The majority of autosomal recessive nanophthalmos and posterior microphthalmia can be attributed to biallelic sequence and structural variants in MFRP and PRSS56

This study aimed to genetically and clinically characterize a unique cohort of 25 individuals from 21 unrelated families with autosomal recessive nanophthalmos (NNO) and posterior microphthalmia (MCOP) from different ethnicities. An ophthalmological assessment in all families was followed by targeted MFRP and PRSS56 testing in 20 families and whole-genome sequencing in one family. Three families underwent homozygosity mapping using SNP arrays. Eight distinct MFRP mutations were found in 10/21 families (47.6%), five of which are novel including a deletion spanning the 5′ untranslated region and the first coding part of exon 1. Most cases harbored homozygous mutations (8/10), while a compound heterozygous and a monoallelic genotype were identified in the remaining ones (2/10). Six distinct PRSS56 mutations were found in 9/21 (42.9%) families, three of which are novel. Similarly, homozygous mutations were found in all but one, leaving 2/21 families (9.5%) without a molecular diagnosis. Clinically, all patients had reduced visual acuity, hyperopia, short axial length and crowded optic discs. Retinitis pigmentosa was observed in 5/10 (50%) of the MFRP group, papillomacular folds in 12/19 (63.2%) of MCOP and in 3/6 (50%) of NNO cases. A considerable phenotypic variability was observed, with no clear genotype-phenotype correlations. Overall, our study represents the largest NNO and MCOP cohort reported to date and provides a genetic diagnosis in 19/21 families (90.5%), including the first MFRP genomic rearrangement, offering opportunities for gene-based therapies in MFRP-associated disease. Finally, our study underscores the importance of sequence and copy number analysis of the MFRP and PRSS56 genes in MCOP and NNO.


Methods
Patients and clinical assessment. Twenty-five patients out of 21 unrelated families from different ethnicities with either isolated or complex NNO or MCOP were recruited from three centers: Ghent University Hospital, Belgium (n = 11); Moorfields Eye Hospital, London, UK (n = 7); and Birmingham Children's Hospital, Birmingham, UK (n = 7) ( Table 1). A full ophthalmological examination included BCVA measurement and dilated fundus examination. Retinal fundus imaging was obtained by conventional 30-degree fundus colour photographs (Topcon Great Britain Ltd, Berkshire, UK) or via ultra-wide field confocal scanning laser imaging (Optos plc, Dunfermline, UK), near-infrared and blue light fundus autofluorescence (FAF) imaging (Spectralis, Heidelberg Engineering Ltd, Heidelberg, Germany), and spectral domain optical coherence tomography (OCT) scans (Spectralis, Heidelberg Engineering Ltd, Heidelberg, Germany). Full-field electroretinography (ERG) was performed according to the International Society for Clinical Electrophysiology of Vision (ISCEV) standards 27 (using either Roland Consult, Brandenburg an den Havel, Germany or Diagnosys, Cambridge, UK equipment). Genomic DNA was extracted from EDTA blood using standard procedures. This study was conducted following the tenets of the Declaration of Helsinki.
Homozygosity mapping. Homozygosity mapping using genome-wide single-nucleotide polymorphisms (SNP) arrays was performed in three patients originating from a self-reported consanguineous marriage, using HumanCytoSNP-12 BeadChips (Illumina, San Diego, CA). Homozygous regions (>1 Mb) were identified using PLINK 28 software integrated in ViVar 29 . Resulting homozygous regions were ranked according to their length and number of SNPs, as described 30 . Mutation screening by Sanger sequencing and by whole genome sequencing. Primers for PCR amplification of the coding region and splice site junctions of MFRP and PRSS56 were designed (available upon request). Sanger sequencing was performed according to the manufacturer's instructions (BigDyeTerminator v3.1 Cycle Sequencing Kit, ABI 3730XL genetic analyzer, Thermo Fisher Scientific, Waltham, MA).
