Abstract
The inner ear sensory neurons play a pivotal role in auditory processing and balance control. Though significant progresses have been made, the underlying mechanisms controlling the differentiation and survival of the inner ear sensory neurons remain largely unknown. During development, ISL1 and POU4F transcription factors are co-expressed and are required for terminal differentiation, pathfinding, axon outgrowth and the survival of neurons in the central and peripheral nervous systems. However, little is understood about their functional relationship and regulatory mechanism in neural development. Here, we have knocked out Isl1 or Pou4f1 or both in mice of both sexes. In the absence of Isl1, the differentiation of cochleovestibular ganglion (CVG) neurons is disturbed and with that Isl1-deficient CVG neurons display defects in migration and axon pathfinding. Compound deletion of Isl1 and Pou4f1 causes a delay in CVG differentiation and results in a more severe CVG defect with a loss of nearly all of spiral ganglion neurons (SGNs). Moreover, ISL1 and POU4F1 interact directly in developing CVG neurons and act cooperatively as well as independently in regulating the expression of unique sets of CVG-specific genes crucial for CVG development and survival by binding to the cis-regulatory elements including the promoters of Fgf10, Pou4f2, and Epha5 and enhancers of Eya1 and Ntng2. These findings demonstrate that Isl1 and Pou4f1 are indispensable for CVG development and maintenance by acting epistatically to regulate genes essential for CVG development.
Significance Statement
Inner ear neurons play critical roles in hearing and balance by transmitting auditory information in the form of electric signals from sensory hair cells to the brain. To study the genetic basis underlying their development, we identified the direct interaction and roles of transcription factors ISL1 and POU4F1 in regulating the development and maintenance of cochleovestibular ganglion (CVG) neurons. Moreover, our expression analysis uncovered that Isl1 and Pou4f1 are involved in the feedback regulation of CVG pioneer genes and control CVG differentiation genes. Furthermore, we identified and validated the function of the cis-regulatory elements bound by ISL1 and POU4F1 in vivo. Together, these findings provide insights into the gene regulatory network controlling the development of the inner ear neurons.
Introduction
The mammalian inner ear consists of auditory and vestibular systems responsible for hearing and balance, respectively (Torres and Giráldez, 1998). Among the diverse cell types in the inner ear, hair cells and sensory ganglion neurons play key roles in hearing and balance function and their dysfunction results in sensorineural hearing loss (SNHL) (Fettiplace and Hackney, 2006). While hearing aids and cochlear implants improve the hearing abilities by amplifying the received sound (NIDCD, 2013) or stimulating the sensory neurons electrically (NIDCD, 2016), these therapies fundamentally rely on functional sensory neurons. The protection and regeneration of sensory neurons provide therapeutic strategies that support hearing restoration. Therefore, it is essential to identify and characterize the gene regulatory network (GRN) of sensory neuron development.
In the mouse inner ear, sensory neurons originate from neuroblasts delaminating from the ventral region of the otic vesicle (Kelley, 2006; Appler and Goodrich, 2011). Once delaminated, the neuroblasts coalesce adjacent to the developing inner ear to form cochleovestibular ganglion (CVG) neurons. As development continues, CVG neurons segregate to form spiral ganglion neurons (SGNs) and vestibular ganglion neurons (VGNs). These processes are highly controlled by a hierarchical network of transcription factors (TFs). SIX1 and EYA1, both crucial neuronal determination factors, interact directly to drive neurogenesis and cooperate with SOX2 to regulate auditory neuron differentiation (Ahmed et al., 2012; Steevens et al., 2019). With the downstream of these TFs, the basic helix-loop-helix (bHLH) TF NEUROG1 is required for the initial specification of neuronal precursors (Ma et al., 1998, 2000) and NEUROD1, another bHLH TF downstream of NEUROG1, is necessary for the differentiation and maintenance of auditory neurons (Kim et al., 2001). Besides bHLH TFs, other TFs play important roles in the development of CVG neurons. For example, POU4F1, a Class IV POU-homeodomain TF, is required for the development of CVG neurons, including regulating terminal differentiation, mediating pathfinding, promoting axon outgrowth and cell survival (Ryan, 1997; Huang et al., 2001; Sherrill et al., 2019). Though POU4F1 plays a pivotal role in regulating CVG development, only about 30% SGNs are lost in Pou4f1KO mice (Huang et al., 2001), suggesting that additional factors could be required for regulating the development of CVG. ISL1, a LIM-homeodomain (LIM-HD) TF, is essential for the differentiation of a variety of neurons, including motor neurons (Pfaff et al., 1996), trigeminal ganglion (TG) neurons and dorsal root ganglion (DRG) neurons (Sun et al., 2008), and retina ganglion cells (RGCs) (Mu et al., 2008; Pan et al., 2008). ISL1 and POU4F1 are mostly co-expressed in developing CVG neurons whereas the onset of ISL1 expression slightly precedes that of POU4F1 (Deng et al., 2014), suggesting a functional interaction of these two TFs in CVG neurons. More importantly, a published study has shown that Pou4f1 and Isl1 act epistatically to regulate the gene expression program of sensory differentiation in DRG (Dykes et al., 2011). Collectively, these studies suggested that Isl1 and Pou4f1 could act together to regulate the development of the inner ear neurons.
To assess the function of ISL1 and POU4F1, we carried out the inner ear-specific knock-out of Isl1 and Pou4f1 using Pax2-Cre (Ohyama and Groves, 2004). We found that deletion of Isl1 alone resulted in profound defects in CVG development and axon pathfinding of SGNs. Additionally, deletion of both Isl1 and Pou4f1 caused more severe CVG defects, with nearly complete ablation of SGNs. Transcriptome analysis by next-generation sequencing (NGS) revealed that the loss of Isl1 or Pou4f1 or both resulted in a broad change in the expression of genes involved in axon guidance, axonogenesis, and other neurodevelopmental pathways. Moreover, our analyses of ISL1-POU4F1 interaction, CUT&Tag, snATAC-Seq and in vivo enhancer deletion identified the downstream effector genes that ISL1 and POU4F1 directly bind to and regulate in the developing CVG neurons. Our results demonstrate that Isl1 and Pou4f1 act epistatically to regulate the differentiation and survival of auditory neurons and reveal the underlying molecular mechanisms of their direct interaction during neurogenesis.
Materials and Methods
Mice
The Pou4f1KO, Isl1lacZ, Isl2V5, and Isl1loxP mice lines were generated in our laboratory and have been described previously (Xiang et al., 1996; Elshatory et al., 2007; Pan et al., 2008). All of the analyses were conducted in mice of either sex (n ≥ 3 for each genotype). Tg(Pax2-cre)1Akg transgenic mice express Cre recombinase throughout the inner ear, the midbrain, the cerebellum, the olfactory bulb, and the kidney (Ohyama and Groves, 2004) and have been used previously as an effective inner ear-specific deleter (Chang et al., 2008; Luo et al., 2013). Thus, Pax2-Cre transgenic mice were used to specifically delete Isl1 in the inner ear during developmental stages (Hayashi and McMahon, 2002; Ohyama and Groves, 2004). Embryos were identified as E0.5 at noon on the day at which vaginal plugs were observed. Mice of both sexes were used in this study. All animal procedures used in this study were approved by Institutional Animal Care and Use Committee (IACUC) at Augusta University.
Immunohistochemistry, X-Gal staining, BrdU pulse labeling, in situ hybridization, and RNAScope
Based on their developmental stages, embryos or heads of postnatal mice were fixed in 4% paraformaldehyde (PFA) in phosphate-buffered saline (PBS, pH7.2) at 4°C for different times. After three washes of PBS, the tissues were dehydrated in 20% sucrose in PBS at 4°C overnight, embedded in O.C.T. compound (Tissue-Tek, Sakura Finetek), and cryosectioned at a thickness of 14–20 μm. Immunolabeling of the sections was then conducted as previously described (Deng et al., 2010; Luo et al., 2013). Primary antibodies and concentrations used for this study were: mouse anti-POU4F1 (1:200, Santa Cruz Biotechnology (SCBT), Inc.), goat anti-POU4F2 (SCBT, 1:200), rabbit anti-Caspase-3 (1:500, R&D Systems, Inc.), rabbit anti-ISL1 (1:500, Developmental Studies Hybridoma Bank), goat anti-NEUROD1 (SCBT, 1:200), rabbit anti-GATA3 (1:500, ab199428, Abcam PLC.), and goat-anti TRKC (R&D, 1:100). Alexa-conjugated secondary antibodies were obtained from Molecular Probes and were used at a dilution of 1:1,000. Images were captured with a Zeiss 710 META confocal microscope or Leica STELLARIS confocal.
