Introduction
Permanent damage to the hair cells of the inner ear results in sensorineural hearing
loss, leading to communication difficulties in a large percentage of the population.
Hair cells are the receptor cells that transduce the acoustic stimulus. Regeneration
of damaged hair cells could potentially yield a cure for a condition that currently
has no therapies other than prosthetic devices. Although hair cells do not regenerate
in the mammalian cochlea, new hair cells in lower vertebrates are generated from epithelial
cells, called supporting cells, that surround hair cells (Balak et al., 1990; Raphael,
1992; Stone and Cotanche, 1994; Stone and Rubel, 2000; Warchol and Corwin, 1996).
In this study, we found newly formed hair cells after loss of the original hair cells
in the newborn mammalian cochlea. The differentiation of new hair cells in response
to damage occurred spontaneously and without significant proliferation of supporting
cells. Previous work had shown generation of supernumerary hair cells in embryonic
and neonatal mammals (Doetzlhofer et al., 2009; Hayashi et al., 2008; Kiernan et al.,
2005; Takebayashi et al., 2007; Yamamoto et al., 2006) after inhibition of Notch signaling,
which blocks differentiation of hair cells in the embryo (Lanford et al., 1999; Yamamoto
et al., 2006; Zine et al., 2000).
The production of new hair cells from supporting cells could be increased by inhibition
of Notch signaling in the damaged cochlea. We found, by the use of lineage tracing
and confocal microscopy in the newborn cochlea after damage and Notch inhibition,
that the capacity for supporting cell transdifferentiation to hair cells was not equally
shared but rather occurred preferentially in a subset of these cells. In previous
work, we had shown that supporting cells expressing Lgr5, a known marker of intestinal
stem cells and a target of the Wnt pathway (Barker et al., 2007), had the capacity
to differentiate into hair cells (Shi et al., 2012). In that study, we were not able
to show that the cells identified retrospectively as progenitor cells after sorting
had the capacity to regenerate hair cells in a damaged organ of Corti. Here, we demonstrate
regenerative potential in Lgr5-positive supporting cells, using Sox2 and Lgr5 lineage
tracing, in a damage model in the newborn cochlea. These results confirm that an Lgr5-expressing
subset of supporting cells in the cochlea act as progenitors for hair cells after
damage.
Results
Supporting Cells Transdifferentiated into Hair Cells at Low Levels following Damage
Because hair cell number can be increased by inhibition of Notch in the neonatal organ
of Corti, we decided to employ lineage tracing using Lgr5 and Sox2-Cre expressing
cells to identify cell populations within the mammalian organ of Corti that could
generate these new hair cells. We tested whether the two Cre lines accurately reflected
Lgr5 and Sox2 expression after crossing to reporters (Figure S1 and Table S1 available
online). We chose to use newborn tissue with drug-induced hair cell damage as a model
for hair cell regeneration that could be combined with lineage tracing. Organ of Corti
explant cultures treated with 50 μM gentamicin overnight and examined 72 hr later
showed significant outer hair cell (OHC) damage in the middle and basal regions, limited
damage in the apex, and limited inner hair cell (IHC) loss (Figure S2).
We first tested whether the model we had selected for lineage tracing was viable by
assessing the fate of the lineage-tagged cells in organs of Corti treated with tamoxifen
at postnatal day 1 (P1) and exposed to gentamicin at P2 in the absence of Notch inhibition.
Unexpectedly, we observed MYO7A-expressing cells in the damaged organ of Corti that
were positive for Sox2 and Lgr5 lineage tags. The number of hair cells that expressed
the lineage tag was small, and the presence of the reporter and unusual location in
the pillar cell region suggested that some of the MYO7A-expressing cells were not
simply surviving hair cells but had differentiated from supporting cells (Figures
1A and 1B). In addition, unlike native hair cells, these cells exhibited antibody
staining for SOX2 in their nuclei (Figures 1C and 1E), consistent with immature hair
cells (J.S. Kempfle et al., 2012, Molecular Biology of Hearing and Deafness, conference).
