|Home | About | Journals | Submit | Contact Us | Français|
Mechanosensitive sensory hair cells are the linchpin of our senses of hearing and balance. The inability of the mammalian inner ear to regenerate lost hair cells is the major reason for the permanence of hearing loss and certain balance disorders. Here we present a stepwise guidance protocol starting with mouse embryonic stem (ES) and induced pluripotent stem (iPS) cells, which were directed toward becoming ectoderm capable of responding to otic-inducing growth factors. The resulting otic progenitor cells were subjected to varying differentiation conditions, one of which promoted the organization of the cells into epithelial clusters displaying hair cell-like cells with stereociliary bundles. Bundle-bearing cells in these clusters responded to mechanical stimulation with currents that were reminiscent of immature hair cell transduction currents.
Our inner ear harbors about 15,000 cochlear and about the same number of vestibular sensory hair cells, which are the mechanoreceptors of our senses of hearing and balance. Because of their paucity, molecular studies on hair cells have been limited, and consequently, the inner ear shelters the last of our senses for which the molecular basis is unknown. Aside from being scarce, hair cells are also sensitive to mechanical and chemical insults. Acoustical overstimulation, chemotherapy, aminoglycoside drug side effects, the effects of aging and increasingly noisy environments contribute to the deterioration of hearing over time. As a result, hundreds of millions of patients worldwide are permanently debilitated by hearing loss and balance problems. The main reason for the permanence of these chronic disorders is the fact that mammalian cochlear hair cells do not spontaneously regenerate and that the limited regeneration observed in the vestibular system is inadequate to restore function (Forge et al., 1993; Warchol et al., 1993).
Probably the most suitable renewable source for the generation of sensory hair cells are pluripotent stem cells, such as ES cells and iPS cells (Beisel et al., 2008; Brigande and Heller, 2009). Whereas detailed and efficient protocols for the stepwise differentiation of retinal photoreceptors exist (Ikeda et al., 2005; Lamba et al., 2006; Osakada et al., 2008), only a single labor-intensive and protracted protocol has been devised for the generation of hair cell-like cells from mouse ES cells (Li et al., 2003). In all these previous approaches, stem cell-generated retinal photoreceptor and hair cell-like cells were mainly defined by the expression of multiple marker proteins. However, a recent report shows that transplanted ES cell-derived retinal cells restored light responses in a blind mouse model, indicating that the stem cell-generated photoreceptors can be functional when the progenitor cells are maturing in an enabling environment (Lamba et al., 2009).
The main goal of our study was to employ principles of early embryonic development and otic induction to generate a population of otic progenitor cells capable of differentiating into mechanosensitive sensory hair cells in vitro. ES and iPS cells isolated from an identical murine model were used in parallel to show that both pluripotent cell types differentiate along the otic lineage without major differences. We searched for conditions that promoted the differentiation of otic progenitors into hair cell-like cells that expressed a battery of marker genes and displayed protrusions that are highly reminiscent of stereociliary hair bundles. Finally, we were able to show that ES and iPS cell-generated hair cell-like cells were responsive to mechanical stimulation and that these responses displayed transduction currents and adaptation reminiscent of immature hair cells.
The transgenic mouse strain Math1/nGFP expresses a nuclear variant of enhanced green fluorescent protein (nGFP) that is driven by an Atoh1 enhancer (Lumpkin et al., 2003). All sensory hair cells of the Math1/nGFP inner ear express nGFP from the time when they differentiate into nascent hair cells until adulthood (Figure S1A,B), which makes stem cells isolated from this mouse line useful for guidance studies because stem cell-derived hair cell-like cells can be identified by nGFP expression (Diensthuber et al., 2009; Oshima et al., 2007). From Math1/nGFP blastocysts, we isolated four lines of ES cells that expressed typical ES cell markers and displayed ES cell colony morphology when grown on mouse embryonic fibroblast (MEF) feeders in the presence of leukemia inhibitory factor (LIF) (Figure S1C-G). Interestingly, all four Math1/nGFP ES cell lines expressed the nGFP reporter, which was not unexpected because Math1 expression has been previously reported in ES cells (Azuara et al., 2006).
To generate iPS cell lines, we infected Math1/nGFP embryonic fibroblasts with retroviruses expressing Oct4, Sox2, Klf4, and cMyc (Takahashi and Yamanaka, 2006). Primary colonies were picked, subcloned, and expanded on MEF feeder cells (Figure S2A). The iPS cell lines expressed typical ES cell marker genes as well as the Math1/nGFP reporter (Figure S2B,C).