Detection of β-galactosidase activity was determined by X-Gal staining (Gan et al., 1999). Briefly, embryos were fixed in 4% PFA in PBS at 4°C for 1 h. Whole-mount embryos or 20 µm frozen sections were stained overnight at room temperature with 0.1% X-Gal, 5 mM potassium ferricyanide, 5 mM potassium ferrocyanide, and 2 mM MgCl2 in PBS.
For bromodeoxyuridine (BrdU) (Sigma) pulse-labeling experiments, pregnant females were injected intraperitoneally with 100 µg BrdU/gram body weight 1 h before they were sacrificed. Embryo were dissected out at E10.5 and E11.5, fixed in 4% PFA for 4 h, cryosectioned at 14 μm thickness, and stained with mouse anti-BrdU antibody (Becton Dickson, 1:200) following the regular immunolabeling procedure. Finally, the BrdU signals were visualized by Alexa-conjugated anti-mouse secondary antibody under confocal microscope.
For section in situ hybridization, embryos were dissected out in PBS, fixed in ice cold 4% PFA in PBS, dehydrated with gradient sucrose, embedded in O.C.T. medium (Tissue-Tek), and sectioned at 20 µm thickness. Hybridization was carried out with specific digoxigenin-labeled RNA probes and detection was done with anti-CIP/NBT color development as previously described (Li and Joyner, 2001).
For RNAScope detection of gene expression, E11.5 embryos were collected and fixed in ice cold 4% PFA in PBS, dehydrated with gradient sucrose, embedded in O.C.T. medium (Tissue-Tek), and sectioned at 16 µm thickness. Sections were subjected to a combination of IHC labeling and RNAScope in situ hybridization following the instruction of “RNAScope Multiplex Fluorescent v2 Assay Combined with Immunofluorescence: Integrated Co-Detection Workflow (ICW)” (MK 51-150) from Advanced Cell Diagnostics. Briefly, the sections were blocked, target retrieved, and then labeled with primary antibody followed by proteinase treatment and hybridization with probe. After that, signals were amplified on sections and probes were labeled with respective channels. Lastly, sections were treated with secondary antibodies, mounted and imaged under a Leica STELLARIS confocal microscope. Myt1 probe was obtained from Advanced Cell Diagnostics (S1047201-C2).
Neuronal tracing
These experiments were performed as described previously (Fritzsch, 2003; Fritzsch et al., 2005). The PFA-fixed heads of embryos or postnatal mice were sagittally cut at the midline into two halves, filter strips with lipophilic dye were inserted into the brainstem or the cerebellum. The samples were then incubated at 37°C in 4% PFA. After appropriate length of incubation when the dye had diffused into the fine terminals, inner ears were dissected and flat-mounted in glycerol. Inner ears were imaged with a Nikon E800 using a Bio-Rad 2000 confocal system or a Zeiss LSM 510 confocal system (Maklad and Fritzsch, 2002).
Proximity ligation assay (PLA)
For PLA, E10.5 embryos were dissected out and fixed in 4% PFA at 4°C for at least 2 h, then washed with PBS three times, dehydrated with gradient sucrose, embedded in O.C.T. medium, and sectioned at 5–10 µm thickness. PLA assay was performed with Duolink PLA kit (DUO92101, Millipore Sigma) according to the manufacturer's instruction. Briefly, on day 1, the slides were warmed at 37°C for 1 h, and then 1 drop (∼40 µl) of Duolink Blocking Solution was applied to each section. After the slides were incubated at 37°C for 1 h in a humidity chamber, Duolink Blocking Solution was tapped off. Primary antibodies diluted in the Duolink Antibody Diluent (rabbit anti-ISL1, Abcam#ab20670, 1:1,000 and mouse anti-POU4F1, Millipore#MAB1585, 1:1,000) were added to the sections and slides were incubated at 4°C overnight. On day 2, slides were washed twice with Wash Buffer A for 5 min each at room temperature. Diluted PLA PLUS and MINUS probe solutions (1:5 dilution in Duolink Antibody Diluent) were applied, and slides were incubated at 37°C for 1 h in a humidity chamber. Afterwards, slides were washed twice with Wash Buffer A for 5 min each at room temperature and were treated in ligation solution (1:8:31 ligase:5× Duolink Ligation buffer:water) at 37°C for 30 min in a humidity chamber. After two washes as above in Wash Buffer A, the slides were incubated in Amplification buffer (1:16:63 polymerase:5× Amplification buffer:water) in the humidity chamber for 100 min at 37°C. Finally, the slides were washed twice in 1× Wash Buffer B for 10 min each at room temperature and once in 0.01× Wash Buffer B for 1 min and mounted in Duolink-In Situ Mounting Medium with DAPI for 15 min before analyzing under a Leica STELLARIS confocal microscope.
RNA-Seq and qRT-PCR
For each pair of replicates, CVG neurons at E11.5 were dissected out from nine embryos of each genotype and frozen in liquid nitrogen immediately. The fresh frozen CVG neurons in Trizol were shipped to Azenta Life Science to perform RNA extraction, library preparation and sequencing. Briefly, total RNA was extracted using Trizol (ThermoFisher Scientific) following the manufacturer's protocol. The RNA-Seq libraries were prepared using NEBNext Ultra II Directional RNA Library Prep Kit (NEB) following the manufacturer's instruction. Sequencing was performed on Hiseq 4000 (Illumina, Inc.) with 2 × 150 bp paired-end reads and the depth was 20–30 million reads per sample. Low quality sequencing reads were filtered and trimmed with the Trimmomatic software. The quality of sequence reads was evaluated using the FASTQ analysis, following which the high-quality sequence reads were aligned to the mouse mm10 reference transcriptome using the HISAT2. The HISAT2 outputs were converted with SAMtools and then loaded into StringTie for transcript quantification. Finally, the count tables or matrices were input into the LEPSeq statistical package to determinate the differential expression gene between controls and mutants (cutoff fold-change >2 and adjusted p value <0.05). Genes of clusters were selected to perform cluster heatmap and GO terms analysis (https://www.bioinformatics.com.cn/srplot). The accession number for the RNA-Seq data is GSE233755.
For qRT-PCR, CVG neurons from six embryos of each genotype at E11.5 were isolated under microscopy and collected into tubes with 350 μl Lysis Buffer provided by the RNeasy Mini Kit (Qiagen). Then total RNA was extracted following the manufacturer's instruction and resuspended in 30 μl RNase-free water. cDNA was synthesized with iScript cDNA Synthesis Kit following the manufacturer's instruction and measured by Nanodrop 2000. qPCR was run with SsoAdvanced Universal SYBR Green Supermix (Bio-Rad) following the manufacturer's instruction on CFX96 PCR System. Actb was used as the housekeeping gene control. The primers are listed in Extended Data Table 5-1.
