Disease progression of retinitis pigmentosa caused by PRPF31 variants in a Nordic population: a retrospective study with up to 36 years follow-up

ABSTRACT Background/aims To investigate the natural history of PRPF31-related retinitis pigmentosa (RP11). Materials and methods We identified individuals with RP11 and collected retrospective data from disease onset to present date including genetics, demographic data, Goldmann visual field areas, and visual acuity measurements. Visual fields were evaluated as summed squared degrees and best-corrected visual acuity was converted to logMAR. We performed linear mixed model regression analysis to evaluate annual disease progression, and survival analysis to evaluate the age of legal blindness. Results We included 46 subjects with RP11. Median age of disease onset was 10 years (range 5–65). Follow-up spanned from 0 to 36 years with a median of 8 years. Median Goldmann visual field areas decreased by 10.0% per year (95% CI 7.5%−12.4%) with target IV4e, 7.9% (95% CI 4.5% − 11.2%) with target III4e, and 9.3% (95% CI: 7.0% −11.5%) when combining target sizes. Individuals with RP11 maintained good visual acuity until late stage of disease. Legal blindness was reached at a median age of 57 years (95% CI 50–75 years). Conclusions PRPF31 variants cause autosomal dominant retinitis pigmentosa that most commonly manifests in childhood with a variable disease progression. Visual field area deteriorates faster than visual acuity and was the major cause of legal blindness in our study population. This study characterizes disease progression in retinitis pigmentosa caused by PRPF31-variants and demonstrates the importance of differentiation between specific genotypes when counselling patients and conducting natural history studies of RP.


Introduction
Retinitis pigmentosa (RP) is a clinically and genetically heterogeneous group of inherited retinal dystrophies where degeneration of photoreceptors, lead to a phenotype characterized by attenuated retinal arterioles, a pale optic disc, and pigmentary changes. It is a relatively rare disorder with a worldwide prevalence of 1:4000 and can follow any inheritance pattern (1). Affected patients experience poor night vision, constriction of visual fields followed by loss of visual acuity, and eventually some may experience blindness (2).
All genetic subtypes of RP are progressive, but progression rates and other clinical features differ between individuals and genetic subtypes. However, it is often difficult to predict the course of disease due to limited knowledge of that specific genotype. With the advancement of molecular genetic analyses, genetic findings have become key in confirming and refining the clinical diagnosis (3,4). To date, 280 genes have been implicated in retinal diseases (https://sph.uth.edu/retnet/). Disease-causing variants in the same gene may present with different clinical findings and degrees of disease progression, depending on the type of variant, other elements such as modifiers, epigenetic factors, and wild-type-allele protein expression. Even individuals from the same family with identical variants may show significant variation in disease severity, which further complicates the counselling of families. This study describes the progression of RP caused exclusively by PRPF31 variants.
Heterozygous pathogenic variants in PRPF31 can cause autosomal dominant RP and are reported to account for approximately 5-10% of all dominant RP cases (5). PRPF31 is located on chromosome 19 and encodes the PRPF31 protein which is a pre-mRNA splicing factor involved in coupling of the spliceosome (6). The spliceosome is a vital part of the protein synthesis and pathogenic variants in PRPF31 cause less efficient splicing and thus an impaired protein synthesis (7). The disease mechanism is haploinsufficiency (8). PRPF31 is ubiquitously expressed in all tissues in the body, but only retina-specific disease has been described, most likely due to high splicing demands in the retina (7). Cases of nonpenetrance with asymptomatic carriers are well described in PRPF31-related retinitis pigmentosa, known as retinitis pigmentosa type-11 (RP11) (9).
Inherited retinal dystrophies are excellent targets for gene therapy because of the relatively easy access to affected tissues (10). Gene therapy trials have shown promising results as seen with voretigene-neparvovec in RPE65-related retinal dystrophy (11), and other genetically based therapies are being tested in ongoing clinical trials (12)(13)(14). Disease natural history studies are often used by health-care administrative bodies to weigh the effect of gene therapy against disease burden (15). As disease progression patterns may vary between genotypes and as many emerging therapies are genetically based, exact genotyping and natural history studies are necessary to investigate the specific genotype-phenotype relations. In addition, natural-course studies of RP subtypes are useful and essential in planning rehabilitation and conduction of future prospective, and ultimately, possible treatment trials.
In this study, using retrospective analyses, we aim to characterize disease progression and natural history in retinitis pigmentosa caused by PRPF31 variants in a Nordic population. This study contributes to improve options for counselling of individuals and families with this type of RP.

