Fast recovery of hearing function after sisomicin-induced deafness in Xenopus laevis
Aminoglycoside antibiotics, like sisomicin, can specifically induce hair cell death and result in deafness. The chicken hair cells were specifically dead after sisomicin treatment 17. Therefore, in this study we successively injected the sisomicin into middle ear from tympanum to induce hair cell death in Xenopus, until loss of all hair cells and the ABR threshold reaches to 90 dB (complete deafness). Then, sisomicin treatment was stopped to allow hair cells to regrow and restore the hearing. Our ABR recording showed that the most sensitive auditory threshold of Xenopus, which normally is about 55 dB at 1.4 kHz (Figure S1A) 32, was increased to 90 dB after the sisomicin treatment (Fig. 1A and S1A). Remarkably, the hearing function started to be fast recovered after one week duration of complete deafness status (90 dB threshold). After two weeks, the ABR threshold almost recover to normal level (Fig. 1A, only showing 1.4 kHz frequency). The fast recovery of hearing function indicates its remarkable capacity for hair cell regeneration and hearing restoration in Xenopus.
Xenopus species holds a pivotal evolutionary position among vertebrate animal models, bridging the gap between fishes, birds and mammals as illustrated in the maximum likelihood phylogenetic tree (Fig. 1B). The regenerative capacities of these animal models are in concordance with their positions (Fig. 1B, the green color indicates the regeneration potentials) 33. The inner ear structure is conserved among vertebrates, especially the vestibular parts, which include three semicircular canals, utricle and saccule. However, the auditory parts exhibit more evolutionary variations from amphibians to mammals (Fig. 1C), corresponding to their diversified hearing functions for better adaption. The conservation and diversity of each animal models make them valuable tools for studying hair cell regeneration.
Comparative transcriptome analyses across vertebrates unveiled a natural feature of inner ears from species with high regenerative capacity
To understand the differences of gene expression programs between species with differential regenerative capacities, we conducted bulk RNA-seq analysis and cross-species profiling of transcriptomes for inner ears of Xenopus, mouse, and chicken. We found that approximately half of the genes, 12,605 genes, are co-expressed in the inner ears of both Xenopus and mouse, and 11,360 co-expressed genes between Xenopus and chicken. Importantly, we identified 10,858 genes co-expressed in these three species (Fig. 1D), indicating highly conserved transcriptomes in inner ears across species. Notably, the number of co-expressing genes was expected to be higher but was impaired by quality issues for Xenopus genome annotation and identification of homologous genes across species. For examples, 16,361 unannotated genes in Xenopus were named and initiated by LOC (Dataset S1). Comparing to Human Deafness Database 34, 226 deafness genes (nearly 90% of total deafness genes in the database) were found in the 12,605 co-expressed genes between Xenopus and mouse (Figure S1B). Additionally, 206 deafness genes were overlapped with the 10858 common genes among Xenopus, mouse, and chicken (Fig. 1D, rightmost), suggesting the importance and conservation of these human deafness genes among vertebrates. Most of the deafness genes were detected among three species, excepted several hair cell-specific deafness genes ESPN, SLC17A8, POU4F3, TMC1, and KCNQ4 (Dataset S2) because of the relatively low abundancy of hair cells in whole inner ear. Therefore, the common conserved gene list (10,652, Fig. 1D, rightmost) provides a valuable resource for exploring unknown and novel deafness-associated genes.