CUT&Tag
The CUT&Tag experiment was performed using reagents and protocol provided by EpiCypher, Inc. with minor modifications (Ge et al., 2023). E11.5 CVGs were isolated and nuclei were extracted immediately by incubating in the NE Buffer for 10 min on ice. For each experiment, greater than 1 × 105 nuclei were used. The nuclei were incubated with the activated Concanavalin A-coated magnetic beads (11 μl/sample, EpiCypher #21-1401) at room temperature for 10 min. The nuclei-beads were collected using magnetic stand and resuspended in 50 μl ice cold Antibody Buffer. 0.5 μg antibody against ISL1 (rabbit, AB4326, Millipore; rabbit, ab20670, Abcam), POU4F1 (mouse, MAB1585, Millipore), H3K27ac (EpiCypher #13-0045), H3K27me3 (C15410195, Diagenode), H3K4me3 (EpiCypher #13-0041), or control IgG (mouse, C15400001, Diagenode; rabbit, EpiCypher #13-0042) was added to each sample and mixed on a rotator at 4°C overnight or at room temperature for 2 h. The antibody-bound nuclei-beads were isolated and resuspended in 50 μl of ice cold Digitonin 150 Buffer. 0.5 μg of anti-rabbit or anti-mouse IgG secondary antibody (EpiCypher #13-0047 and #13-0048) was added and samples were incubated for 30 min at room temperature on a rotator. After two washes in ice cold Digitonin 150 Buffer, nuclei-beads were resuspended in 50 μl of ice cold Digitonin 300 Buffer, 2.5 μl of CUTANA pAG-Tn5 solution (EpiCypher #15-1017) was added, and nuclei-beads were incubated for 1 h at room temperature on a rotator. The samples were again washed twice in ice cold Digitonin 300 Buffer, resuspended in 125 μl of cold Tagmentation Buffer, and incubated at 37°C for 1 h. Then, 2.5 µl of 0.5 M EDTA, 1.25 µl of 10% SDS and 1.1 µl of 20 mg/ml Proteinase K were added to each sample and samples were incubated at 70°C for 10 min or 55°C for 30 min. After this, DNA fragments were purified and collected in 30 µl of ddH2O using Cycle Pure Kit (Omega, D6492-02) following the manufacturer's instruction. To prepare the library, 21 µl of DNA fragments were transferred to a PCR tube, to which 2 µl of barcoded i5 and i7 primers and 25 µl CUTANA High Fidelity 2× PCR Master Mix (EpiCypher, #15-1018) were added. PCR was performed using the following conditions: 72°C for 5 min, 98°C for 30 s, and 20 cycles of 10 s at 98°C and 10 s at 63°C, followed by an extra 1 min extension at 72°C. The PCR product was analyzed on an Agilent Fragment Analyzer. Combined DNA libraries were cleaned up using AMPure XP beads (#A63880, Beckman Coulter) following the manufacturer's instructions. Each pooled DNA library of 10 samples was resuspended into 40–80 μl water and quantified using the NanoDrop (Thermo Scientific). Library fragments were sent for sequencing commercially at Azenta Life Science. Sequencing was performed on Hiseq 4000 (Illumina) with 2 × 150 paired-end reads and a depth of 10 million reads per sample. Each antibody had two or more replicates. Raw data of each replicate was analyzed following the CUT&Tag Data Processing and Analysis Tutorial (https://yezhengstat.github.io/CUTTag_tutorial/). Briefly, raw data was pre-processed for quality control using FastQC. Qualified peaks were aligned to the mouse mm10 reference transcriptome using the Bowtie2. Peaks were called using MACS2 and SEACR. At last, peak annotation and functional analysis were performed using ChIP seeker and visualized by IGV. The accession number for the CUT&Tag-Seq data is GSE233840.
snATAC-Seq
Cochleae of C57BL/6J embryos at E12.5 were isolated in cold PBS and nuclei were immediately isolated according to 10× Genomics protocols (CG000375 Rev C) without FACS sorting. The libraries for Chromium Single Cell Multiome ATAC + Gene Expression were produced exactly as written (CG000338 Rev A, 10× Genomics Inc.) and sequenced using Illumina Novaseq 6000 S1 (100 cycles, 2 × 50 pair ending). Sequences were aligned to the Ensembl mouse mm10 assembly using CellRanger-ARC-2.0.0 (10× Genomics) analysis software. After removing doublets with AMULET (Thibodeau et al., 2021) for snATAC data and scDblFinder (Germain et al., 2021) or scRNA data, quality control (QC) of single-nuclei transcriptome and chromatin profiles were conducted by Seurat (v4.3.0) in R package (v4.2.2). Weighted nearest neighbors (WNN) (Hao et al., 2021) were applied to integrate batch-effect corrected (Harmony (Korsunsky et al., 2019)) snRNA-Seq and snATAC-seq data. Cells with expression of Nrxn1 and Sox2 were selected as spiral ganglion cells and prosensory cells, respectively. The accession number for the snATAC-Seq data is GSE233706.
Enhancer deletion by genome editing with CRISPR/Cas9 in mice
To delete the selected enhancers, two single guide RNAs (sgRNAs) flanking each peak region were identified using CRISPOR (Concordet and Haeussler, 2018) and synthesized by Synthego Corp. The sgRNAs and Cas9 protein (Alt-RS. p. Cas9 Nuclease V3 from Integrated DNA Technologies, Inc.) were introduced into C57BL/6J zygotes (Jackson Laboratory, stock#009086) by electroporation. The zygotes were then transferred into pseudopregnant Swiss recipient mice. Embryos were collected at E11.5 and embryos with homozygous deletion were verified by PCR genotyping and Sanger sequencing. Wild-type control and homozygous deletion littermates were selected for expression analysis by in situ hybridization. Sequences of sgRNAs and PCR genotyping primers used in this study are shown in Extended Data Table 8-1.
Quantification and statistical analysis
Statistical analyses of apoptotic cell numbers were performed by using a two-tailed Student t test to determine the differences. Due to the morphologically disorganized mutant tissues and lack of clear demarcation of early CVG, we averaged the counts of the CASP3+ cells in CVG per section for the CASP3+ cell quantification. For each E11.5 to E13.5 specimen, four to six serial sections were co-labeled with anti-CASP3 and anti-NEUROD1 (to mark the CVG). For each E14.5 specimen, six to eight serial sections were co-labeled with anti-CASP3 and anti-TUJ1 (to mark SG and VG). Once segregated, SGNs are in different sectional levels from VGNs. We analyzed six embryos for each group and compared the number of CASP3+ cells for CVG, SG, and VG separately. To quantify the changes in the number of NEUROD1−/ISL1/2+ postmitotic CVG neurons in the central CVG region, we have analyzed more than three embryos of each group and performed ANOVA analysis. All graphs are constructed via Prism 6.0.
Results
Isl1 inactivation leads to defective SGNs and VGNs
To investigate the role of Isl1 in CVG development, we used two modified Isl1 alleles: a conditional knock-out allele (Isl1loxP) and a nuclear lacZ reporter knock-in allele (Isl1lacZ) (Pan et al., 2008). We first analyzed the spatiotemporal expression of Isl1-lacZ knock-in reporter in developing inner ear neurons of Isl1lacZ/+ heterozygous mice by X-Gal staining for β-galactosidase activity. Consistent with previous studies of Isl1 expression (Radde-Gallwitz et al., 2004; Deng et al., 2014), Isl1-lacZ expression was detected in all delaminated and migrating CVG neurons as well as in facial ganglion (FG) cells at E10.5–E12.5 (Fig. 1A). In SGNs, Isl1-lacZ expression was strong at E13.5 and persisted to adulthood (Fig. 1A). In addition to the high expression level in SGNs, strong Isl1 expression was also detected in VGNs as well as in the saccule, utricle, and crista (Fig. 1A). Collectively, these results indicate that Isl1 is continuously expressed in early CVG neurons as well as mature SGNs and VGNs.