We randomly differentiated ES and iPS cell lines by generation of embryoid bodies, removal of LIF, and culturing the embryoid body cells before analyzing expression of endo-, meso-, and ectodermal markers. We found upregulation of transcripts for GATA6, Brachyury, and microtubule-associated protein 2 (MAP2), which was confirmed by immunocytochemistry (Figure S3A-F). In differentiated cell populations, expression of the nGFP reporter was reduced or absent, and cells that expressed germline-specific markers were consistently nGFP-negative. This observation indicates that the Math1/nGFP reporter is active in ES and iPS cells and downregulated upon differentiation of the cells. When ES and iPS cell lines were injected subcutaneously into immunodeficient mice, we found formation of typical teratomas. The teratomas consisted of tissues that could be assigned to all three germ layers, indicative of the pluripotency of the ES and iPS cell lines (Figure S4).
It has been hypothesized that inhibition of primitive streak cell identities during embryoid body formation will suppress the induction of endo- and mesoderm from uncommitted epiblast cells. Establishment of primitive streak cells upon differentiation of ES cells depends on the presence of active Wnt and TGF-ß/nodal/activin signaling, which recapitulates early events that lead to germ-layer induction in the mammalian embryo (Gadue et al., 2006). We anticipated that interference with Wnt- and TGF-ß-signaling would strongly suppress the formation of primitive streak cells and concomitantly increase presumptive ectoderm. In addition, we presumed that activation of IGF signaling would promote the formation of anterior ectoderm (Pera et al., 2001), which is more competent to otic induction than trunk ectoderm (Groves and Bronner-Fraser, 2000). Similar strategies were used to generate ectoderm that is capable of differentiating into retinal cell types (Ikeda et al., 2005; Lamba et al., 2006; Osakada et al., 2008).
Embryoid bodies, generated from ES and iPS cells, were treated with the Wnt-inhibitor Dkk1 (Glinka et al., 1998), the selective inhibitor of Smad3 (SIS3) that interferes with TGF-ß-signaling (Jinnin et al., 2006), and IGF-1, either alone or in combinations (Figure 1A). Embryoid body-derived cells were attached to culture dishes and stained with antibodies to Brachyury and GATA6, indicators of differentiation along the meso- and endodermal lineages, respectively. We observed that treatment of ES cell-derived embryoid bodies with either Dkk1 or SIS3 alone significantly reduced the number of Brachyury-positive cells (Figure 1B). For iPS-derived embryoid bodies, only Dkk1 alone was able to significantly reduce the Brachyury-expressing cell population (Figure 1C). Combination of Dkk1 and SIS3 significantly reduced the number of Brachyury-positive cells from 65.0±14.9% to 20.9±6.9% in ES cell-derived populations and from 44.1± 9.6% to 15.3±6.5% in iPS cell-derived populations (Figure 1B,C). These two factors also led to significant reduction of the GATA6-positive cell population from 24.9±3.1% to 10.1±3.7% (ES cell-derivatives) and from 34.3±7.1% to 13.9±8.4% (iPS cell-derivatives). Combining Dkk1, SIS3, and IGF-1 (D/S/I) was most effective, leading to a reduction of Brachyury-positive cells to 20.8±13.1%, and of the GATA6-expressing cell population to 9.8±5.0% in ES cell-derived cell populations (Figure 1B,D). Likewise, iPS cell-derivatives displayed reduction to 14.3±5.8% (Brachyury) and 12.1±6.8% (GATA6) (Figure 1C,E). The reduction of Brachyury and GATA6 expression was also detectable at the transcript level, where the mRNA for the ES cell marker Nanog was also reduced most in D/S/I-treated cultures (Figure 1F,G).