Many of the new hair cells in the pillar region stained for PRESTIN (Zheng et al.,
2000), a motor protein expressed only in OHCs (Figures 1D and 1F). The new hair cells
were found in the apex and middle turns of the cochlea, but not in the base (Figure 1H),
and the number of new hair cells was significantly increased relative to the undamaged
control. The expression pattern of Lgr5 (inner pillar cells, third Deiters cells,
inner border cells) and location of the new hair cells indicated that they were derived
from inner pillar cells.
To confirm the presence of new hair cells in the pillar cell region, Pou4f3 was employed
for hair cell lineage tracing (Figure S1). Following gentamicin damage, reporter-negative
hair cells were observed in the pillar cell region of Pou4f3 lineage-tagged cultures
(Figure 1G). The absence of reporter expression in these hair cells, in contrast to
native hair cells, again indicated that they had differentiated to hair cells from
reporter-negative cells.
The appearance of new MYO7A-expressing cells in the newborn organ of Corti was surprising
considering the previously observed resistance of the postnatal sensory epithelium
to hair cell regeneration. Previous studies had showed no regeneration and in fact
no supporting cell mitosis after birth.
In Vitro Damage followed by Notch Inhibition Resulted in Transdifferentiation of Supporting
Cells into Hair Cells
The results showing newly generated MYO7A-positive cells in the damaged organ of Corti
suggested that the postnatal organ of Corti had some regenerative capacity. However,
because the number of new MYO7A-positive cells was small, we proceeded with Notch
inhibition to study the lineage tracing of transdifferentiated cells.
Treatment of gentamicin-damaged cochlea with LY411575, a γ-secretase inhibitor, increased
hair cell numbers (Figure 2A). The restoration of hair cells was not complete (80.5%
and 36.3% of the normal hair cell numbers in the middle and base, respectively), but
the increase was significant (28.3% and 154.7% in the middle and base, respectively;
Figure 2B).
Lineage tracing (exposure to tamoxifen at P0 and P1) showed an increase in reporter-labeled
hair cells in gentamicin-damaged (at P2), γ-secretase inhibitor-treated (at P3) cultures
(Figure 3A; culture is shown at P6). The treatment resulted in significantly more
Sox2 lineage-tagged hair cells in all cochlear regions and Lgr5 lineage-tagged hair
cells in the apical and middle regions (Figure 3B). Based on the location of the reporter-positive
hair cells and the expression pattern of Lgr5, the new hair cells were likely descended
from Lgr5-positive inner pillar and third Deiters cells. Quantification of the percentage
of total red OHCs located in the pillar region (corrected for red OHCs in the undamaged
controls) indicated that reporter-positive OHCs occurred most frequently in the pillar
cell region (62.4% ± 15.6% of the total in the apex, 88.1% ± 11.6% in the middle,
and 40% ± 24.5% in the base). These data confirmed that the additional hair cells
were newly differentiated and suggested that inner pillar cells were the source of
the new hair cells.
After correction for the number of recombined OHCs in undamaged controls, a comparison
of the total Lgr5 and Sox2 lineage-tagged OHCs following gentamicin/LY411575 treatment
indicated no significant difference between the number of OHCs from supporting cells
labeled by the two markers in any cochlear region (Figure 3C). If a supporting cell
type other than inner pillar and third Deiters cells had contributed to the production
of new hair cells, we would have expected more Sox2 than Lgr5 lineage-tagged OHCs
due to the relatively restricted expression of Lgr5. The similarity in the number
of Lgr5 and Sox2 lineage-tagged OHCs suggested that Lgr5-positive supporting cells
were the source of all new OHCs.
Many new hair cells in both the pillar cell and OHC regions stained for PRESTIN, suggesting
that they were OHCs (white arrows, Figures 4A, 4C, 4E, and 4F). Other new hair cells
did not appear to be PRESTIN-positive (yellow arrows, Figure 4C). Unlike preexisting
hair cells, many hair cells with reporter expression were positive for SOX2, confirming
that they were new (Figures 4B and 4D). Some new hair cells were observed in areas
of extensive damage (Figures 4E and 4F). Immature stereocilia bundles were noted on
many new hair cells in the pillar cell region, further indicating they were newly
generated (Figures 4G–4G″).
