Single‐cell RNA sequencing: Inhibited Notch2 signalling underlying the increased lens fibre cells differentiation in high myopia

Abstract High myopia is the leading cause of blindness worldwide. It promotes the overgrowth of lens, which is an important component of ocular refractive system, and increases the risks of lens surgery. While postnatal growth of lens is based on the addition of lens fibre cells (LFCs) supplemented by proliferation and differentiation of lens epithelial cells (LECs), it remains unknown how these cellular processes change in highly myopic eyes and what signalling pathways may be involved. Single‐cell RNA sequencing was performed and a total of 50,375 single cells isolated from the lens epithelium of mouse highly myopic and control eyes were analysed to uncover their underlying transcriptome atlas. The proportion of LFCs was significantly higher in highly myopic eyes. Meanwhile, Notch2 signalling was inhibited during lineage differentiation trajectory towards LFCs, while Notch2 predominant LEC cluster was significantly reduced in highly myopic eyes. In consistence, Notch2 was the top down‐regulated gene identified in highly myopic lens epithelium. Further validation experiments confirmed NOTCH2 downregulation in the lens epithelium of human and mouse highly myopic eyes. In addition, NOTCH2 knockdown in primary human and mouse LECs resulted in enhanced differentiation towards LFCs accompanied by up‐regulation of MAF and CDKN1C. These findings indicated an essential role of NOTCH2 inhibition in lens overgrowth of highly myopic eyes, suggesting a therapeutic target for future interventions.

Research Center of Laser and Autostereoscopic 3D for Vision Care, Grant/Award Number: 20DZ225500; Construction of a 3D digital intelligent prevention and control platform for the whole life cycle of highly myopic patients in the Yangtze River Delta, Grant/Award Number: 21002411600 These findings indicated an essential role of NOTCH2 inhibition in lens overgrowth of highly myopic eyes, suggesting a therapeutic target for future interventions.

| INTRODUCTION
High myopia, defined as eye axial length larger than 26 mm or spherical equivalent greater than À6.00 D, is now a leading cause of blindness and becomes increasingly prevalent across the world. 1 Lens is an important component of ocular refractive system, and changes in its transparency or size would lead to vision impairment. We previously identified enlarged lens sizes in human highly myopic eyes associated with excessive buildup of lens structural proteins (β/γ-crystallin), 2,3 which was recapitulated in two independent highly myopic mouse models. The enlarged lens size in high myopia contributes to higher incidences of surgical complications, such as intraocular lens malposition after lens replacement surgery, impairing visual outcomes. [3][4][5] Though postnatal growth of lens is achieved through addition of fibre cells supplemented by the lifelong proliferation and differentiation of lens epithelial cells (LECs), 2,6,7 it remains largely unknown about how these cellular processes get changed and play roles in highly myopic eyes.
LECs are the single layer of cells that locate underneath the anterior and equatorial part of the lens capsule. LECs underwent heterogeneous cellular processes under the same temporal dimension, including proliferation, migration, and fibre cells differentiation. 8 Meanwhile, since they are the outermost layer of cells in the lens to get exposed to the ocular environment, LECs play essential roles for lens biology and pathology.
However, there currently does not exist a characterization of transcriptional landscape involved in the heterogeneous lens epithelium biology at single-cell levels, not to mention that how it would change in the pathological environment of highly myopic eyes and consequently induce lens abnormalities. Recently developed single-cell RNA sequencing (sc-RNA seq) technology now provides a potent tool for characterization of the underlying mechanism at single-cell resolution.
In this study, we obtained mouse lens epithelium from both highly myopic and contralateral control eyes for sc-RNA seq analysis. We outlined the single-cell transcriptome atlas and the lineage differentiation trajectory for LECs, and found increased differentiation towards fibre cells in highly myopic eyes which was regulated by Notch2 down-regulation, providing a novel perspective for understanding lens abnormalities, especially lens overgrowth, in highly myopic eyes.

| Animals
During the experiment, mice were housed in specific pathogen free (SPF) barrier facilities at a constant temperature of 21 C with 40%-60% humidity and under a regular 12 h light/dark (L/D) cycle (light on at 7:00 AM and off at 7:00 AM).
The defocus-induced high myopia mouse model was established by attaching a À25.00 D lens to the skin around the right eye of a 4-week-old male C57BL/6J mouse, while the fellow eye served as control. The SLAC Laboratory Animal Co. Ltd. (China) supplied all the mice used in this study. The refraction of both eyes was measured with a mouse infrared photorefractor (Steinbeis Transfer Center, Germany) at the beginning and end of the study. Mice with baseline refraction of more than 1.00 D between eyes were excluded. The lenses were checked every morning to ensure their attachment. Four weeks later, only mice with the right eye exhibiting more than À6.00 D of myopia compared to the left eye were regarded as effective models of high myopia.