To test for competence to otic induction, we plated the D/S/I-treated embryoid bodies into gelatin-coated culture dishes and exposed them to FGFs, which have been shown to be both sufficient and necessary for otic induction (Freter et al., 2008; Ladher et al., 2005; Phillips et al., 2004; Pirvola et al., 2000; Pirvola et al., 2002). We used bFGF as a general otic inducer because it activates several different FGF receptor subtypes and has been previously used to substitute for the proposed natural otic-inducing FGF3 and FGF10 (Groves and Bronner-Fraser, 2000; Pauley et al., 2003; Vendrell et al., 2000; Wright and Mansour, 2003). As a marker for otic induction, we used antibodies to Pax2 (Li et al., 2004), and we quantified the number of Pax2-positive cells after 3 day treatment with bFGF (Figure 2A). In both ES- and iPS cell-derived populations, we observed the largest increase of Pax2-positive cells in cultures that were previously exposed to D/S/I, reaching 29.8±7.1% for ES cell-derivatives and 19.6±5.6% for iPS cell-derivatives (Figure 2B-E). Comparable results were obtained when we used FGF3 and FGF10 instead of bFGF, which resulted in 24.6±4.0% Pax2-positive cells for ES cell derivatives and 16.3±4.0% for iPS cell derivatives (n=3). Neither of the initial factors alone nor combinations of two factors were as effective as the triple combination; therefore all three factors/compounds are needed to generate a cell population that is most responsive to FGF treatment. Dkk1 and SIS3 are mainly effective in suppressing endo- and mesodermal lineages, whereas the effect of IGF-1 only became obvious after FGF induction, where D/S/I-treatment resulted in an increased number of Pax2-positive cells when compared with D/S treatment (p values (paired t-test) for these experiments were 0.04 for ES-derived cells and 0.1 for iPS-derived cells indicative of significance in case of ES-derived cells and a possible trend for iPS-derived cells) (Figure 2B,C). Control cultures not treated with any of the three initial factors, but treated with bFGF, displayed only a few Pax2-expressing cells (0.05±0.04% for ES- and 0.1±0.09% for iPS cell derivatives).
RT-PCR confirmed the strong upregulation of Pax2 in ES and iPS cell cultures after D/S/I-treatment and exposure to bFGF (Figure 2F,G). Transcripts for other genes that are expressed in the developing inner ear, such as Pax8, Dlx5, Six1, and Eya1 (Brown et al., 2005; Groves and Bronner-Fraser, 2000; Ohyama et al., 2006; Xu et al., 1999; Zou et al., 2004) were also most abundant in D/S/I & bFGF treated cultures. Double immunostaining revealed that 56.0±5.3% of the Pax2-positive cells in ES cell-derived cultures co-expressed the otic marker Dlx5 (Figure 2H). Conversely, 73.2±10.3% of Dlx5-positive cells co-expressed Pax2. In the native developing inner ear, Pax2 expression precedes Dlx5 expression (Brown et al., 2005), and it is therefore not surprising to find only partial co-expression. Likewise, 64.1±5.6% of Pax2-expressing cells co-labeled with antibody to Pax8 (Figure 2I); 43.0±8.3% of Pax8-positive cells co-expressed Pax2. Pax8 is induced prior to Pax2 in the native developing inner ear (Hans et al., 2004; Heller and Brandli, 1999) and well before Dlx5, therefore we did not expect to detect complete co-expression of these markers because their temporal expression periods during native otic development do not completely overlap.
Pax2 is not an inner ear-specific marker. For example, it is also expressed in neural progenitors in close vicinity to the otic vesicle at the midbrain/hindbrain boundary, where it is co-expressed with engrailed 1 (Rowitch and McMahon, 1995) (Figure S5A). Engrailed 1, however, is not associated with Pax2-expressing otic progenitor cells in the developing otic vesicle (Figure S5A,B). D/S/I+bFGF treatment only resulted in 1.2±0.8% of engrailed 1-positive cells, which all expressed Pax2, indicating that the vast majority of Pax2-positive cells were not midbrain/hindbrain boundary neural progenitors.
Our guidance strategy utilizes similar steps as retinal cell guidance protocols (Ikeda et al., 2005; Lamba et al., 2006; Osakada et al., 2008). As a result, we would expect to find retinal progenitors in ES and iPS cell-derived cultures. Indeed, Pax6-expressing cells were detectable in D/S/I+bFGF-treated cultures and were clearly distinct from the Pax2-positive cell population (Figure S5C).
Native otic induction is blocked by inhibition of FGF signaling (Alsina et al., 2004; Martin and Groves, 2006). Blockade of FGF signaling with the FGF receptor inhibitor SU5402 resulted in abolishment of Pax2 induction (Figure 2J), which shows that also in guidance experiments, FGF signaling is essential for otic induction from presumptive ectodermal cells. Overall, the strong upregulation and co-expression of multiple early inner ear markers suggests that D/S/I followed by bFGF treatment sufficiently mimics, in a culture dish, the events leading to otic induction during normal embryonic development.
