Investigation of Aberrant Splicing Induced by AIPL1 Variations as a Cause of Leber Congenital Amaurosis

Abstract

Purpose: Biallelic mutations in AIPL1 cause Leber congenital amaurosis (LCA), a devastating retinal degeneration characterized by the loss or severe impairment of vision within the first few years of life. AIPL1 is highly polymorphic with more than 50 mutations and many more polymorphisms of uncertain pathogenicity identified. As such, it can be difficult to assign disease association of AIPL1 variations. In this study, we investigate suspected disease-associated AIPL1 variations, including nonsynonymous missense and intronic variants to validate their pathogenicity.

Methods: AIPL1 minigenes harboring missense and intronic variations were constructed by amplification of genomic fragments of the human AIPL1 gene. In vitro splice assays were performed to identify the resultant AIPL1 transcripts.

Results: We show that all nine of the suspected disease-associated AIPL1 variations investigated induced aberrant pre-mRNA splicing of the AIPL1 gene, and our study is the first to show that AIPL1 missense mutations alter AIPL1 splicing. We reveal that the presumed rare benign variant c.784G>A [p.(G262S)] alters in vitro AIPL1 splicing, thereby validating the disease-association and clarifying the underlying disease mechanism. We also reveal that in-phase exon skipping occurs normally at a low frequency in the retina, but arises abundantly as a consequence of specific AIPL1 variations, suggesting a tolerance threshold for the expression of these alternative transcripts in the retina normally, which is exceeded in LCA.

Conclusions: Our data confirm the disease-association of the AIPL1 variations investigated and reveal for the first time that aberrant splicing of AIPL1 is an underlying mechanism of disease in LCA.

Leber congenital amaurosis (LCA) is the most severe inherited retinal degeneration and the most common cause of congenital blindness.¹ Leber congenital amaurosis is diagnosed in infancy and is characterized by a severely reduced or nonrecordable ERG. Leber congenital amaurosisis is genetically heterogeneous and typically inherited in an autosomal recessive manner with more than 19 genes reported to be involved (Retinal Information Network), including AIPL1.² 

The AIPL1 gene is comprised of six exons coding for the Aryl Hydrocarbon Receptor Interacting Protein-Like 1. AIPL1 is a photoreceptor-specific molecular cochaperone that interacts specifically with the molecular chaperone HSP90 to modulate the stability and assembly of the HSP90 substrate, retinal cGMP phosphodiesterase (PDE6).^3–5 

Presently, the Human Gene Mutation Database (HGMD) records 52 AIPL1 mutations.⁶ Several of these AIPL1 mutations are known benign single nucleotide polymorphisms (SNPs), while several others can be reclassified as benign rare variants in the light of recent evidence, including the detection of these sequence variations in the homozygous state in unaffected control populations.^7,8 A number of AIPL1 nonsynonymous missense sequence variations reported as LCA-associated, but not recorded in the HGMD, are of uncertain pathogenicity.⁷ Moreover, several hundred AIPL1 sequence variations of different functional classes are reported throughout the AIPL1 gene in the Single Nucleotide Polymorphism Database (dbSNP). Ascribing the disease-causing status of LCA-associated AIPL1 sequence variations is therefore a challenge confounded by the polymorphic nature of the gene. Confirmation of disease-associated AIPL1 variations is important for patient diagnosis, counselling, and potential treatments, with the main therapeutic strategy currently focusing on AIPL1-targeted gene replacement therapy. 

Thorough in vitro investigations are necessary to differentiate true disease-associated variations from rare polymorphisms. The functional consequences of only a handful of missense and nonsense changes in the AIPL1 coding sequence have been investigated by introducing the variation in the cDNA sequence and testing the impact on protein function.^4,9–14 These assays have primarily investigated the consequences of AIPL1 variations on domain-mediated AIPL1 protein interactions. The AIPL1 carboxy-terminal tetratricopeptide repeat (TPR) domain mediates the interaction with the molecular chaperone HSP90.⁴ The AIPL1 amino-terminal FK506 binding protein (FKBP)-like domain interacts directly with a farnesyl motif in vitro and is therefore predicted to interact with the farnesylated alpha subunit of retinal PDE6.¹⁴ A number of in vitro functional studies investigating AIPL1 domain-dependent interactions,^4,9–11,14 have proven useful in validating disease-causing AIPL1 mutations and understanding the role of AIPL1 as an HSP90 cochaperone for retinal PDE6. 