Inactivation of Isl1 results in abnormal distribution and reduction of inner ear sensory neurons. A, X-Gal staining of Isl1lacZ/+ inner ears reveals the spatiotemporal expression pattern of Isl1-lacZ in developing sensory neurons. B, ISL1/2 immunolabeling shows the ablation of ISL1 expression in Isl1-deficient CVG neurons at E11.5. Note that ISL1 expression remains in the facial ganglion and that the signals in the central CVG region likely show the expression of ISL2. C, Co-labeling of Isl2-V5 knock-in mice with antibodies against V5 and NEUROD1 showing ISL2 expressing in the central CVG region. D, Relative fold change of Isl1 mRNA expression in E11.5 Isl1CKO inner ears by RT-PCR using primers in exons 2 and 3. E, Whole mount X-Gal staining of the P0 inner ear shows that inactivation of Isl1 disturbs the spatial distribution of inner ear sensory neurons. F,G, X-Gal staining of the P0 cochlea whole mount (F) and sagittal section (G) reveal the altered distribution of SGNs in Isl1CKO. H, Flat-mounts of the boxed area in E show the dramatic loss of inferior vestibular ganglion (ivg) cells in Isl1CKO mice. Asterisks in E and H mark ectopically located cells that do not appear to be part of the SGNs, inferior or superior vestibular ganglion in the Isl1CKO inner ear. I, DiI labeling (P0) shows the morphological and projection defects of the spiral and vestibular ganglion neurons in Isl1CKO mice. Asterisks show the absence of SGNs in the hook region in Isl1CKO cochlea. J,K, Flat-mount DiI labeling shows the defects of fiber projection and reduction of VGNs in Isl1CKO inner ear. Arrows indicate fiber projected to the saccule and arrowheads indicate fiber projected to the posterior crista. oe, otic epithelium; cvg, cochleovestibular ganglion; fg, facial ganglion; cd, cochlear duct; sg, spiral ganglion; vg, vestibular ganglion; s, saccule; u, utricle; oC, organ of Corti; svg, superior vestibular ganglion; ivg, inferior vestibular ganglion; m, modiolus; hc, hair cell; ac, anterior crista; lc, lateral crista; pc, posterior crista. Scale bars, 50 µm in (A); 100 µm in (B,C,E–K).
To investigate the impact of Isl1 deletion on CVG development, we crossed Isl1loxP and Isl1lacZ mice with Tg(Pax2-Cre) mice, which express Cre recombinase throughout the inner ear, midbrain, cerebellum, olfactory bulb, and kidney (Ohyama and Groves, 2004), to generate Isl1 conditional knock-out (Isl1loxP/lacZ; Pax2-Cre, hereafter referred to as Isl1CKO) mice to specifically delete Isl1 in the inner ear. We evaluated the efficiency of inner ear-specific Isl1 inactivation by Pax2-Cre using immunolabeling with anti-ISL1 antibody, which recognizes both ISL1 and ISL2 in neurons of CVG and FG. Compared to the control, there was a significant reduction of proliferating CVG cells positive for anti-ISL1 immunolabeling in the Isl1CKO, while the postmitotic neurons in the central CVG remained positive for anti-ISL1 immunolabeling (Fig. 1B). These postmitotic neurons in the central CVG expressed ISL2 and were NEUROD1-negative (Fig. 1C), suggesting that the anti-ISL1 labeling signal detected in the central CVG of the Isl1CKO inner ear should result from the expression of ISL2 but not of ISL1. Consistently, quantitative RT-PCR of Isl1 mRNA using primers in exons 2 and 3 demonstrated that Isl1 mRNA expression was specifically ablated in the developing inner ear of Isl1CKO embryos (Fig. 1D).
In contrast to control Isl1lacZ/+ inner ears in which X-Gal labeled SGNs were densely organized and coiled along the length of the cochlear duct at P0, X-Gal staining for Isl1-lacZ expression in the inner ear whole mounts and sections showed that the Isl1-deficient SGNs formed clump and failed to distribute along the coiled cochlear duct (Fig. 1E–G). We also assessed VGNs that consist of superior vestibular ganglion (SVG) and inferior vestibular ganglion (IVG) cells. In contrast to the orderly and compact arrangement of SVG neurons observed in controls, SVG neurons in Isl1CKO mice showed a scattered and disorganized distribution (Fig. 1E,H). To further verify the developmental defects of SGNs and VGNs in Isl1CKO mice, we performed DiI-labeling to highlight the neurons and their processes. Consistent with X-Gal staining results, DiI-labeling revealed similar defects of SGNs and VGNs in Isl1CKO inner ear at P0 (Fig. 1I–K). While SGNs were regularly distributed along the cochlea in controls, the regular distribution pattern of SGNs was disrupted in Isl1CKO mice, and SGNs were found closer to the modiolus (Fig. 1I). Likewise, DiI-labeling showed that in Isl1CKO mice, the number of IVG and SVG neurons was dramatically decreased, and the fiber projection to the posterior crista was completely lost (Fig. 1J,K). Collectively, these results demonstrate that inactivation of Isl1 disrupts the orderly arrangement of SGNs and VGNs and results in their dramatic loss.
Isl1 inactivation disrupts the differentiation of CVG neurons
To determine how Isl1 inactivation leads to SGN defects, we analyzed the generation and differentiation of CVG neurons at various developmental stages using three well-characterized CVG markers, NEUROD1, POU4F1, and POU4F2 (Kim et al., 2001; Deng et al., 2014). In mice, CVG neurons originate from the otic placode at E9.5 and most CVG neurons are formed between E10.5 and E11.5 (Fritzsch et al., 1999). NEUROD1 is one of the earliest CVG markers (Kim et al., 2001; Deng et al., 2014), and its expression is downregulated in the central region where CVG neurons exit the cell cycle, start to differentiate and express POU4F2 (Davies, 2007; Deng et al., 2014). Compared with the expression patterns of NEUROD1, POU4F1, and POU4F2 in the controls, the NEUROD1− central region was not well defined at E10.5 (Fig. 2A), and no POU4F2+ CVG cell was detected in Isl1CKO mice (arrow in Fig. 2B), implying that the differentiation of CVG neurons is affected at this stage. However, POU4F1+ CVG cells showed no significant difference between Isl1CKO and control inner ears (n = 6) (Fig. 2B,D). Interestingly, the downregulation of NEUROD1 and the onset of POU4F2 expression were detected in a subset of CVG neurons in Isl1CKO mice at E11.5 (arrows in Fig. 2C,D), implying that Isl1 inactivation leads to the delayed differentiation of CVG neurons. To further define the effect of Isl1CKO on the differentiation of CVG, we performed immunolabeling for ISL1 along with NEUROD1 labeling. Immunolabeling of ISL1 antibody detected both ISL1+ proliferating CVG cells and ISL1+/ISL2+ postmitotic CVG neurons in the controls (Fig. 2E,F). In comparison, the expression of ISL1 was absent and anti-ISL1 immunolabeling marked the ISL2+ postmitotic CVG neurons in Isl1CKO. Consistently, co-labeling with NEUROD1 and ISL1 antibodies revealed that the number of NEUROD1−/ISL2+ differentiated CVG neurons in the central region was reduced in Isl1CKO compared to the controls at E10.5 and E11.5 (Fig. 2E,F,K), suggesting the disruption of CVG differentiation in Isl1CKO. We also labeled mitotic cells at S-phase with 5-bromo-2'-deoxyuridine (BrdU) and found that compared to the controls, the BrdU− central CVG region was smaller in Isl1CKO at E10.5 (Fig. 2G), suggesting that CVG neurons fail to exit cell cycle on time in Isl1CKO mice. Taken together, these results indicate that Isl1 is not required for the initial generation of CVG neurons and that Isl1 inactivation disrupts the cell cycle exit and differentiation of CVG neurons.
Inactivation of Isl1 causes a delay in the differentiation and an increase in apoptosis of CVG neurons. A–D, Immunolabeling of inner ear sections for NEUROD1, POU4F1 and POU4F2. Formation of NEUROD1−/POU4F2+ central CVG region is hardly seen in Isl1CKO at E10.5 (A,B) but is comparable to the control at E11.5 (C,D). Arrows indicate differentiated CVG neurons in central region. Arrowheads indicate POU4F2+ differentiating CVG and FG neurons. Dashed boxes indicate CVG region. E,F, NEUROD1 and ISL1/2 co-labeling of inner ear sections at E10.5 (E) and E11.5 (F) show the reduced ISL2+ central CVG region (arrows) in Isl1CKO compared to the controls. G, BrdU labeling of inner ear sections. At E10.5, a BrdU− central CVG region (arrows) is readily identified in the controls but not found in Isl1CKO at E10.5. No difference in cell proliferation is seen between control and Isl1CKO at E11.5. Dashed boxes indicate CVG region. H, DiI-labeling of afferent projection to the basal turn of cochlea shows the reduction in fiber density (asterisks) in the Isl1CKO mice compared with the organized and dense fiber innervation in the control. I, At E13.5, NEUROD1 and CASP3 immunostaining reveal increased cell death in the SGNs of the Isl1CKO mice compared with control. J, At E14.5, POU4F2+ SGNs (arrow) are significantly decreased in the Isl1CKO mice compared with control. K, Quantitation of ISL1/2+/NEUROD1− cells in the central region of CVGs at E10.5 and E11.5. Data are expressed as mean ± SEM (n > 3). ****p < 0.001. L, Quantitation of CASP3+ cells show that cell death is significantly increased in the CVG neurons of the Isl1CKO mice from E12.5. Data are expressed as mean ± SEM (n = 6). *p < 0.05; **p < 0.01. m, modiolus; cd, cochlear duct; sg, spiral ganglion; cvg, cochleovestibular ganglion; fg, facial ganglion; oe, otic epithelium. Scale bars, 50 µm in (A–G,I,J), 100 µm in (H).