Withdrawal of growth factors and serum-free culture on gelatin is an effective way to initiate differentiation of ES cell-generated otic progenitors (Li et al., 2003). This approach led to upregulation of hair cell markers, but in culture, the generated hair cell-like cells did not adopt typical hair cell morphology. We tested four different substrates for differentiation of ES and iPS cell-generated otic progenitors, generated by D/S/I+bFGF treatment. When ES and iPS-derived cells were plated onto fibronectin, gelatin, or MEF feeders, we detected nGFP-positive cells (Figure 3A,B). Per 104 plated cells, 955±153, 857±240, and 520±95 nGFP-positive cells were found in ES cell-derived cultures, as well as 670±110, 597±170, and 360±66 nGFP-positive cells in iPS cell-derived cultures (on fibronectin, gelatin, MEFs, respectively; n=3). A subpopulation of the nGFP-positive cells was immunopositive for the hair cell marker myosin VIIa: 37±5, 12±8, and 33±12 (ES cell-derived) and 24±6, 8±5, 25±6 (iPS cell-derived). We detected neither cytomorphological specializations nor expression of hair bundle markers, such as espin (Zheng et al., 2000).
When we plated the ES and iPS cell-derived otic progenitors onto a layer of mitotically-inactivated chicken utricle stromal cells, we observed a different behavior. The progenitors formed defined patches of cells that harbored nGFP-positive cells, which co-expressed the hair cell marker myosin VIIa and the actin filament bundling protein espin (Figure 3C,D), which is abundantly expressed in the stereocilia of the mechanosensitive hair bundle where it is necessary for hair cell function (Zheng et al., 2000). Per 104 plated cells, 1186±150 (ES-derived) and 908±209 (iPS-derived) cells were nGFP-positive, 139±49 and 113±24 cells were nGFP- and myosin VIIa-positive, and 36±7 and 24±19 cells expressed both markers plus espin (n=4). When we plated 104 control cells that were not subjected to D/S/I, but otherwise treated identically, we only found a few (135±84 and 30±18, ES and iPS-derived) nGFP-positive cells, and no myosin VIIa or espin-expressing cells (n=4). These results show that D/S/I treatment is a specific requirement for hair cell differentiation from ES and iPS cells.
Interestingly, the cells surrounding the nascent hair cell-like cells displayed nuclear immunoreactivity for p27Kip1 (Figure 3E,F), a cell cycle regulator that is initially expressed in the nuclei of all cells of the prosensory domains of the developing inner ear and later becomes restricted to supporting cells (Chen and Segil, 1999). In nascent hair cells, p27Kip1 translocates from the nucleus to the cytoplasm before the protein is no longer detectable in fully differentiated hair cells. We observed that after 12 days of differentiation culture, the majority of nGFP/myosinVIIa-double positive cells displayed cytoplasmic p27Kip1 immunoreactivity.
Supporting cells isolated and expanded from embryonic chicken utricle have previously been used to generate hair cell-like cells (Hu and Corwin, 2007). We performed a series of control experiments to ensure that hair cell-like cells that differentiated in ES- and iPS cell-derived cultures were neither chicken hair cells nor the product of fusion of a murine cell with a chicken hair cell. First, it is unlikely that chicken hair cells will develop from the non-sensory stromal cell layer. When we cultured inactivated chicken utricle stromal cells for up to three weeks, we never observed cells with hair cell morphology or cells that expressed hair cell markers (Figure S6A). Furthermore, ES- and iPS cell-derived myosin VIIa and nGFP-positive cells did not stain with a monoclonal antibody specific to the chicken isoform of hair bundle protein tyrosine phosphatase receptor Q (Ptprq, also known as hair cell antigen (Goodyear et al., 2003)) (Figure S6B,C). Conversely, chicken hair cells, derived from dissociated otic vesicle cells that were seeded onto stromal cells, displayed strong Ptprq-immunoreactivity, but lacked nuclear green fluorescence (Figure S6D).