To better understand the differences of transcriptomic features between Xenopus and mouse inner ears, we classified genes into four groups into according to the differential gene expressional level in the overlapped genes. We are particularly concerned about the group 1 and 2 genes, which show higher expression levels in Xenopus and lower levels in mouse, or vice versa. Interestingly, many genes crucial for inner ear development and hair cell regeneration, such as Atoh1, Tbx2, Insm1, Bdnf, Gdnf, and Sox15 in group 1, showed higher transcriptional levels in Xenopus than that in mice (Fig. 1E, labeled in red). The genes Atoh1 and Tbx2 are widely recognized for their robust function in inducing hair cell regeneration 8,9,35. Notably, Atoh1 is spatiotemporally expressed at the early developmental stage of inner ears rather than the adult stage in mice 36 and is hardly detected in our adult mouse inner ear bulk RNA-seq dataset as expected (Dataset S2). Strikingly, the expression level of Atoh1 in adult Xenopus is over 100-fold higher than that in mice (Log2(FC) = 6.77, P value = 1.37E-21) (Dataset S2). The expression level of Tbx2 was also higher in Xenopus than that in mouse (Dataset S2). Consistently, the group 1 genes are highly enriched in the pathway of “inner ear development, axonogenesis, cell fate commitment, neural precursor cell proliferation, and regeneration” revealed by the top terms of gene ontology (GO) analysis (Fig. 1F, upper). In contrast to group 1, the group 2 genes showed lower expression level in Xenopus than that in mice (Fig. 1E), significantly associating with the biological pathway of “negative regulation of cell cycle, autophagy, regulation of neuron death, and neuron apoptotic process” (Fig. 1F, lower). For example, Cdkn1a, Cdkn1b, and Rb1 genes, displayed the inhibitory function in the cell cycle or proliferation (Dataset S2). Importantly, Cdkn1b (p27Kip1) gene, which regulates cell cycle exit and cell differentiation of prosensory progenitor cells during inner ear development 37, showed 20.6-fold higher expression level in mouse than Xenopus (Log2(FC) = 4.37, P value = 3.20E-172) (Dataset S2). The transcriptomic analyses among the three species revealed the high expression level of putative regeneration-related genes in both Xenopus and chicken (Figure S1C). These findings revealed the underlying mechanisms for the hair cell regeneration in species with high regeneration capacity.
Single cell atlas of Xenopus and mouse inner ear
To further characterize the Xenopus inner ear with high regeneration capacity at the cellular and molecular levels, we performed single-nucleus RNA-seq (snRNA-seq) experiments with the whole inner ears to comprehensively explored the cell identity (Fig. 2A). Additionally, snRNA-seq experiment of mouse inner ear was performed as a technical control for comprehensive across-species comparison. After filtration of doublets or nucleus with low quality, a total number of 12,646, 11,275 and 10,977 single nuclei from the whole inner ears, auditory parts of Xenopus and mouse cochleae were obtained, respectively.
As a reference, the mouse snRNA-seq data were analyzed, and the cell atlas of mouse inner ears was profiled (Fig. 2B and 2C). The uniform manifold approximation and projection (UMAP) algorithm revealed a total of 24 distinct cell types in the mouse inner ear. All the well-known inner ear cell types were detected that include hair cells, supporting cells, basal cells, marginal cells, intermediate cells, fibrocytes, neurons, and Schwann cells identified by the reported gene markers 38 (Fig. 2B, 2C and S2A). The recently discovered cell types 38 were also identified in our snRNA-seq dataset, including the Scala Vestibuli Border cells (SVBs) (integrated with the basal cell population, Figure S2C). Especially, the well-characterized mouse spiral ganglion neurons subtype I (SGN type I, including IA, IB, and IC three subtypes) and II were clearly distinguished in our snRNA-seq dataset (Fig. 2B and 3C), consistent with previous single-cell studies 29,39. Remarkably, a new neuronal cell type was identified with a differential expression pattern of neuronal genes in our mouse cochlear atlas, distinct from the known neuronal types I and II (Fig. 2B, 2C, 3E and S3B) 29,39. This newly discovered Type III neuron showed a stronger transcriptional feature with neuronal marker genes Nefh, Nefm, Nefl, and Chga (Fig. 2C and 3E) than types I and II. However, the type I or II neurons specially expressed genes Cacna1b, Grm8, Pou4f1, Th, Cilp, Grik3, and Ryr2 were all silenced or slightly expressed in type III neurons (Fig. 3E). The discovery of new neuronal cell types indicated that the snRNA-seq approach in this study is a sensitive and suitable method for the inner ear characterization and cell atlas profiling, and technically different from the previous scRNA-seq approach using the cell suspension 9,31,39.