Materials and methods
We performed a retrospective study on patients with retinitis pigmentosa caused by variants in PRPF31. The individuals were identified among patients who had been examined at the Department of Ophthalmology, Rigshospitalet-Kennedy Centre and via the Danish Family Archive of Hereditary Eye Diseases. The Kennedy Centre is a national tertiary referral center for patients with rare and inherited eye diseases. Historically, all patients in Denmark with a retinal dystrophy have been seen at the Kennedy Centre at some point in their disease process and some have been followed regularly throughout life. Patients diagnosed and examined at the Department of Ophthalmology, Oslo University Hospital, Norway, were also included. The individuals were identified through the patient registry of inherited retinal disease at Oslo University Hospital.
Medical charts were accessed to extract information on Goldmann visual fields, visual acuity measurements, and demographic data including age of onset and age of diagnosis. Furthermore, descriptions of fundoscopy in medical files, and evaluation of fundus photographs, when available, were used to perform a fundoscopic characterization. Disease onset was defined as the age when symptoms were first noticed by the patient. If a subject specified onset as "before starting school" the age of disease onset was set to 5 years. Disease duration was derived as the time interval from onset of symptoms to examination. The probable disease-causing PRPF31-variant was noted in every case.

Molecular genetic analysis
The majority of the molecular genetic testing had been performed in a previous study (16) with a next-generation sequencing (NGS) panel of 125 genes previously shown to be associated with retinal dystrophies. Some samples were analyzed as part of clinical routine of 186 retinal dystrophy genes using a custom made targeted SureSelect panel (Agilent, Santa Clara, CA, USA) followed by NGS with Illumina technology and Miseq (Illumina, San Diego, CA, USA). Bioinformatics (alignment and variant calling) were performed using the SureCall program (Agilent) and default parameters. Annotation and filtering were performed using VarSeq (Golden Helix, Bozeman, MT, USA). A few samples were analyzed by external laboratories. Variants were classified according to the American College of Medical Genetics and Genomics (ACMG) classification (17) and additional guidelines from the ClinGen sequence variant interpretation working group (https://www.clinicalgenome.org/working-groups /sequence-variant-interpretation/). Array Comparative genomic hybridization (aCGH) was performed using Agilent 1 M Sureprint G3 Human CGH micromatiser (Agilent).

Goldmann kinetic visual fields
Visual fields from all visits were evaluated as summed degrees squared (deg 2 ) using the software Fiji ImageJ (18) using the method described by Zahid et al (19). Areas with scotoma were subtracted, and non-attached areas of residual visual field were summarized. The horizontal diameter of central visual field was registered. Available isopters for the different Goldmann targets were quantified at all available time points. Primary outcome was the yearly decline in visual field area, corresponding to the slope of the fitted model. We fitted models to observations with Goldmann target III4e, IV4e, and both targets combined.

Visual acuity
Best-corrected monocular and binocular visual acuity (VA) from all visits were registered. Decimal, fraction and ETDRS visual acuity measures were converted to logMAR values, and off-chart visual acuity values were converted according to Holladay (20). VA measurements of the better seeing eye were grouped according to individual's age in the following age groups: 0-17 years, 18-39 years, 40-59 years, and 60+ years.

Legal blindness
We used the United States definition of legal blindness corresponding to a best-corrected visual acuity of 20/200 or less or a constricted visual field with a diameter no greater than 20 degrees (21). For each individual who met the criteria, their age when first meeting the criteria was registered.

Statistical methods
We used median and range to describe age distributions, number of follow-up visits, and follow-up times. Summed visual field areas in squared degrees were log-transformed prior to analysis to approximate linear decay with time. Rates of decay were estimated using a linear mixed model including disease duration as a fixed effect. Subject-and family-specific random intercepts and slopes were included to account for variation between families and individual study participants. Finally, an assessment-specific effect of subject was added to the model to account for strong correlation between assessments on the left and right eye. Goodness of fit was evaluated using residual diagnostics. Age of legal blindness was analyzed using a Kaplan-Meier curve. All analyses were performed using R statistical software (22). The lme4-package (23) was used for linear mixed model analyses. The level for statistical significance was set to 5%.
This study was approved by Danish Patient Safety Authority (file no. 3-3013-3054/1) and The Danish Data Protection Agency (file no. P-2019-323), and it adhered to tenets of the Declaration of Helsinki.

Results
We identified 49 individuals with PRPF31-related retinal dystrophy (45 Danish and 4 Norwegian). Three individuals were excluded from the statistical analyses presented below because no clinical information was available. Of the 46 included individuals, 29 were females (63%) and 17 were males (37%). Follow-up ranged from 0 to 36 years (median: 8 years; IQR: 0-21 years), with up to 20 visits for one individual (median: 3 visits; IQR: 1-6 visits). The age of disease onset ranged from 5 to 65 years (median: 10 years; IQR: 5-16 years), see supplementary Figure S1. Age of diagnosis ranged from 7 to 65 years (median: 10 years; IQR: 17-37 years). The natural disease course of one individual is illustrated by observations at three time points in Figure 1. Additionally, Figure 1 illustrates the typical fundoscopic phenotype in RP11. The most frequent finding was a classical RP with attenuated vessels, a pale optic disc, and variable degrees of pigmentary changes in the form of bone spicules. However, a few adults also with advanced disease, had close to none or very sparse retinal pigmentary changes. The genetic distribution of PRPF31 variants is shown in Table 1. The included individuals came from 18 different families representing 17 unique PRPF31 variants. Two unrelated families carried the same variant with deletion of exons 2-14. All identified variants were potentially disease-causing variants.