In contrast, the effects of noncoding variations in AIPL1 are unknown, and missplicing of AIPL1 as an underlying disease mechanism has not been experimentally investigated. In this study, we investigated missense and intronic AIPL1 variations identified in LCA patients that are predicted to alter AIPL1 pre-mRNA splicing. We determined the outcome of the variations on AIPL1 splicing and confirm that aberrant alternative transcription of AIPL1 could be an underlying cause of LCA. 

Materials and Methods

AIPL1 Sequence Variations and Nomenclature

AIPL1 sequence variations detected in previously genotyped LCA patients were investigated in this study.^7,15,16 The AIPL1 sequence variations investigated were selected on the basis that the disease-causing status was either unknown or uncertain,⁷ and on the basis that we had previously performed clinical investigations of the patients harboring the AIPL1 variations.^7,16 The study was conducted in accordance with the tenets of the Declaration of Helsinki and approved by the Moorfields and Whittington Hospitals' local ethics committee. Ensembl and UCSC Genome Browsers were used for analyses of the Human GRCh37/hg19 (February 2009) and GRCh38/hg38 (December 2013) assemblies. The AIPL1 cDNAs are numbered according to the Ensembl Transcript ENST00000381129 (RefSeq NM_014336, NP_055151). Complementary DNA nucleotide numbering uses +1 as the ATG translation initiation codon in the reference sequence, with the initiation codon as codon 1. The coordinates for the genomic AIPL1 sequence used are chr17:6,327,059-6,338,519 (hg19) and chr17:6,423,737-6,435,185 (hg38). Nomenclature was according to HGVS standards.¹⁷ 

AIPL1 In Silico Analysis

AIPL1 variations were analyzed using three in silico software prediction programmes: SIFT (Sorting Intolerant From Tolerant; in the public domain, http://sift.jcvi.org),¹⁸ PolyPhen-2 (Polymorphism Phenotyping v2; in the public domain, http://genetics.bwh.harvard.edu/pph/index.html),¹⁹ and BLOSUM62 (Blocks Substitution Matrix; in the public domain, http://www.uky.edu/Classes/BIO/520/BIO520WWW/blosum62.htm). In silico analysis of splice-site confidence levels was performed with the NetGene2 server (in the public domain, http://www.cbs.dtu.dk/services/NetGene2/)^20,21 and the Berkeley Drosophila Genome Project Neural Network splice site prediction algorithm NNSPLICE (in the public domain, http://www.fruitfly.org/seq_tools/splice.html)²² using default settings. 

Construction of Minigenes

Primers were designed to amplify genomic fragments of the human AIPL1 gene. Restriction sites were added to the 5′end of the primers such that the amplified product could be directionally cloned into the pBK-CMV expression vector (Agilent Technologies, Waldbronn, Germany). The 5′ end of each minigene was cloned into the BmtI (NheI) site, thereby deleting the lac promoter, while the corresponding 3′ ends were cloned into either the XhoI or NotI site. 

Three AIPL1 minigenes were constructed. Minigene 1 encompasses exons 1 through 4 and excludes the start codon to lessen the potential effects of nonsense-mediated decay (NMD). This minigene was amplified in two fragments that were cloned sequentially into pBK-CMV. Fragment 1 primers were designed to amplify a 1418 bp fragment and tagged with either NheI or SalI: AIPL1_Ex1_1F, 5′-gctagcGATGCCGCTCTGCTCCTGAACGTGGAAG-3′ and AIPL1_Int2_1R, 5′-gtcgacACTCCTGCTGGTCATAGCCCTGCTCC-3.′ Fragment 2 primers were designed to amplify a 2126 bp fragment and were tagged with either SalI or NotI: AIPL1_Int2_1F, 5′-gtcgacCAGCACTGCCAGGACACCAAAGCGACTCTCTTGG-3′ and AIPL1_Ex4_1R, 5′-gcggccgcTTGGTCTGCAGGTTCCTTAGGCAGATGATGG-3.′ The SalI sites introduce an artificial junction to allow deletion of approximately 4.7 kb of the central region of intron 2 of AIPL1. 

