Abstract

   Trinucleotide repeat (TNR) expansion in the transcription factor 4
   (TCF4) gene represents the most prevalent genetic risk factor for Fuchs
   endothelial corneal dystrophy (FECD) and may cause dysfunction of
   splicing regulators. We investigated differential alternative splicing
   (DAS) events in corneal endothelial cells (CECs) from FECD patients
   with and without TCF4 TNR expansion through RNA-Seq analysis. We
   identified distinct splicing profiles among control subjects, FECD
   patients with TNR expansion, and FECD patients without TNR expansion.
   Skipped Exon events constituted approximately 50% of all DAS events
   across all comparisons, with the remaining events distributed among
   alternative 3’ splice site, alternative 5’ splice site, mutually
   exclusive exon, and retained intron categories. Motif analysis in FECD
   patients with TNR expansion revealed several RNA-binding proteins,
   including MBNL1, as potential regulators of these splicing alterations.
   Computational analysis demonstrated that 34% of Skipped Exon events in
   the TNR expansion group significantly impacted protein structure. This
   comprehensive analysis revealed distinct alternative splicing
   signatures in FECD, particularly in cases with TNR expansion,
   suggesting a crucial role for aberrant splicing in FECD pathogenesis.

   Subject terms: RNA splicing, Genetics research, Corneal diseases

Introduction

   Fuchs endothelial corneal dystrophy (FECD) is a progressive, hereditary
   eye disease that affects the corneal endothelium and leads to vision
   impairment and potential blindness^[40]1–[41]3. It affects
   approximately 4–11% of individuals and represents the leading
   indication for corneal transplantation^[42]4. Epidemiological studies
   have revealed that the clinical manifestation of FECD varies
   considerably, ranging from individuals who may remain asymptomatic
   throughout their lives to those who experience severe vision impairment
   requiring one or repeat transplants^[43]5,[44]6. The pathogenesis of
   FECD is complex, involving both genetic and environmental
   factors^[45]1–[46]3.

   One of the most common genetic factors associated with FECD is the
   expansion of trinucleotide repeats (TNR) in the transcription factor 4
   (TCF4) gene^[47]2,[48]7. This expansion is present in approximately
   40–80% of FECD patients, making it the most prevalent genetic
   background for the disease^[49]2. The presence of TNR expansion in TCF4
   has led researchers to propose various hypotheses regarding its role in
   FECD pathogenesis. For example, the presence of intronic TNR expansion
   in FECD is reminiscent of myotonic dystrophies, suggesting potential
   mechanistic parallels between these disorders. In myotonic dystrophies,
   the expanded TNR sequence affects multiple RNA-binding proteins (RBPs)
   through distinct mechanisms. While the mutant RNA primarily sequesters
   muscleblind-like (MBNL) proteins, it also impacts other splicing
   regulators through different pathways, including CUG-binding protein
   elav-like family member 1 (CELF1; also referred to as CUGBP1),
   Staufen1, and heterogeneous nuclear ribonucleoprotein H (hnRNP
   H)^[50]8,[51]9. The dysregulation of these splicing factors,
   particularly the sequestration of MBNL proteins, leads to widespread
   mis-splicing of target genes^[52]10,[53]11. Studies utilizing a mouse
   model lacking functional RNA-binding activity of MBNL1 protein have
   demonstrated alterations in hundreds of splicing events in muscle,
   heart, and brain tissues^[54]12. These findings underscore the
   extensive involvement of MBNL in the regulation of a wide array of
   splicing events and they support the critical role of MBNL in myotonic
   dystrophy^[55]13.

