Conserved expression of Hsd17b7 in zebrafish and mice sensory hair cells.

(A) The UMAP analysis of the zebrafish scRNA-seq data. LLHCs, lateral line hair cells; MHCs, macula hair cells; CHCs, crista hair cells; (B, C) Feature plot and violin plot of hsd17b7 expression across cell types. (D) Quantitative comparison of average expression levels of hsd17b7 within LLHCs. (E) hsd17b7 expression increased along the pseudotime trajectory of LLHC formation. (F) Heatmap of marker genes along the pseudotime trajectory. (G) The expression of the hsd17b7 in 72 hpf and 96 hpf embryos was detected by whole-mount in situ hybridization analysis. Dashed circles indicate the otic vesicle and the neuromast. (H, I) Representative images of the CHCs and LLHCs in Tg(myo6b: hsd17b7-egfp) at 4 dpf. (J) The average expression of Hsd17b7 and HC marker genes across mouse HC subtypes from published scRNA-seq datasets. UHCs, utricular hair cells. (K) Immunostaining and quantification analysis of HSD17B7 in dissected mouse organ of Corti sections at P1, P4, and P7. Dashed circles indicate outer hair cells (OHCs) and inner hair cells (IHCs).

Loss of hsd17b7 impaired acoustic startle responses and mechanotransduction activity in zebrafish.

(A) Left, schematic of the experimental setup used to assess acoustic startle responses in zebrafish larvae. Right, representative locomotor trajectories showing C-start responses to a single acoustic stimulus (9 dB re. 1 m·s⁻², 60 Hz tone burst) in control, hsd17b7 knockout (KO), and hsd17b7 mRNA–rescued larvae at 5 dpf. Scale bars, 10 mm. (B, C) Quantification of peak swimming velocity (B) and total movement distance (C) in response to acoustic stimulation shown in (A) (n = 20). One-way ANOVA followed by Tukey’s multiple comparisons, *P < 0.05, **P < 0.01, ***P < 0.001, ns, non-significant (p > 0.05), mean ± SEM. (D) Representative confocal images of lateral line hair cells (LLHCs, green) in Tg(Brn3c:mGFP) larvae at 4 dpf from control, hsd17b7 KO, and hsd17b7 mRNA–rescued groups. Scale bars, 20 μm. (E) Quantification of HC number per neuromast (n = 30). One-way ANOVA followed by Tukey’s multiple comparisons, *P < 0.05, **P < 0.01, ns, non-significant (p > 0.05), mean ± SD. (F) Representative images of LLHCs (green) and FM4-64–labeled functional HCs (red) in single neuromasts of Tg(Brn3c:mGFP) larvae at 5 dpf from the indicated groups. White dashed outlines indicate LLHCs. Scale bars, 20 μm. (G) Quantification of relative FM4-64 fluorescence intensity per neuromast (n = 25). One-way ANOVA followed by Tukey’s multiple comparisons, ****P < 0.0001, ns, non-significant (p > 0.05), mean ± SD.

Hsd17b7 deficiency disrupted cholesterol-associated transcriptional states in hair cells.

(A) Schematic overview of the scRNA-seq workflow. mGFP+ hair cells were isolated from control and hsd17b7 mutant zebrafish by fluorescence-activated cell sorting (FACS) and subjected to 10× Genomics single-cell RNA sequencing. (B) UMAP visualization of scRNA-seq data from control and hsd17b7 mutant samples, colored by annotated cell types, including inner ear (IE) and lateral line (LL) hair cells (HCs) and supporting cells (SCs). (C) Volcano plot showing pseudobulk differential gene expression analysis of LLHCs from hsd17b7 mutants relative to controls. Genes with adjusted p < 0.05 are highlighted, with upregulated genes shown in red and downregulated genes shown in blue. (D and E) Gene Ontology (GO) biological process enrichment analysis of genes upregulated and downregulated in hsd17b7 mutant HCs. A Sankey diagram illustrates representative genes contributing to each enriched GO term. (F) Heatmap and module score analysis of gene sets related to tip-link structure, mechanotransduction (MET), and cholesterol metabolism in LLHCs and IEHCs. Module scores were calculated using predefined gene sets, and statistical significance was assessed using the Wilcoxon rank-sum test. (G) Ridge plots showing the distribution of z-scored module scores associated with tip link structure, cholesterol uptake, and cholesterol efflux in control and hsd17b7 mutant LLHCs and IEHCs, revealing population-wide transcriptional shifts rather than discrete subpopulation changes. (H) Gene set enrichment analysis (GSEA) demonstrating significant downregulation of cholesterol biosynthesis–associated pathways in hsd17b7 mutant LLHCs.

Hsd17b7 deficiency reduced cholesterol in hair cells.