One family (F7.GC20271) was included in a whole genome sequencing (WGS) study. Genome enrichment was performed using the Illumina TruSeq DNA PCR-Free Sample preparation kit (Illumina, Inc.), followed by sequencing on an Illumina HiSeq 2500 with a minimum coverage of 15x for approximately 95% of the genome. The Isaac Genome Alignment Software (version 01.14; Illumina, Inc.) was used for reads mapping against the Genome Reference Consortium human genome build 37 (GRCh37) 31 . Standard variant filtering was performed as previously described 32,33 . Structural variant (SV) assessment was done by interrogation of copy number variation (CNV), SV calls and visual inspection of the individual split and chimeric reads (Integrated Genome Viewer, IGV) across the breakpoints as described by Carss et al. 32 (Fig. S1). Subsequent testing of the MFRP and PRSS56 genes revealed three distinct homozygous mutations in these families, two of which are novel ( Table 2). Direct testing of MFRP and PRSS56 in 17 families revealed 11 additional distinct mutations in MFRP and PRSS56, five of which are novel ( Table 2). WGS in one family (F7) revealed two novel heterozygous mutations in MFRP: a coding nonsense variant c.955C > T p.(Gln319*) and genomic rearrangement consisting of a deletion of 6.2 kb and an insertion of 7 nucleotides c.−6087_54 +40delinsTAGTGGT p.(?). This genomic rearrangement encompasses the 5′UTR and the coding part of exon 1 and is predicted to abolish the transcription initiation site (Fig. 1A). The breakpoints of this deletion were characterized by a junction PCR followed by sequencing (Fig. 1B). An assessment of the breakpoints at the nucleotide level showed the insertion of seven base pairs (TAGTGGT), representing a potential information scar. As no microhomology was detected at the breakpoints and only one breakpoint overlapped with an Alu repeat, non-allelic homologous recombination (NAHR) and microhomology-based mechanisms are unlikely. Altogether, non-homologous end joining (NHEJ) is the most likely mechanism underlying this rearrangement [35][36][37] .
Overall, genetic defects were found in 19 of the 21 families (90.5%): eight distinct MFRP mutations in ten families (10/21, 47.5%) and six distinct PRSS56 mutations in nine families (9/21, 42.9%). Biallelic mutations were found in 18 families, while a monoallelic MFRP mutation was found in one family, with an undiscovered second mutation. With mutations neither in MFRP nor PRSS56, two Belgian families (2/21, 9.5%) remained molecularly unaccounted for. A summary of all variants identified, their in silico assessment and ACMG variant classification can be found in Table 2. Flowchart of the molecular workflow and outcomes is provided in Fig. S2.
Phenotypic characteristics. Twenty-five patients from 21 unrelated families were investigated, six with a diagnosis of NNO and 19 with MCOP. The age of diagnosis varied from 3-75 years. Overall, all eyes had an axial length of <18 mm and hyperopia of >8 diopters with crowded discs and foveal hypoplasia. The clinical features of all studied individuals are summarized in Table 1. Based on the genotypes found, the families were divided into two genetic subtypes: a MFRP-and PRSS56-associated group.

Discussion
This study characterizes 21 unrelated families with NNO (n = 6) or MCOP (n = 19). Fourteen distinct MFRP and PRSS56 variants were identified in the majority of the studied families (19/21, 90.5%). Eight different MFRP variants were found in 10/21 (47.6%) of the families, five of which are novel. Biallelic pathogenic variants were found in 9/10 families, supporting autosomal recessive inheritance. In line with the previously reported 19 distinct MFRP variants, mainly causing the introduction of a premature termination codon 20,38-44 , frameshift and nonsense variants represent the majority of the MFRP mutation spectrum of this study. Some of these variants were found to be recurrent: a novel mutation c.1090_1094del p.(Thr364Glnfs*26) in two unrelated Belgian families (F1 and F9) and a known mutation c.498del p.(Asn167Thrfs*25) in two Moroccan (F2, F10) and one Italian family (F8). The latter variant was previously reported amongst other consanguineous and non-consanguineous families of Spanish, Palestinian, Mexican and Japanese origin 9,22,38,42,45,46 . Interestingly, the previously reported reciprocal duplication c.498dup p.(Asn167Glnfs*34) 8,43 , which might point to a mutational hotspot, was found in a Dagestanian family (F5) in our cohort.