   Similar molecular mechanisms have been proposed for FECD, in which the
   expanded CTG repeat in TCF4 may likewise sequester MBNL proteins,
   potentially contributing to disease pathogenesis through a disruption
   of RNA processing^[56]14,[57]15. Wieben and colleagues identified
   differential splicing in the NUMA1, PPFIBP1, MBNL1, MBNL2, VEGFA, and
   FGFR2 transcripts in corneal endothelial cells (CECs) of patients with
   FECD harboring TNR expansion in TCF4, although larger numbers of
   potentially differential splicing events were also found^[58]16. Chu
   and colleagues also reported more than one thousand splicing event
   alterations in FECD with TNR expansion when compared with non-FECD
   control susbjects^[59]17. However, the genetic heterogeneity, varying
   disease stages, and racial diversity in FECD makes the accumulation of
   further evidence through validation studies crucial for obtaining a
   better understanding of the splicing events, as these may vary in
   healthy individuals as well as in the disease state.

   The motivation for this study was to provide additional evidence that
   would elucidate the comprehensive landscape of altered splicing events
   in FECD, with the overall aim of offering new insights into the disease
   mechanism. To address the current knowledge gap, we analyzed splicing
   events in FECD using RNA-Seq data obtained from CECs of affected
   patients. Our study was designed to identify MBNL-regulated splicing
   events but also to uncover other splicing alterations that may
   contribute to FECD pathogenesis.

Results

Aberrant splicing in MBNL1 and MBNL2

   We analyzed alternative splicing events among three groups: 6 FECD
   patients with TNR expansion (Expansion group), 4 FECD patients without
   TNR expansion (No Expansion group), and 7 non-FECD controls (Control
   group). We first investigated the splicing patterns of MBNL1 and MBNL2
   genes, as previous studies have demonstrated their splicing
   dysregulation in FECD patients carrying TNR expansions in TCF4^16,18.
   MBNL proteins are known to regulate alternative splicing of their own
   pre-mRNAs. In both myotonic dystrophy and FECD with TNR expansions,
   altered MBNL function leads to increased inclusion of MBNL1 exon 6,
   MBNL2 exon 6, and MBNL2 exon 9 compared to controls, suggesting a
   disruption in the autoregulatory splicing
   mechanism^[60]13,[61]16,[62]18. This enabled corroboration of these
   findings and validated the consistency for our RNA-Seq data. The gene
   sequences of MBNL1 and MBNL2 were examined with a particular focus on
   regions where splicing alterations had been previously identified
   (Fig. [63]1A). In MBNL1, exon 6 demonstrated increased inclusion in the
   Expansion group compared to both the Control and No Expansion groups
   (Fig. [64]1B). A similar pattern was observed for exon 6 of MBNL2, with
   elevated inclusion in the Expansion group relative to both Control and
   No Expansion groups (Fig. [65]1C). For exon 9 of MBNL2, increased
   inclusion was observed in the Expansion group compared to the No
   Expansion group, although the inclusion pattern was comparable between
   the Expansion and Control groups (Fig. [66]1D).

Fig. 1.

   [67]Fig. 1
   [68]Open in a new tab

   Alternative splicing events in MBNL1 and MBNL2 genes in Fuchs
   endothelial corneal dystrophy (FECD). (A) Schematic representation of
   the MBNL1 and MBNL2 gene structures. Previously reported alternatively
   spliced regions in FECD corneal endothelial cells are highlighted in
   purple, blue, and red. Gray regions indicate untranslated regions.
   Arrows indicate the positions of the RT-PCR primer pairs used for
   validation, summarized in (Supplementary Table [69]2). (B–G) Sashimi
   plots show increased inclusion of MBNL1 exon 6 (B), MBNL2 exon 6 (C),
   and MBNL2 exon 9 (D) among the three groups. Representative gel images
   show RT-PCR validation results of splicing alterations in corneal
   endothelial cells (CECs) for increasing exon inclusion in the Expansion
   group compared to the Control and No Expansion group for (E) MBNL1 exon
   6, (F) MBNL2 exon 6, and (G) MBNL2 exon 9.