Hair cells without the lineage tag were also observed in the pillar and OHC regions
of damaged organ of Corti cultures treated with LY411575 after Pou4f3 lineage tracing
(Figures 4H and 4I), again indicating that new hair cells had been generated. SOX2-positive
hair cells in the pillar cell region were not PROX1 positive (Figure 4J), unlike surrounding
supporting cells that were both SOX2 and PROX1 positive. This suggests that the SOX2/MYO7A-positive
cells in the pillar cell region had begun to lose some supporting cell markers during
their differentiation to hair cells.
Generation of New Hair Cells after γ-Secretase Inhibition was Dependent on Wnt Signaling
To ask whether transdifferentiation of the Lgr5-positive supporting cells into hair
cells in the cochlea could be Wnt dependent, we inactivated β-catenin, the intracellular
mediator of canonical Wnt signaling. Organ of Corti cultures from β-catenin
flox(exon2–6)
;Sox2-Cre-ER pups were treated with tamoxifen at P0 to delete β-catenin in supporting
cells. The organ of Corti was treated in vitro with gentamicin as in the damage model
described above and then with LY411575, and treated mutant tissues were compared to
their β-catenin
flox(exon2–6)
-negative littermates. β-catenin
flox(exon2–6)
;Sox2-Cre-ER cultures showed significantly fewer MYO7A/SOX2 double-positive cells
in the apical region than control cultures (Figure 5). This indicates that β-catenin
and Wnt signaling are necessary for transdifferentiation of supporting cells into
hair cells following damage and Notch inhibition. Thus, the effects of Notch inhibition
are β-catenin dependent and the response of Lgr5-positive cells to damage is attributable
to Wnt signaling.
A Small Number of New Hair Cells Were Generated through Proliferation after Damage
Organ of Corti cultures treated with gentamicin followed by bromodeoxyuridine (BrdU)
exhibited a small number of MYO7A-positive cells in the pillar cell region that were
also labeled for SOX2 and BrdU (Figure 6A). These cells were seen only in the apical
end of the cochlea and often occurred in pairs. In 16 organ of Corti cultures, MYO7A/SOX2/BrdU-positive
cells were observed with a mean frequency of 0.62 ± 0.22 per organ of Corti, which
is significantly more (p < 0.01) than in undamaged controls, which had no BrdU-positive
cells. Cultures damaged with gentamicin and then treated with LY411575 also displayed
occasional MYO7A/SOX2/BrdU/-positive cells in the pillar cell region (Figure 6B).
Among seven explant cultures, an average of 15.57 ± 12.66 triple-positive cells were
observed per organ of Corti. The variability was high, with some cultures lacking
any BrdU-positive hair cells. These results suggest that cell division can precede
transdifferentiation in inner pillar cells in response to damage.
Discussion
We have demonstrated generation of new hair cells in the neonatal mouse cochlea following
gentamicin damage. The new hair cells were derived from supporting cells by transdifferentiation
and limited proliferation. Appearance of new hair cells has not been observed in studies
that have assessed regeneration in the mammalian cochlea after damage. In contrast,
spontaneous transdifferentiation of supporting cells in response to aminoglycoside
damage resulted in vestibular hair cell replacement, and hair cell number could be
increased by Notch inhibition (Lin et al., 2011), consistent with previous studies
showing some capacity for regeneration in the vestibular, as opposed to the hearing,
organs of the mammalian inner ear (Forge et al., 1993; Rubel et al., 1995; Warchol
et al., 1993).
The sensitive lineage tracing tools used in this study allowed for the unexpected
discovery that hair cells regenerated spontaneously in the postnatal mammalian cochlea.
New hair cells were detected by the presence of both the reporter from the tagged
supporting cell and hair cell markers in the same cell. Supporting cell transdifferentiation
was thus shown to be the mechanism for the generation of hair cells, whether after
damage alone or after damage and blocking Notch signaling by treatment with a γ-secretase
inhibitor. The organ of Corti’s regenerative response, whether occurring spontaneously
or after pharmacological inhibition of Notch, appeared to be restricted to early postnatal
times and gave rise to OHCs, even when the new hair cells were derived from cells
closer to IHCs. Surprisingly, we saw less generation of new hair cells in areas with
the most damage. The damage effect was thus not confined to the areas of cell loss
but extended to undamaged regions. Apical areas may maintain their capacity for transdifferentiation
of supporting cells longer because of their later development, as compared to the
base and middle regions.