| Lens epithelial cell isolation and single-cell RNA sequencing
Lens epithelium (n = 90 for highly myopic eyes, n = 90 for contralateral control eyes) were isolated in DMEM (Thermo Fisher Scientific, USA) on ice. After digestion in TrypLE (Thermo Fisher Scientific) at 4 C for 4 h and then at 37 C for 15 min, cells were suspended and filtered through a 40 μm cell strainer (Bel-Art, USA), centrifuged (300 g,

| Data pre-processing and quality control
CellRanger software (version 4.0.0) was used to pre-process raw sequencing data. After demultiplexing, fastq-files were generated and aligned to mm10 mouse reference genome and transcriptome to produce gene versus cell expression matrixes. The expression matrixes were then processed using Seurat package (version 4.0.6) in R software (version 4.0.5). After quality control, cells with >200 genes and <2500 genes and <10% mitochondrial RNA were retained.

| Single-cell analysis
Using Seurat package in R software, the filtered expression matrixes were integrated, scaled, and normalized. Highly variable genes were detected and used for the principal component analysis (PCA). Cell clustering was subsequently performed and visualized with the top 10 principal components (PC) using uniform manifold approximation and projection (UMAP) at a resolution of 0.6. Then, we used wellestablished marker genes to annotate each cell type based on their average expression.

| Cluster markers identification and functional enrichment analysis
Differentially expressed genes (DEGs) or marker genes for each cluster were generated after running the 'FindAllMarkers' function, and filtered by PCT >25% (at least 25% of cells in either of the two populations compared express the gene), Log 2 FC (Fold Change) >0.25 and p value (Bonferroni adjust) <0.05. The DAVID database (https://david. ncifcrf.gov/home.jsp) and the Metascape tool (http://metascape.org) was used for functional enrichment analysis. 9,10 2.7 | Cell cycle and differentiation analysis and pseudotime transcriptional trajectory analysis Based on the expression of genes related to the G2/M and S phases as well as differentiation states, the 'CellCycleScoring' function was used to determine the cell cycle and differentiation score for each cell before matching it to the metadata. 11,12 Then, cells are categorized into specific cell cycle and differentiational stages according to their scores.
We employed 'Monocle2' package (version 2.18.0) for pseudotime analysis using 400 marker genes generated from 'differential-GeneTest' function. 13 RNA counts in all cells from LEC and LFC clusters were chosen as input for downstream analysis. Lineage differentiation trajectory among LEC and LFC clusters is performed using default parameters of 'Monocle' after DDRTree-based dimensionality reduction and cell ordering. The proliferation state suggested the start point of the pseudotime when running the 'orderCells' function. Visualization was realized using the 'plot_cell_trajectory' function.

| Primary human and mouse LEC culture and siRNA transfection
For primary human LEC culture, human lens epithelial samples from highly myopic eyes were placed with LECs facing upward and cultured in DMEM (20% FBS) at 37 C with 5% CO 2 . The medium was changed every second day. The same method was used for primary mouse LEC culture with lens epithelial samples obtained from

| Statistics analysis
Experimental data were presented as mean ± standard error (SD).
Paired t-test was used for the comparison of refraction between highly myopic and fellow control eyes. Independent-sample t-test was used for the comparison of other experimental data between two groups. Differences with p value <0.05 were considered to be statistically significant.