   To confirm these aberrant splicing events, we further validated the
   splicing alterations using RT-PCR. This analysis was performed to
   verify the splicing alterations observed in MBNL1 exon 6 (Fig. [70]1E),
   MBNL2 exon 6 (Fig. [71]1F), and MBNL2 exon 9 (Fig. [72]1G). The results
   from the RT-PCR experiments corroborated the findings from the Sashimi
   plots, demonstrating consistent splicing alterations in the Expansion
   group across these exons. Original gels are presented in (Supplementary
   Fig. [73]1).

Identification of differential alternative splicing (DAS) events

   We subsequently evaluated DAS events occurring across all genes
   detected by RNA-Seq. The principal component analysis (PCA) revealed
   distinct clustering patterns in splicing profiles among the control, no
   expansion, and expansion groups. Each cohort formed discrete clusters,
   indicating substantial differences in alternative splicing patterns
   (Fig. [74]2A). Hierarchical clustering analysis corroborated these
   findings, demonstrating distinct patterns of splicing events across the
   groups (Fig. [75]2B).

Fig. 2.

   [76]Fig. 2
   [77]Open in a new tab

   Global analysis of alternative splicing patterns in Fuchs endothelial
   corneal dystrophy (FECD). (A) Principal component analysis (PCA) of
   alternative splicing events across the study groups. Control (black),
   No Expansion (green), and Expansion (yellow) samples are plotted in PC1
   versus PC2 space. The percentage of total variance explained by each
   principal component is indicated on the respective axes. (B)
   Hierarchical clustering heatmap depicting relative inclusion levels of
   differentially alternatively spliced (DAS) events across Control
   (black), No Expansion (green), and Expansion (yellow) groups,
   demonstrating distinct splicing signatures between FECD patients (No
   Expansion and Expansion groups) and non-FECD controls.

   Comparison of the expansion and control groups for all genes in RNA-Seq
   data yielded 1,816 DAS events among five patterns (Skipped Exon (SE):
   974, Alternative 3’ Splice Site (A3SS): 161, Alternative 5’ Splice Site
   (A5SS): 155, Mutually Exclusive Exons (MXE): 200, and Retained Intron
   (RI): 326). The Expansion versus No Expansion group comparison
   identified 809 DAS events (SE: 404, A3SS: 84, A5SS: 91, MXE: 74, RI:
   156). Lastly, the No Expansion versus Control group comparison revealed
   2,299 DAS events (SE: 1,131, A3SS: 202, A5SS: 171, MXE: 239, RI: 556)
   (Fig. [78]3A). The distribution of event types for the Expansion versus
   Control comparison was as follows: SE: 53.6%, A3SS: 8.9%, A5SS: 8.5%,
   MXE: 11.0%, and RI: 18.0% (Fig. [79]3B). For the Expansion versus No
   Expansion comparison, the distribution was SE: 49.9%, A3SS: 10.4%,
   A5SS: 11.3%, MXE: 9.2%, and RI: 19.3% (Fig. [80]3C). Similarly, for the
   No Expansion versus Control comparison, events were distributed as SE:
   49.2%, A3SS: 8.8%, A5SS: 7.4%, MXE: 10.4%, and RI: 24.2% (Fig. [81]3D).
   SE was consistently the most frequent splicing event type across all
   group comparisons.

Fig. 3.

   [82]Fig. 3
   [83]Open in a new tab

   Analysis of differential alternative splicing (DAS) events in Fuchs
   endothelial corneal dystrophy (FECD). (A) Schematic diagram of five
   splicing patterns of an alternative splicing event. Distribution of
   significant DAS events and affected genes across five splicing
   patterns: skipped exon (SE), alternative 3’ splice site (A3SS),
   alternative 5’ splice site (A5SS), mutually exclusive exon (MXE), and
   retained intron (RI). Significance criteria: FDR < 0.05 and
   |IncLevelDifference| > 0.2. (B–D) Pie charts showing the relative
   proportions of five types of differential splicing patterns (SE, A3SS,
   A5SS, MXE, and RI) identified between Expansion versus Control groups
   (B), Expansion versus No Expansion groups (C), and No Expansion versus
   Control groups (D). The percentages represent the proportion of each
   splicing pattern type among all DAS events detected in each comparison.

