Characterizing Adult cochlear supporting cell transcriptional diversity using single-cell RNA-Seq: Validation in the adult mouse and translational implications for the adult human cochlea

Hearing loss is a problem that impacts a significant proportion of the adult population. Cochlear hair cell loss due to loud noise, chemotherapy and aging is the major underlying cause. A significant proportion of these individuals are dissatisfied with available treatment options which include hearing aids and cochlear implants. An alternative approach to restore hearing would be to regenerate hair cells. Such therapy would require recapitulation of the complex architecture of the organ of Corti, necessitating regeneration of both mature hair cells and supporting cells. Transcriptional profiles of the mature cell types in the cochlea are necessary to can provide a metric for eventual regeneration therapies. To assist in this effort, we sought to provide the first single-cell characterization of the adult cochlear supporting cell transcriptome. We performed single-cell RNA-Seq on FACS-purified adult cochlear supporting cells from the LfngEGFP adult mouse, in which supporting cells express GFP. We demonstrate that adult cochlear supporting cells are transcriptionally distinct from their perinatal counterparts. We establish cell type-specific adult cochlear supporting cell transcriptome profiles, and we validate these expression profiles through a combination of both fluorescent immunohistochemistry and in situ hybridization co-localization and qPCR of adult cochlear supporting cells. Furthermore, we demonstrate the relevance of these profiles to the adult human cochlea through immunofluorescent human temporal bone histopathology. Finally, we demonstrate cell cycle regulator expression in adult supporting cells and perform pathway analyses to identify potential mechanisms for facilitating mitotic regeneration (cell proliferation, differentiation, and eventually regeneration) in the adult mammalian cochlea. Our findings demonstrate the importance of characterizing mature as opposed to perinatal supporting cells.


175
Outlier identification 176 Cells that appeared unhealthy, as noted by lack of GFP or fragmented cellular appearance in the  Table S1).  Clusters were then visualized using a t-distributed stochastic neighbor embedding (t-SNE) plot.

212
Differential gene expression analysis. Differential expression analysis was performed in Seurat 213 utilizing the FindAllMarkers function with the default settings except that the "min.pct" and 214 "thresh.use" parameters were utilized to identify broadly expressed (min.pct = 0.8, thresh.use = 215 0.01) and subpopulation-specific (min.pct = 0.5, thresh.use = 0.25) expression profiles. The 216 parameter "min.pct" sets a minimum fraction of cells that the gene must be detected in all 217 clusters. The parameter "thresh.use" limits testing to genes which show, on average, at least X-218 fold difference (log-scale) between groups of cells. The default test for differential gene 219 expression is "bimod", a likelihood-ratio test (McDavid et al. 2013 protocol (PN 100-6117 G1). Mixes for lysis, RT, and specific target amplification were prepared 240 from the Single Cell-to-Ct qRT-PCR kit (Ambion, Austin, TX) and pre-designed Delta Gene 241 assays (Fluidigm, South San Francisco, CA). In addition to known markers, putative adult cochlear 242 supporting cell markers were selected for analysis in an arbitrary manner.

243
After 18 cycles of preamplification, expression levels using single cell qPCR were 244 performed on the Fluidigm Biomark HD system as previously described (Honda et al. 2017). 245 cDNA from single cells was selected for qPCR in the same way as it was selected for RNA-Seq.

246
A total of 170 single cells from four C1 captures were profiled using 2 Dynamic Array IFCs.

247
Empty wells with primers were utilized for negative controls. The threshold of cycles (Ct) values 248 were calculated with Fluidigm Real-time PCR analysis software with the following settings: 249 quality threshold of 0.65; a linear (derivative) baseline correction; and auto (detectors) method. 250 We defined gene expression levels as log2 expression = LOD -Ct, in which Ct = 24 was set as the 251 LOD. We used the log2 expression dataset for hierarchical clustering. The SINGuLAR package 252 v3.6.1 was utilized to display and analyze single-cell qPCR data. For immunohistochemistry and in situ hybridization of cochlear sections, fixed adult 275 mouse inner ears were decalcified in 150 mM EDTA for 5-7 days, transferred to 30% sucrose, and then embedded and frozen in SCEM tissue embedding medium (Section-Lab Co, Ltd.; Hiroshima, 277 Japan). Adhesive film (Section-Lab Co, Ltd.; Hiroshima, Japan) was fastened to the cut surface of 278 the sample in order to support the section and cut slowly with a blade to obtain 6 µm thickness 279 sections. The adhesive film with section attached was submerged in 100% EtOH for 60 seconds, 280 then transferred to distilled water. The adhesive film consists of a thin plastic film and an adhesive 281 and it prevents specimen shrinkage and detachment. This methodology allows for high quality 282 anatomic preservation of the specimen. Sections were cut to a thickness of 10 micrometers.