Minigene 2 encompasses exons 3-5. Primers were designed to amplify an 1892 bp fragment and were tagged with either BmtI or XhoI: AIPL1_Ex3_1F, 5′-gctagcCACACGGGGGTCTACCCCATCCTATCC-3′ and AIPL1_Ex5_1R, 5′-ctcgagCTGGGTGGTGCCGGAGAATATCACTGGTGTGC-3.′ 

Minigene 3 encompasses exons 4-6. Primers were designed to amplify a 1693 bp fragment and were tagged with either BmtI or XhoI: AIPL1_Int3_1F, 5′-gctagcGGGGTCCCTGCCTCACTGACCTGCAGC-3′ and AIPL1_Ex6_3UTR_1R, 5′-ctcgagACCAGAAGTGACCAGGCCACTTGCTCC-3.′ 

Genomic AIPL1 fragments were PCR amplified using Q5 Hot Start High-Fidelity 2X Master Mix (New England Biolabs, Hitchin, UK) in 25-μL reactions containing 10 pmol of each primer (Sigma-Aldrich, Dorset, UK) and 10 ng of genomic DNA, with the following cycling conditions: Td = 98°C for 30 seconds, then 30 cycles of Td = 98°C for 10 seconds, Ta = 72°C for 10 seconds, and Te = 72°C for 60 seconds. Polymerase chain reaction products were resolved on a 1.2% agarose gel (TAE buffer). Amplicons were excised and gel purified (QIAquick Gel Extraction Kit; QIAGEN, Crawley, UK), prior to cloning into pSC-Bamp/kan (StrataClone Blunt PCR Cloning Kit; Agilent Technologies). Sequence identity and integrity was confirmed by direct Sanger sequencing. Where patient gDNA was not readily available, mutations were engineered into w/t alleles cloned in the pSB-Bamp/kan vector using site-directed mutagenesis (SDM; Q5 Site-Directed Mutagenesis Kit; New England Biolabs). Primers were designed using NEBaseChanger software (in the public domain, http://nebasechanger.neb.com), and SDM PCRs undertaken using suggested conditions. For each mutation introduced by SDM, the entire allele was sequenced to confirm mutation and sequence integrity. Sequence verified AIPL1 fragments were cloned into pBK-CMV using standard protocols (enzymes supplied by New England Biolabs). Briefly, after digestion with appropriate enzyme pairs, digestion products were separated on a 1.2% agarose gel and relevant fragments excised, gel purified and ligations prepared (T7 DNA Ligase). Ligations were transformed into DH5α competent cells (New England Biolabs). Completed expression constructs were sequence checked to confirm mutational identity. 

In Vitro Splice Assays

Minigenes were transfected into HEK293 cells plated at 2.5 × 10⁵ cells/well in 6-well plates with TransIT-LT1 Transfection Reagent (Cambridge BioScience for Mirus Bio LLC, Cambridge, UK) following the manufacturer's instructions. Twenty-four hours post transfection, cells were rinsed twice with PBS and total RNA extracted (RNeasy Mini Kit and QIAshredder, QIAGEN) with on-column DNAse treatment (Promega, Hampshire, UK). Poly-T primed cDNA was synthesized in 20-μL volume from 2.5 μg of total RNA (Tetro cDNA Synthesis Kit, Bioline Reagents Ltd., London, UK). Splice products were amplified from 1 μL of cDNA using the following primer pairs: minigene 1, AIPL1_Ex1_1F and AIPL1_Ex4_1R - expected w/t amplicon = 652 bp; minigene 2, AIPL1_Ex3_1F and AIPL1_Ex5_1R - expected w/t amplicon = 797 bp; minigene 3, AIPL1_Int3_1F and AIPL1_Ex6_3UTR_1R - expected w/t amplicon = 520 bp. Polymerase chain reaction conditions were similar to those used for minigene generation except that the extension time was reduced to 10s/cycle. Polymerase chain reaction products were resolved on 2% to 3% agarose gels. Amplicons were excised, gel purified, cloned, and sequenced as above. All sequence analysis was conducted using MacVector 11.1.2 (MacVector, Inc., Cary, NC, USA). 