   To determine whether DAS events were specifically induced by TNR
   expansion in TCF4 or by other factors such as unknown genetic
   backgrounds in the No Expansion group or secondary changes associated
   with FECD development, we compared DAS events identified from two
   distinct analyses: Expansion versus Control and No Expansion versus
   Control. Venn diagrams revealed overlapping events as follows: SE: 383,
   A3SS: 51, A5SS: 39, MXE: 41, and RI: 124. Although these events
   overlapped between the Expansion and No Expansion groups, higher
   numbers of unique DAS events were found in both the Expansion versus
   Control and No Expansion versus Control comparisons (Fig. [84]4).

Fig. 4.

   [85]Fig. 4
   [86]Open in a new tab

   Comparison of alternative splicing events between Fuchs endothelial
   corneal dystrophy (FECD) groups. Venn diagrams comparing differential
   alternative splicing (DAS) events identified between two distinct
   analyses: Expansion versus Control and No Expansion versus Control. The
   blue circles represent DAS events specific to the expansion versus
   control comparison, while purple circles indicate DAS events specific
   to the No Expansion versus Control comparison. Overlapping regions show
   shared DAS events between both comparisons. Separate diagrams are shown
   for each splicing pattern: (A) skipped exon (SE), (B) alternative 3’
   splice site (A3SS), (C) alternative 5’ splice site (A5SS), (D) mutually
   exclusive exon (MXE), and (E) retained intron (RI). All DAS events met
   significance criteria of FDR < 0.05 and |IncLevelDifference| > 0.2.

Identification of RBPs

   We explored the regulatory mechanisms underlying the differential
   splicing patterns by performing motif analyses on SE events that
   differed between the Expansion and No Expansion groups. Through rMATS
   analysis, we identified 404 significant differential SE events between
   the Expansion and No Expansion groups (Fig. [87]3A). Of these, 184
   events demonstrated increased exon inclusion, while 220 events showed
   increased exon skipping in the Expansion group compared to the No
   Expansion group. To better understand the regulatory mechanisms
   underlying these splicing changes, we performed motif enrichment
   analysis to identify potential RBPs associated with each splicing
   pattern separately. Among the 404 total differential SE events
   identified, 184 events showed increased inclusion while 220 events
   exhibited increased skipping in the Expansion group compared to the No
   Expansion group. Analysis of the 184 inclusion events revealed
   enrichment of binding motifs for several RBPs, including HuR, PTB,
   SFPQ, SNRPA, and YBX2, as illustrated in the heatmap (Fig. [88]5A). For
   the 220 skipping events, we identified potential binding sites for 11
   distinct RBPs, with MBNL1 emerging as a key regulatory factor in these
   events (Fig. [89]5B).

Fig. 5.

   [90]Fig. 5
   [91]Open in a new tab

   Motif enrichment analysis of RNA-binding proteins (RBPs) in Skipped
   Exon (SE) patterns between expansion and no expansion groups. The
   heatmaps show the motif enrichment analysis for RBPs potentially
   binding near the introns and exons of genes with altered splicing
   patterns. The left side of each heatmap lists RBPs, with their
   corresponding binding motifs shown beside them. The color intensity of
   the heatmap indicates the statistical significance (P-value), with
   darker shades representing more significant enrichment. The schematic
   diagram depicts two types of SE events: (A) exons that undergo
   increased skipping and (B) exons that show increased inclusion. For
   each event type, we analyzed the following regions: 50 bp from both
   ends of the differentially regulated exon, 50 bp from the edges of
   adjacent upstream and downstream exons, and 250 bp from both ends of
   the flanking introns. The analysis focuses on exons included and
   skipped in the Expansion group compared to the No Expansion group, both
   within the context of SE patterns.