Next, snRNA-seq experiments of the whole inner ear in Xenopus were performed. The Xenopus inner ear cells can be easily identified and divided into 21 cell populations by using the reported and well-recognized markers from mice, such as hair cells (Otof and Loxhd1), spiral ganglion neuron (Snap25, Pvalb, and Calb), Schwann cells (Mpz and Pmp22), fibrocytes (Coch), marginal cells (Kcnq1), and basal cells (Tjp1) (Fig. 2D, 2E, and S2B). Due to the complicate heterogenicity and evolutionary variations, the identification and distinguishing of supporting cell-like subtypes, such as inner phalangeal cells, Deiter’s cell, Hensen’s cell, Claudius cells, inner and outer pillar cells in Xenopus, was quite difficult. According to the reported supporting cell markers in mouse, the supporting cells were identified in Xenopus and subdivided into supporting cell-like type I, II, III and IV (Fig. 2D and 2E). Cells that did not support by any markers are named as undefined cell type 1 and 2 and remain to be further characterized.
The amphibian papilla (AP) and basilar papilla (BP) are the auditory organs of the amphibians, responsible for detection of the low-frequency and high frequency sounds, respectively. For better understanding the auditory organs of Xenopus, we further conducted the snRNA-seq experiment specifically for the AP-BP parts of Xenopus and identified a total of 19 cell populations (Fig. 2F and 2G). The UMAP algorithm revealed a good overlapping of most of cell types of the whole inner ear and AP-BP part in Xenopus (Figure S3-1A). In contrast to three neuronal cell types found in the whole inner ear, only one neuronal cell type was identified in AP-BP (Fig. 2E and 2G). However, a few of unknown cell types can be hardly defined by currently available markers (Fig. 2G).
Cross-species comparison studies confirmed an exceptionally conserved neuronal cell type
To explore cell type similarity and evolution across species, we integrated the Xenopus AP-BP and mouse cochlear cell atlas using canonical correlation analysis (CCA) and dimensional reduction by UMAP algorithm. The integrated cell atlas exhibited overlapped major cell types and separated species-specific cell types between mouse and Xenopus (Fig. 3A, S3-1 B and D). The SAMap algorithm 40, which calculates similarity coefficients between cell types, further confirmed cell type similarity across species. Auditory hair cells, the most critical cell type in inner ear, shared a 73.42% similarity coefficient between Xenopus AP-BP and mouse cochlea. Similarities coefficients of other cell types, such as Schwann cells (88.27%), marginal cells (85.93%), and endothelial cells (75.19%) (Fig. 3B), also suggested conservation of cell atlas features across vertebrates. Surprisingly, the Xenopus neuron cells showed the highest similarity (89.11% for AP-BP or 94.98% for whole inner ear) (Fig. 3B and S3-1C) with the mouse neuronal type III cells, which were first identified in this study (Fig. 3). However, the SGN type I and II in mouse show very low similarity with all other Xenopus neuronal cell types (Fig. 3B). Only a portion of the neuronal cells were clustered together when merged by UMAP algorithm (Figure S3-1E). The relative low similarity of SGN type I or II suggested that they are newly evolved in mammals. In contrast, the extreme conservation of SGN type III cross species revealed its importance and ancient function in hearing.
