The detailed procedure and variant filtering were conducted as previously described. Targeted panel sequencing (TPS) of 256 known retinal disease genes and whole exome sequencing (WES) were used to identify pathogenic variants. Briefly, Genomic DNA of all probands and their available relatives was isolated from peripheral leukocytes and captured via SeqCap EZ Choice XL Library (Roche NimbleGen), then sequenced on the Illumina HiSeq platform. The raw data were analyzed using NextGene V2.3.4 software, and the reads were compared to the reference sequence of hg19 from the UCSC Genome Browser. A comparison was conducted using the 1000 Genomes Project database, dbSNP database, and Exome Aggregation Consortium database to exclude nonpathogenic polymorphisms.

Putative pathogenic RS1 variants (NM_000330.3) were verified by Sanger sequencing and segregation analysis when family members were available. Variants were classified according to the standards and guidelines of the American College of Medical Genetics and Genomics (ACMG), and the nomenclature of all variants was adjusted to the Human Genome Variation Society guidelines. ^, The protein model of human RS1 (PDB: 3JD6 ) was applied to visualize the distribution of the variants within RS1 on the protein structure. The rendering of the protein structure and annotation of interesting regions were achieved by the PyMOL Molecular Graphics System (Version 2.0, Schrödinger, LLC.).

For genotype-phenotype correlation, RS1 variants were grouped by inferred molecular mechanism into protein truncation (PT), misfolding with endoplasmic reticulum retention (ER), oligomerization failure (OF), and mislocalization (ML), following the mechanistic framework summarized by Molday et al. [10].

At each visit, detailed medical and family histories were obtained, followed by a standardized ophthalmic examination. Assessments included VA, slit-lamp biomicroscopy, dilated indirect ophthalmoscopy, color fundus photography (Topcon, Tokyo, Japan), fundus autofluorescence imaging (Heidelberg Spectralis, Heidelberg Engineering, Germany), 633-nm scanning laser ophthalmoscopy (Heidelberg Engineering), spectral-domain OCT (Topcon or Heidelberg Engineering), and full-field electroretinography (RetiPort system). VA was measured with Snellen charts and converted to logMAR. For low-vision levels, counting fingers, hand motions, and light perception were assigned logMAR 2.6, 2.7, and 2.8, respectively. ERG recordings used corneal ERG-Jet contact lens electrodes (Fabrinal, Neuchâtel, Switzerland) or DTL electrodes (Roland Consult, Brandenburg an der Havel, Germany) and adhered to International Society for Clinical Electrophysiology of Vision standards. Microperimetry was performed with the MP-3 microperimeter (Nidek, Gamagori, Japan). All ocular complications were extracted from slit-lamp examination notes and operative records. Complicated cataract was defined as posterior subcapsular cataract or cataract with onset before 50 years, and in patients aged 50 years or older we used morphology (posterior subcapsular vs nuclear sclerosis) to distinguish complicated from age-related cataract after excluding alternative causes documented in the medical record. For OCT test, macular volume scans (5.6 × 4.0 mm, centred on the fovea) comprised 25 horizontal OCT B-scans per eye, acquired by a trained technician using the eye-tracking and averaging features per manufacturer recommendations ( Figure 1 E). The scanning pattern area was 5.6 mm (W) × 5.6 mm (L) in in the X-Y plane, centered at the fovea, generating a set of 25 horizontal cross-sectional B-scan images. Each OCT scan was 5.6 mm (W) × 1.9 mm (H) in in the X-Z plane. To accurately quantify the macular cyst cavities in SD-OCT images, we defined and calculated CCV as the sum of the areas of cyst cavities across 25 sequential horizontal cross-sectional B-scans, multiplied by the distance between these scans. Mathematically, the CCV can be approximated by:

Longitudinal clinical follow-up of 107 patients in our cohort. (A-C) Age-stratified timelines of all 107 patients, divided into 3 groups: (A) youngest, (B) middle, and (C) oldest. Each horizontal bar represents a single patient, with the muted segment indicating age at first visit and the pastel segment extending to age at last visit. Superimposed symbols mark individual follow-up encounters, color-coded by assessment type: visit with VA, ERG, OCT. (D) Histogram of age at first clinical visit across the entire. (E) Workflow of the DEEP-OCT-CCSEG system, a deep learning–based image segmentation pipeline that automatically computes cyst cavity volume (CCV) from sequential OCT B-scans.