Functional enrichment of splicing variants

   A total of 404 SE events showing splicing alterations between the No
   Expansion and Expansion groups were analyzed to determine their
   potential impact on the resulting proteins. Overall, 34% of the SE
   events affected the resulting proteins, while the remaining 66% had no
   impact on the protein. Among these 34% events, 85% were located in
   protein domains, 10% affected residues, and 5% influenced linear motifs
   (Fig. [92]6A). Enrichment analysis was performed on the genes harboring
   the 34% of exons that were predicted to impact the proteins. Reactome
   pathway analysis revealed that these exons were associated with
   pathways such as the immune system, adaptive immune system, signaling
   by receptor tyrosine kinases, and FGFR1 ligand binding and activation
   (Fig. [93]6B). KEGG pathway analysis indicated associations with
   pathways including the MAPK signaling pathway, PI3K-Akt signaling
   pathway, Ras signaling pathway, focal adhesion, and apoptosis
   (Fig. [94]6C).

Fig. 6.

   [95]Fig. 6
   [96]Open in a new tab

   Functional and pathway analysis of skipped exon (SE) patterns altered
   between the expansion and no expansion groups. (A) The pie chart shows
   the proportion of exons, identified through SE patterns, that affect
   protein structure or function when splicing changes occur. The bar
   graph further categorizes these exons into domains, residues, and
   linear motifs, representing their relative proportions. (B) Reactome
   pathway analysis of genes with SE events that were altered in the
   Expansion group compared to the no expansion group and affect protein
   structure or function. (C) KEGG pathway analysis for the same set of
   genes. In both the Reactome and KEGG pathway analyses, the top 12
   pathways with the highest NEASE scores are displayed. The color
   intensity represents the statistical significance (adjusted P-value).

Discussion

   This study examined splicing events in the CECs of patients with FECD
   and revealed distinct splicing patterns among Expansion, No Expansion,
   and Control groups. As anticipated, expansion patterns diverged from
   both control and No Expansion groups, although the No Expansion group
   also unexpectedly exhibited splicing patterns that differed from the
   Control group. Among the alternative splicing events observed in the
   Expansion group, approximately half were Skipped Exons compared to the
   No Expansion and Control groups. Among 184 events involving exon
   skipping in which typically included exons were skipped in the
   Expansion group, 11 splicing regulatory factors influencing these
   skipped exons, including MBNL1, were identified. Current research
   supports the hypothesis that TNR expansion-induced MBNL dysregulation
   leads to splicing abnormalities in multiple genes, potentially
   contributing to corneal endothelial dysfunction.

   Alternative splicing is a fundamental mechanism that generates
   proteomic diversity from a limited genome, thereby enabling the
   generation of multiple protein isoforms from a single gene. This
   process fosters cellular heterogeneity and tissue-specific functions,
   allowing precise responses to developmental and environmental cues. Its
   evolutionary conservation underscores its importance in biological
   adaptation and resilience^[97]19–[98]21. However, while alternative
   splicing plays a crucial role in generating biological diversity,
   aberrations in this process can lead to pathological conditions^[99]22.

   Splicing errors can result in the loss of protein function, production
   of deleterious protein isoforms, or disruption of cellular
   homeostasis^[100]23. These abnormalities have been implicated in a wide
   range of diseases, including neurodegenerative disorders, cancers, and
   genetic syndromes^[101]24–[102]26. These types of aberrations may arise
   from mutations in splice sites, dysregulation of splicing factors, or
   dysfunction of the spliceosome machinery^[103]27,[104]28. The
   significance of alternative splicing in both normal physiology and
   pathological states highlights its importance as a subject of ongoing
   research in molecular biology and medicine. Understanding the
   mechanisms and consequences of splicing alterations may provide
   valuable insights into disease pathogenesis and potential therapeutic
   strategies.