Goldmann visual fields
In total, 332 Goldmann visual fields from 34 individuals were available with a median of 7 (range 2-52) measurements per individual. Six visual fields were excluded from the analyses as they were deemed unreliable due to young age and larger visual fields at subsequent visits. Goldmann standard target IV4e was the most frequently used (208 visual fields) followed by standard target III4e (96 visual fields), see supplement table S1. Thus, the statistical models were fitted using Goldmann target IV4e and III4e measurements. The individual observations of visual field loss are presented in Figure 2, with the curves of the mixed-effect linear regression models overlain. The annual decay in median visual field measured in summed square degrees was 7.9% (95% CI: 4.5% − 11.2%) for target size III4e (VF III ) and 10.0% (95% CI: 7.5% − 12.4%) for target size IV4e (VF IV ) based on the exponential functions below: Here VF is the visual field area in deg 2 , and t is reported disease duration in years. Combining visual fields irrespective of target size, we found an annual decline of 9.3% (95% CI: 7.0% − 11.5%). See Supplement Tables S2, S3, S4, and Supplement Figure S2 for the detailed full output of the statistical analysis and the fitted combined model plotted.

Legal blindness
Seventeen patients (37%) met the criteria for legal blindness. Median age of legal blindness in RP11 was estimated to 57 years (95% CI 50-75 years) from the Kaplan-Meier plot, see Figure 4. All individuals met the legal blindness criteria due to visual field loss; one person also met the visual acuity criterium.

