Lineage-tracing and translatomic analysis of damage-inducible mitotic cochlear progenitors identifies candidate genes regulating regeneration

Cochlear supporting cells (SCs) are glia-like cells critical for hearing function. In the neonatal cochlea, the greater epithelial ridge (GER) is a mitotically quiescent and transient organ, which has been shown to nonmitotically regenerate SCs. Here, we ablated Lgr5+ SCs using Lgr5-DTR mice and found mitotic regeneration of SCs by GER cells in vivo. With lineage tracing, we show that the GER houses progenitor cells that robustly divide and migrate into the organ of Corti to replenish ablated SCs. Regenerated SCs display coordinated calcium transients, markers of the SC subtype inner phalangeal cells, and survive in the mature cochlea. Via RiboTag, RNA-sequencing, and gene clustering algorithms, we reveal 11 distinct gene clusters comprising markers of the quiescent and damaged GER, and damage-responsive genes driving cell migration and mitotic regeneration. Together, our study characterizes GER cells as mitotic progenitors with regenerative potential and unveils their quiescent and damaged translatomes.

PLOS Biology | https://doi.org/10.1371/journal.pbio.3001445 November 10, 2021 1 / 30 a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 In nonmammalian vertebrates, SCs also act as transit amplifying cells and HC precursors by proliferating and regenerating lost HCs [9][10][11]. By contrast, the mature mammalian cochlea neither regenerate nor proliferate, thus both HC and SC degenerations lead to permanent hearing loss. While the neonatal cochlea harbors SCs, particularly those expressing the Wnt target gene Lgr5, which are capable of proliferating and regenerating lost HCs, they are limited in numbers and spatially restricted, with regenerated cells being short-lived [12][13][14]. One prior study showed nonmitotic regeneration of the SC subtype inner phalangeal cells (IPhCs) following selective ablation, and that broadening the extent of ablation to include the greater epithelial ridge (GER) reduced regeneration, suggesting the presence of SC precursors therein [15].
Here, we selectively ablated Lgr5 + SCs using the Lgr5 DTR/+ mouse line [16] and found robust proliferation in the GER in the neonatal cochlea. As a transient structure during neonatal stages of cochlear development [17], the GER's role in development is incompletely understood [18]. We fate-mapped GLAST-Cre + GER cells in vivo and observed that they proliferated and migrated to replenish IPhCs. Regenerated cells expressed markers of nascent and mature IPhCs, displayed spontaneous calcium activity, and remained present in the mature cochlea. Using GLAST-RiboTag mice, a method to enrich transcripts in the GER region, we identified differentially expressed genes (DEGs) upon depletion of Lgr5 + cells. Together, our results show that severe cochlear SC loss stimulates mitotic regeneration by GER cells, thereby providing a framework that may guide regeneration of the mammalian cochlea.