   In FECD, splicing abnormalities have become a significant focus in
   research into its pathogenesis. The discovery of CTG trinucleotide
   repeat expansions in the TCF4 gene as a major genetic mutation led to
   investigations into RNA-mediated toxicity mechanisms. Studies have
   revealed that expanded CUG-containing transcripts form nuclear RNA foci
   sequestering splicing factors, particularly MBNL
   proteins^[105]2,[106]14,[107]18. RNA-Seq analyses have demonstrated
   widespread aberrant splicing events correlated with CTG18.1 expansion
   status^[108]16,[109]17. Our study provides additional evidence for
   aberrant splicing events in numerous genes with multiple alternative
   splicing patterns in patients with TCF4 TNR expansion. Analysis of exon
   skipping regulators revealed 11 genes, including MBNL1, aligning with
   the hypothesis that MBNL1 impairment, induced by TNR expansion,
   precipitates aberrant splicing. Furthermore, these findings suggest a
   potential involvement of other splicing modulators in patients with
   TCF4 repeat expansions. Further research is needed to elucidate which
   splicing factors induce pathological aberrant splicing events that
   result in FECD.

   The observed differential splicing patterns between the No Expansion
   and Control groups, while unexpected, warrant careful interpretation
   within the context of disease pathophysiology. Our analysis revealed
   that the number of DAS events was paradoxically higher in the No
   Expansion versus Control comparison than in the Expansion versus
   Control comparison. However, this quantitative difference should be
   interpreted cautiously, as aberrant splicing events can function as
   both etiological factors and downstream effects of pathological
   processes^[110]29. We propose that the distinct splicing patterns
   observed in these groups may reflect different underlying molecular
   mechanisms. In the Expansion group, splicing alterations likely arise
   from a combination of MBNL1 sequestration by expanded TCF4 TNR and
   secondary effects of disease progression. Conversely, in the No
   Expansion group, the splicing changes may primarily result from disease
   progression effects, potentially in conjunction with yet unidentified
   genetic factors (Fig. [111]7). The toxic effects of TCF4 TNR
   expansion-mediated alterations in MBNL-dependent splicing are likely
   central to the pathogenesis in the Expansion group. Therefore, future
   research should focus not only on the quantity of DAS events but also
   on identifying and characterizing specifically toxic splicing
   alterations. This distinction is crucial for understanding the
   pathogenic mechanisms and developing targeted therapeutic strategies.
   This hypothesis highlights the complex interplay between genetic
   predisposition, molecular mechanisms, and disease progression in the
   pathogenesis of this condition. Further studies are needed to elucidate
   the precise mechanisms underlying these differential splicing patterns
   and their respective contributions to disease etiology and progression.

Fig. 7.

   [112]Fig. 7
   [113]Open in a new tab

   Proposed model of splicing aberrations based on TCF4 Trinucleotide
   repeat (TNR) expansion status. Schematic representation of splicing
   aberration mechanisms in Fuchs endothelial corneal dystrophy (FECD)
   based on the TCF4 TNR expansion status. In FECD patients with TNR
   expansion, disease progression is initiated by MBNL sequestration,
   which leads to widespread aberrant splicing patterns. By contrast, FECD
   patients without TNR expansion exhibit splicing alterations that likely
   occur as a secondary response to disease progression triggered by
   currently unknown pathogenic mechanisms.

   Mis-spliced TCF4 transcripts were identified in patients harboring TNR
   expansion, potentially leading to pathogenic TCF4 isoforms or altered
   expression levels^[114]30–[115]32. While dysregulated TCF4 expression
   has been proposed as a pathological mechanism, some studies have
   reported inconsistent results, likely due to qPCR limitations given
   numerous isoforms of TCF4. Indeed, RNA-Seq analyses revealed
   differential expression of multiple TCF4 isoforms in FECD patients with
   TNR expansions, suggesting TNR contribution to dysregulated TCF4
   expression^[116]30,[117]32,[118]33. Sirp et al. demonstrated
   downregulation of TCF4 isoforms from alternative 5’ exons near TNR,
   while other isoforms were upregulated^[119]30. Long-read RNA-Seq
   confirmed lower expression of TCF4 isoforms downstream of TNR in FECD
   patients^[120]34. We previously identified TCF4-277 as a commonly
   dysregulated isoform in FECD with TNR expansion^[121]32. However,
   whether MBNL1 abnormalities cause TCF4 splicing defects in FECD remains
   unclear, and further investigation is anticipated to clarify this
   question.