The occurrence of asymmetrically distributed espin immunoreactivity toward one side of the presumptive hair cells (Figure 3C,D) raised the question whether the cells were developing hair bundle-like structures. To answer this question, we analyzed clusters containing nGFP-positive cells by scanning electron microscopy (SEM) (Figure S7). Protruding from the surface of the clusters, we visualized structures that were highly reminiscent of stereociliary hair bundles at different stages of maturation (Figure 4A-C,H-J) (Tilney et al., 1992). The hair bundle-like structures displayed single acentric protrusions reminiscent of kinocilia, which were consistently located toward the side of the bundle that featured the tallest stereocilia-like protrusions. Cytoskeletal stereocilia cores consist of F-actin, cross-linked by espin, whereas kinocilia are tubulin-filled. When we visualized F-actin and espin in bundles protruding from ES and iPS cell-derived clusters, we found that stereocilia-like extensions were labeled with phalloidin and antibody to espin (Figure 4D-E,K-L). The longer kinocilia remained unlabeled with both reagents, but they displayed immunoreactivity for tubulin (Figure 4F-G,M-N).
We further noticed many inter-stereociliary links as well as links between the tips of stereocilia and the sides of taller neighboring stereocilia (Figure 5A-C). Multiple links are reminiscent of nascent hair cells, which transiently display many inter-stereociliary links that can be visualized with antibodies to cadherin 23 (Boeda et al., 2002; Kazmierczak et al., 2007; Michel et al., 2005; Siemens et al., 2004). We found that the protrusions were labeled with antibodies to cadherin 23 (Figure 5E,F), further indicating that the bundles correspond to immature hair cell stereociliary bundles. The tops of short stereocilia that are connected with tip links to their taller neighbors usually appear pointed and asymmetric (Lin et al., 2005), which could be an indication of tension in the link. ES and iPS cell-derived hair cell-like cells displayed asymmetric or pointed stereociliary tips that appeared to be linked by thin filaments to the sides of the next tallest neighbors (Figure 5B,C). Finally, we observed that stereocilia of stem cell-derived hair cell-like cells were tapered at their bases (Figure 5D), which is a hallmark of hair cell stereocilia (Tilney et al., 1983).
These results show that mitotically inactivated utricle stromal cells provide one or a combination of signals that induce the formation of hair bundles. Albeit unknown, we hypothesize that these signals are not entirely secreted because stromal cell conditioned medium was unable to evoke hair bundle differentiation (no espin-positive cells, n=3). Conversely, plating ES and iPS cell-derived cells on paraformaldehyde-fixed utricle stromal cells also resulted in abolishment of the hair bundle-inducing activity (n=3). Future identification of these signals could provide important clues about the mechanisms controlling the initiation of hair bundle growth during embryonic development. Our results indicate that hair bundles only grow in cells that express Math1 and myosin VIIa, but that expression of Math1 and myosin VIIa is not sufficient to induce hair bundle formation.
The occurrence of hair bundle-like structures with asymmetric tips and inter-stereociliary links raised our curiosity whether the cells were responsive to mechanical stimulation. A total of 52 cells were successfully recorded, with 42 being derived from ES cells and 10 being derived from iPS cells. No statistical differences were observed in any measured parameter. Cell capacitance was 3.7±1.0pF (n = 52) and series resistance was 14±7MΩ (n =52) prior to compensation of up to 50%. Resting potentials were -45±7mV (n = 7). Mechanosensitivity was probed in 45 of these cells with 24 positive responses. Figure 6 shows examples of responses from both ES- and iPS-derived cells. The mean current amplitude was 74±82pA with responses ranging from 14pA to 370pA. Normalized current-displacement plots are shown in Figure 6E for ES (n=5) and iPS (n =6) cells. Single Boltzmann functions of the form: I/Imax = 1/(1 + e -(x-x0)/dx) where x0 is the half activating displacement and dx is the slope, found no differences between populations. Values for x0 of 198±9 and 224±18 nm and dx values of 125±11 and 125±18 nm-1 were obtained for ES- and iPS-derived cells, respectively.