CFT was measured on the central B-scan (ILM-to-RPE). CCV was computed automatically with an in-house deep-learning(DL)-based OCT cavity-segmentation system (DEEP-OCTCCSEG). We have designed a DL-based automatic schisis cavity segmentation pipeline in OCT images of XLRS patients. We incorporated a reinforcement learning-based automatic data augmentation into our framework to boost the generalization and robustness of the DL segmentation model. Further, we trained U-Net++ model as shown in Figure 1 E. This algorithm detects and segments intraretinal hyporeflective cavities and integrates the cavity areas identified on individual B-scans into a 3-dimensional volumetric metric. All automated outputs were reviewed for quality assurance by graders who were masked to clinical information. Image quality assessment included verification of adequate signal strength and contrast, appropriate scan centration with sufficient macular coverage, and the absence of substantial motion- or shadow-related artifacts. Segmentation plausibility was assessed by confirming that schisis cavities were consistently captured across B-scans and that the delineated boundaries were anatomically consistent. Scans with suboptimal image quality or clearly erroneous segmentation were flagged and handled according to a predefined protocol: outputs were manually corrected/re-segmented when feasible; otherwise, the scan was excluded from CCV-based analyses.

All statistical analyses were conducted in Python. Data management and visualization were performed using pandas and Matplotlib/Seaborn, and statistical modelling was undertaken with statsmodels and scikit-learn. P value of less than 0.05 was considered statistically significant. Continuous variables are presented as mean ± SD (SD) and as median with IQR (IQR), whereas categorical variables are summarized as counts. Independent-samples T-tests were applied to normally distributed variables with homoscedasticity, and Mann-Whitney U tests were used otherwise. For within-subject longitudinal analyses, the data were evaluated by fitting simple linear regressions of each parameter against time, with the slope defined as the per-eye annual progression rate. Population-level effects of age were assessed using both ordinary least squares regression (OLS) or restricted cubic spline (RCS) mixed-effects models. Interocular symmetry was evaluated using Bland-Altman analysis to estimate the mean bias, 95% limits of agreement, and the proportion of pairs within the limits, while intraclass correlation coefficients were calculated to quantify between-eye concordance.

In this longitudinal cohort of 107 patients with XLRS, VA was recorded for 105 patients, full-field ERG for 89, and SD-OCT for 90 (Supplementary Table 1). The total number of visits is 440. A total of 70% of patients presented at ≤ 10 years of age, with 40% younger than 5 years. At first visit, 28% patients’ VA is at the level of or severe impairment (1.00 logMAR) or blindness (1.3 logMAR). The age at first assessment averaged 15.09 years for VA (SD 16.34; median 7.26; range 1.86-78.45), 16.33 for ERG (SD 16.99; median 7.51; range 0.00-79.08), and 14.96 for OCT (SD 15.74; median 7.36; range 2.35-79.00). The number of visits per patient was 2.96 for VA (SD 2.95; median 2.00; range 1.00-19.00), 1.36 for ERG (SD 0.69; median 1.00; range 1.00-3.00), and 2.68 for OCT (SD 2.45; median 2.00; range 1.00-19.00). Correspondingly, follow-up duration was 1.91 years for VA (SD 4.64; median 0.16; range 0.00-31.82), 1.11 years for ERG (SD 2.80; median 0.00; range 0.00-16.17), and 1.56 years for OCT (SD 2.75; median 0.23; range 0.00-15.79). Figure 1 A-D maps each patient’s age span observed and the timing of visits. Collectively, these metrics indicate denser longitudinal sampling for VA and OCT than for ERG and a right-skewed follow-up distribution in which a subset of patients contributes extended age-related pattern suitable for progression analyses.

Interocular symmetry was a notable feature in our cohort. Using Bland-Altman analysis, baseline measurements showed good interocular symmetry across all parameters, with > 90% of eye pairs falling within the 95% limits of agreement (LOA): 93.3% for VA, 94.3% for CFT, 95.5% for CCV, and 92.9% for scotopic ERG metrics (Supplementary Table 2). Mean OD-OS differences were all small and centered near zero: VA 0.019 ± 0.705 LogMAR (LOA − 1.363 to 1.400), CFT − 55.5 ± 217.8 µm (−482.3 to 371.3), CCV − 0.073 ± 0.859 mm³ (−1.756 to 1.611), St a-wave + 5.70 ± 59.24 µV (−110.41 to 121.80), St b-wave + 4.85 ± 69.10 µV (−130.59 to 140.29), and St b-a ratio − 0.028 ± 0.139 (−0.300 to 0.244). Mean values for OD and OS were highly comparable (e.g., VA 0.829 vs. 0.811 LogMAR; CFT 466.8 vs. 522.4 µm; CCV 1.503 vs. 1.576 mm³), indicating minimal baseline lateral bias and supporting the use of one eye to track longitudinal change when needed (Supplementary Table 2). Across 10 ophthalmic measures of VA, ERG and OCT, inter-eye distributions were broadly overlapping (Supplementary Fig. 1ab), indicating comparable functional and structural status of OD and OS in XLRS. Across ten ophthalmic parameters, right-left eye distributions were broadly concordant in XLRS. Mean VA was similar between eyes, and electrophysiology showed closely matched central tendencies with slightly larger scotopic amplitudes in OD, while the scotopic b-a ratio was marginally higher in OS. Structural metrics favored a modest OS predominance: CFT and CCV were higher on average in, with violin plots indicating greater dispersion in OS for CFT. Photopic ERG measures, including b-wave and 30-Hz flicker amplitudes, exhibited overlapping distributions with minimal lateral skew.