The new SGN subtype III exhibited the strongest Nefh expressing compared to other known neuronal subtypes in mouse cochlea (Fig. 2C and 3E). Immunostaining experiments indicated that cells with the strongest expressing of Nefh are all Tuj1 + cells (β-Tubulin 3, a common marker of neurons), confirming that this cell type belongs to neuronal cells (Fig. 3G). In addition, RNA in situ experiments detected an uneven distribution of Nefh RNAs in spiral ganglions, consistent with the distinct Nefh expressing features for each SGN subtypes revealed by single-cell atlas (Fig. 3E). Interestingly, the highest Nefh expressing cells are predominantly located in the spiral ganglions at the regions close to the sensory epithelium (hair cells) (Fig. 3G, section direction: vertical to circular plane of cochlea). Section in different direction confirmed the differential expression pattern of the Nefh gene in ganglion regions (Figure S3-2A, as indicated by arrows). A few of sporadic and loose Nefh high-expression cells were found near the central modiolus region (Figure S3-2A). RNA in situ experiments of nefh were performed in Xenopus in parallel. The nefh gene displayed differential expression pattern among neuronal cell types, especially showing strongest expressed in ganglion neuron type III, despite the failure of immunodetection with Tuj1 in Xenopus due to species-specific antibody issue (Fig. 3H and S3-2B, weak signals indicated by arrow). Other neuronal cell types in Xenopus were further analyzed and confirmed by detection of RNAs of cemip gene (Fig. 3I). In contrast of the three neuronal cell types in whole Xenopus inner ear, only one neuronal cell type was identified in AP-BP, which is analogous to the mouse SGN type III cells (Fig. 2E, 2G and 3B). In total, across species analysis of the new SGN cell type provided us new insights into the evolution and function of SGNs.
Characterization of hair cells in Xenopus inner ear revealed their high conservation in vertebrates
The bulk RNA-seq revealed that canonical hair cell markers in mice were commonly expressed in Xenopus inner ear (Fig. 1E and Dataset S2). Besides, a high similarity coefficient was detected between Xenopus and mouse hair cells (Fig. 3B). The canonical hair cell marker genes were highly conserved and commonly expressed across species (Fig. 4C). Therefore, the auditory hair cell markers Otof, Myo6, Myo7a, Myo1h, Slc17a8, Tmc1, Espn, and Calb2 were found specifically expressed in hair cells of Xenopus inner ear, distinguishing hair cells from other cell types (Fig. 4A, 4B, 4C, and S4A). MYO7A is a well-known classic marker for hair cells. Strikingly, an anti-MYO7A antibody derived from human MYO7A protein can successfully mark all the inner ear hair cells from both Xenopus (Fig. 4D, 4E, and S4C), mouse (Fig. 4F), and chicken 41, indicating exceptional conservation of hair cell at molecular level. In addition, otof mRNAs, a well-known human deafness gene, were universally detected in Xenopus cristae, saccule and amphibian papilla hair cells (Figure S4B), revealing a similar expression pattern with myo7a. Many less common or new hair cell specifically expressed genes were also identified in Xenopus, including lmod2.S, srrm3.S, and sh2d4b.L (Figure S4A).
Two clusters of the total 21 cell atlas in whole Xenopus inner ear were identified as the hair cells marked by calb2.L, atp8a2.L, c6h10orf6.L and dclk3.L gene expression (Figure S5-1 A-B). These two clusters were further extracted (total 856 hair cells from the 11275 total single cells) and sub-clustered into 5 hair cell subtypes (0–4) (Figure S5-1 C-F). Particularly, the auditory hair cells from AP-BP were further identified with the canonical marker genes myo6 and otof. A total of 315 auditory hair cells from the 12,646 AP-BP single cells (Fig. 5A) were subsequently sub-clustered into 4 hair cell subtypes (0–3) indicated by top differential expression genes (Fig. 5B, 5C, and 5D). Briefly, ptprz1.S shows strongest expression but not specific in subcluster 0. In contrast, myo1h.L/loxhd1.L, gata2.S, and abi3bp.L are specifically expressed in subcluster 1, 2, and 3, respectively. The RNA in situ detections of myo1h, loxhd1, and gata2 genes were carried out to confirm their specificity of expression in hair cell (Fig. 5E and 5F). The gata2 was strongly expressed in most of hair cells in epithelium of AP but relatively weak expressed in the hair cells of marginal region (Figure S5-2A). RNA in situ detections of genes ush2a, dclk3 and calb2 in whole inner ear also confirmed their hair cell specific expression and differential expression in hair cell subtypes (Figure S5-1 and S5-2). These results revealed conservation of hair cell identity in AP-BP or whole inner ear in Xenopus as confirmed by common expression of the canonical hair cell markers from mammals, however, each hair cell subtypes displayed certain diversity.