   A limitation of this study is the uncertainty regarding the extent to
   which MBNL1 sequestration due to TNR expansion contributes to the
   numerous splicing abnormalities. Furthermore, the potential involvement
   of other splicing factors cannot be excluded. One noteworthy point is
   that diverse splicing events also occur in the healthy corneal
   endothelium. The fact that splicing aberrations can be both etiological
   factors and consequences of disease processes complicates the isolation
   and identification of specific splicing abnormalities induced by TNR
   expansion that directly contribute to FECD pathogenesis. However, one
   pertinent observation is that in other fields, such as oncology,
   several therapeutic agents targeting splicing abnormalities have been
   developed and are currently undergoing clinical trials^[122]35,[123]36.
   This precedent suggests that targeting splicing aberrations in FECD
   could potentially offer a novel therapeutic approach. Therefore,
   further investigation into splicing mechanisms in FECD is warranted, as
   it may lead to the development of innovative treatment strategies.

Methods

Ethics statement

   This investigation was conducted in accordance with the ethical
   standards outlined in the Declaration of Helsinki. The study protocol
   received approval from the ethics committees of three institutions: the
   Ethics Committee for Scientific Research Involving Human Subjects at
   Doshisha University (reference number: 20009), the Institutional Review
   Board of Kyoto Prefectural University of Medicine (reference number:
   ERB-G-73), and the Ethics Committee of Friedrich-Alexander Universität
   Erlangen-Nürnberg (reference number: 140_20 B). Participants at
   Friedrich-Alexander Universität Erlangen-Nürnberg provided informed
   consent for the collection of peripheral blood samples and Descemet’s
   membranes, including CECs. All procedures in this study were executed
   in compliance with the guidelines set forth by these ethical
   committees.

Acquisition and processing of RNA-Seq data

   This study utilized RNA-Seq data from CECs of both non-FECD controls
   and FECD patients, as we previously reported^[124]37,[125]38. Raw fastq
   files were accessed from the DNA Data Bank of Japan Sequence Read
   Archive (accession IDs: DRA010237 for non-FECD and DRA015078 for FECD
   subjects). Sample information from these RNA-Seq data is available in
   (Supplementary Table [126]1). Previous studies utilized peripheral
   blood samples to classify FECD patients based on the presence or
   absence of TNR expansion in TCF4^37,38. The threshold of 50 repeats was
   chosen based on its established association with FECD
   pathogenesis^[127]7. In our cohort, all No Expansion samples had fewer
   than 40 repeats, while Expansion samples had 50 or more repeats, with
   no samples falling within the intermediate range of 40–50 repeats. In
   the present study, we investigated RNA-Seq data of CECs derived from
   these previously characterized FECD patients with known TNR expansion
   status. Data processing commenced with elimination of low-quality reads
   from raw fastq files using fastp software (v 0.20.0)^[128]39.
   Subsequently, the processed reads were aligned to the human reference
   genome (Homo_sapiens.GRCh38.111) using STAR aligner (v
   2.7.10a)^[129]40. Gene annotations were based on the Ensembl annotation
   file (GRCh38.111.gtf).