Inactivation of Isl1 affects migration and axon pathfinding of CVG neurons and leads to programmed cell death
During development, CVG neurons go through migration and axon pathfinding processes to position their cell bodies and to establish proper synaptic contacts with hair cells. We tested the role of Isl1 in regulating the migration and axon pathfinding of developing CVG neurons at E14.5. While SGNs of the control mice were orderly distributed and extended their fibers to hair cells, SGNs of Isl1CKO mice were scattered and irregularly distributed, and areas devoid of fibers were frequently observed (Fig. 2H, asterisks), which was similarly observed at P0 (Fig. 1I). In addition, the fibers of VGNs extending to the anterior and posterior crista were dispersed and bifurcated in Isl1CKO mice (Fig. 1J), implying an axon pathfinding defect of Isl1CKO VGNs. Combined with X-Gal staining (Fig. 1E–H), these data suggested that Isl1 could play a pivotal role in regulating the migration and axon pathfinding of CVG neurons.
To address how Isl1 inactivation impacts the number of SGNs and VGNs, we investigated cell apoptosis in Isl1CKO CVG neurons using anti-CASP3 immunolabeling. At E11.5, there was no difference in CVG apoptosis between controls and Isl1CKO (Fig. 2L). At E12.5, apoptosis was significantly elevated in the CVG neurons of Isl1CKO (Fig. 2L). The peak of apoptosis of SGNs and VGNs appeared at E13.5 and lasted to at least E14.5 (Fig. 2I,L). Consistently, anti-POU4F2 immunolabeling of sections showed a significant loss of SGNs in Isl1CKO inner ears (Fig. 2J). These data indicate that Isl1 is required for the survival of CVG neurons.
ISL1 and POU4F1 interact directly to co-regulate the differentiation and survival of CVG neurons
Previously, studies have shown that ISL1 and POU4F TFs function in the same regulatory pathways to regulate neuronal differentiation (Pan et al., 2008; Dykes et al., 2011). We thus characterized the functional relationship of Isl1 and Pou4f1 in developing SGNs. We first examined SGNs in Isl1CKO, Pou4f1KO, and Isl1/Pou4f1 double knock-out mice (DKO) at P0 using immunostaining with antibodies against SGN markers NEUN and CALRETININ. Compared to the control, the immunostaining revealed the reduction and disorganization of SGNs in the inner ear lacking either Isl1 or Pou4f1 and strikingly, the loss of almost all of the SGNs in the DKO inner ear (Fig. 3).
Targeted inactivation of Isl1 and Pou4f1 results in the reduction of spiral ganglion neurons at P0. Inner ear cryosections of the control, Isl1CKO, Pou4f1KO and Isl1/Pou4f1 DKO at P0 were immunolabeled with antibodies against NEUN (A) and CALRETININ (B), both makers of SGNs, showing a reduction of SGNs in the inner ears lacking either Isl1 or Pou4f1 and almost complete loss of SGNs in the inner ear deficient of both Isl1 and Pou4f1. cd, cochlear duct; sg, spiral ganglion. Scale bar, 100 µm.
We further examined SGNs in Isl1CKO, Pou4f1KO, and Isl1/Pou4f1 DKO mice at E14.5 and P0 using DiI labeling. Compared with controls, disorganization and reduction in the number of the SGNs were readily seen in Isl1CKO, Pou4f1KO, and DKO inner ears at E14.5 (Fig. 4A). While DiI labeling showed the reduction of SGNs in Isl1CKO and Pou4f1KO mice, almost all of the SGNs were lost in the DKO inner ear at P0 (Fig. 4B). In addition, compared with the densely and orderly organized afferent fibers observed in controls, the overall fiber density was decreased in Isl1CKO or Pou4f1KO cochleae and only a very limited number of afferent fibers were detected in DKO mice at P0 (Fig. 4B). The arrangement of remaining fibers was also profoundly disturbed in DKO mice (Fig. 4B). The severe phenotype seen in DKO mice suggests that Isl1 and Pou4f1 might co-regulate the migration, axon pathfinding and survival of SGNs.
Isl1 and Pou4f1 cooperatively regulate the development of CVG and survival of SGNs. A,B, DiI-labeling of SGN fibers shows the loss of SGNs and innervation defects in Isl1CKO, Pou4f1KO, and DKO mice at E14.5 (A) and P0 (B). C, Combined immunolabeling of NEUROD1 and Myt1 RNAScope ISH of the inner ear sections at E10.5. D,E, Immunolabeling of the inner ear sections with antibodies against SOX2 at E10.5 (D) and NEUROD1 at E11.5 (E). F, Quantitation of CASP3+ apoptotic cells of anti-CASP3 immunolabeled inner ear sections at E13.5. Data are expressed as mean ± SEM (n = 6). *p < 0.05. G, Proximity Ligation Assay using no primary antibody, anti-ISL1, anti-POU4F1, and anti-ISL1/anti-POU4F1 antibodies at E10.5 CVG transverse sections. Dashed boxes indicate CVG region. m, modiolus; sg, spiral ganglion; hc, hair cell; oe, otic epithelium; cvg, cochleovestibular ganglion; fg, facial ganglion. Scale bars, 100 μm (A–E), 30 μm (G).
To understand the SGN defect seen in DKO mice, we examined the initial generation of CVG neurons by labeling of two early neuronal markers, NEUROD1 and SOX2, and a differentiated CVG marker, Myt1 (Sun et al., 2022). At E10.5, the expression of NEUROD1 was detected in CVG neurons of the Isl1CKO, Pou4f1KO, and DKO mice, indicating that CVG neurons are generated and delaminated normally in these mutants (Fig. 4C). Consistently, in contrast to the controls, NEUROD1 downregulated central region was smaller in Isl1CKO mice. However, NEUROD1 was hardly downregulated in the central region of CVG neurons in Pou4f1KO and DKO mice at E10.5 (Fig. 4C). Similarly, co-labeling with anti-NEUROD1 antibody and Myt1 RNAScope probe did not detect differentiated CVG neurons (NEUROD1−/Myt1+) in Pou4f1KO and DKO at E10.5 (Fig. 4C). Similar to NEUROD1 expression, SOX2 downregulation in CVG neurons was increasingly attenuated in an order of control < Isl1CKO < Pou4f1KO < DKO (Fig. 4D). Moreover, compared to Pou4f1KO and Isl1CKO mice at E11.5, compound mutation of Isl1 and Pou4f1 further blocked NEUROD1 downregulation (Fig. 4E). Consistently, we evaluated the effect of DKO on CVG apoptosis using anti-CASP3 immunolabeling at E13.5 and found that loss of Isl1 significantly increased the number of CASP3+ apoptotic SGNs while inactivation of Pou4f1 had no significant effect (Fig. 4F). Intriguingly, compound mutation of Isl1 and Pou4f1 led to a greater loss of SGNs than that in Isl1CKO mice (Fig. 4F).