Adaptation is a complex process in which hair bundle dynamic range is enhanced and sensitivity maintained over large displacements (Eatock, 2000). It likely involves multiple mechanisms and has several distinct temporal components in mature hair cells (Wu et al., 1999). Adaptation matures in a stepwise manner, so that immature cells show little adaptation while mature cells have robust adaptation (Lelli et al., 2009; Michalski et al., 2009; Waguespack et al., 2007). Our results show a broad range of responses. 18% of the cells showed no adaptation, 45% showed a single time constant for decay of the current and 37% showed the more mature double exponential decay in currents (Figure 6D,F). The fast time constant measured was 0.5±0.4ms (n=5) and the slow was 11±5ms (n=5). No relationship to current amplitude was observed. Directional sensitivity also matures over time and appears to correlate with the alignment of tip-links orienting in one direction along the stereocilia (Waguespack et al., 2007). Immature hair cells do not show directional sensitivity. In a population of cells tested here, directional sensitivity was also ambiguous as shown in Figure 6G. This example demonstrates that either pushing or pulling on the hair bundle elicited an increase in current. Hair cell mechanotransduction currents are blocked by aminoglycosides (Kroese et al., 1989; Marcotti et al., 2005; Ricci, 2002). Mechanically induced currents were tested for pharmacologic sensitivity by locally applying 1mM dihydrostreptomycin (DHS) to hair bundles and then stimulating mechanically. Both ES- and iPS-derived hair cell-like cell responses were antagonized reversibly by DHS (Figure 6H,I), supporting the argument that the elicited current was comparable to that evoked in native sensory hair cells.
Voltage-dependent currents were also investigated in the same group of cells. Again no differences between ES- and iPS-derived cells were observed. A great deal of diversity was observed in the cell responses measured. Figure 7 shows representative examples of the types of responses observed. Current magnitudes with K+ as the major intracellular ion ranged from 397pA to 4982pA with a mean of 2190±1595pA (n=24). Cells shown in figure 7A are distinguished by the presence of an inward current, likely Na+, where 10/30 cells tested were positive for this current. Figure 7C shows an expanded view of the initial current response for a cell with an inward current. Additionally, two major types of outward currents were observed, those that activated rapidly and showed some level of inactivation and those that activated more slowly with little inactivation. The predominant response was a slowly activating, non-inactivating conductance with no inward current. Most cells showed components of each to different degrees. Steady state activation properties also varied considerably with half activating voltages ranging from -17mV to 23mV with the more negative activation associated with the inactivating currents, and the more depolarized with the more slowly activating, non-inactivating currents. Experiments with Cs+ replacing K+ revealed two kinetically distinct components that were carried by Cs+ (Figure 7B). About half of the cells had a rapidly activating component while about 20% had a more slowly activating current and 30% had no Cs+-permeant component at all.
In this study, we utilized principles of early development to suppress the differentiation of ES and iPS cells along endo- and mesodermal lineages. The resulting presumptive ectoderm displayed competence to respond to otic-inducing FGFs. The generated otic progenitor cells were capable of differentiation into hair cell marker-expressing cells, independent of the substrate they were cultured on. The development of cytomorphological specializations, such as hair bundle-like protrusions, however, required co-culture with fibroblast-like cells that were isolated from embryonic chicken utricles after removal of the sensory epithelial layers. In these cultures, the hair cell-like cells were organized in clusters, displayed hair bundle-like protrusions, and were surrounded by cells that showed features of inner ear supporting cells. Upon mechanical stimulation of bundles the cells responded with currents reminiscent of immature hair cell transduction currents. Other currents detected in the young hair cell-like cells were variable in type and size. This observation suggests that voltage-dependent currents that are diagnostic for specific mature hair cell subtypes develop independently from hair bundles and mechanoelectrical transduction. We found no substantial differences between ES and iPS cells with respect to their ability to differentiate along the otic lineage or their differentiated hair cell-like function.
Pluripotent cells were guided in a step-wise manner toward an otic fate. Exposure to bFGF or FGF3/10 revealed that embryoid body-derived cultures, which were treated with Dkk1, SIS3, and IGF-1 were substantially more responsive to otic inducers than cultures that were only treated with Dkk1 and SIS3. IGF-1 therefore seems to anteriorize the ectoderm that was generated during embryoid body formation, increasing the number of cells capable of responding to otic-inducing FGFs. This anteriorizing effect of IGF-signaling was previously observed in developing Xenopus embryos (Pera et al., 2001), used to promote anterior development of ES cells (Lamba et al., 2006), and increased responsiveness of anterior ectoderm to otic induction was reported in chicken embryos (Groves and Bronner-Fraser, 2000). It is interesting that the same logic that we applied to generate ectoderm that is responsive to inner ear inducers was utilized to guide ES cells toward retinal fate (Ikeda et al., 2005; Lamba et al., 2006; Osakada et al., 2008). In the case of retinal development, Pax6-positive precursors, often organized in neural rosettes, were observed. Our cultures also harbored Pax6-expressing cells that often occurred in rosettes (Figure S5C). These Pax6-positive cells were clearly distinct from the Pax2-expressing inner ear progenitor cells, indicating that retinal and otic lineages appeared to develop independently in D/S/I and bFGF-treated cultures.