Supporting cell regeneration after ablation of Lgr5 + cells
Sensory HCs and SCs in the cochlea are marked by expression of Myosin7a and Sox2, respectively [19,20]. SC subtypes can be grouped as those residing in the medial (GER and IPhCs) and lateral compartments (pillar and Deiters' cells (DCs)) (S1A Fig) [21]. To establish a model of SC ablation, we first examined cochleae from untreated Lgr5 DTR-EGFP/+ (Lgr5-DTR) mice [16]. This is a knock-in model in which human diphtheria toxin receptor (DTR) and EGFP are driven by endogenous Lgr5 expression, and DT administration induces rapid degeneration of Lgr5 + cells [16,22]. In the Lgr5-DTR cochlea, we found organized rows of Myosin7a + outer and inner hair cells (OHCs and IHCs) intercalated by Sox2 + SCs (Fig 1A). At postnatal day (P) 1, Lgr5-EGFP expression is restricted to specific SC subtypes (the third row of DCs, inner pillar cells, IPhCs, and lateral GER) (Fig 1A-A' , S1A Fig). These data corroborated previous in situ hybridization in wild-type cochlea and GFP expression in Lgr5 EGFP-CreERT2/+ knock-in cochleae [23].
To selectively ablate Lgr5 + SCs in the neonatal cochlea, we treated P1 Lgr5-DTR mice with DT (4 ng/g, IM or IP) and harvested cochleae 1 to 20 days later (Fig 1B). In control cochlea (saline-treated Lgr5 DTR/+ or DT-treated wild-type mice), we did not detect any cell loss and rarely observed pyknotic nuclei among DCs, pillar cells, or IPhCs from P4 to P21 (Fig 1C, 1G and 1I, S2A, S2C, S2E, S2G, S2I and S2K Fig, S1 and S2 Tables). In the P4 DT-treated Lgr5 DTR/+ cochlea, we found loss of Sox2 + SCs and EGFP expression in both the lateral and medial compartments in all 3 turns, with degeneration most severe among DCs, pillar cells (both >60%), and IPhCs (>50%) (Fig 1D and 1L, S2B and S2D Fig, S1 Table). SC loss in the lateral compartment was preceded by ectopic EGFP expression in the first and second row of DCs and outer pillar cells (S1B-S1I Fig). The IPhCs are defined as 2 rows of Sparcl1 + SCs medial and subjacent to the inner HCs (Fig 1C and 1E) [15,24]. After DT-induced damage, we detected many pyknotic nuclei and loss of Sparcl1 expression in the IPhC region, indicating cell death in all 3 cochlear turns at P4 (Fig 1D" and 1F, S2B" and S2D" Fig, S2 Table).
Three days later (P7), Sox2 + IPhCs along the whole length of DT-treated Lgr5 DTR/+ cochleae significantly increased and were replenished to control levels (Fig 1H, 1M and 1O, S2F and S2H Fig, S1 Table). On the other hand, Sox2 + SCs in the lateral compartment (DCs and pillar cells) modestly increased only in the apical and middle turns but remained significantly fewer than controls (Fig 1H, S2F, S2H, and S2M Fig, S1 Table). At P7, P14, and P21, Sox2 + IPhCs of DT-treated Lgr5 DTR/+ cochleae remained comparable to those in controls, whereas Sox2 + SCs in the lateral compartment degenerated in all 3 turns (Fig 1H, 1J and 1O, S2F, S2H, S2J, S2L and S2M Fig). Also, we found progressive degeneration of OHCs throughout the DTtreated Lgr5 DTR/+ cochleae beginning at P14 and P21 (Fig 1N and 1P). No degeneration of IHCs was detected (S2N Fig). These results indicate spontaneous regeneration of IPhCs and survival of IHCs in the medial compartment and progressive degeneration of SCs and OHCs in the lateral compartment, suggesting a compartmentalized regenerative response to ablation of Lgr5 + SCs.
Three days later at P7, the number of EdU + Sox2 + cells in the GER significantly decreased, while EdU + cells in the IPhC region significantly increased in each cochlear turn (Fig 2L-2O, S4 Table). As the GER normally degenerates between P7 and P10, these results suggest that replenished, but only few Sox2 + SCs in the PC/DC region remained in the DT-treated Lgr5 DTR/+ cochlea. Similarly, many Sox2 + IPhCs remained in the P21 DT-treated Lgr5 DTR/+ cochlea, while few Sox2 + SCs were found in PC/DC region. (K-N) Orthogonal views of control cochlea with Sox2 + SCs and Myosin7a + HCs at P4 cochlea (K). In the P4 DT-treated Lgr5 DTR/+ cochlea, there was loss of SCs including IPhCs (L). Progressive loss of Myosin7a + OHCs in the lateral compartment and regeneration of IPhCs and survival of IHCs in the medial compartment (M, N). (O) Normalized Sox2 + IPhC counts (per 160 μm) of DT-treated Lgr5 DTR/+ cochleae showing a significant loss at P4 followed by regeneration by P7 in each turn (n = 6 at P4, n = 8 at P7, n = 6 at P14, and n = 8 at P21). (P) There were significant losses of OHCs in all 3 turns of the DT-treated Lgr5 DTR/+ cochleae between P7 and P14. DT-treated wildtype and saline-treated Lgr5 DTR/+ mice served as undamaged controls. Data represent mean ± SD. �� p < 0.01, ��� p < 0.001. Two-way ANOVA with Tukey's multiple comparisons test. n = 4-8. See S1 Data for O and P. DCAU : Abbreviationli , Deiters' cell; DT, diphtheria toxin; GER, greater epithelial ridge; HC, hair cell; IHC, inner hair cell; OHC, outer hair cell; IPhC, inner phalangeal cell; PC, pillar cell; SC, supporting cell.
https://doi.org/10.1371/journal.pbio.3001445.g001 DT was injected into P1 wild-type or Lgr5 DTR/+ mice. EdU was injected daily from P3 to P5, and cochleae were examined at P4 or P7. (B-D, I-K) Representative images of the apical, middle, and basal turns of DT-treated wild-type cochleae showing no EdU + or Ki67 + Sox2 + SCs at P4 or P7. IPhC region is outlined by dashed lines. (E-G) In each turn of the P4 DT-treated Lgr5 DTR/+ cochlea, EdU + and/or Ki67 + Sox2 + cells were detected in the GER. Some EdU + and/or Ki67 + Sox2 + cells were also found in the IPhC and PC/ DC regions. E'-G""represent single channel images. Insets in E and E' are high magnification images of a cell in metaphase. G"' and G""are high magnification images from G' and G". (H) Significantly more EdU + Sox2 + cells were found in the P4 GER regions throughout the DT-treated Lgr5 DTR/+ cochlea relative to controls. Numbers decrease in an apical-basal gradient. (L-N) In all 3 turns of the P7 DT-treated Lgr5 DTR/+ cochlea, EdU + Sox2 + cells were primarily found in IPhC region, and only few were found in the GER or PC/DC region. Ki67 + Sox2 + cells were rarely detected at this age. L'-N""represent single channel images. N"' and N""are high magnification images from N' and N". (O) some EdU + GER cells have degenerated, while others migrated to repopulate lost IPhCs. EdUlabeled IPhCs remained present in the P14 and P21 cochlea, demonstrating survival of regenerated cells (S3B, S3D, S3F, S3H, S3J, S3L and S3M Fig, S4 Table). To pinpoint the timing of damage-induced proliferation, we immunostained and found that Ki67 + Sox2 + cells in both the GER and IPhC regions decreased at P7 relative to P4 and were not detected in any regions at P14 or P21 (Fig 2E-2G and 2L-2N, S3B, S3D, S3F, S3H, S3J and S3L Fig, S3 Table). Moreover, when we delayed EdU injection to P7 to P9, we found no EdU-labeled cells in the sensory epithelia (S3Q- S3S Fig). Collectively, these results indicate that SC loss stimulates robust, yet transient proliferation in the GER and damage-activated proliferative cells may migrate laterally to replace lost IPhCs within the first week after insult.
As Lgr5 + cells include both the IPhCs and lateral GER cells, the proliferative response observed in the GER of damaged Lgr5-DTR cochleae can be attributed to both the broader and overall more severe depletion of SCs.
To test whether proliferative cells in the GER serve as the source of regenerated IPhCs, we fate-mapped and compared the GLAST and Plp1 lineages, which represent cells that reside in the GER and IPhC regions (GLAST-Cre + ) and the IPhC region only (Plp1-Cre + ) (Fig 3J, S5A  Fig). We first performed fate mapping using the GLAST CreERT/+ mice [28]. Tamoxifen was administrated at P1 to induce labeling by the Cre reporter line Rosa26R-tdTomato, followed Significantly more EdU + Sox2 + cells were detected throughout the P7 IPhC regions of DT-treated Lgr5 DTR/+ cochlea relative to controls, decreasing in an apical-basal gradient. Data represent mean ± SD. �� p < 0.01, ��� p < 0.001. Twoway ANOVA with Tukey's multiple comparisons test. n = 4-7. See S1 Data for H, O. DC, Deiters' cell; DT, diphtheria toxin; GER, greater epithelial ridge; IPhC, inner phalangeal cell; PC, pillar cell; SC, supporting cell.
https://doi.org/10.1371/journal.pbio.3001445.g003 into Lgr5 DTR/+ ; Atoh1-mCherry; Pax2-Cre; R26 GCaMP3/+ and Atoh1-mCherry; Pax2-Cre; R26 GCaMP3/+ mice at P1 and cochleae were examined at P7. (B-C') Representative still images from the apical turn are shown (see S1 Video). Without damage, periodic calcium (EGFP) signals (ROI outlined in colors that correspond to the bars in the time line plotted) were detected in lateral GER cells and IPhCs surrounding the Atoh1-mCherry + IHCs. In the damaged cochlea, more EGFP + events were detected adjacent to IHCs, with each event spanning a wider area and appearing more intense when compared to controls. (D, E) Representative tracings of EGFP signals measured from individual ROIs in undamaged and damaged organs. Colors correspond to those in B'-C'. Tracing from 4 events from C' were ( Fig 4C and 4E, S1 Movie). Additionally, spontaneous calcium transients in the damaged cochleae were more intense and spanned larger areas relative to controls (Fig 4D-4H). These properties are reminiscent of SCs in the perinatal cochlea [30] and suggest the presence of an SC network incorporating the newly regenerated cells. Taken together, these results indicate that regenerated SCs exhibit characteristics of neonatal IPhCs and may be connected to native SCs.