The more severe phenotypes observed in DKO mice along with the co-expression of ISL1 and POU4F1 in developing CVG neurons prompted us to test whether ISL1 directly interacts with POU4F1 during CVG development. We performed the proximity ligation assay (PLA), which enables the detection of direct protein-protein interaction in situ in cells and tissues (Fredriksson et al., 2002; Hegazy et al., 2020), to evaluate the relationship between ISL1 and POU4F1 in the CVG neurons. Compared to an absence of signals in controls without primary antibody or with antibody against either ISL1 or POU4F1 alone, addition of both anti-ISL1 and anti-POU4F1 antibodies revealed strong signals in the nuclei of CVG neurons at E10.5 (Fig. 4G), implying a direct interaction between ISL1 and POU4F1. Collectively, these data show that ISL1 and POU4F1 interact directly to regulate the differentiation and survival of CVG neurons.
Isl1 and Pou4f1 epistatically regulate CVG-specific genes
To elucidate the GRN of Isl1 and Pou4f1 and to investigate how they regulate CVG development, we performed RNA-Seq analysis of the control, Isl1CKO, Pou4f1KO, and DKO CVG neurons at E11.5 when CVG defects are evident but CVG apoptosis has not started yet. We identified 761 downregulated genes and 445 upregulated genes (p adj <0.05, Log2FC > 2) in CVG neurons null for Isl1 or Pou4f1 or both (Fig. 5A). We noted that these differentially expressed genes (DEGs) were affected by the inactivation of Isl1 or Pou4f1 alone or both. We classified these DEGs into 3 subgroups and 8 clusters based on the genotypes for further overlapping analysis. The main group (cluster 1–4) consists of genes whose expression were altered in DKO CVG neurons, e.g., Ntng2, Meis2, Eya1, Shh, and Sall3 (Fig. 5B). The other two groups (cluster 5–8) were mainly impacted by inactivation of either Isl1 (e.g., Gata3, Fgf10, Ntrk3, and Epha5 of cluster 5) or Pou4f1 (e.g., Pou4f2 and Isl2 of cluster 7) (Fig. 5B). Importantly, many of these DEGs have known roles in CVG development (Schimmang et al., 1995; Fritzsch et al., 1998; Luo et al., 2013), indicating that Isl1 and Pou4f1 cooperatively as well as independently regulate genes associated with CVG differentiation.
Isl1 and Pou4f1 epistatically or cooperatively regulate genes enriched in CVG neurons. A, A Venn diagram indicating the overlap of DEGs down- or upregulated in Pou4f1KO, Isl1CKO, and DKO CVG neurons. B, Clustered heatmaps of the 8 clusters of DEGs from A. C, Top 10 GO terms of cluster 1 and 3. Genes of cluster 3 are also enriched in the terms that cluster 1 is involved in. D, Top 10 GO terms of cluster 5 and 7 show the common biological processes. E, Changes in the expression of selected DEGs revealed by qRT-PCR and RNA-Seq analysis. qRT-PCR primers used were listed in Extended Data Table 5-1. Data are expressed as mean ± SEM (n = 3). *p < 0.05, **p < 0.01, ***p < 0.001 and so forth.
Table 5-1
PCR Primers Used in qRT-PCR Analysis. Download Table 5-1, DOCX file.
Gene Ontology (GO) terms analysis showed that the DEGs positively regulated by both Isl1 and Pou4f1 were highly enriched in the biological processes of axonogenesis, axon guidance, neuron projection guidance and synapse organization (e.g., Ntng2, Epha10, Sema3e, Ngfr, Shh, and Ntrk2) (Fig. 5C). In addition, Isl1 and Pou4f1 independently activated genes commonly involved in biological processes of ear development, axonogenesis, and neuronal migration (e.g., Gata3, Fgf10, and Ntrk3 of cluster 5 and Pou4f2, Isl2, and Prrxl1 of cluster 7) (Fig. 5D). To confirm the RNA-Seq results, we performed qRT-PCR of 13 selected genes using control and mutant CVG neurons and normalized the expression levels to an internal control, Actb. Consistent with the RNA-Seq data, we noted that Epha5, Gata3, and Fgf10 were down regulated in Isl1CKO and DKO, Isl2 and Pou4f2 downregulation was more severe in Pou4f1KO than in Isl1CKO, and Sall3, Ntng2, Meis2, and Eya1 were significantly downregulated in all three mutants (Fig. 5E). We further confirmed the changes in the expression of 11 genes mentioned above by in situ hybridization or immunolabeling in E11.5 CVG neurons. While all genes were dramatically downregulated in DKO, the downregulation of Fgf10 and Gata3 were largely caused by Isl1 inactivation and conversely, Isl2 and Pou4f2 downregulation could be attributed mostly to the loss of Pou4f1 (Fig. 6). Thus, Isl1 and Pou4f1 co-regulate a common set of genes and also act independently to control separate groups of genes essential for CVG development.
Isl1 and Pou4f1 regulate the expression of a common set as well as unique sets of CVG-specific genes. Confirmation of gene expression changes in CVG neurons by ISH and immunolabeling of inner ear cryosections of control, Isl1CKO, Pou4f1KO and DKO mice at E11.5. (A) Fgf10. (B) Epha5. (C) Ntng2. (D) Eya1. (E) Meis2. (F) Sall3. (G) Isl2. (H) Pou4f2. (I) Shh. (J) NTRK3. (K) GATA3. Dashed lines indicate CVG regions. Scale bars, 100 µm.
ISL1 and POU4F1 directly bind to the regulatory elements in DEGs
To determine how Isl1 and Pou4f1 regulate the DEGs identified by RNA-Seq analysis, we identified genomic regions occupied by ISL1 and POU4F1 using the CUT&Tag approach (Kaya-Okur et al., 2020), which is developed from ChIP-Seq and CUT&RUN (Skene and Henikoff, 2017) for its high specificity with a low number of cells, and compared these regions to those bound by H3K4me3 and H3K27ac or by H3K27me3, which mark active or silenced chromatin regions, respectively. Data from duplicate experiments of each protein were combined and peaks were called against IgG controls. The locations of ISL1- and POU4F1-bound peaks showed a significant enrichment at the proximity of transcriptional start site (TSS) (Fig. 7A), suggesting their roles in initiating transcription. Genomic feature annotation showed more than half of ISL1 (78.5%) and POU4F1 (80%) peaks were distal, intergenic, and intronic, which is similar with those of H3K27ac (62.4%), but different from those of H3K27me3 (36.5%) and H3K4me3 (7.7%). The majority of H3K27me3 (53%) and H3K4me3 (91.1%) bindings are enriched in promoter region, which is consistent with their known roles (Fig. 7B). These results suggest that ISL1 and POU4F1 could activate the transcription potentially via engaging in the promoter-enhancer contacts.
Epistatic and combinatorial regulation of CVG-specific genes by ISL1 and POU4F1 at E11.5. A, Genomic distribution of POU4F1-bound (top) and ISL1-bound (bottom) peaks relative to the distance from transcriptional start site (TSS). TTS indicates the transcriptional termination site. B, Genomic annotation of peaks identified by CUT&Tag analysis using antibodies against POU4F1, ISL1, H3K27ac, H3K27me3, and H3K4me3. C, A Venn diagram showing the overlaps of DEGs identified by RNA-Seq and genes found by CUT&Tag analysis. 103 downregulated and 30 upregulated genes are ISL1 and POU4F1 co-bound. D, Biological processes of POU4F1-bound (left) and ISL1-bound (right) genes are often related to neuronal system, nervous system development and axon guidance. E–L, Integrative Genomics Viewer (IGV) track visualization of peaks identified by snATAC-Seq of prosensory cells, SGNs and VGNs, and CUT&Tag analysis using antibodies against POU4F1, ISL1, H3K27ac, H3K27me3, and H3K4me3. Colored boxes mark the genomic regions co-bound by ISL1 and POU4F1 (black boxes) or bound by just ISL1 (red boxes) or POU4F1 (green boxes). Y-axis value of each track indicates track height. The direction of transcription is shown by arrows. Asterisks indicate the peaks selected for functional tests in vivo.