Beside FGF signaling, activation of the canonical Wnt pathway has been proposed to further promote otic commitment of Pax2-expressing cells of the FGF-dependent pre-otic field of the chicken embryo (Ohyama et al., 2006). We did not observe an increase in the ability or efficiency of D/S/I and bFGF-treated cultures to generate hair cell-like cells when we supplemented the differentiation cultures with recombinant Wnt3a or LiCl (data not shown). An explanation for this result is that the cultures might already produce sufficient levels of Wnts. This speculation is supported by the observation that otic commitment, revealed by differentiation of hair cell marker-expressing cells, happens in the cultures without adding additional factors.
We previously have generated Pax2-expressing otic progenitors using a protracted protocol that was based on selective survival of progenitors (Li et al., 2003). These cells were able to differentiate along the otic lineage after withdrawal of growth factors and they displayed hair cell morphology when they were grafted into the developing inner ear of chicken embryos. In general, sensory cell types such as hair cell-like and photoreceptor-like cells generated by in vitro guidance of ES or, more recently iPS cells (Hirami et al., 2009; Meyer et al., 2009) were characterized by immunocytochemistry. In-depth ultrastructural analysis of cytomorphological specializations and direct functional testing has not been applied to these cells. To test for these specializations and for function, we needed to determine culture conditions that promote the generation of hair bundles. Maintaining D/S/I and bFGF-treated ES and iPS cells on various substrates including fibronectin, gelatin, and MEF feeder cells confirmed that in vitro-generated otic progenitors are able to upregulate the hair cell marker myosin VIIa. Development of hair bundle-like structures and expression of hair bundle proteins such as espin (Zheng et al., 2000), however, did not occur on these substrates. We hypothesized that the cells need additional signals and we tested whether expanded embryonic chicken utricle stromal cells would be able to provide such signals. Both ES- and iPS cell-derived otic progenitor cell cultures responded to stromal cells by organizing into clusters that were reminiscent of inner ear sensory epithelia (Figure S7). Experiments with stomal cell conditioned medium and coculture with fixed stromal cells indicate that the hair bundle-inducing activity is not present in conditioned media and that it is abolished by paraformaldehyde fixation. These results are compatible with a surface-linked and fixation-sensitive signal, but they do not exclude multiple factors or other more complex scenarios.
Coculture with utricle stromal cells led to formation of F-actin-filled protrusions that were immunopositive for the hair bundle protein espin and single tubulin-filled kinocilia in cells that co-expressed myosin VIIa and nGFPAtoh1. When we analyzed clusters of nGFP-positive cells by SEM, we found an organization of hair bundle-bearing hair cell-like cells surrounded by cells that displayed short microvilli, reminiscent of hair and supporting cells. Hair bundles displayed many other features such as interciliary links, asymmetric stereciliary tops and filamentous links from stereociliary tops to the neighboring stereocilia, tapering at the base, and immunoreactivity to antibodies to cadherin 23. Although the hair bundles were of various shapes, we did not detect typical mature cochlear bundle morphologies. The bundle morphologies appeared more generic as if specificity had not yet been assigned.
Current responses obtained from mechanically stimulated bundles were similar to those obtained from immature hair cells where the currents were small, current-displacement functions were broad, the presence of adaptation and the rates measured were quite variable and directional sensitivity was often absent (Lelli et al., 2009; Michalski et al., 2009; Waguespack et al., 2007). The time course of maturation of mechanotransduction varies depending both on end organ and on location within the end organ such that in mammalian cochlea basal cells mature 2-3 days earlier than apical outer hair cells. Mechanotransduction in basal outer hair cells begins at postnatal day 0. Vestibular hair cells mature in waves but begin neonatally around E16 (Geleoc and Holt, 2003). A common feature of the maturation is that the current amplitudes begin small, less than 100pA, adaptation is nonexistent or slow, progressively becoming faster and more complete, and directional sensitivity is initially absent, becoming progressively more apparent (Waguespack et al., 2007). Maturation of the current responses takes about 5 days (Waguespack et al., 2007). Measurements presented here would suggest that mechanotransduction was within 2 days of the maturation process, with the variability in responses indicating a range of maturation of up-to about two days.