Compartmentalized regeneration, cell survival, and maturation
Regenerated HCs in the neonatal cochlea are short-lived and undergo delayed degeneration [12,14]. In the lateral compartment, CD44 marks outer pillar cells and the lesser epithelial ridge (LER) cells, and Sox2 labels DCs and pillar cells (S6A Fig) [20,31]. The lateral compartment of P7 DT-treated Lgr5 DTR/+ cochlea displayed a loss of Sox2 + and CD44 + SCs and abnormally clustered Myosin7a + outer HCs (S6B Fig), confirming degeneration without significant regeneration. These results are consistent with previous results when DCs and pillar cells in the lateral compartment were ablated [32].
On the other hand, regenerated IPhCs in the medial compartment have been shown to survive in the mature cochlea [15]. To assess whether regenerated IPhCs in the Lgr5 DTR/+ mice survive and undergo maturation, we examined P14 and P21 cochleae. At these ages, IHCs have matured to express Vglut3, and IPhCs have begun to express the mature markers Na + /K + ATPase α-1 and GLAST (Fig 5A, 5C, 5E and 5G, S6C and S6E Fig) [33][34][35]. By P14, the GER in control cochlea has undergone apoptosis leading to the formation of the inner sulcus. In the lateral compartment of regenerated Lgr5 DTR/+ cochlea, SCs transiently and modestly increased between P4 and P7, before progressively degenerated between P7 to P21 (S2M Fig, S1 Table). Though present at P7, OHCs also degenerated at P14 and onwards (Figs 1M, 1N, 1P, 5B and 5D, S1 Table). Thus, regeneration in the lateral compartment is limited, and cells undergo progressive degeneration after ablation of Lgr5 + SCs. In stark contrast, robust regeneration and survival of regenerated cells were observed in the medial compartment of Lgr5 DTR/+ cochleae. At P14 and P21, there was no detectable loss of IHCs or IPhCs, which matured to express Vglut3 and Na + /K + ATPase α-1/GLAST, respectively (Fig 5F and 5H, S6D and S6F Fig, S9 Table). The numbers of IHCs and IPhCs in Lgr5 DTR/+ cochleae remained comparable to controls at P14 and P21 (Fig 5S, S2N Fig, S9 Table). At P14 and P21, mitotically regenerated tdTomato + IPhCs matured to express GLAST protein in the all 3 turns, indicating differentiation and long-term survival (Fig 5I, 5J-5J', and 5K-5L, S6G and S6H-S6H' Fig, S7 and S8 Tables). These results suggest that regenerated IPhCs are at least partially mature and remained viable in the mature cochlea.
Degeneration of HCs during the neonatal period causes retraction of the innervating fibers [36]. In the lateral compartment, Tuj1 + fibers project laterally to innervate OHCs (Fig 5M). In the damaged Lgr5 DTR/+ cochlea, no Tuj1 + fibers were detected in the lateral compartment, likely as a result of OHC degeneration (Fig 5N). However, in the medial compartment of DTtreated animals, there was no detectable loss of Vglut3 + IHCs (Fig 5S, S9 Table), which appeared to remain innervated and juxtaposed to the inner spiral plexus with radial fibers projecting from the spiral ganglia neurons (Fig 5M-5P). While the plexus was somewhat shown in E. (F-H) Frequency, areas, and relative intensity change of spontaneous calcium activities in damaged cochleae were significantly greater than those in controls. Data represent mean ± SD. � p < 0.05, �� p < 0.01, ��� p < 0.001. Unpaired Student t test or Mann-Whitney test. n = 5-239 cells from 5-7 cochleae. See S1 Data for F-H. DT, diphtheria toxin; GER, greater epithelial ridge; IHC, inner hair cell; IPhC, inner phalangeal cell; OHC, outer hair cell; ROI, region of interest.
To probe whether synapses remained present, we examined the presynaptic and postsynaptic markers CtBP2 and GluR2 (Fig 5Q) [37]. In the P21 DT-treated Lgr5-DTR cochleae, IHCs showed fewer CtBP2 + GluR2 + synapses relative to controls (Fig 5Q, 5R and 5T), suggesting synaptopathy despite cell survival. Consistent with these histological findings, auditory brainstem responses (ABRs) and distortion product otoacoustic emissions (DPOAEs) were absent across all frequencies in P21 DT-treated Lgr5 DTR/+ mice (S6I and S6J Fig). Collectively, these results indicate that ablation of Lgr5 + SCs caused limited regeneration and progressive degeneration in the lateral compartment, while inducing robust regeneration and survival of IPhCs in the medial compartment of the mature cochlea. In addition, ablation of Lgr5 + SCs caused synapse loss in IHCs and loss of OHCs, resulting in profound hearing loss.