283
Fluorescent immunohistochemistry was performed as described above. Sections were mounted 284 with SCMM mounting medium (Section-Lab Co, Ltd.; Hiroshima, Japan).

285
Fluorescent in situ hybridization was performed using RNAscope® probes against       glass slides (Superfrost Plus glass slides, Thermo Scientific). Bibulous paper soaked with 10% 319 formalin was placed over the section. A small roller was used to flatten the sections. A 4" x 4" 320 block of wood was placed over each slide. Sections were allowed to dry for 1 day, and the weight 321 and the bibulous paper were removed. Celloidin removal was performed by immersing the sections 322 in sodium ethoxide diluted in ethanol (1:3 ratio) for 30 minutes followed by sequential immersions 323 in 100% acetone, methanol (100%, 70%, 50% for 5 minutes each), distilled water, 5% hydrogen 324 peroxide (10 minutes) and then washed with distilled water prior to antigen retrieval. Antigen 325 retrieval was performed as described previously (Lopez et al. 2016). Briefly, sections were heated 326 in a microwave oven using intermittent heating methods of two 2-minute cycles with an interval     To examine whether the SC1 and SC2 might represent distinct supporting cell types, 474 expression of four genes, including one known (Tuba1b) and three novel (S100a6, Spry2, 475 Pla2g7) genes, which showed differential expression between SC1 and SC2 ( Figure 3C 499 adult cochlear supporting cell scRNA-Seq. While validation using smFISH and/or 500 immunohistochemistry provides valuable spatial information regarding transcriptional expression, 501 these methods are labor intensive and low throughput. In our case, 2 cochlear supporting cell 502 subpopulations were identified in the data and there was not necessarily a way to specifically 503 isolate the exact combination of cells that created these subpopulations in our data from Lfng EGFP 504 adult mice. Therefore, we wanted to determine whether digital droplet PCR (ddPCR) and single-505 cell qPCR (sc-qPCR) could be used in combination as higher throughput methods for validating 506 scRNA-Seq data. Specifically, the presence or absence of the genes of interest in FACS-purified 507 GFP-positive adult cochlear supporting cells could be determined by ddPCR and the differential 508 expression between two supporting cell subpopulations could be confirmed with sc-qPCR. For  Figure 5C with an arrow pointing to the region of the inner 547 hair cell and a bracket outlining the outer hair cell region. S100A6 and LCP1 expression are 548 demonstrated in adult human cochlear supporting cells and co-localize with acetylated tubulin, a 549 known adult supporting cell marker ( Figure 5D-D", 5F-F"). Unlike S100A6 protein in mouse, 550 S100A6 protein in humans appears to be expressed by all cochlear supporting cells. Between  Figure S7). 556 These data demonstrate that S100A6 and LCP1 protein are expressed by adult human cochlear  Figure S10). Using ddPCR, we confirmed the presence of the transcripts for a selected group of these cell cycle genes in FACS-purified GFP-591 positive adult cochlear supporting cells ( Figure 6B). Differential expression of these cell cycle 592 genes was then validated by sc-qPCR ( Figure 4D, Figure 6C). Results from the sc-qPCR from 593 Figure 4D are redisplayed in Figure 6C to show correlation between scRNA-Seq violin plots 594 ( Figure 6D) and the sc-qPCR violin plots ( Figure 6C). Transcripts identified as being enriched in  Figure 6G). Overall, these analyses suggest that 627 these data could be utilized as a resource to explore mechanisms related to the maintenance of 628 supporting cell quiescence.  Table S2).  can be used to examine expression of specific transcripts within each data set, with a higher level 715 of sensitivity by comparison with scRNA-seq (Figure 4, Figure 6). The results of our analysis of 716 supporting cells using ddPCR and/or sc-qPCR yielded results that were largely consistent with the 717 scRNAseq results (Figure 4, Figure 6). However, some differences were also observed. For 718 instance, while S100a6 transcript expression is increased in SC1 versus SC2, a result that was 719 observed by smFIHS as well, single cell qPCR (sc-qPCR) for S100a6 demonstrates relatively  Adult mouse cochlear supporting cell-specific genes are expressed in human inner ears. An 726 important consideration in any biomedically-related study using an animal model is the 727 applicability to humans. For this study, we examined the expression of two candidate adult mouse SC genes in human cochlear SCs. S100A6 and LCP1 demonstrated slightly different patterns of 729 expression in humans versus mice. Specifically, S100A6 protein appears to be expressed in most 730 human supporting cells ( Figure 5D-E") unlike the mouse where S100A6 protein expression is 731 more highly expressed in medial SCs ( Figure 5A-A'). In addition to transcriptional differences 732 between mouse and humans, S100A6 is known to be secreted and taken up by other cells, which 733 may explain this apparent disparity in protein expression between mouse and human SCs