In one individual with MCOP (F4) a monoallelic MFRP variant was found c.491_492insT p.(Asn167Glnfs*34) 43 , with an as yet unidentified second mutation. A structural variant including a copy number variant (CNV), deep intronic mutation of regulatory mutation are possible underlying causes to be explored further. Indeed, this study provided evidence for the occurrence of CNVs affecting MFRP with the identification of the 6.2 kb deletion c.6087_54 +40delinsTAGTGGT.
To date, nine distinct PRSS56 pathogenic variants have been described, mostly frameshift and nonsense mutations 17,24,47,48 . In our cohort, we detected six different PRSS56 mutations in 9/21 (42.8%) families, some of which are known as recurrent mutations. Specifically, the previously reported missense variant c.1555G > A p.(G-ly519Arg), which was found in a Saudi Arabian family 41,48 , was identified in four unrelated families of Pakistani origin in this study (F14, F15, F16 and F17), suggestive of a founder mutation in this population. Another recurrent mutation is c.1066dup p.(Gln356Profs*152) found in a Pakistani family (F12), which was previously reported in six Tunisian and five Saudi Arabian families 17,48,49 . Furthermore, two novel missense mutations were found in the PRSS56 group in patients with isolated MCOP (F13, F18) and NNO (F19).
Overall, isolated cases with NNO or MCOP were found in 13/19 (68.4%) of the families with a molecular diagnosis, mostly in the PRSS56-associated group (11/13, 84.6%). Complex cases with retinal involvement (RP features or ERG changes) were found in 6/19 (31.6%) of the families with a molecular diagnosis, mostly due to MFRP mutations (5/6, 83.3%). The retinal involvement in the MFRP-associated group is in line with previous phenotypic studies 39 and is in agreement with its expression pattern in human, mouse and zebrafish eyes www.nature.com/scientificreports www.nature.com/scientificreports/ including neural and pigmentary retina 3,20,21 and its role in photoreceptor outer segment maintenance 50 . An additional explanation could be the fact that MFRP and a gene implicated in late-onset retinal dystrophy, C1QTNF5 (encoding C1q and tumor necrosis factor related protein 5) are both expressed as a bicistronic transcript and found to co-localize to the same tissues with a clear functional relationship in the retina 51,52 .
The presence of papillomacular folds was found to be more frequent in MCOP (12/19, 63.2%) than in NNO (3/6, 50%). It has been proposed that these papillomacular folds result from the disparity between the retinal and scleral growth which seems to be more prominent in MCOP cases, although no clear genotype-phenotype correlations have been established yet 16,21,48 .
Finally, no clear genotype-phenotype correlations could be established in our studied cohort. For instance in the MFRP group both F2 and F3, having a 'null' allele and a missense variant respectively, displayed posterior microphthalmos without RP-like changes. On the other hand, clinical heterogeneity was observed in the PRSS56 group, illustrated by a missense variant c.766T > C, p.(Cys256Arg) identified in F18 with a clinical diagnosis of MCOP and in F19 with a diagnosis of NNO.
Finally, a definite genetic diagnosis opens up opportunities for gene-based therapies in MFRP-associated retinal disease. Indeed, studies in two mouse models have demonstrated that MFRP-retinopathy is a potential target for gene-based therapy: Mfrp rd6 /Mfrp rd6 described by Dinculescu et al. 53 and Mfrp KI/KI described by Chekuri et al. 54 .