Translatomic analysis of GLAST-Cre + cells
To begin to determine the molecular mechanisms directing mitotic regeneration, we first analyzed the arrays of genes expressed by undamaged and damage-activated GLAST-Cre + cells using the transgenic RiboTag mouse [38,39]. This allele permits the immunoprecipitation (IP) of ribosomes from Cre + cells, enriching for cell type-specific actively translated mRNAs, or translatomes. We generated GLAST CreERT/+ ; Rpl22 HA/+ (GLAST-RiboTag, from hereon "control") and Lgr5 DTR/+ ; GLAST CreERT/+ ; Rpl22 HA/+ (GLAST-RiboTag-DTR, from hereon "DTR") for experiments. Control and DTR mice were injected with DT and tamoxifen at P1, and cochleae were harvested at P4, a time point when damaged-induced proliferation was robust (Fig 6A-6C ). RNA was extracted both from whole sensory epithelia (input) as well as ribosome immunoprecipitated samples (IP) to enrich for translatomes of GLAST-Cre + cells and allow calculation of an enrichment factor (EF) for each gene [40,41]. This index was then used to gauge whether signal originates in the cell type immunoprecipitated (EF > 2) or primarily from other cell types.
Unsupervised hierarchical clustering analysis identified striking translatomic differences between sample types (input versus IP) and treatment groups (control versus DTR), which, in a principal component analysis (PCA), correspond to PC1 and PC2, respectively, while genetic convergence among the biological replicates was seen within each experimental group (Fig 6D  and 6E, S10 Table).
https://doi.org/10.1371/journal.pbio.3001445.g005  Student t test). n = 4. See S1 Data for G and Cdkn1b, and Dlx5) using nCounter (Fig 6G, S7A Fig). In addition, we performed in situ hybridization and found Igf2 and Igf2pb1 to be spatially enriched in the control and damaged GER, thereby confirming that GLAST-Cre + cells are representative of GER cells (Fig 6F, S7B-S7D Fig). While our data on Igf2 and Igfbp3 are consistent with previous results [42], other genes have not been characterized in the cochlea and therefore serve as novel markers of the GER. These data indicate that GLAST-Cre + GER cells have unique molecular signatures in both the undamaged and damaged cochleae.