| Single-cell expression atlas and cell types of lens epithelium
To investigate the cell profiling of lens epithelium in both highly myopic and emmetropic eyes, the defocus-induced highly myopic mouse models were established using À25.00 D lens attached to the right eyes ( Figure S1). After 4 weeks of induction, the refraction of the defocus-induced eyes (right eyes) was significantly more myopic than that of untreated fellow eyes (left eyes) (0.62 ± 0.97 D vs. 12.89 ± 1.60 D, p value <0.001, paired t-test). The lens epithelium was then dissected from the highly myopic and emmetropic eyes respectively, and prepared for sc-RNA seq ( Figure 1A). Using the 10Â Genomics    Figure S1), which represented contamination of non-lens-epithelium cells and were then excluded from the downstream analysis. Finally, a total of 10 clusters which highly expressed established pan-LEC markers, such as Cryab ( Figure S1), 14 were identified as LECs.
By comparing gene expression patterns, we found that these cell clusters can be distinguished by specific gene sets (Figure 2A). The proportions of each LEC cluster in highly myopic and control lens epithelium were exhibited in Figure 2B,C. We firstly scored each cell based on their expression levels of G2M or S phase marker genes. We found that all the cells in cluster 10 exhibited high G2M or S scores, and were therefore classified to be in either G2M or S phase of cell cycle, whereas the majority of cells in other clusters were identified to be in the G1 phase ( Figure 2D,E). Cluster 10 was then identified as a proliferating LEC subpopulation, as confirmed by its enriched function for cell proliferation and specific expression of relevant markers (e.g., Top2a and Cdk1) ( Figure 3F,G). Since LECs undergo distinct cellular states under natural conditions, functional enrichment analysis was performed for each cluster using their marker genes ( Figure S2). Roles of genes specific for cluster 1, 3, 4, and 6 were related to regulation of cell morphogenesis (e.g., Anxa1, S100a10, and Myl12a), collagen formation (e.g., Ctsb, Ctsl, Pcolce, and Ppib), cell junction organization (e.g., Cdh2, Hdac7, and Gjd3), and extracellular matrix organization (e.g., Ccn2, Nid1, and Prdx4), respectively, indicating that they performed basic functions of maintaining cell polarity and constituting lens capsular matrix. Genes of cluster 2 was enriched for regulation of cell population proliferative and developmental process (e.g., Notch2, Ctsl, Gja1, and Ecrg4), suggesting its possible regulative effects on lens growth and development. Cluster 8 specifically expressed type IV collagen genes (Col4a1, Col4a2, Col4a3, and Col4a4) (Figure 2A), and was enriched for epithelial cell migration (e.g., Efnb2, Itgb1, and Slit2), suggesting that there existed a proportion of LECs related to epithelialmesenchymal transition, cell migration and lens fibrotic diseases. 19,20 Remarkably, cluster 7 and 9 highly expressed crystallin β/γ genes (e.g., Cryba2, Crybb1, and Crygn) (Figure 2A) and could therefore be further annotated as fibre cells as a result of LEC differentiation. 21 On the contrary, genes for cluster 5 were enriched for negative regulation of epithelial cell differentiation (e.g., Notch1, Hes1, Tgfbr1, and Vegfa) and homeostasis maintenance of cell numbers within the tissue.

| Pseudotime analysis exhibited differentiation trajectory of lens epithelial cells
To characterize the differentiation trajectory of LECs, pseudotime analysis was further conducted, where cells were gathered along a trajectory according to their gene signatures and calculated pseudotime without prior clustering information ( Figure 3A). Cluster 1 was presented throughout the pseudotime trajectory and composed the majority of 2 minor branches, indicating that maintaining the morphogenesis of polarized epithelium was needed during all these processes. Except for that, the cells were ordered into 1 root and 3 major branches by the unsupervised algorithm. Mitotic and post-mitotic LECs (cluster 10 and cluster 2) characteristically populated at the root of the trajectory, while the three major branches were mostly composed of maturely differentiated LFCs (branch 1; cluster 7, and 9), homeostasis-maintaining cells (branch 2; cluster 5) and static or migrating cells performing basic functions (branch 3; cluster 3, 4, 6, and 8), respectively. These results were consistent with the expression atlas defined by UMAP, further supporting our prior observations. Figure 3C exhibited the average pseudotime of all cells in highly myopic and control lens epithelium, and Figure 3D exhibited that of each LEC cluster. In consistence with functional enrichment analysis, proliferating population (cluster 10) appeared at early periods, while differentiating population (cluster 7 and 9) appeared at terminal periods ( Figure 3D, E). We further analysed the dynamic gene expression patterns along the trajectory of LEC differentiation. Of note, genes enriched for Notch signalling pathway (Notch2, Psenen, Psen2, Aph1b, Aph1c, and Dll1) were found to be significantly inhibited during lineage differentiation trajectory of LECs ( Figure 3F). The expression patterns of all genes related to Notch signalling pathway in each LEC cluster was exhibited in Figure S3. Of these genes, Notch2 showed the highest level of association with pseudotime (q value = 4.96 Â 10 À41 ) and remarkably decreased at early periods ( Figure 3G,H), indicating its potential role in regulating the differential process of LECs.