Across the cohort, we identified 64 distinct RS1 variants among 107 patients/alleles. Among these, we found 5 novel variants: a hemizygous deletion at chrXp22.13(18665260-18665502), c.593delT ( p .F198Sfs*39), c.329dup ( p .C110Wfs*11), c.569T > A ( p .L190Q), and c.403G > A ( p .G135R). The majority were missense substitutions (70/107, 65.4%), followed by nonsense variants (12/107, 11.2%), small deletions (9/107, 8.4%), gross exon-level deletions (5/107, 4.7%), splice-site variants (4/107, 3.7%), and less frequent insertions (2/107, 1.9%) and duplications (2/107, 1.9%) ( Figure 2 C). Variants were unevenly distributed by location within the gene, with the majority mapping to exons 4–6 that encode the discoidin (DS) domain, fewer observed in exons 1-3 encompassing the signal peptide/leader sequence (LS) and Rs1 domain, and only sporadic changes detected within the short C-terminal tail ( Figures 2 A and B). Recurrent cDNA changes included c.214G > A ( n = 11, 10.3%), c.637C > T ( n = 8, 7.5%), and c.598C > T ( n = 5, 4.7%), with additional clusters at c.304C > T , c.422G > A , c.638G > A , and c.288delG (each n = 3, 2.8%).

Distribution of RS1 variants identified in this study. (A) Structural representation of the RS1 protein homology model (PDB-3JD6), with the RS1 domain shown in cyan, the discoidin (DS) domain in lemon, and the C-terminal region in pink. The locations of amino acid substitutions are indicated. (B) Spectrum of RS1 variant types in the cohort, with proportions indicated in a pie chart. (C) Distribution of identified variants in the RS1 gene and corresponding protein. The upper panel shows the positions of variants on the genomic structure, with exons indicated as yellow boxes. The lower panel depicts the positions of amino acid changes along the protein, with functional domains annotated.

For genotype–phenotype correlation, our date of RS1 variants were grouped by: PT ( n = 31), ER ( n = 69), OF ( n = 1), ML ( n = 6). Comparative analyses demonstrated measurable distinctions among these groups (Supplementary Figure 1C). PT variants were associated with significantly worse visual acuity and reduced ERG amplitudes compared with ER, OF, and ML variants, consistent with a near-complete loss of retinoschisin function. Differences were also detected in retinal structure, as CFT varied significantly across mechanisms.

An age-related pattern of VA was observed in the analyses of the cohort (Figure 3a1). RCS modeling of VA vs age demonstrated that VA remained relatively stable, and in average slight improvement throughout childhood and early adolescence. From the late teenage years through the third decade, VA plateaued at approximately 0.9 logMAR on average. Beyond age 40, the spline curve exhibited a clear downward inflection, indicating accelerated VA loss in later adulthood. By ages 40 to 60, mean VA exceeded 1.2 logMAR. This non-linear age-related pattern was statistically significant (RCS model: age effect P <.001; non-linearity P <.01). These findings suggest that XLRS patients generally maintain relatively better VA during youth, followed by a prolonged plateau through early to mid-adulthood, after which a more rapid decline becomes evident. Kaplan-Meier analyses further characterized the timing of functional decline ( Figure 3 B). The median ages at which patients reached mild (0.30 logMAR), moderate (0.48 logMAR), and severe (1.00 logMAR) impairment were approximately 12, 34, and 61 years, respectively. By age 60, about 40% of patients were projected to have reached the blindness threshold (1.30 logMAR). Longitudinal data were also examined in a small long-term follow-up subgroup (≥5 years; n = 9). In this subgroup, the mean annual VA change was + 0.015 logMAR/year, indicating a very slow decline. Individual patient time courses are illustrated in Figure 3c1, where most eyes showed nearly flat slopes with minimal change over time. Because the subgroup represents a small fraction of the cohort and may be prone to selection bias, these findings are descriptive and should not be generalized to the entire cohort.

Only gold members can continue reading. Log In or Register to continue

Share this:

• X
• Facebook

Related

Related posts:

1. Outer Retinal Erosion and Outer Retinal Pinching at the Origin of Subretinal Fluid in Central Serous Chorioretinopathy
2. Effects of Optic Nerve Head Structures on Retinal Nerve Fiber Layer Defects in Primary Open Angle Glaucoma Eyes: Analyses Using Retinal Optical Texture Analysis
3. Reply to Comment on Outcomes of Penetrating Keratoplasty and Boston Type 1 Keratoprosthesis in Aniridia-Associated Keratopathy: A Systematic Review and Meta-Analysis
4. Development and Validation of a New Formula for Intraocular Lens Power Calculation in Patients With Myopic Corneal Refractive Surgery

Stay updated, free articles. Join our Telegram channel

Join