To identify the genes directly regulated by ISL1 and POU4F1, we first filtered the peak regions for predicted binding sites by Bedtools intersect and identified a total of 1,459 ISL1-bound and 5,446 POU4F1-bound genes (Fig. 7C). By integrating these genes with the DEGs identified by RNA-Seq analysis, we obtained 103 downregulated and 30 upregulated genes co-bound by ISL1 and POU4F1, 48 ISL1-bound and 220 POU4F1-bound downregulated genes and 24 ISL1-bound and 132 POU4F1-bound upregulated genes (Fig. 7C). GO term enrichment analysis showed that ISL1 and POU4F1 targeting genes were often related to neuronal system, nervous system development and axon guidance (Fig. 7D).
The CUT&Tag peaks of ISL1 and POU4F1 were further examined by intersecting with snATAC-Seq peaks identified from SGNs and VGNs at E12.5. The snATAC-Seq peaks identified from a cluster of prosensory cells (Ps cells) were plotted as non-neuronal controls. The representative Integrative Genomics Viewer (IGV) genome track snapshots showed the binding sites of ISL1 or POU4F1 or both at the verified downregulated genes: a highly active distal enhancer in Eya1 with a specificity in SGNs and VGNs and with H3K27ac-deposition was occupied by both ISL1 and POU4F1 (Fig. 7E); two highly active distal enhancers in Ntng2 more dynamic in SGNs and VGNs and with H3K27ac-deposition were bound by ISL1 only (Fig. 7E); and a distal enhancer and an intronic enhancer with H3K27ac-deposition in Ntrk3 were bound by ISL1 and POU4F1, respectively (Fig. 7F). We also identified quiescent regions such as the ISL1-bound Epha5 promoter region and an ISL1-bound Fgf10 promoter region with H3K27ac-, H3K27me3-, and H3K4me3-deposition. In addition, a distal enhancer with H3K27ac and H3K27me3 deposition in Fgf10 was bound by ISL1 only (Fig. 7G). In Pou4f2, we found a POU4F1-bound biovalent promoter region with H3K4me3- and H3K27me3-deposition and a biovalent distal enhancer co-bound by ISL1 and POU4F1 (Fig. 7G). These two regions in Pou4f2 have been previously reported to regulate the expression of Pou4f2 in the retina (Ge et al., 2023). Furthermore, we found the occupancy of ISL1 and POU4F1 in other genes associated with axon guidance. For instance, Sema3e was bound by ISL1 at a neuron-specific distal enhancer (Fig. 7H), Gap43 was co-bound by ISL1 and POU4F1 at a distal, an intronic, and a downstream highly active enhancers (Fig. 7I), and Nrcam, a gene expressed in SGNs and required for the innervation between SGNs and hair cells (Harley et al., 2018), was co-bound by ISL1 and POU4F1 at two intronic enhancers (Fig. 7J). In addition, Ppp1r1c and Runx1, two DEGs previously identified in DKO DRG neurons (Dykes et al., 2011), showed different patterns of ISL1 and POU4F1 occupancy that in Ppp1r1c, ISL1 and POU4F1 co-bind to an exonic and downstream enhancer (Fig. 7K), whereas ISL1 and POU4F1 individually bind to two distal enhancers in Runx1 (Fig. 7L). Thus, these discoveries imply that ISL1 and POU4F1 could regulate the axon guidance and survival of the CVG neurons by shaping the epigenetic landscape of associated genes through binding to the active promoters or enhancers.
Enhancers bound by ISL1 and POU4F1 are functionally important in CVG neurons
As a first step to gain further insight into the mechanisms of ISL1 and POU4F1 function and to validate the biological significance of these cis-regulatory elements (CREs) in vivo, we used genome editing by CRISPR/Cas9 approach to delete selected CREs in mice and analyzed the effects on gene expression in homozygous deletion mutants by in situ hybridization. The expression of the neuron-enriched gene Fgf10 is greatly downregulated in Isl1CKO (Figs. 5E, 6A), and two adjacent active CREs of Fgf10 were bound by ISL1 (Fig. 7G). We designed two single-guide RNA (sgRNAs) to delete the 965 bp enhancer of Fgf10 (Chr13: 118711166–118712131) (Figs. 7G, 8A,B) and found that the deletion of this Fgf10 enhancer led to a significant reduction in Fgf10 expression in CVG neurons at E11.5 (Fig. 8C), strongly arguing for its biological function in regulating Fgf10 expression. Similarly, the downregulation of Ntng2 in DKO mice (Figs. 5E, 6C) and occupancy of ISL1 at its two distal enhancers (Fig. 7E) prompted us to test their functional significance in vivo, too. We used two pairs of sgRNAs to remove the two distal enhancers containing multiple ISL1 binding sites predicted from JASPAR database, 805 bp at Chr2: 29272157–29272961 or ~24 kb upstream of the TSS and 1,563 bp at Chr2: 29285119–29286682 or ~35 kb upstream of the TSS (Fig. 8A,B). In addition, the ~35 kb Ntng2 enhancer contains two predicted POU4F1 binding sites (Fig. 8B). Deletion of the ~24 kb upstream enhancer notably reduced the expression of Ntng2 while the deletion of the ~35 kb upstream enhancer also resulted in, though to a lesser extent, a decrease in Ntng2 expression (Fig. 8C). These findings further show that ISL1 and POU4F1 function epistatically as well as independently to regulate the CVG differentiation program.
Deletions of ISL1-bound enhancers attenuate the expression of Fgf10 and Ntng2. A, Deletion strategies of ISL1-bound peaks in Fgf10 and Ntng2 using two Cas9 sgRNAs to flank each peak. sgRNAs and PCR genotyping primers used for enhancer deletions by CRISPR/Cas9 were listed in Extended Data Table 8-1. B, Top panels are predicted ISL1 and POU4F1 motifs from JASPAR database. Lower panels show ISL1 and POU4F1 motifs in selected peaks. C, ISH showing the downregulation of genes in CVG neurons of enhancer deletion mutants at E10.5. Dashes indicate the CVG region. cvg, cochleovestibular ganglion. Scale bar, 100 µm. D, A graphic conclusion of the roles of Isl1 and Pou4f1 in the gene regulatory network during CVG differentiation.
Table 8-1
sgRNAs for Enhancer Deletions by CRISPR/Cas9 and Genotyping Primers. Download Table 8-1, DOCX file.
Discussion
It is well established that ISL1 and POU4Fs are co-expressed in developing RGCs and sensory neurons of DRG and TG and regulate their development (Mu et al., 2008; Pan et al., 2008; Sun et al., 2008; Dykes et al., 2011). During RGC development, Pou4f2, which is expressed before Pou4f1 and is required for activating Pou4f1 expression, has a stronger effect than Pou4f1 on RGC differentiation and the expression of RGC-specific genes (Gan et al., 1996; Pan et al., 2008). Pou4f2 and Isl1 function in the same genetic pathway to co-regulate RGC development. While knock-out of either gene results in a partial loss of RGCs, Isl1/Pou4f2 DKO leads to the loss of nearly all RGCs (Pan et al., 2008). In TG and DRG neurons, the regulatory relationship of Pou4f1 and Pou4f2 is reversed compared to that in RGCs, where Pou4f1 is expressed before Pou4f2 and is required for Pou4f2 expression. Similarly, in Pou4f1/Isl1 DKO mice, the initial formation of TG and DRG is not affected but DRG neurons are displaced, and their peripheral projections are disrupted (Dykes et al., 2011). In our current study, we found that Isl1 controls the differentiation of SGNs and is required for their maintenance, axon pathfinding and migration. Without Isl1, a majority of SGNs fail to settle along the length of the cochlear duct and to extend neural fibers to hair cells, and a significant number of SGNs undergo apoptosis. These phenotypes are similar to those reported in a recent study (Filova et al., 2022) and further demonstrate the critical role of Isl1 in SGN development. Moreover, Pou4f1KO leads to disorganization and reduction in SGNs though to a lesser extent than Isl1CKO, and Pou4f1/Isl1 DKO severely disrupts SGN differentiation and results in a nearly complete loss of SGNs. Thus, our study uncovers a remarkable similarity of the role of Pou4f1 and Isl1 in developing SGNs and sensory neurons of TG and DRG, arguing for a conserved functional mechanism of Isl1 and Pou4fs in both developing central and peripheral nervous systems.