(A) Schematic illustration of the cholesterol biosynthesis pathway highlighting the role of HSD17B7. (B) Western blot analysis of Hsd17b7 protein levels in control and Hsd17b7-knockdown HEI-OC1 cells; quantification is shown on the right (n = 6). Significance was determined by an unpaired two-tailed Student’s t-test. ****P < 0.0001, mean ± SD. (C) Quantification of the relative cellular cholesterol levels in control and Hsd17b7-knockdown HEI-OC1 cells (n=12). Significance was determined by an unpaired two-tailed Student’s t-test. ***P < 0.001, mean ± SD. (D) Representative immunofluorescence images showing the cholesterol probe D4H-mCherry in HEI-OC1 cells transfected with pCMV-Flag or pCMV-Flag–HSD17B7 plasmids. White dashed outlines indicate transfected cells. Scale bars, 20 μm. (E) Quantification of relative D4H-mCherry fluorescence intensity for (D) (n = 16). Significance was determined by an unpaired two-tailed Student’s t-test. ****P < 0.0001, mean ± SD. (F) Experimental schematic illustrating expression of D4H-mCherry in control and hsd17b7 knockout zebrafish larvae using the Tg(Brn3c:mGFP) background. (G) Representative confocal images of D4H-mCherry (red) in crista hair cells (CHCs) and lateral line hair cells (LLHCs) of control and hsd17b7 knockout larvae at 4 dpf. White dashed outlines indicate hair cells. Scale bars, 20 μm. (H) Quantification of relative D4H-mCherry fluorescence intensity in vivo for (G) (n = 30). Significance was determined by an unpaired two-tailed Student’s t-test. ***P < 0.001, ****P < 0.0001, mean ± SD.

Identification of a heterozygous nonsense variant in human HSD17B7.

(A) Two-generation family pedigree for the affected individual. The hearing-impaired individual is indicated by a black circle (female). (B) Pure-tone audiometry audiograms for the hearing-impaired individual at 8 years old. Blue represents the results for the left ear and red for the right ear. The affected individual shows severe-to-profound or profound HI. (C) Sequence of HSD17B7 mRNA in wild type and heterozygote. (D) Multiple sequence alignment of 8 different species using the TBtools program. The p.Glu182 residue, as indicated by a red arrow, is evolutionarily conserved from zebrafish to human. (E) The domain structure of human HSD17B7 and HSD17B7E182*. ED, extracellular domain; TM, transmembrane; CD, cytoplasmic domain. The residue numbers are labeled at right.

Wild-type but not HSD17B7E182* restored MET activity and startle responses in hsd17b7 mutants.

(A) Representative images of lateral line hair cells (LLHCs, green) and FM4-64–labeled functional HCs (red) in single neuromasts of Tg(Brn3c:mGFP) at 5 dpf from control, hsd17b7 knockout (KO), KO injected with HSD17B7E182* (p.E182*) mRNA, and KO injected with HSD17B7 (WT) mRNA groups. White dashed outlines indicate LLHCs. (B) Quantification of the relative FM4-64 intensity per HC (n=25). One-way ANOVA followed by Tukey’s multiple comparisons, **P < 0.01, ****P < 0.0001, ns, non-significant (p > 0.05), mean ± SD. (C) Representative locomotor trajectories of 5 dpf larvae exhibiting behavioral responses to a single acoustic stimulus (9 dB re. 1 m·s⁻², 60 Hz tone burst) in the indicated groups. (D, E) Quantification of total movement distance (D) and peak swimming velocity (E) in response to acoustic stimulation (n = 20). One-way ANOVA followed by Tukey’s multiple comparisons, ****P < 0.0001, ns, non-significant (p > 0.05), mean ± SEM.

Subcellular localization of wild-type and p.E182* mutation HSD17B7.

(A) Immunostaining shows the subcellular localization of Flag-HSD17B7 and Flag-HSD17B7E182* in HEI-OC1 cells. Calnexin was used as an endoplasmic reticulum (ER) marker. White arrows indicate the line-scan positions used for intensity profile analysis in (B) and (C). Scale bars, 20 μm. (B, C) Line-scan intensity profiles showing the spatial distribution of HSD17B7 (B) or HSD17B7E182* (C) relative to the ER marker Calnexin along the indicated axes in (A). (D) Representative immunofluorescence image of a HEI-OC1 cell co-expressing HSD17B7 and HSD17B7E182*. White arrows indicate the positions used for line-scan analysis. Scale bars, 20 μm. (E) Line-scan intensity profiles showing the intracellular distribution of HSD17B7 and HSD17B7E182* along the indicated axis in (D).

Expression of HSD17B7E182* impaired mechanotransduction activity, auditory-related behavior, and cholesterol organization.