| Higher proportion of differentiated fibre cells in lens epithelium of highly myopic eyes
As was shown in Figure 2C, the proportion of LFCs (cluster 7 and 9) was significantly higher in highly myopic eyes. In consistence, the average pseudotime of LECs in highly myopic eyes was higher than that in control eyes ( Figure 3C

| Down-regulation of Notch2 promoted differentiation towards fibre cells in the lens epithelium of highly myopic eyes
To determine the possible mechanism underlying the increased fibre cells in the lens epithelium of highly myopic eyes, we combined the results from pseudotime analysis and biological functions underlying significantly altered clusters in these eyes. Prior pseudotime analysis revealed the down-regulation of Notch signalling along lens differentiation trajectory. Considering that predominant changes could exhibit at the tissue level, we performed qPCR to confirm the specific molecules that have the greatest potential to get involved in the LEC differentiation process of highly myopic eyes. While the mRNA levels of Psenen, Psen2, Aph1b, Aph1c, and Dll1 exhibited no significant differences ( Figure S4), mRNA and protein levels of NOTCH2 were both remarkably down-regulated in lens epithelium dissected from both human and mouse highly myopic eyes compared to that of control eyes ( Figure 5A,B). This result was consistent with differential gene expression analysis that Notch2 was the top down-regulated gene identified in highly myopic lens epithelium compared with control eyes (log 2 FC = À0.26, adjusted p value = 2.75 Â 10 À296 ) ( Figure S5A).
On the other hand, as was shown in Figure 2C, the proportion of cluster 2 was significantly lower in highly myopic eyes, indicating its vulnerability to high myopia. Notch2 was significantly activated in cluster 2 compared with other LEC clusters ( Figure 5C). Therefore, immunohistochemistry was performed to confirm the distribution of these NOTCH2-expressing cells. As was shown in Figure 5D, LFCs. 26 Of note, consistent with the significant reduction in the cluster 2 proportion, NOTCH2-expressing cells were significantly reduced in highly myopic lens epithelium ( Figure 5D). The expression level of NOTCH2 signalling was also remarkably down-regulated in these cells of highly myopic eyes. These results indicate that NOTCH2 signalling inhibition may accelerate the process towards differentiation of these NOTCH2-expressing cells, therefore participating in promoting fibre cell differentiation in highly myopic eyes.
Next, we assessed the biological role of Notch2 on fibre cell differentiation. As the immortalized cell line lacked the ability to differentiate, human and mouse primary lens epithelial cells were used. We F I G U R E 5 Notch2 signalling inhibition promotes differentiation towards lens fibre cells in highly myopic lens epithelium. (A) Down-regulated protein level of NOTCH2 in human highly myopic lens epithelium. (B) Down-regulated protein level of NOTCH2 in mouse highly myopic lens epithelium. (C) Volcano plot displaying differentially expressed genes detected between cluster 2 and any other LEC cluster. Notch2 was significantly activated in cluster 2 compared with other LEC clusters (Log 2 FC = 0.42, adjusted p value = 1.29 Â 10 À9 ). (D) Immunohistochemistry for NOTCH2 showing the distribution of NOTCH2-expressing cells in the transitional zone (the area between two dotted lines) around the lens equator and a significant reduction in the number of NOTCH2-expressing cells in highly myopic lens. (E) Western blotting analysis of downstream molecules in primary cultured human lens epithelial cells in response to NOTCH2 knockdown with siRNA. (F) Western blotting analysis of downstream molecules in primary cultured mouse lens epithelial cells in response to Notch2 knockdown with siRNA. Following NOTCH2 knockdown, its effector HES1 was significantly down-regulated, while MAF, CDKN1C and fibre cell markers (CRYBB1 and CRYG) were significantly up-regulated in both human and mouse primary lens epithelial cells. Data are expressed as mean ± SEM. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001; Student's t-test.
found that NOTCH2 inhibition by siRNA increased the mRNA levels ( Figure S5) and protein levels ( Figure 5E,F) of fibre cell markers in both human and mouse primary LECs, indicating that NOTCH2 could inhibit secondary lens fibre differentiation. Furthermore, the downstream NOTCH effector HES1 was simultaneously down-regulated following NOTCH2 inhibition ( Figure 5E,F), while other known downstream molecules including Hes5, Hey1, and Hey2 were not significantly altered ( Figure S5). Meanwhile, the transcription factor MAF and cyclindependent kinase inhibitor CDKN1C was significantly up-regulated ( Figure 5D,E, Figure S5). These results were further confirmed by repeated transfection experiments using a different siRNA molecule targeting a different region of NOTCH2 ( Figure S5).