The hallmarks of Pou4f1/Isl1 DKO defects are the delayed cell cycle exit of CVG neurons, the disorganization and reduction in SGNs and neural fibers, and SGN death. As TFs, ISL1 and POU4F1 could promote cell cycle exit and neuronal differentiation by repressing neurogenic genes and activating genes essential for axon pathfinding and migration. Previous studies have shown that both Isl1 and Pou4f1 repress neurogenic bHLH genes of the Neurog and Neurod classes, e.g., Neurod1, Neurod4, and Neurog1 in DRG and TG (Eng et al., 2007; Lanier et al., 2007; Dykes et al., 2011). Interestingly, we found that Isl1 and Pou4f1 co-repress Neurod1 expression in CVG neurons (Fig. 4), providing an insight into how Isl1 and Pou4f1 regulate neurogenesis and cell cycle exit. Intriguingly, the SGN defects observed in the DKO mutants show some similarity to the phenotypic defects observed in Neurod1-null mice, in which approximately 90% of CVG neurons were lost (Kim et al., 2001). However, unlike Neurod1-null mice, Isl1 and Pou4f1 DKO do not affect the initial generation and delamination of CVG neurons, suggesting different roles of Neurod1 in CVG neurogenesis and of Isl1 and Pou4f1 in CVG differentiation. Our previous study demonstrates a sequential expression order of NEUROD1 > ISL1 > POU4F1 > POU4F2 during the inner ear neurogenesis and shows that ISL1 expression in CVG neuron is downregulated in Neurod1-null mice but is not affected in Pou4f1KO CVG neuron (Deng et al., 2014). Here, we further show that Isl1CKO does not affect the expression of Pou4f1 or the onset of Neurod1 expression (Fig. 2A–D). These results suggest that Isl1 and Pou4f1 act in parallel and immediately downstream of Neurod1 in a regulatory hierarchy of NEUROD1 > ISL1⬄POU4F1 to drive the differentiation of CVG neurons into SGNs.
To understand how ISL1 and POU4F1 regulate the differentiation, migration and axonogenesis of SGNs, we found through expression analysis that Isl1/Pou4f1 DKO results in the downregulation of neurotrophin receptor genes (Ntrk2 and Ntrk3) and RUNX gene (Runx1) (Figs. 5, 6), consistent with the finding that Isl1 and Pou4f1 positively regulate the expression of Ntrk and Runx in DRG and TG (Eng et al., 2007; Lanier et al., 2007; Sun et al., 2008). Additionally, Ntng2 plays an essential role in axon guidance and homozygous missense mutation of NTNG2 leads to developmental delay globally and other neurodevelopmental disorders in humans (Dias et al., 2019). Inactivation of Isl1 or Pou4f1 or both leads to the downregulation of Ntng2 (Figs. 5, 6), which might contribute to defective SGN axonal processes observed in these mutant mice. Moreover, we discover that Isl1 and Pou4f1 activate Eya1, a pioneer gene upstream of Neurog1 and essential for the survival of placodally delaminated sensory neurons (Zou et al., 2004). Interestingly, interaction between Eya1 and Six1 also plays a role in determining the neuronal fate and subtype identities of early auditory neurons (Zou et al., 2004; Ahmed et al., 2012). The downregulation of Eya1 is first seen in the central CVG region in Isl1/Pou4f1 DKO mice (Fig. 6D), suggesting Eya1's late role downstream of Isl1 and Pou4f1 in the differentiation of postmitotic CVG neurons. Besides their many shared downstream target genes, Isl1 and Pou4f1 each regulate unique sets of genes. Isl1 is largely responsible for the activation of Fgf10 and Gata3, while Isl2 and Pou4f2 expression mostly rely on Pou4f1 (Fig. 6). These data show that Isl1 and Pou4f1 epistatically regulate many of their target genes as well as play the predominant or exclusive role in regulating other unique sets of genes during CVG development.
In addition to its expression in early CVG neurons starting at E9.5 (Fedtsova and Turner, 1995), Pou4f1 expression is found to be restricted in type I SGNs later (Shrestha et al., 2018; Sherrill et al., 2019; Petitpre et al., 2022). On the other hand, Isl1 expression persists in both SGNs and VGNs. Many of the downstream DEGs found in this study are neuronal subtype-featured genes, for example, Meis2, Gata3, Runx1, Epha5, Ntrk2, Ntrk3, and Casz1 (Shrestha et al., 2018; Petitpre et al., 2022). Considering their function of subtype specification in other system or species, potential roles of Isl1 and Pou4f1 in regulating the specification of auditory neuronal subtypes will be of great interest for future studies.
In our genome-wide search for the downstream effector genes directly regulated by Isl1 and Pou4f1, we combined CUT&Tag and snATAC-Seq analyses to obtain the location and identity of ISL1 and POU4F1 binding sites at the downstream genes. These elements consist of promoters and active enhancers. Interestingly, we found that ISL1 and POU4F1 could co-regulate gene expression through binding to the same or different sites. The majority of ISL1 and POU4F1 binding sites are distal and intronic enhancers, suggesting their transactivating function through these enhancers (Fig. 7). It is known that the promoter-enhancer contacts orchestrate the genome-wide gene transcription (van Arensbergen et al., 2014); ISL1 and POU4F1 might regulate the cross-talking between promoters and enhancers by binding to these enhancers and bringing them together with promoters. Previous study in Caenorhabditis elegans reports that unc-86 (a Pou4f1 ortholog) and mec-3 (an Isl1 ortholog) form an UNC-86/MEC-3 heterodimer to bind to the promoters of mec-4 and mec-7 (Duggan et al., 1998). The co-expression of ISL1 and POU4F1 in developing CVG neurons and their epistatic interaction on their many target genes suggest that ISL1 and POU4F1 could form a heterodimer in CVG neurons. Indeed, our PLA assay provides the first in vivo evidence of the direct interaction between ISL1 and POU4Fs in the developing CVG neurons (Fig. 4G) and further suggests similar interactions between ISL1 and POU4Fs in other neurons.
Using CRISPR/Cas9 approach to delete enhancers bound by ISL1 and POU4F1, we found that the deletion of Ntng2 enhancers shows different degree of effects on its expression in CVG neurons, and these effects are less severe than those observed in CVG null for Is11 or Pou4f1 or both, suggesting other factors or multiple regions are involved in these transcription processes. It is also possible that the interaction between ISL1 and POU4Fs could be far more complex than ISL1-POU4F dimer, such that they might be involved in “transcription factory” for co-regulated subgroup of genes to share regulatory factors in a large subnuclear transcription structure within the nucleus (Edelman and Fraser, 2012). Future studies are needed to gain a further insight into the mechanism of ISL1 and POU4F1 in regulating CVG differentiation by fully characterizing ISL1 and POU4F1 binding sites and their functional significance.
Taken together, we have uncovered the essential role of ISL1 in the development of CVG neurons and SGNs. More importantly, we have revealed the novel direct interaction of ISL1 and POU4F1 and have identified the GRN of ISL1 and POU4F1 during the development of auditory neurons using a combined approach of genetic, protein-protein interaction, and multiomic methods. Our studies establish the role of Isl1 and Pou4f1 in developing CVG and define the diverse relationship among Isl1, Pou4f1 and their target genes (Fig. 8D). Importantly, our findings shed light into the general principle of ISL1 and POU4Fs function in the development of various nervous systems.
Footnotes
We thank Dr. Xiuqian Mu at the University of Buffalo for thoughtful discussion and assistance with CUT&Tag experiments and Drs. Min Deng and Ling Pan for technical assistance. We gratefully acknowledge the Genome Editing Core at Augusta University for generating the mouse models necessary to support this research and The Integrated Genomics Shared Resource in the Georgia Cancer Center for providing snATAC-Seq service. Lin Gan is supported by the NIH funding (DC008856 and DC021456), Augusta University Medical College of Georgia, and the Georgia Research Alliance.
The authors declare no competing financial interests.
- Correspondence should be addressed to Lin Gan at ligan{at}augusta.edu.