Discussion
Cellular plasticity in several mammalian organs, such as the adult liver, pancreas, and brain, facilitates regeneration in response to injury [54]. While the mature cochlea lacks the ability to regenerate, the neonatal mouse cochlea harbors Lgr5 + progenitors that act to replace lost HCs [12,14] and can serve as a renewable source of sensory cells [55,56]. Previously, using multiple CreER lines to express DTA, Mellado Lagarde and colleagues reported that the GER cells nonmitotically regenerate the SC subtype IPhCs [15]. Here, through the use of fate mapping and EdU pulse-chase experiments in the Lgr5-DTR mouse line, we reveal that loss of Lgr5 + SCs activates GER cells to proliferate and migrate into the organ of Corti to replace the lost IPhCs. Notably, these newly regenerated cells mature to display features of differentiated SCs (i.e., expression of Glast, Na + /K + ATPase α-1, and Fabp7) and survive in the mature cochlea and likely represent the source of SC regeneration noted previously [15]. Lastly, we have unveiled the translatomes of quiescent and damage-activated GER cells and candidate genes, which may regulate mitotic regeneration in the neonatal cochlea.

Discovery of a hidden progenitor cell population in the cochlea
Cochlear SCs also serve the critical function of maintaining ionic homeostasis, mediating purinergic signaling required for spontaneous calcium activity, and supplying neurotrophic support essential for afferent innervation [5][6][7][8]. Despite the multiple roles of SCs, the mechanism by which SCs are regenerated is largely unknown.
The GER has long been known as a transient structure located adjacent to the organ of Corti; however, its function remains incompletely understood [17,57]. Previously, it has been shown that ablation of GER cells limits the degree of IPhC regeneration, suggesting that the GER harbors precursors that can regenerate lost IPhCs [15]. Using lineage tracing, our study unambiguously shows that GLAST-Cre + GER cells mitotically regenerate SCs in the organ of Corti after injury. In the previous and current study, proliferation of GER cells was not detected in the damaged Plp1Cre ERT/+ ; Rosa26 DTA/+ cochleae likely because cell loss was less severe and restricted only to the IPhCs. In contrast, both IPhCs and lateral GER cells were concurrently ablated in the Lgr5-DTR model, as Lgr5 is more broadly expressed. This is evinced by the escalating degrees of damage leading to progressively more proliferation (S6 Table). The higher degree of damage in the Lgr5-DTR model also likely accounted for the synapse loss in IHCs and profound hearing loss, neither of which was observed using the Plp1-DTA model [15]. Despite proliferation and subsequent differentiation into IPhCs, GER cells appeared lineage-restricted and did not give rise to HCs, as noted by the absence of any EdU-labeled HCs. This suggests that GER cells act primarily as progenitors for SCs. In support of this notion, after HC ablation, fate mapping experiments also failed to detect regenerated HCs deriving from the GER [58]. On the other hand, isolated GER cells can proliferate and form HC-like cells [59,60], underscoring their context-dependent ability to act as HC progenitors. In the zebrafish lateral line system, 2 independent studies have demonstrated that SCs in the dorsal/ ventral poles proliferate and act as transit amplifying cells and that mantle cells on the periphery divide to replenish SCs after severe injury [61,62]. As mitotic progenitors for SCs, GER cells share features with these SCs and mantle cells from the neuromasts.
In our study, although the IPhCs were replenished, we did not observe a return of Lgr5-EGFP signal. This result contrasts with those obtained with self-renewing organs where Lgr5 + cells reemerge after damage [16,22]. This difference may be because Lgr5 expression in the medial compartment decreases in the neonatal cochlea and regenerated IPhCs have likely matured past this developmental stage [23] and is also in agreement with the diminishing regenerative capacity as the neonatal cochlea matures [12,14]. Furthermore, regenerated IPhCs displayed several molecular markers (e.g., Fabp7, GLAST, Na + /K + ATPase α-1) and corresponding physiological functions, suggesting some degree of differentiation and maturation. Of note, calcium transients appeared highly active in the damaged cochlea despite a down-regulation of Ano1. This is surprising since Ano1 knockout mice were reported to exhibit significantly fewer spontaneous calcium activities [7]. One possibility is that the hyperactive calcium transients resulted from high levels of extracellular potassium released by degenerating cells. An alternative explanation is that the expanded calcium transients may be as a result of the damage itself. Both of these possibilities warrant further investigation.