(A) Representative images of lateral line hair cells (LLHCs, green) and FM4-64–labeled functional hair cells (red) in single neuromasts of Tg(Brn3c:mGFP) larvae at 5 dpf following injection of HSD17B7 or HSD17B7E182* mRNA. The white dashed indicates LLHCs. Scale bars, 20 μm. (B) Quantification of the FM4-64 relative intensity per LLHCs (n=25). One-way ANOVA followed by Tukey’s multiple comparisons, *P < 0.05, ****P < 0.0001, ns, non-significant (p > 0.05), mean ± SD. (C-E) Moving traces and quantification of 20 embryos were filmed in 150 frames before and after stimulation (9 dB re. 1 m·s⁻², 60 Hz tone burst). Scale bars, 10 mm. One-way ANOVA followed by Tukey’s multiple comparisons, *P < 0.05, **P < 0.01, ns, non-significant (p > 0.05), mean ± SEM. (F) Immunostaining of the cholesterol probe D4H in HEI-OC1 cells transfected with pCMV-Flag, pCMV-Flag-HSD17B7, and pCMV-Flag-HSD17B7E182* constructs, respectively. White dashed outlines indicate transfected cells; arrows denote line-scan positions used for intensity profile analysis. Scale bars, 20 μm. (G–I) Line-scan intensity profiles showing the spatial distribution of Flag, HSD17B7, or HSD17B7E182* relative to D4H signal along the indicated axes in (F). (J) Representative images of crista and lateral line hair cells in Tg(myo6b: D4H-mCherry) larvae following expression of myo6b-driven HSD17B7-EGFP or HSD17B7E182*-EGFP at 4 dpf. White dashed outlines indicate HCs. Scale bars, 10 μm.

HSD17B7 interacts with RER1 to ensure ER localization.

(A) Co-immunoprecipitation of Flag-tagged HSD17B7 and HSD17B7E182* from HEI-OC1 cells transfected with pCMV-Flag, pCMV-Flag-HSD17B7, or pCMV-Flag-HSD17B7E182*. Cell lysates were immunoprecipitated using anti-Flag beads. Tubulin and IgG light chains served as loading and immunoprecipitation controls, respectively. (B) Venn diagram showing the overlap and uniqueness of interacting proteins identified for HSD17B7 (red) and HSD17B7E182* (blue) by LC–MS/MS analysis. (C) Gene Ontology (GO) Biological Process enrichment analysis of proteins specifically associated with HSD17B7 or HSD17B7E182*. (D) GO Cellular Component enrichment analysis of HSD17B7- and HSD17B7E182*-specific interacting proteins. (E) Top 20 ER-localized proteins (ranked by iBAQ intensity) specifically associated with HSD17B7. (F) In vitro GST pull-down assays demonstrating direct interaction between HSD17B7 and RER1. The upper panel shows Coomassie blue staining of purified GST and GST-HSD17B7 proteins; the lower panel shows immunoblot detection of RER1-His. (G) Immunofluorescence staining of RER1-Myc in HEI-OC1 cells co-transfected with Flag-HSD17B7 or Flag-HSD17B7E182*. White arrows indicate positions used for fluorescence intensity profiling. Scale bars, 20 μm. (H, I) Fluorescence intensity profiles showing co-localization of RER1 with HSD17B7 (H) but not with HSD17B7E182* (I).

Validation of HSD17B7 antibody specificity in HEI-OC1 cells.

(A) HEI-OC1 cells were transfected with pCMV-Flag-HSD17B7 or pCMV-HSD17B7-EGFP constructs. Immunofluorescence staining using an anti-HSD17B7 antibody showed strong co-localization with both FLAG and EGFP signals, confirming the specificity of the antibody for HSD17B7. (B) Western blot analysis of cells transfected with the constructs further verified the specific detection of HSD17B7 protein by the anti-HSD17B7 antibody.

Evolutionary conservation of Hsd17b7.

(A) Evolutionary conservation of Hsd17b7. The phylogenetic tree was reconstructed using Phylogeny. fr (v3.1/3.0, aLRT) by maximum likelihood, and its graphical representation and editing were performed with TreeDyn (v198.3). (B) The amino acid sequence was analyzed using TBtools multiple-sequence alignment of Hsd17b7 orthologs.

Generation of hsd17b7 mutant using the CRISPR/Cas9 system in zebrafish.

(A) Schematic diagram showing sgRNA and the target site on the exon 3 of the hsd17b7 gene. (B) Schematic diagram showing the hsd17b7 mutant establishment process. (C) Mutation pattern of sgRNA and Cas9-injecting embryos. (D) Sequencing Chromatograms of -4 bp, -6 bp, and -6 bp deletion mutant line. (E) Three types of mutations were identified by sequencing and screening. (F) Schematic diagram of the predicted proteins encoded by the -4 bp deletion mutant line.