In addition to demonstrating spontaneous supporting cell transdifferentiation to hair
cells after damage, we showed that the capacity to generate hair cells was limited
to a subset of supporting cells: inner pillar cells and third-row Deiters cells (Figure 7).
This was a striking set of supporting cells, because the same cells were identified
as Lgr5-expressing progenitor cells by their capacity for self-renewal and generation
of hair cells in vitro compared to a general supporting cell population (Shi et al.,
2012). Inner pillar cells were the source of most of the hair cells after damage in
the absence of Notch inhibition. Inner pillar cells and third-row Deiters cells were
the major contributors to new hair cells after treatment of the damaged cochlea with
a γ-secretase inhibitor. These cell types showed the highest expression of Lgr5 in
neonatal animals in the previous study (Shi et al., 2012). They also maintained expression
into adulthood.
A recent study on hair cell differentiation in adult animals with hair cell loss after
Notch inhibition did not identify the supporting cells that responded (Mizutari et al.,
2013). The results we describe here are surprising, because we did not expect that
the transdifferentiation of supporting cells to hair cells after Notch inhibition
would be related to Lgr5 expression. The current work, however, implies that hair
cell differentiation after Notch inhibition was not only due to overcoming the inhibition
imposed by Notch on supporting cell conversion to hair cells but also was dependent
on Wnt. Both cell division and differentiation resulted from forced expression of
β-catenin in Lgr5-positive cells in the newborn cochlea in vivo (Shi et al., 2013),
consistent with the known relationship between LGR5 expression and Wnt signaling (de Lau
et al., 2011). The lack of a proliferative response despite the dependence on Wnt
signaling seen here suggested a low level of signaling, because proliferation was
predominant only at high levels of Wnt expression (Shi et al., 2013).
The capacity of inner pillar cells to transdifferentiate into hair cells appears to
contradict findings that pillar cells, as compared to other supporting cells, resisted
transdifferentiation in response to Notch inhibition (Doetzlhofer et al., 2009). Cells
that remained after treatment with a γ-secretase inhibitor in that paper were identified
as pillar cells (Figure 2A; yellow bracket in Doetzlhofer et al., 2009); however,
the markers used to identify pillar cells (PROX1, p75ntr and HEY2) did not distinguish
between inner and outer pillar cells. In addition, PROX1 is expressed in both pillar
and Deiters cells (Bermingham-McDonogh et al., 2006). Our results would be consistent
if the two rows of resistant cells were outer pillar and first-row Deiters cells rather
than two rows of pillar cells. Due to their apparent disappearance following Notch
inhibition, supporting cells have been assumed to be the source of these new hair
cells (Doetzlhofer et al., 2009; Kiernan et al., 2005; Takebayashi et al., 2007),
and identification of inner pillar cells as the source of most new hair cells is in
agreement with previous data showing increased hair cell numbers and evidence of pillar
cell division but no concomitant increase in pillar cell numbers in Notch mutants,
suggesting possible transdifferentiation of pillar cells into new hair cells (Kiernan
et al., 2005). Similarly, a population of sensory epithelial cells enriched for pillar
and Hensen cells gave rise to significantly more hair cells than an unfractionated
population of sensory epithelial cells (White et al., 2006), even without Notch inhibition.
Hair cells express Notch ligands, interacting with Notch receptors on neighboring
cells, thus activating Hes family genes and blocking supporting cell conversion to
a hair cell fate. Previous studies on the generation of new cochlear hair cells in
embryonic and neonatal mammals have relied on deletion or inhibition of Notch (Doetzlhofer
et al., 2009; Hayashi et al., 2008; Kiernan et al., 2005; Takebayashi et al., 2007;
Yamamoto et al., 2006). Extra hair cells come from a combination of transdifferentiation
and proliferation (Kiernan et al., 2005; Takebayashi et al., 2007), although proliferation
was not observed in neonatal mice (Doetzlhofer et al., 2009; Kiernan et al., 2005;
Takebayashi et al., 2007). The limited proliferation that we saw in neonatal explant
cultures may have been facilitated by damage, which was not used in the studies on
Notch inhibition. Although loss of Notch ligands due to damage or loss of hair cells
could theoretically induce hair cells from neighboring cells, this has not appeared
to be the case in the postnatal mammalian cochlea.