| DISCUSSION
High myopia is a blinding eye disease with high prevalence across the globe. It was estimated to affect nearly 1 billion people by 2050. 1 The extensive elongation of eyeballs dramatically increases the risk of a variety of ocular pathologies, including lens diseases such as earlyonset cataract and lens dislocation, 27,28 resulting in poor visual prognosis. Our previous studies further demonstrated a pathologically increased lens size in highly myopic eyes which may account for their higher IOL malposition incidence. 2 Here, using a sc-RNA seq strategy, we uncovered the gene-expression changes associated with the increased LECs differentiation towards fibre cells in highly myopic lens epithelium. Our data also demonstrated that the NOTCH2 signalling pathway plays an essential role in LEC fate determination and lens size control even in adulthood, and its inhibition participates in increased fibre cell differentiation in highly myopic eyes.
We previously verified the pathological growth of lens in both human and mouse highly myopic eyes using magnetic resonance imaging (MRI). 2  together make up about 90% of lens structural proteins, 32 and these two families account for more than 77% of all crystallins. 2,32 They are expressed specifically in LFCs, and therefore serve as LFC molecular markers. 21,31 This is consistent with a previous bulk-RNA sequencing result showing up-regulation of β/γ-crystallins in highly myopic lens epithelium, 2 and has been confirmed through our validation experiments. The accumulation of differentiated fibre cells therefore provides a structural foundation for increased lens size in highly myopic eyes.
In the current study, we utilized an unsupervised method to outline the pseudotime trajectory of LEC differentiation and visualize the dynamic alternations of genes of interest. We found that the NOTCH signalling pathway was significantly inhibited during differentiation towards secondary fibre cells. Particularly, NOTCH2 rather than other NOTCH receptor subtypes exhibited predominant down-regulation.
Moreover, we found that the proportion of the second largest LEC In this study, we revealed a significantly decreased expression of NOTCH2 in the lens of human and mouse highly myopic eyes, confirming the results identified from sc-RNA seq analysis. To validate the specific role of NOTCH2, it was inhibited using NOTCH2 siRNA, following which its effector molecule HES1 was down-regulated, while MAF, CDKN1C and fibre cell markers were up-regulated. These experiment results indicated that NOTCH2 affects LEC differentiation, and MAF and CDKN1C are possible mediators. MAF is a transcription factor which has been reported to control embryonic lens development. 42 It remains activated after birth and could regulate postnatal lens growth by directly interacting with promoters of β/γcrystallin genes and activating their expressions. 2 Similar to our findings, Eom identified increased expression of MAF in the mouse pancreas resulting from Notch downregulation. 43 CDKN1C is a cell cycle-dependent kinase inhibitor that promotes withdrawal from cell cycle and initiating differentiation. 44 Previous investigations reported CDKN1C activation in the lens of Notch2 or Rbpj CKO mice. 7,45 Meanwhile, CDKN1C have been shown to be directly regulated by Notch2 downstream transcriptional factor HES1 in pancreas and intestine. 45,46 Taken together, we suppose that in high myopia, NOTCH2 signalling inhibition could accelerate secondary lens fibre cell differentiation by activating CDKN1C, the differentiation initiator, and promote accumulation of β/γ crystallin, the specific structural proteins of LFCs, by activating MAF (Figure 6). These mechanisms collectively contribute to the overgrowth of lens in highly myopic eyes. As the expression level of HES1 decreases following NOTCH2 knockdown, it was speculated that its transcriptional repression on MAF and CDKN1C was relieved. Importantly, using primary LECs cultured from adult human lens epithelium, our results indicated that NOTCH2 could affect LEC differentiation and lens structure even in grown-ups. In the future, with persistently increasing evidence, the functions of cross-talk between these pathways in high myopia associated lens diseases worth further investigation.
In conclusion, our study uncovered the transcriptome atlas and lineage differentiation trajectory of LECs, and revealed the gene expression signatures of LEC subpopulations. We validated the role of NOTCH2 signalling inhibition underlying the increased differentiation towards fibre cells in highly myopic eyes, providing a novel perspective and potential therapeutic target for lens overgrowth in highly myopic eyes. F I G U R E 6 Schematic illustration of NOTCH2 inhibition underlying the increased differentiation towards lens fibre cell in highly myopic lens epithelium. In highly myopic eyes, NOTCH2 signalling inhibition may relieve the transcriptional repression of HES1 on MAF and CDKN1C. MAF activation facilitates accumulation of β/γ crystallins which are specific structural proteins of lens fibre cells, while CDKN1C activation initiates differentiation. They together accelerate secondary lens fibre cell differentiation in highly myopic eyes. This figure was created by Figdraw.