Identification and analysis of DAS events

   The analysis of DAS events was performed using rMATS-turbo
   (v4.3.0)^[130]41,[131]42. Five types of alternative splicing events
   were examined: SE, A3SS A5SS, MXE, and RI. Statistical significance was
   determined by adjusting P-values using the Benjamini-Hochberg method to
   control for the false discovery rate (FDR). DAS events were considered
   significant if they met the following criteria: FDR < 0.05 and
   |IncLevelDifference| > 0.2. The significant DAS events identified from
   the comparisons of the No Expansion group with the Control group, and
   the Expansion group with the Control group, were displayed as Venn
   diagrams using the “VennDiagram” package (v1.7.3) in R (version 4.4.1)
   to visualize the overlap between these datasets. Visualization of
   specific genes with significant differences was accomplished using
   rmats2sashimiplot (v3.0.0).

   To identify potential splicing regulators, we applied rMAPS2
   (v2.2.0)^[132]43,[133]44 to assess the enrichment of RBPs motifs near
   alternatively spliced exons in DAS events identified by rMATS-turbo.
   This analysis was performed on both exon inclusion and skipping events,
   aiming to elucidate the regulatory mechanisms underlying the splicing
   alterations observed in FECD. The results were also visualized using a
   heatmap generated with the “ggplot” function from the “ggplot2” package
   (v3.5.1) in R.

Assessment and confirmation of data profile

   PCA was performed using the “prcomp” function from the “base” package
   in R to assess the overall distribution of samples based on their
   splicing patterns. Heatmap analysis was conducted using the “heatmap.2”
   function from the “gplots” package (v3.1.3.1) in R. Hierarchical
   clustering was applied using the “ward.D2” option in the “hclust”
   function from the “stats” package in R to group samples and splicing
   events based on similarity. We visualized the distribution of different
   types of alternative splicing events by generating a pie chart based on
   the counts of each splicing event type (SE, A3SS, A5SS, MXE, and RI)
   across all samples by using the “ggplot” function from the “ggplot2”
   package in R.

Domain and pathway enrichment analysis in DAS events

   NEASE (v1.3)^[134]45 software was used to assess whether the exons
   involved in DAS events affect the resulting protein structure. This
   analysis focused on the impact of these exons on protein domains,
   residues, or linear motifs. Splicing events were classified based on
   their potential to disrupt or alter these functional features, and the
   proportion of affected events in each category was calculated. For DAS
   events predicted to impact protein domains or functional features,
   pathway enrichment analysis was performed within the NEASE analysis
   using Reactome^[135]46 and KEGG databases^[136]47,[137]48. The results
   were ranked based on NEASE scores, with significantly enriched pathways
   identified by adjusted P-values < 0.05, thereby providing insights into
   the broader biological consequences of the splicing alterations.

Validation of splicing alterations with RT-PCR

   Corneal endothelial tissue of patients with FECD was obtained during
   corneal transplantation. For controls, we utilized residual healthy
   corneal endothelium from non-FECD donor corneas collected during
   corneal transplantation procedures. The TCF4 TNR status of all samples
   was previously characterized in our studies^[138]49, confirming that
   none of the non-FECD donor corneas harbored TNR expansion. Total RNA
   was extracted using the RNeasy Micro Kit (Qiagen, Valencia, CA). cDNA
   was synthesized from the extracted RNA using the ReverTra Ace -α-
   (Toyobo, Osaka, Japan) according to the manufacturer’s instructions.
   PCR amplification was performed using KOD FX Neo (Toyobo) with the
   following conditions: 90 °C for 2 min (1 cycle), 98 °C for 10 s, 60 °C
   for 30 s, and 68 °C for 15 s (32 cycles), followed by a final extension
   at 68 °C for 5 min. The PCR primers for MBNL1 and MBNL2 were designed
   based on previously published studies (Supplementary Table
   [139]2)^[140]50,[141]51. The amplification products were separated on a
   2.5% agarose gel in TBE running buffer and imaged using the Amersham
   Imager 600 (Cytiva, Tokyo, Japan).

Electronic supplementary material

   Below is the link to the electronic supplementary material.
   [142]Supplementary Material 1^ (29.5KB, pdf)
   [143]Supplementary Material 2^ (36.3KB, pdf)
   [144]Supplementary Material 3^ (686.5KB, pdf)

Acknowledgements