Cross-species comparative analyses of auditory hair cells confirmed that the OHCs are newly evolved hair cell subtypes existing exclusively in mammals
Aiming to provide evolutionary insights into hair cell functions, the hair cell features of mouse, Xenopus, and chicken, were analyzed and compared. The hair cells were extracted from our mouse and Xenopus datasets and public chicken scRNA-seq datasets 41. The mouse auditory hair cells have two distinct well-known subtypes, inner hair cell (IHC) and outer hair cell (OHC) (Fig. 6A). IHCs are dominantly marked by Tbx2 and Slc17a8 gene expression, while OHCs are mainly marked by Slc26a5 and Ikzf2 gene expression (Fig. 6A and S6A). However, these markers all display an overall chaotic expressing pattern in four subtypes of AP-BP hair cell in Xenopus (Fig. 6B). Interestingly, one population of mouse hair cells, identified as OHCs with dominant expression of Slc26a5 and Ikzf2, were obviously separated aside according to merged UMAP of mouse and Xenopus (Fig. 6D). The rest of mouse hair cell, IHC, mostly overlapped with Xenopus hair cells. Despite the detection of OHC-specific gene expression (like slc26a5, ikzf2) in whole hair cell population, there are no OHC-like cells in Xenopus, providing the evidence that OHC is an exclusive hair cell type in mammals, consistent with the traditional viewpoint. The hair cell subtypes from Xenopus and chicken were overall clustered together in the UMAP, suggesting their similarity (Figure S6F). In chicken, four subtypes of auditory hair cells were defined according to their morphology and single-cell atlas as reported previously (short1, short2, tall1 and tall2) 41. But none of them display similar features with mouse IHC or OHC, as indicated by expression pattern of TBX2, SLC17A8, SLC26A5, and IKZF2 genes (Fig. 6C and S6C). Consistently, the mouse OHCs were separated aside and clearly distinguished from chicken hair cells, as revealed by merged UMAP (Figure S6H). It suggested that the chicken neither has the OHC-like cells, similar to Xenopus. However, the chicken hair cell showed closer distance to mouse OHC in the UMAP compared to Xenopus (Fig. 6D). The merged UMAP of hair cells from three species Xenopus, chicken, and mouse (Fig. 6E), further confirmed the pairwise species comparison above (Fig. 6D, S6F, and S6H). The mouse OHCs still distributed separately, and chicken hair cells distributed in the middle of mouse and Xenopus (Fig. 6E). Consistently, the expression of mammal IHC and OHC marker genes confirmed that Xenopus hair cells specifically expressed IHC marker (slc17a8) rather than OHC marker gene (slc26a5), confirming no OHC like cells in Xenopus (Fig. 6F). In total, these data revealed that IHC is an ancient hair cell type in vertebrates and the OHC is the newly evolved cell type existing exclusively in mammals. Our results implicate that OHC may gradually evolve and emerge from amphibians to mammals in agree with their evolutionary relationship (Fig. 1B).
Evolution of outer hair cell
Furthermore, the similarity coefficients were calculated for each hair cell subtype. The overall similarities of hair cell subtypes between Xenopus, chicken, and mouse are moderate (Fig. 7A), compared to the entire hair cell populations (73.42%) from mouse and Xenopus (Fig. 3B). The Xenopus hair cell subcluster 1 showed relatively high similarity (68.72%) to mouse IHC subtype, while hair cell subclusters 0 and 2 display relatively low similarities with mouse OHC (42.89% and 31.29%, respectively). The chicken tall1 hair cell subtype most resembles the mouse IHC subtype (60.23%), and the rest hair cell subtypes short1, short2, tall2 were all similar to mouse OHC but with low similarity coefficients (25.77%, 20.71%, and 31.40%, respectively). These data further confirmed that the OHC is the newly evolved hair cell subtype in mammals.