Discussion
We present the natural history of PRPF31-related retinitis pigmentosa disease progression using three functional parameters: Goldmann visual field, visual acuity, and the age of legal blindness. The study was based on 46 individuals who had been followed for up to 36 years that makes our cohort the largest population of RP11 patients published to date. The study was based on retrospective data; thus, a patient could have one or several visual field or visual acuity measurements performed over the years and the number of follow-up visits varied greatly. We used advanced statistical analyses to overcome this issue and we found that half of patients had become legally blind by the age of 57 years, and that each year patients had lost around 9% of the residual visual field (in summed squared degrees) but with large interindividual variations.
We found disease onset wide-ranging from 5 to 65 years with a median of age 10, suggesting that most individuals with RP11 have noticed symptoms, often night blindness, in early Table 1. Genetic variants of PRPF31 identified in this study. Three individuals were excluded from further analyses because clinical information could not be retrieved. 7 individuals were asymptomatic carriers and were not included in the study.    childhood. In two other isolated RP11 populations with 12 and 26 patients, the disease onset ranged from 4 to 19 years and 6-63 years, respectively (24,25). Large age-span in disease onset is well known in retinal dystrophies in general, exemplified by a study based on 370 RP-patients reporting onset age from 1 to 89 years (26). In theory, the variations in disease onset in PRPF31-related retinal disease may be linked to the degree of haploinsufficiency, and thereby reflect expression level of the healthy PRPF31-allele. However, our observations did not indicate that those individuals with a late disease onset, or those with a delayed decline in visual field area, progressed with a slower rate of visual field loss. Neither was there a clear relationship between individuals with identical gene-variants, see supplement Figure S3. In one family, another genetic variant, besides the PRPF31-variant, was also detected in another RP-gene (IMPDH1), but since the clinical course in all family members was comparable to that observed in other PRPF31-ADRP subjects, we chose to include the individuals in the study. The fundoscopic characteristics were those of classic RP with narrowed arterioles, a pale optic disc and bone spicules as the most noticeable changes. In the individuals with sparse pigmentary changes we still found attenuated vessels, a pale optic disc, and a retinal atrophy compatible with generalized retinal dystrophy. The differences in retinal fundoscopic phenotype did not relate to specific PRPF31-variants.
Visual field loss was the only cause of legal blindness in our cohort, except for one individual who in addition met the visual acuity criterium. Goldmann kinetic perimetry had been used consistently to follow patients with retinal dystrophies the last five decades, and kinetic visual fields act as a useful examination to detect gross visual field loss (27). We found an annual decline in visual field area of 10.0% with Goldmann target IV4e, and a 7.9% decline with target IIIVe. Observations with both targets combined in one statistical model allowing different intercepts, resulted in a 9.3% yearly decline in visual field area. These results are in line with two previous, smaller studies on RP11. One cross-sectional study (n = 26) reported a yearly loss of 6.9% for target size V4e (28) and a second study, based on visual field data from 12 patients of whom 6 had serial visual fields available, reported an annual decline of 8.0% (95% CI: 5.6-10.0) for target size V4e, 8.1% (5.6-10.6) for target size III4e, and 8.4% (6.2-10.7) for target size I4e (25). The studies did not use mixed model statistical analyses to account for repeated measurements or the correlation between eyes of each one individual, which makes a direct comparison to our data difficult.
Other time-course studies, mainly retrospective (29-38) and a few prospective (39)(40)(41), have reviewed the progression of retinitis pigmentosa, and most have used advanced statistical models similar to our approach (29)(30)(31)33,34,36,37,40). These studies were based on genetic heterogenic populations (including unknown genetic variants) or populations with other specific gene-variants than PRPF31. Therefore, our calculations of the progression rates of the kinetic visual fields contribute to the existing knowledge for RP11. When comparing the progression rates computed in this study to those found in methodically similar reports, we see both higher and lower rates. One mixed genotype study (n = 23) found an annual progression rate (target III4e) of 14.5% (95% CI: 9.2-19.4) (34) and a larger mixed genotype population (n = 52) showed a decline with same target size of 10.7% (95% CI: 5.7-15.5) and a 7.5% (95% CI: 3.8-11.1) decrease with target V4e (29). An annual visual field loss of 13.6% (95% CI: 11.7-15.5) with target V4e has been reported in autosomal recessive Usher type II (30). However, when looking at autosomal dominant RP, like RP11, two studies reported an annual decline in visual fields (Goldmann target V4e) of only 2.6% (95% CI: 1.5-3.7) in RHO-related RP (37), and 4.6% (SD 3.5) in IMPDH1-related RP (42). These pervasive differences illustrate the importance of conducting natural history studies of isolated genotypes and call for some carefulness when evaluating different cases, and not solely rely on clinical diagnosis or even the inheritance pattern when informing of prognosis and disease time course. Moreover, using the advanced statistical models gives statistical strength to the estimates and confidence intervals (43), and the exponential model of visual field loss also reflects the exponential decay of photoreceptors (40,44).
We found that visual acuity remained fairly stable until a late stage of disease in the majority of patients. This is in line with three other reports on patients with RP11, that found minimal change in BCVA in a three-year follow-up (45), an annual decline in visual . Kaplan-Meier plot showing age distribution for legal blindness with 95% confidence intervals. Legal blindness probability refers to the percentage of the individuals who have met legal blindness criteria at a given age. The numbers below the plot refer to how many individuals from the study population who were at risk at a given time (not blind and not lost to follow-up). acuity corresponding to 2 EDTRS letters lost every 5 years (28), and 1 ETDRS letter lost every second year (25), respectively. In contrast, in a large mixed genotype study (n = 1.085), legal blindness due to visual acuity loss was reported to be at least 20% and 7.3% had a visual acuity of counting fingers or worse (32). These findings suggest that RP11 patients may have a slower decline of visual acuity than the average RP-patient. Due to the retrospective nature of our study, we were not able to differentiate if the decline in visual acuity with age was related to progression of RP with loss of foveal function, or if it was related to other eye disorders of an age-dependent nature, e.g. age-related macular degeneration or cataract.
Furthermore, we found that legal blindness for half of patients in our study population of RP11, occurred already in late midlife at age 57. This very likely affects social life, employment options and other activities of daily living (46). Other RP-studies reported median ages of legal blindness spanning from 44 to 77 years (29,(35)(36)(37), the differences again advocating for genetic subtyping of RP to differentiate and identify different natural histories.
This study does have limitations. The retrospective study design based on existing patient files, entails that the data volume, content, and consistency vary between included individuals and between visits for the single individual. Some patients were only seen once and time intervals between follow-up was very long for others. Moreover, the quality and validity of the clinical observations is hard to assess and the risk of intra-and interobserver variation is present, however a limited number of staff have been involved in managing the patients over the years. The self-reported disease onset is associated with some uncertainty. Conversely, RP11 is a rare and a relatively slowly progressive condition, meaning it is difficult to gather a larger and more consistent population, and especially to do a prospective trial with the same follow-up duration as our study. Therefore, we believe, that this study has a significant value both to the scientific and to the clinical field.
In conclusion, we characterized the progression of PRPF31related retinitis pigmentosa. This study contributes to improved counselling of patients with this subtype of RP and describe the largest population to date. The progression of visual field loss in these individuals will aid to a better prediction of prognosis in a patient with RP11. Ultimately, natural history data are useful in future treatment trials, and our study suggests that kinetic visual fields could be useful in monitoring subjects with RP11 in such trials.