Both the morphological and electrophysiological data suggest a common signaling pathway to trigger the development of a mechanosensitive hair bundle, however additional signaling is required to specialize the bundle as well as to specify hair cell subtypes such as auditory or vestibular, inner or outer hair cell, or type I or type II hair cell. Supporting the argument that additional signaling is required to further specialize cells to specific phenotypic hair cells were the wide range of basolateral responses observed. The array of voltage-dependent currents measured suggest a distinct lack of appropriate signaling needed to promote complete maturation into specific hair cell subtypes. In both auditory and vestibular hair cells there is a pattern of maturation where cells have a particular set of outward currents that include outward potassium (though limited selectivity) and inward sodium currents (Geleoc et al., 2004; Marcotti et al., 1999; Marcotti et al., 2003; Marcotti and Kros, 1999; Oliver et al., 1997). Both of these are transiently expressed and replaced by more selective channel types that vary depending on hair cell type and location within the end organ. Because of this diversity of channels, present data does not allow for the type of hair cell to be identified. As already pointed out, the lack of specificity in the basolateral conductances does suggest that hair bundle formation and development of mechanoelectrical transduction occurs independently of basolateral subtype specification.
Our findings provide a useful assay to study signals involved in hair cell subtype specification, a topic that is largely unexplored, particularly in mammals. Likewise, the guidance method outlined here offers a platform for molecular studies on hair cells, which are difficult to obtain in large numbers. A single retina, for example, harbors more than 120 million photoreceptors that can be isolated fairly easily, whereas a single mammalian inner ear only yields a few tens of thousand hair cells, which are difficult to dissect. The fact that in vitro-generated hair cell-like cells display mechanosensitivity demonstrated that generation of replacement hair cells from pluripotent stem cells is feasible, a finding that justifies the development of stem cell-based treatment strategies for hearing and balance disorders.
ES cells were isolated from blastocysts and iPS cells were generated from fibroblasts of Math1/nGFP mice (Lumpkin et al., 2003). Details and culture procedures, including ectodermal guidance and otic induction are described in the Supplemental Data.
Cells were cultured in 4-well tissue culture plates (Greiner 35/10), harvested by lysis in the dish for RNA isolation and RT-PCR or fixed and subjected to immunocytochemical analysis. Details are available in the Supplemental Data.
Data are presented as mean values±S.D. with the number of independent experiments (n) indicated. Statistical differences were determined with paired two-tailed t-tests using Aabel 3 (Gigawiz) on a Macintosh computer (Apple) running OS X.
The cells were fixed for 2hr with 2.5% glutaraldehyde / 4% paraformaldehyde with 50mM CaCl2 and 20mM MgCl2 in 0.1M HEPES buffer (pH=7.4), and treated with 1% OsO4 in the same buffer, 1% tannic acid in water, 1% OsO4 in water, followed by 1% tannic acid in water for 1hr each. The specimens were washed 3× between each treatment step and then dehydrated in a graded ethanol series, and finally dried by critical point drying. Specimens were viewed with a Hitachi S-3400N variable pressure SEM operated under high vacuum at 5–10kV at a working distance of 7–10mm. All chemicals were supplied by Electron Microscopy Sciences (Hatfield, PA).
nGFP-expressing hair cell-like cells were identified by fluorescence microscopy and nearby hair bundle-like protrusions were imaged with a 100× objective using brightfield optics. Recordings were conducted with an Axoclamp 200a (Axon Instruments) amplifier, interfaced with a Digidata 1332 board (Axon), and jClamp Software (Scisoft). Mechanical stimulation was done with a stiff glass probe attached to a piezo stack. Details are described in the Supplementary Data.
We thank H. Zeng (Stanford Transgenic Research Center) for help with ES cell line derivation; J. Johnson (UT Southwestern) for Math1/nGFP mice; U. Mueller (Scripps) for cadherin 23 antibody; G. Richardson for hair cell antigen antibody. F. Salles (NIDCD) for advice in electron microscopy sample preparation and L. Joubert (Stanford EM core facility) for expert assistance. M. Yazawa, R. Dolmetsch, and A. Ootani for help with teratoma assays. K. Masaki, V. Starlinger, E. Corrales, M. Herget, Z. Guo, T. Jan, and S. Sinkkonen for fruitful discussions. M.D. is a fellow of the Alexander-von-Humboldt Foundation. This work was supported by NIH grants DC006167 and P30DC010363, by the California Institute for Regenerative Medicine (grant RC1-00119-1), and by a Neuroscience of Brain Disorders Award from the McKnight Endowment Fund for Neuroscience.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.