using a 3′ external probe (see Fig. 1A). One clone exhibiting homologous recombination was aggregated with morula-stage embryos from NMRI mice to generate mouse chimeras (Wood et al., 1993). Male chimeras were bred with NMRI female mice. Heterozygous (Gata2c/+) mice were intercrossed to produce Gata2c/c homozygotes. Adult Gata2c/c mice were fertile and normal. The Gata2c/c mice used in the phenotypic analyses were bred for at least 6 generations into C57Bl/6 background.
Cre-mediated excision was established using Foxg1-Cre mice (Hébert and McConnell, 2000), which were held in a NMRI background. Cre transgene was detected with primer pair CreF 5′-CGA TGC AAC GAG TGA TGA GGT TC-3′ and CreR 5′-GCA CGT TCA CCG GCA TCA AC-3′ (lower panel in Fig. 1D). Cre-mediated deletion was verified by PCR analysis using the following primers: (a) 5′-CTT TCC ACC CTC CTT GGA TT-3′, (b) 5′-TTT TTC CCC AAA GTC ACC TG-3′, (c) 5′-ACA GAG GCG CGG GGA ATA CA-3′ and (d) 5′-CGT ACG TCG ACC AGC CTT CGC TTG GGC TTG AT-3′ (Fig. 1A). Genotyping of wild-type Gata2+/+, heterozygous Gata2c/+ and homozygous Gata2c/c mice was performed by PCR that amplified a region where a single loxP sequence was inserted (primers a and b in Fig. 1A; Fig. 1C). In all comparative phenotypic analyses, wild-type Gata2c/c or Gata2c/+ littermates without Cre-transgene or wild-type and Gata2c/+ littermates with Cre-transgene were used as control embryos and they all showed normal inner ear morphology during embryonic development. The R26R reporter mouse line (Soriano, 1999) was bred into a NMRI background. All mouse work has been approved by the Helsinki University ethical review board.
X-gal Staining, RNA In Situ Hybridization, Hematoxylin-Eosin Staining, and Paintfill Method
X-gal staining to visualize β-gal activity in whole-mount embryos was performed as before (Salminen et al., 1998). Whole-mount RNA in situ hybridization with digoxigenin-UTP (Roche) labeled RNA probes was performed according to Wilkinson (1993). The 4% paraformaldehyde-fixed stained specimens were embedded in gelatin and cut with Vibratome into 25-μm sections. Radioactive in situ hybridization was performed on 10-μm serial paraffin sections and the obtained images were treated as described before (Lilleväli et al., 2004). The sense probes used as controls did not present any specific signals (data not shown).
Before injecting paint in the inner ears, mouse embryo heads were cut in two, fixed in Bodian's fixative, dehydrated in ethanol series, and cleared in 100% methyl salicylate and the injections were performed with a Hamilton syringe and a capillary needle as in Kiernan (2006). The injected ears were dissected out and photographed with Leica MZFL-III camera attached to Olympus DP50-CU microscope.
Detection of Cell Proliferation and Apoptosis
For cell proliferation analyses, BrdU (Sigma) was injected intraperitoneally into pregnant females 2 hr before killing and proliferating cells were detected in the embryos as described before (Salminen et al., 2000). The proliferation index (average proportion of BrdU-labeled nuclei from all nuclei) was calculated from 10-μm serial paraffin sections from four ears of E14.5–E16.5 stages for both control and Gata2cko genotypes. Apoptosis detection was performed on 10-μm paraffin sections with ApopTag Fluorescein In Situ Apoptosis Detection Kit (Chemicon) according to Cecconi et al. (2004). The proportion of TUNEL-positive cells from all otic epithelium cells was calculated from serial sections of E14.5–E16.5 from four ears of each genotype. The nuclei in the samples were visualized using bisBenzimide (Hoechst No. 33258, Sigma). The stained sections were mounted in Vectashield (Vector) mounting medium and photographed with Olympus DP70 CCD-camera attached to Olympus AX70 microscope.
The statistical analysis of the obtained data was performed with the Student's t-test when the distribution was normal. In the case of the TUNEL analysis, the data was treated with the SAS system (Release 9.1.3) using statistical procedure GENMOD that uses generalized linear statistical analysis. 95% confidence limits were used and P values calculated for the comparison of control and Gata2cko values.
Frozen inner ear sections from E18.5 stage were used for immunohistochemistry with anti-EphB2 antibody. The antigen retrieval was performed by boiling sections for 20 min in antigen unmasking solution (Vector). Sections were blocked for 30 min at room temperature with 10% horse serum and 0.3% Triton X-100 in phosphate buffered saline, followed by incubation with primary antibody overnight at +4°C. Antibody dilution for EphB2 was 1:200 (R&D Systems). The enhancement of fluorescent signal was carried out by using Tyramide Signal Amplification Plus Fluorescence System according to the manufacturer's instructions (Perkin Elmer). For anti-Bcl-X and anti-Gata2 stainings, 10-μm paraffin sections were used. Immunohistochemical analysis was performed using 1:200 dilutions of polyclonal anti-Gata2 (Santa Cruz) and anti-Bcl-X (Sigma) antibodies and 1:500 dilution of Alexa Fluor 546 goat anti-rabbit secondary antibody (Molecular Probes). All immunohistochemical samples were counterstained with bisBenzimide (Hoechst No. 33258, Sigma), and the stained sections were mounted in Vectashield (Vector) mounting medium. The anti-Eph2 stained sections were photographed with Leica TCS SP 5 confocal system, and image analysis was made with Bitplane Imaris software (Bitplane Inc.). The anti-Gata2 and anti-Bcl-X stained sections were photographed with Olympus DP70 CCD-camera attached to Olympus AX70 microscope.
We thank Thomas Perlmann, Giovanni Levi, and Irma Thesleff for providing in situ probes, for Bernhard Saeger and Annette Neubueser for the help with the paintfill method, Anne Eichmann, Susan Ackerman, and Patrick Mehlen for Unc5b−/−, Unc5c−/−, and Neo1−/− embryos, respectively and Tõnu Möls for the help with the statistical analysis. Alar Karis is acknowledged for valuable support in the beginning of the project. We are especially grateful for Raija Savolainen and Viikki Mouse Transgenic Unit for the help in generation the Gata2 conditional mouse line and for the Viikki Laboratory Animal Centre for mouse husbandry. Suzan Cingi is acknowledged for expert technical assistance and Juha Partanen for critical reading of the manuscript. Viikki Garduate School and the FinGerDev-Network are thanked for the travel-fellowship for M.H.