Prediction of myopia-associated ECM genes

To predict myopia-associated ECM genes, previously reported genes were identified through a literature review and the Consortium of Refractive Error and Myopia (CREAM) report, serving as guide genes (Haarman et al. 2021). These guide genes were used to search for novel candidate genes by leveraging the direct linkages of HumanNet V3, an integrated network of human genes for disease studies based on supervised machine learning techniques (Kim et al. 2022). For the inference algorithm model, we selected HumanNet-XC, which incorporates advanced models such as co-functional links, co-citation, co-expression, pathway database, domain profile association, genetic interaction, gene neighborhood, phylogenetic profile association, and protein-protein interaction (PPI) network predictions. HumanNet-XC has connected 18,462 genes with 1,125,494 connections, enabling researchers to gain a comprehensive understanding of complex biological processes and identify potential disease-related genes. To assess the discriminative power of gene prediction, we performed Area under the Receiver Operating Characteristic Curve (AUROC) analysis which was based on HumanNet database and exported the top 5% ranked potential genes for further analysis. AUROC represents the prediction power of candidate genes in top ranks.

Functional enrichment analysis and validation of hub genes

To validate the association of the gene list with myopic scleral matrix remodeling again, we performed Reactome pathway and functional enrichment analyses using g: Profiler and visualization tools based on the R platform’s clusterProfiler package (Raudvere et al. 2019; Li et al. 2022). The gene list obtained from the inference algorithm was imported into the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING), a database that combines known and predicted protein-protein interactions (Szklarczyk et al. 2021). STRING incorporates direct and indirect associations derived from computational prediction, knowledge transfer between organisms, and aggregation from other primary databases. We exported the PPI network as a TSV document and reconstructed it using Cytoscape software (Kohl et al. 2011). To rank the nodes in the predicted myopia ECM network, we utilized the cytoHubba plugin in Cytoscape, which combines multiple algorithms to determine the importance of nodes in the overall network (Chin et al. 2014). Based on previous literature reports, we employed the Maximal Clique Centrality (MCC) algorithm, which has demonstrated better accuracy performance among the 11 algorithms available in cytoHubba, to sort the potential hub genes (Chin et al. 2014). In this network, nodes with higher MCC algorithm scores have higher centrality. The centrality of each node was represented by shades of color, with darker colors indicating higher values.

Animal administration

Male C57BL6J mice were housed in a controlled environment consisting of standard transparent cages maintained at a temperature of 24 ± 2 °C and a humidity range of 40–60%. The mice were kept in a clean room operating on a 12-hour light-dark cycle. During the entire experimental period, the mice were provided ad libitum access to a standard rodent diet and water.

Ethical approval for all animal experiments conducted in this study was obtained from the Animal Experimental Committee of Keio University (A2022-242). The study adhered to the Institutional Guidelines on Animal Experimentation at Keio University, the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research, and the Animal Research: Reporting of In Vivo Experiments (ARRIVE) guidelines for the use of animals in research.

Establishment of LIM and measurement of ocular biometric characteristics in mice

According to previous reports by our laboratory, monocular LIM was induced in mice (Jiang et al. 2018). The left side of the glasses used in this study was affixed with a 0D lens as an internal control, and the right side of the glasses was affixed with a -30D lens (Fig. 1).

Fig. 1

Construction of lens-induced myopia (LIM) mice models. Monocular LIM was induced by -30D lens attached in the right eyes while 0D lens with left eyes as the internal control. The time course of LIM started from three-week-old mice and lasted 2 days, 5 days, 8 days and 24 days

The glasses were removed and washed at least twice a week. Ocular biometric characteristics, including the refractive state measured using an infrared photorefractor (Steinbeis Transfer Center) and axial length (AL) and choroid thickness analyzed using an SD-OCT system (Envisu R4310, Leica) designed for mice, were measured according to previous reports.

In this study, a time-course LIM operation was performed on male wildtype C57BL/6J mice at 3 weeks of age for durations of 2 days, 5 days, 8 days, and 24 days. 3-week-old male C57BL/6 mice (n = 20) were randomly divided into 4 groups of different durations of LIM.

AAV-based CRISPR/Cas9 system for THBS1 disruption in mouse sclera

SaCas9 or guide RNA (Supplementary Table T2) against THBS1 expression cassettes were constructed and packaged into AAV-DJ vectors (titer > 1.0E + 13 GC/ml) by Vector Builder as described in previous reports (Ikeda et al. 2022; Zhao et al. 2018). The mice were randomly assigned to three groups: a sham-operation group (anaesthetized and had a ‘dry’ needle inserted with no injection), a group injected with scramble guide RNA, and a group injected with gThbs1. A 1:1 mixture of AAV-packaged SaCas9 and guide RNA was injected into mice following anesthesia and the application of Scopisol to prevent reflux. Subtenon’s injections were administered at two locations around each eye, spaced 180 degrees apart (while avoiding blood vessels) as previous reported (Supplementary Figure S1) (Ikeda et al. 2022; Zhao et al. 2018). A 33 G needle was used for the injections, and the eyes were covered with 0.2% polyacrylic acid for the 3 days following the injection.

To evaluate the efficacy of subtenon’s AAV injection for gene transfer into scleral tissue, the expression of green fluorescent protein (GFP) in sclera flat mounts was checked 28 days after AAV-DJ-GFP or AAV-DJ-Vector injection using a Keyence BZ-800 fluorescence microscope.

Western blotting

After administering anesthesia, mice were euthanized, and two sclera samples were extracted from each mouse. Subsequently, the two sclera samples obtained from a single mouse were pooled to create a unified sample. Each lane of western blot corresponded to a different independent sample. Sclera samples were homogenized in RIPA buffer supplemented with Halt protease inhibitor cocktail (ThermoFisher Scientific, USA). Protein concentration was determined using the BCA protein assay, and samples were adjusted with Laemmli sample buffer (Nacalai Tesque, Japan). The protein extracts were separated by SDS-PAGE, transferred to PVDF membranes (Merck Millipore, MA, USA), and blocked with Blocking One (Nacalai Tesque, Japan). Membranes were then incubated overnight at 4 °C with specific antibodies, followed by incubation with corresponding secondary antibodies. Visualization was performed using SuperSignal West Femto Maximum Substrate (Thermo Fisher Scientific, USA). SDS-PAGE was conducted on 10% acrylamide gels with protein size markers (MagicMark XP Western Protein Standard, Thermo Fisher Scientific, USA). The primary and secondary antibodies used for western blot were listed as follows: THBS1 (1:1000 dilution, Invitrogen, #39-9300), COL1A1 (1:1000 dilution, Cell Signaling Technologies, #84,336), LRP1 (1:1000 dilution, Cell Signaling Technologies, #64,099), TGFB1 (1:1000 dilution, Abcam, ab179695), MMP9 (1:1000 dilution, Abcam, ab228402), α-SMA (1:100 dilution, Invitrogen, #14-9760-82), and β-ACTIN (1:5000 dilution, Cell Signaling Technologies, #3700).

Statistical analysis

To determine the statistical significance of comparisons, independent sample t-tests (Fig. 4) and analysis of variance (ANOVA) with Tukey post hoc test (Figs. 5 and 6) were performed individually using GraphPad Prism 9. Figures 4, 5 and 6 are average of experiments and Fig. 7 is representative or of experiments. Choroidal thickness and histogram analysis of Western blot were conducted using Image J (version 1.53t; NIH). A significance level of P-value < 0.05 was considered statistically significant.