To elucidate the molecular differences between mammal OHCs and IHCs, or between hair cell subtypes among vertebrates, the transcriptomic profiling of hair cells cross species was further performed. Firstly, the top DEGs between mouse IHC and OHC, which were enclosed in the conserved 10,858 genelist among Xenopus, chicken, and mouse (Fig. 1D and Dataset S3), were selected for further profiling studies. The Xenopus (subtype 0 and 1) or chicken (tall1 and tall2) hair cell subtypes showing highest similarities with the mouse IHC or OHC were selected for cross species comparison (Fig. 7B, left). Interestingly, the expression of the IHC/OHC DEGs showed an overall similar feature of Xenopus hair cell with mouse IHC (Fig. 7B, middle) in agreement with the merged hair cell feature in UMAP (Fig. 6D) as well as Sankey diagram (Fig. 7A). The comparative analysis between mouse and chicken revealed that neither mouse IHC or OHC resembles the molecular profiles of chicken hair cell types (Fig. 7B, right). The chicken hair cells displayed a mixture feature of IHC and OHC, suggesting an intermediate transition state from IHC to OHC. Slc26a5 is one of most critical markers for OHC cell identity and function. Strikingly, our snRNA-seq data reveal that the slc26a5 is universally expressed within most of Xenopus inner ear cell types (Fig. 6F). In addition, the hair cell percentage in Xenopus total single cells (7.6%, 856/11275 for whole inner ear; 2.5%, 315/12646 for AP-BP) is higher than that of mouse cochlea (0.8%, 92/10977). Despite low expression of slc26a5 at the single-cell level in Xenopus (Fig. 6F and 7B), its universal expression accumulated an overall high expression level in Xenopus than mice (Log2(FC) = 3.57, P value = 2.62E-81), as indicated by bulk RNA-seq dataset (Dataset S2). Expression of slc26a5 gene was not specific and cell type restricted that may result in no OHC-like cells in these lower vertebrates. Even that the tbx2- and slc26a5-expressing cells were isolated from the total mosaic hair cell population in Xenopus, the tbx2-positive or slc26a5-positive cells displayed obviously different from the mouse IHCs (Tbx2-positive) or OHCs (Slc26a5-positive) (Figure S7B). These results suggest a complex mechanism for hair cell subtype identity and evolution.
Next, the genomic structure and regulatory program of Slc26a5 gene were analyzed. Remarkedly, despite a huge divergence of hundreds of millions years ago between Xenopus and mice 42, the Slc26a5 gene locus showed an extremely conserved genomic arrangement and collinearity with nearby Psmc2, Dnajc2, and Pmpcb gene regions (Fig. 7C). Other critical hair cell marker gene loci, like Otof, also displayed extremely conservated genomic structure and collinearity (Figure S7C). The H3K27ac ChIP-seq signals exhibited a relatively stronger H3K27ac enrichment in Slc26a5 gene locus in Xenopus but very weak in mice, consistent with that of higher Slc26a5 gene expression in Xenopus (Dataset S2). The Slc26a5 expression might be precisely regulated by transcription factors such as Six1, a well-known master transcription factor for inner ear development (Fig. 7C). Finally, a hypothesis was made according to these data, and a schematic model figure was drafted to demonstrate the hair cell evolution across vertebrates and the process of how OHC originates and evolves during the evolution from amphibian to mammal. The ancient hair cells (Xenopus, green color) gradually adapted to change their morphology, restriction, and specification of their gene expression patterns into a mediate stage (chicken, yellow), finally evolved into the mammalian specific OHC cell type (mouse, red) (Fig. 7D).