Compartmentalized degeneration/regeneration in the neonatal cochlea
After HC ablation in the neonatal cochlea, mitotic regeneration is limited to the apical turn, and regenerated HCs are short-lived [12,13]. In contrast, mitotic regeneration of IPhCs occurred along the length of the cochlea, and regenerated cells remain present in the mature organ. Interestingly, SCs in the lateral compartment in each cochlear turn also proliferated (S4 Table), though to a lesser degree than the medial compartment and regenerated cells underwent delayed degeneration.
Similarly, ablation of Prox1 + SCs in the lateral compartment induced secondary OHC degeneration, without affecting cell survival in the medial compartment [32]. The degeneration of the lateral compartment SCs and OHCs observed in our model may be due to a previously unreported up-regulation of Lgr5 expression in the first and second rows of DCs and outer pillar cells in P2 and P3 Lgr5-DTR mouse cochlea (S1 Fig). Alternatively, the degeneration of the third row of DCs may lead to a secondary degeneration of the remaining DCs and outer pillar cells, and subsequently OHCs in a noncell autonomous fashion. Interestingly, the time course of delayed degeneration is similar to that observed following HC ablation, suggesting that the lateral compartment lacks of survival factors following regeneration [12,13]. While the robust proliferation observed in the medial compartment can be attributed to the innate regenerative capacity of GER cells, it is also possible that injury targeting SCs provokes a different and broader response than HC loss alone.
Among the damage-up-regulated genes (clusters 3, 4, 8, and 9), it is notable that Egr1related gene expression overlaps spatially with proliferative cells, particularly in the GER. In the liver, loss of function of Egr1 partially attenuated regeneration [50]. Egr4 acts upstream and Atf3 downstream of Egr1 in other contexts [63,64]. Whether they play a role in mitotic regeneration in the damaged cochlea should be of interest in future studies. Of note, cluster 2 consists of genes differentially expressed in input relative to IP samples, with the former collected from whole cochlea and the latter enriched GER cells. As such, cluster 2 genes contained numerous HC genes (S13 Table).
During development, both HCs and SCs arise from a common precursor domain [65,66]. Though GER cells express some molecular markers similar to organ of Corti SCs (e.g., Sox2, Jag1), they are morphologically, functionally, and molecularly distinct [18,24]. The current study reveals that GER cells display a unique molecular profile and that they are capable of dividing and migrating laterally to replace lost IPhCs. Regeneration is restricted to the medial compartment with no migration to the lateral compartment detected, implicating that the regenerative and migratory potentials for GER cells may be domain specific.
In summary, our work validates previous studies and further characterizes a hidden progenitor cell population in the GER that may serve as an endogenous source for SC progenitors. The unique translatomes of the quiescent and activated GER cells also set the foundation for future mechanistic studies on mitotic regeneration in the mammalian cochlea.
To activate Cre recombinase, tamoxifen dissolved in corn oil (0.075 mg/g for Plp1 CreERT/+ mice and 0.2 mg/g for GLAST CreERT/+ mice, IP, Sigma) was administered at P0 to P1. EdU (25 μg/g, IP, Invitrogen) was injected to label proliferative cells. Institutional Animal Care and Use Committee of Stanford University School of Medicine approved all procedures in accordance with NIH guidelines (protocol #18606).

Genotyping
Genomic DNA was isolated from collected tail tips by adding 180 μl of 50 mM NaOH and incubating at 98˚C for 1 hour, followed by the addition of 20 μl of 1 M Tris-HCl. PCR was performed to genotype transgenic mice with the listed primers in S14 Table. Immunohistochemistry Methods modified from those previously reported [13]. Briefly, isolated cochleae were fixed in 4% paraformaldehyde (PFA) (in phosphate buffered solution (PBS) (pH 7.4) Electron Microscopy Services) for 40 to 60 minutes at room temperature. Cochleae isolated from P10 or older animals were decalcified in 0.125 M EDTA for 48 hours at 4˚C. Cochleae processed for cryosection were immersed overnight in 30% sucrose, then flash frozen in optimal cutting temperature (OCT) compound, and sliced into 10 μm thick sections. Tissues were permeabilized with 0.5% TritonX-100 (in PBS) for 1 hour at room temperature, and then blocked with 5% donkey serum, 0.1% TritonX-100, 1% bovine serum albumin, and 0.02% sodium azide (NaN3) in PBS (pH 7.4) for 1 hour at room temperature. This was followed by incubation with primary antibodies in the same blocking solution overnight at 4˚C. The following day, tissues were washed with PBS 3 times at 5-minute intervals and then incubated with secondary antibodies diluted in PBS containing 0.1% TritonX-100, 1% bovine serum albumin, and 0.02% NaN3 for 2 hours at room temperature. After washing with PBS 3 × 5 minutes, tissues were mounted in ProLong Gold Antifade Mountant (Invitrogen) and coverslipped.

Ribosome immunoprecipitation and RNA extraction
Cochleae were isolated from P4 Lgr5 DTR-EGFP/+ ; GLAST CreERT/+ ; Rpl 22HA/+ and GLAST CreERT/+ ; Rpl 22HA/+ mice (spiral ganglia, lateral wall, and Reissner's membrane were removed) in cold Hanks' Balanced Salt Solution (HBSS) and then immediately flash-frozen using liquid nitrogen prior to RNA isolation. About 2 to 8 cochleae were pooled and processed for ribosome IP followed by RNA extraction as described [40]. Briefly, cochlear ducts were homogenized and incubated with 5 μg purified anti-HA.11 (BioLegend) for 6 hours at 4˚C before the addition of the equivalent of 300 μl Dynabeads protein G (ThermoFisher Scientific) and further incubation overnight. Additionally, 5% of the homogenate was kept before addition of the antibody to be used as input. RNA was extracted from the IP and input samples using the RNeasy Plus Micro kit (Qiagen) following the manufacturer's instructions. RNA quality was confirmed with a BioAnalyzer 2100 picochip (Agilent Technologies) performed at the Genomics Core Facility in the Center for Innovative Biomedical Resources (University of Maryland School of Medicine). All RNA integrity numbers (RINs) were above 8.