Notch and Wnt signaling are important for self-renewal and differentiation of adult
stem cells in a variety of tissues. Notch signaling is required for the generation
and maintenance of neural stem cells and proper control of neurogenesis and preserves
neural stem cell characteristics in the adult brain (Piccin et al., 2013; Tanigaki
et al., 2001). Notch and Wnt signaling regulate proliferation as well as acquisition
of a mature phenotype by muscle, intestinal, and cardiac stem cells (Hirata et al.,
2013; Klaus et al., 2012; Noah and Shroyer, 2013). The accurate expansion of Lgr5-expressing
intestinal stem cells, with the correct timing of the switch to differentiation of
goblet cells, in addition to being regulated by β-catenin (Barker et al., 2007; van
Es et al., 2005a), is under Notch control (Stamataki et al., 2011; van Es et al.,
2005b). The Lgr5-expressing progenitor cells identified in the cochlea are activated
by Wnt signaling (Shi et al., 2012). Indeed, Wnt signaling has been shown to increase
expression of a basic helix-loop-helix transcription factor, Atoh1, which is required
for hair cell differentiation (Shi et al., 2010), and Wnt is necessary for upregulation
of Atoh1 in response to Notch inhibition.
Inner pillar cells express Fgfr3, which may contribute to their capacity for regeneration
(Jacques et al., 2007). Supporting cells in the chicken basilar papilla express Fgfr3,
which is transiently downregulated during hair cell regeneration (Bermingham-McDonogh
et al., 2001). Fgfr3
−/− mice lack inner pillar cells in the apical and middle regions of the cochlea but
exhibit an extra row of OHCs in these areas (Hayashi et al., 2007; Puligilla et al.,
2007). Expression of both Lgr5 and Fgfr3 may be important for the transdifferentiation
of cochlear cells after damage. We did not find any evidence of transdifferentiation
in supporting cells that did not express Lgr5, suggesting that signaling through Lgr5
or Fgfr3 may be necessary to generate new hair cells with Notch inhibition. These
studies therefore confirm the identification of Lgr5-expressing supporting cells as
hair cell progenitors and show that damage to the newborn cochlea results in hair
cell regeneration initiated by these Lgr5-positive cells.
Experimental Procedures
Animals
The Lgr5-EGFP-IRES-Cre- ER (Barker et al., 2007), β-catenin
flox(exon2–6)
(Brault et al., 2001), CAG-tdTomato (Madisen et al., 2010), and CAG-tdTomato-EGFP
(Muzumdar et al., 2007) mice were obtained from The Jackson Laboratory. The Sox2-Cre-ER
mouse was described previously (Arnold et al., 2011). The Pou4f3-Cre mouse (Sage et al.,
2006) was a gift from Dr. Douglas Vetter (Tufts University, Boston, MA). The Sox2-Cre-ER
and Pou4f3-Cre mice were genotyped with PCR using primers to amplify the Cre gene
(forward: 5′-TGGGCGGCATGGTGCAAGTT-3′; reverse: 5′-CGGTGCTAACCAGCGTTTTC-3′). The Lgr5-EGFP-IRES-Cre-ER,
CAG-tdTomato and CAG-tdTomato-EGFP mice were genotyped with PCR according to Jackson
Laboratory recommendations. The β-catenin
flox(exon2–6)
mice were genotyped as described previously (Brault et al., 2001). All animal use
was approved by the Massachusetts Eye and Ear Infirmary Animal Care Committee.
In Vitro Lineage Tracing
Sox2-Cre-ER mice were crossed with CAG-tdTomato-EGFP or CAG-tdTomato reporter mice.
The CAG-tdTomato-EGFP reporter construct contains two loxP sites flanking both a STOP
sequence and the red fluorescence gene tdTomato. Administration of tamoxifen induced
Cre excision of the floxed region, allowing for permanent expression of EGFP. Lgr5-EGFP-IRES-Cre-ER
and Pou4f3-Cre mice were crossed with CAG-tdTomato reporter mice. At P1 (or both P0
and P1 for Lgr5 lineage tracing), an intraperitoneal injection of 100 μl tamoxifen
dissolved in corn oil (50 mg/ml) was administered to mothers of double-transgenic
litters and transferred to the pups via the milk.