891
"Immunolocalization of the Calcium Binding S100A1, S100A5 and S100A6 Proteins in the   Note that P1 and mature cochlear supporting cells cluster within their respective groups but exhibit distinct clustering from each other. B, Comparison of averaged gene expression between FACS-purified mature (P60, P120) and P1 cochlear supporting cells indicates both equivalent (genes expressed on or near the red line) and differential (genes located closer to either axis) expression between the two cell stages. C, Feature plots of select known cochlear supporting cell genes (Dstn, Notch1, S100a1, Tuba1b) demonstrate distinct differences between P1 and mature cochlear supporting cells.      A-A', S100A6 protein expression in a representative mid-modiolar cross section of P60 Lfng EGFP organ of Corti. S100A6 protein (red) is localized to GFP-positive supporting cells (A) and grayscale single channel image demonstrates S100A6 protein expression (A'). B-B', LCP1 protein expression in a representative mid-modiolar cross section of P60 Lfng EGFP organ of Corti. LCP1 protein (red) is localized to GFP-positive supporting cells (B) and grayscale single channel image demonstrates LCP1 protein expression primarily in pillar cells (B'). Myosin 7A (MYO7A) and DAPI co-labeling for hair cells and nuclei, respectively. C-C", Representative immunofluorescence of celloidin embedded human organ of Corti sections from a patient with relatively normal hearing demonstrate S100A6 expression (green) that overlaps with acetylated tubulin (AT), a known adult cochlear supporting cell marker (red) (C). Single channel images for S100A6 in green (C') and acetylated tubulin (AT) in red (C") are shown. Arrow and bracket point to the inner hair cell (IHC) and outer hair cell (OHC) region, respectively. D-D", S100A6 expression in human organ of Corti from a patient with age-related hearing loss and hair cell loss. Note that S100A6 expression persists in the absence of hair cells. Representative immunofluorescence demonstrates S100A6 (green) and AT (red) (D) with single channel images for S100A6 protein (D') and AT (D"). Arrow and bracket point to IHC and OHC regions, respectively and are notable for a lack of these cell types in the section. E-E", Representative immunofluorescence of celloidin embedded human organ of Corti sections from a patient with relatively normal hearing demonstrate LCP1 expression (green) overlaps with AT (red) (E). Single channel images for LCP1 in green (E') and AT in red (E") are shown. The location of inner and outer hair cells are marked by the arrow and bracket, respectively. F-F'', LCP1 expression in human organ of Corti from a patient with age-related hearing loss and hair cell loss. Arrow and bracket point to IHC and OHC regions, respectively and are notable for a lack of these cell types in the section. Note that LCP1 expression persists in the absence of hair cells. G-G', S100A6 is expressed by adult human cochlear supporting cells and is not expressed in adult human cochlear hair cells which express calbindin protein (CALB1). Grayscale single channel image of S100A6 protein expression;. H-H', LCP1 is expressed by adult human cochlear supporting cells and is not expressed in adult human cochlear hair cells which express CALB1. I-J'. LCP1 is not expressed in the cuticular plate or stereocilia of the adult human cochlear hair cell. Human inner hair cell (upper panels) and outer hair cell (lower panels) cuticular plate and stereocilia (outlined by dotted white line), demonstrating lack of LCP1 in CALB1-positive human inner and outer hair cell stereociliary bundles, respectively. S100a6 = S100 calcium-binding protein a6; AT = acetylated tubulin; DAPI = 4',6diaminodino-2-phenylindole; MIP = maximal intensity projection; IHC = inner hair cell; OHC = outer hair cell; IPh = inner phalangeal cell; IPC = inner pillar cell; OPC = outer pillar cell. Scale bar in all panels, 20 μm. Lfng EGFP cochlea suggests that these cells may be maintained in a non-proliferative state by a repressive network of genes. All cell cycle genes expressed by adult supporting cells from the dataset, regardless of which cluster of adult cochlear supporting cells expressed these genes, were used as the starting input in Enrichr. GO biological process analysis suggests that genes involved in the G1/S transition of the mitotic cell cycle are prominent in adult cochlear supporting cells. GO molecular function and cellular component analysis point to cyclindependent protein serine/threonine kinase activity and cellular components associated with condensed chromatin at the centromere, respectively. The color of the bar corresponds to the combined score which is calculated by taking the log value of the p-value from the Fisher exact test and multiplying this value by the z-score of the deviation from the expected rank. The longer and lighter colored bars indicate that the term is more significant. G, Use of the STRING database to perform protein-protein interaction analysis identifies a set of interactions that may be related to the persistence of the post-mitotic state in adult cochlear supporting cells. The STRING plot demonstrates the action types and action effects as noted in the accompanying legend. Scale bar in all panels, 20 µm.

Supplementary Tables
Supplemental Table S1. Details of single cell captures on Fluidigm C1 capture system.