Assessment of IP efficiency by real-time RT-PCR
Cell type-specific RNA enrichment following the IP was assessed by reverse transcription (Maxima First Strand cDNA Synthesis Kit for RT-qPCR, ThermoFisher Scientific) using equivalent amounts of IP and corresponding input samples, followed by real-time PCR (Maxima SYBR Green/ROX qPCR Master Mix, ThermoFisher Scientific) on a StepOnePlus Real-Time PCR System (Applied Biosystems). The primers used are listed in S14 Table. RNAseq of RiboTag samples Four individual biological replicates were used for sequencing, with each replicate comprised of 8 cochleae from 4 mice. About 1 to 3 ng of RNA from the input and IP samples were used as template for library preparation with the SMART-Seq v4 Ultra Low Input RNA Kit for Sequencing. Libraries were sequenced on an Illumina HiSeq 4000 at 75 paired-end read length and a depth of sequencing of 74 to 100 million reads per sample. Library preparation and sequencing were performed at the Institute for Genome Sciences, University of Maryland School of Medicine. The RNA sequencing data generated in this paper are available from GEO with accession number GEO: GSE135728. The data are also available at gEAR, a gene Expression Analysis Resource (https://umgear.org), via PERMA-LINK https://umgear.org/p?s= 3296013a.

RNAseq normalization and expression analysis
Quality of Fastq files were evaluated by using FastQC. Reads were aligned to the mouse genome (Mus_musculus.GRCm38) using HiSat (version HISAT2-2.0.4) [69], and the number of reads that aligned to the coding regions was determined using HTSeq [70]. Approximately 171,121 genes were assessed for significant differential expression using DESeq2 with an FDR value �0.05 [71]. The RiboTag immunoprecipitated samples and their corresponding input were compared to generate an EF. Transcripts were considered enriched in the IP if the log2 of the ratio between the IP and the input was �2. For DEGs between the IP samples, the criteria of LFC �1 between DTR IP and Control IP samples was used. Additionally, to avoid overinflation of fold change, all the values were quantile normalized CPM, and values lower than 10% of the value of the dataset were replaced with the 10th quantile. Furthermore, only genes with cutoff >1.5 (min(group1) / max(group2) were considered for differential expression or enrichment. For hierarchical clustering, similarity between samples was measured using Pearson correlation, as such the samples' distance was calculated as 1-Pearson correlation coefficient for each sample pair. The heat map was generated using the R packages pheatmap and RColor-Brewer. PCA was done using DESeq2. Gene Ontology was performed using DAVID [72,73].

NanoString reactions and analysis
The nCounter technology from NanoString measure gene expression at the RNA level. Nano-String reactions were performed in technical duplicates on the 4 biological replicates used for RNA-seq plus 1 independent fifth replicate. About 1 ng of RNA from input and IP samples was used for preamplification with the nCounter Low RNA Input Kit (NanoString Technologies, WA) to obtain sufficient cDNA to be run on the Counter platform (NanoString Technologies) (Primers for preamplification are listed in S14 Table). A Custom CodeSets for 28 targets including 4 housekeeping genes was designed (probes listed in S14 Table), and the samples were run with the nCounter Master Kit (NanoString Technologies) following the manufacturer's protocol. Preamplification, nanoString reactions, and quality check steps were performed at the Institute for Genome Sciences, University of Maryland School of Medicine.
Normalization to housekeeping genes and data analysis was performed using the nSolver 4.0 analysis software (NanoString).

Clustering analysis
Prior to clustering, conditional quantile normalization (CQN) was performed to correct for sample-specific gene length effect in the RiboTag IP samples [74]. Genes were included in the analysis only if (1) their expression level was readily detected in at least one of the biological conditions (i.e., their expression was at least 0.5 cpm in all replicate samples of at least one of the 4 conditions); (2) showed significant differential expression (FDR < 0.05 and FC > 1.5) either between "cell type" (i.e., between input and IP samples) or in response to DT-induced damage; and (3) expression levels of a DEG were fully separated between the 2 conditions the DE was called. Overall, 5,095 genes met these criteria and were subjected to cluster analysis to detect the main expression patterns in the dataset. Cluster analysis was done using the CLICK algorithm implemented in the EXPANDER package [43].

In situ hybridization
Harvested tissues were fixed in 4% PFA overnight at 4˚C, embedded for cryosections, and prepared as 10 μm sections. Tissue sections were hybridized with probes from Advanced Cell Diagnostics (ACDbio) and counterstained with hematoxylin (Sigma-Aldrich) according to the manufacturer's instructions for fixed frozen sections with colorimetric detection. Briefly, sections were washed once in PBS for 5 minutes and then treated with H 2 O 2 for 10 minutes. Next, sections were permeabilized using target retrieval reagent (ACDbio) and proteinase before hybridization.