Organ of Corti Dissection
Mice positive for both Cre and the reporter construct were dissected at P2 and the
cochlea was removed from the temporal bone. Cochleae for organ of Corti culture were
transferred to Hanks’ balanced salt solution (Invitrogen) and the organ of Corti was
isolated. The stria vascularis and basal hook region were removed and the organ of
Corti was plated and cultured as described below. Cochleae not meant for culture were
immediately fixed for 10 min in 4% paraformaldehyde (PFA; Electron Microscopy Sciences).
The organ of Corti was then isolated and the stria vascularis and basal hook region
removed. This was followed by additional fixation in 4% PFA for 30 min and then storage
in 1× phosphate buffer solution at 4°C until staining.
Organ of Corti Culture
The organ of Corti was plated on a glass coverslip coated in poly-L-ornithine (Sigma)
and laminin (BD Biosciences), given 100 μl Dulbecco’s modified Eagle’s medium (DMEM;
Invitrogen) with 10% fetal bovine serum (FBS; Invitrogen) and 25 μg/ml ampicillin
(Sigma), and cultured at 37°C with 5% CO2. One hour after dissection, cultures were
treated with 50 μM gentamicin (Sigma) in DMEM with 10% FBS and ampicillin. The organ
of Corti was cultured overnight (16 hr) and then placed in serum-free DMEM:F12 (Invitrogen)
with 1% B27 supplement (Invitrogen), 25 μg/ml ampicillin, and either 5 μM dibenzazepine
(deshydroxy LY411575; Santa Cruz) or 0.1% DMSO (Sigma) as a control. Undamaged controls
were cultured under the same conditions without gentamicin or LY411575. LY411575 treatment
continued for 72 hr. After 4 days in vitro, the cultures were fixed with 4% PFA (Electron
Microscopy Sciences) and stained for MYO7A to identify hair cells and either DsRed
or GFP to enhance the endogenous signal from the reporter.
To measure proliferation, cultures were treated with gentamicin as detailed above
followed by 10 μM BrdU along with either LY411575 or 0.1% DMSO for 72 hr and then
fixed, treated with 2 N HCl for 30 min at 37°C, and stained for BrdU.
Antibodies
The primary antibodies used in this study are as follows: anti-MYO7A (mouse, Developmental
Studies Hybridoma Bank, 1:500; rabbit, Proteus Biosciences, 1:500), anti-BrdU (rat,
ABD Serotec, 1:100), anti-GFP (rabbit, Invitrogen, 1:500), anti-SOX2 (goat, Santa
Cruz Biotechnology, 1:500), anti-DsRed (rabbit, Clontech, 1:500), anti-PROX1 (mouse,
Chemicon, 1:200), and anti-PRESTIN (goat, Santa Cruz Biotechnology, 1:400). Immunoreactivity
was visualized using Alexa Fluor-conjugated secondary antibodies (Molecular Probes,
1:500).
Cell Counting
The organ of Corti was analyzed using a Leica TCS SP5 confocal microscope. Organ of
Corti cross sections were obtained from confocal z stacks using Amira imaging software
(Visage Imaging). High-magnification fluorescent images of the organs of Corti were
merged in Photoshop and counted manually with ImageJ software (National Institutes
of Health). Each organ of Corti was split into three regions for counting (apex, middle,
and base), and hair cell counts were obtained per 100 μm. Hair cells in the pillar
cell region were counted as OHCs. Each counted region was 1,900 ± 100 μm. Counts of
reporter-positive hair cells for damaged and damaged/Notch-inhibited organs of Corti
were corrected for residual recombination of native hair cells by subtracting the
mean number of hair cells expressing the reporter in undamaged controls for each region
(apex, middle, and base). Hair cell counts for the control and treated groups were
compared using the Student’s t test.
Sox2 lineage tracing recombination rates were estimated by counting the percentage
of inner pillar cells displaying reporter expression in each region. Lgr5 lineage
tracing recombination rates were similarly calculated for inner pillar and third Deiters
cells in each region.