Knockdown of hsd17b7 impairs auditory-evoked startle responses in zebrafish.

(A) Schematic illustration of morpholino targeting the exon1/intron1 splice junction of the hsd17b7 gene. (B) Schematic diagram of one-cell embryo injection of hsd17b7 morpholino (Mo) into wild-type embryos. (C) The efficiency of morpholino was detected by RT-PCR. (D) Western blots showing hsd17b7 protein levels of control, hsd17b7 morphants, and hsd17b7 mRNA-rescued groups at 3 dpf. Quantification of relative hsd17b7 protein is shown on the right (n=7). One-way ANOVA followed by Tukey’s multiple comparisons, **P < 0.01, ns, non-significant (p > 0.05), mean ± SD. (E) Representative locomotion trajectories of larvae exhibiting behavior responses to a single acoustic stimulus (9 dB re. 1 m·s⁻², 60 Hz tone burst) in control, hsd17b7 morphants, and hsd17b7 mRNA rescued groups, respectively. Scale bars, 10 mm. (F, G) Quantification of the mean distance and peak velocity of movement of 5 dpf larvae in response to acoustic stimulation shown in (E) (n=20). One-way ANOVA followed by Tukey’s multiple comparisons, **P < 0.01, ns, non-significant (p > 0.05), mean ± SEM.

Knockdown of hsd17b7 leads to MET dysfunction and decreases cholesterol levels in zebrafish.

(A) Representative images of lateral line hair cells (LLHCs, green) and FM4-64–labeled functional HCs (red) at 5 dpf Tg(Brn3c: mGFP) larvae from control, hsd17b7 morphants, and hsd17b7 mRNA rescued groups, respectively. White dashed indicates LLHCs. Scale bars, 20 μm. (B) Quantification of the FM4-64 relative intensity per HC for (A) (n=23). P values were determined using a one-way ANOVA test followed by Tukey’s multiple comparisons. ****P < 0.0001; **P < 0.01;ns, non-significant (p > 0.05). (C) Representative images of the cholesterol probe D4H–mCherry (red) in crista hair cells (CHCs, green) from control and hsd17b7 morphants at 4 dpf. White dashed outlines indicate CHCs. Scale bars, 20 μm. (D) Quantification of relative D4H fluorescence intensity in CHCs shown in (C) (n = 24). Unpaired two-tailed Student’s t test, ****P < 0.0001, mean ± SD. (E) Representative images of D4H–mCherry (red) in LLHCs (green) from control and hsd17b7 morphants at 4 dpf. White dashed outlines indicate LLHCs. Scale bars, 20 μm. (F) Quantification of relative D4H fluorescence intensity in LLHCs shown in (E) (n = 24). Unpaired two-tailed Student’s t test, ****P < 0.0001, mean ± SD.

Validation of cell type annotation and cholesterol- and MET-related transcriptional states in hair cells.

(A) Nebulosa density plots and violin plots showing the expression of canonical hair cell markers across the integrated scRNA-seq dataset. The pan–hair cell marker myo6b is broadly enriched in hair cell populations, while lhfpl5a and lhfpl5b specifically mark inner ear hair cells (IE_HCs) and lateral line hair cells (LL_HCs), respectively, confirming accurate hair cell annotation. (B) Nebulosa density plots and violin plots of supporting cell marker genes. The general supporting cell marker stm, together with otogl (IE_SCs) and irg1l (LL_SCs), show cell type–restricted expression patterns, validating the identification of supporting cell populations in both inner ear and lateral line systems.

Schematic of the experimental process for immunoprecipitation and LC-MS/MS to identify HSD17B7 and HSD17B7E182* interacted proteins.

E182* mutation decreased HSD17B7 protein levels, mRNA levels, and the half-life periods.

(A) Representative images of HEI-OC1 cells transfected with pCMV-Flag, pCMV-Flag-HSD17B7, or pCMV-Flag-HSD17B7E182* plasmids. (B) Western blots showing the protein levels of exogenously expressed HSD17B7 in HEI-OC1 cells transfected with pCMV-Flag-HSD17B7 (WT) and pCMV-Flag-HSD17B7E182* (p.E182*) plasmids, respectively. Tubulin was used as the loading control. Red stars indicated HSD17B7 and HSD17B7E182*. Quantification of relative protein levels was shown on the right (n=6). Unpaired two-tailed Student’s t test, ****P < 0.0001, mean ± SD. (C) RT-qPCR showing the mRNA level of HSD17B7 and HSD17B7E182* in transfected HEI-OC1 cells (n=11). Unpaired two-tailed Student’s t test, ****P < 0.0001, mean ± SD. (D) RT-qPCR analysis of mRNA stability in transfected HEI-OC1 cells (n=12). (E) Schematic diagram illustrating the experimental workflow for mRNA stability analysis and sample collection.