Auditory physiology measurements
ABRs and DPOAEs were measured as previously described [75]. Briefly, P21 mice were anesthetized with a ketamine/xylazine mixture (100 mg/kg ketamine and 10 mg/kg xylazine, IP) and placed on a heating pad at 37˚C. ABRs were recorded from a needle electrode located inferior to the tympanic bulla, referenced to an electrode on the vertex of the head, and a ground electrode was placed at the hind limb. Tone pip stimuli were delivered with frequencies ranging from 4 to 46 kHz (4.0, 5.7, 8.0, 11.3, 16.0, 23.0, 31.9, 46.1 kHz) up to 80 dB sound pressure level (SPL) in 10 dB steps. At each frequency and SPL, 260 trials were tested and averaged. DPOAEs were recorded with a probe tip microphone placed in the auditory canal. Two 1-second sine wave tones of different frequencies (F2 = 1.22 × F1) were used as the sound stimuli. F2 ranged from 4 to 46 kHz (4.0, 5.7, 8.0, 11.3, 16.0, 23.0, 31.9, 46.1 kHz), and the 2 tones were from 20 to 80 dB SPL in 10 dB steps. The amplitude of the cubic distortion product was recorded at 2 × F1-F2. The threshold was calculated as a cross point of DPOAE signal with the noise floor level above 3 standard deviations at each frequency. For statistical analyses of both ABR and DPOAE thresholds, a lack of a response was defined as the highest sound level, 80 dB SPL.

Statistical analyses
Statistical analyses were performed using Microsoft Excel (Microsoft) and GraphPad Prism 7.03 (GraphPad). For comparison of 2 groups, Student t test and Mann-Whitney test were used. One-way ANOVA was used when comparing more than 2 groups, and a two-way ANOVA was used for comparison with 2 independent variables. p < 0.05 was considered statistically significant. Representative images of the apical, middle, and basal turns of DT-treated wild-type (control) undamaged cochleae showing no EdU + or Ki67 + Sox2 + SCs at P14 or P21. The IPhC region is outlined by dashed lines. In the DT-treated P14 and P21 Lgr5 DTR/+ cochlea, many EdU + Sox2 + cells remained in IPhC region in all 3 turns. No Ki67 + Sox2 + cells were detected in the IPhC or PC/DC regions. (M-P) Quantification of EdU + Sox2 + cells in the apical, middle, and basal turns of control (DT-treated wild type) and damaged (DT-treated Lgr5 DTR/+ ) cochleae at P14 and P21. Wild-type cochleae had almost no EdU + Sox2 + cells at both ages examined. Conversely, there were significantly more EdU + Sox2 + IPhCs in the damaged cochleae with an apex-to-base gradient at both P14 and P21. Some EdU + Sox2 + cells survived in PC-DC regions in the damaged cochleae in all 3 turns at both ages. (Q) Schematic showing DT administration to P1 wild-type or Lgr5 DTR/+ mice. EdU was injected daily from P7 to P9, and cochleae were examined at P10. (R, S) In both DT-treated wild-type (control) and DT-treated Lgr5 DTR/+ cochleae, there were no EdU + or Ki67 + Sox2 + cells. Data represent mean ± SD. � p < 0.05, ��� p < 0.001 (two-way ANOVA with Tukey's multiple comparisons test). n = 3-5. See S1 Data for M-P. DC, Deiters' cell; DT, diphtheria toxin; GER, greater epithelial ridge; IPhC, inner phalangeal cell; PC, pillar cell; SC, supporting cell. DT and tamoxifen were injected into the P1 Plp1 CreERT/+ ; R26R tdTomato/+ (control) or Lgr5 DTR/+ ; Sox2 CreERT2/+ ; R26R tdTomato/+ (damage) mice. EdU was injected daily from P3 to P5, and cochleae were examined at P3, P7, or P14. (C-E) Representative images of the apical turn of control cochleae showing Plp1-tdTomato + Sox2 + or SCs at P3 and P7. At P14, Plp1-tdTomato + IPhCs expressed GLAST. IPhC region outlined by dashed lines. (F-H) In damaged cochleae, Plp1-tdTomato + Sox2 + cells were not detected in the IPhC region or in the GER at any age. Conversely, EdU + Sox2 + Plp1-tdTomato-negative cells were detected in the GER at P3. At P7 and P14, EdU + Sox2 + Plp1-tdTomato + cells were not found in the GER/IS or IPhC regions. Many EdU + Sox2 + Plp1-tdTomato-negative cells were detected in the IPhC region at P7 and P14. Orthogonal views shown in C'-H'. (I) Quantification of Plp1-tdTomato + Sox2 + SCs in the apical turn. (J, K) Quantification of EdU + Sox2 + SCs in the apical turn. In damaged cochleae, there were no Plp1-tdTomato + Sox2 + SCs in the GER or in the IPhC region at any age. There was, however, an increase in Sox2 + EdU + Plp1-tdTomato-negative cells in the GER peaking at P7, followed by a reduction at P14. There were no Sox2 + EdU + Plp1-tdTomato-negative cells in the IPhC region at P3, but many at P7 and P14. Data represent mean ± S.D. ��� p < 0.001 (two-way ANOVA with Tukey's multiple comparisons test). n = 4. See S1 Data for I-K. DC, Deiters' cell; DT, diphtheria toxin; GER, greater epithelial ridge; IHC, inner hair cell; IPhC, inner phalangeal cell; IS, inner sulcus; LER, lesser epithelial ridge; OHC, outer hair cell; PC, pillar cell; SC, supporting cell.