RNA-Seq Data Analysis

Publically available RNA-seq data from three healthy human adult retinas²³ was aligned against the hg19 reference genome. Aligned reads were visualized using the Broad Institute Integrated Genomics Viewer (IGV-2.3.40; Cambridge, MA, USA).^24,25 A Sashimi plot of AIPL1 mRNA sequencing reads from each of the three retinas aligned to the reference genome (human hg19) was generated to visualize differentially spliced exons in normal retina. The transcript abundance was estimated using the FPKM (fragments per kilo bases of exons for per million mapped reads) normalization method. 

Results

In Silico Analysis of AIPL1 Variations

In silico analysis of the AIPL1 variations investigated indicates that all are very rare (allele frequency < 1/10,000) and have not been observed in the homozygous state in the current Exome Aggregation Consortium (ExAC) data set (Table 1). Of the three missense substitutions, the c.784G>A [p.(G262S)] substitution appears to be the least damaging, being SIFT tolerant and PolyPhen-2 benign. This is consistent with previous in vitro studies that do not report a deficit in [p.(G262S)] function compared with the wild-type protein.¹¹ SIFT determines that the c.465G>T [p.(G155H)] and c.642G>C [p.(K214N)] missense substitutions are damaging, and neither are considered benign by PolyPhen-2 and Blosum62. In silico analysis of the effect on splicing indicates that all 9 AIPL1 variations have the potential to result in some form of misspliced AIPL1 transcripts (Table 2). All AIPL1 variations are predicted to reduce or abolish native splice site recognition leading to exon skipping, or in the case of c.276+2T>C and c.785-10_786del, the creation of potential cryptic splice sites. 

Table 1

 

In Silico Analysis of AIPL1 Variations

 

Table 2

 

AIPL1 Splice Site Predictions

 

In Vitro Splicing Assay

We confirmed experimentally that AIPL1 could not be amplified from total RNA purified from whole blood samples from unaffected individuals (data not shown). Therefore, three AIPL1 minigenes were constructed to investigate the splice effects of the AIPL1 variations (Figs. 1A, 1B). The identity of the indicated PCR fragments (Fig. 1C, bands 1–24) was determined by sequence analysis (Supplementary Figs. S2–S22). It can be seen from Figure 1C that the three w/t constructs produced the cleanest PCR fragment profiles with the fewest unique fragments, typically a strong splice product of the expected size for minigene 1 (band 1), minigene 2 (band 16), and minigene 3 (band 19). For w/t minigene 1, in addition to the expected PCR fragment (band 1), a smaller secondary splice product (band 2) was also amplified (see below). All of the minigenes carrying nucleotide variations exhibited PCR fragment profiles that differed from the corresponding w/t minigene. Polymerase chain reaction products were selected for sequence analysis on the basis of size similarity to the w/t product(s) and abundance. High molecular weight bands greater than the expected size of the w/t product and present in all samples, including the w/t control, were excluded from analysis. The results of the in vitro splicing assay are summarised in Table 3, which lists the band size (bp), a detailed description of the corresponding observed splicing effect (also see Supplementary Figs. S2–S22) and the predicted protein product. The resultant splice products and their predicted protein products are also depicted schematically in Figures 2 and 3, respectively. 

Figure 1

 

Transcript analysis of AIPL1 variations mapping within exons or introns close to or within intron/exon boundaries in LCA patients. (A) AIPL1 genomic structure of the longest AIPL1 transcript (RefSeq NM_014336). The exons (1–6) and introns (I1–I5) are numbered from the 5′ to the 3′ untranslated region (UTR). (B) Minigene 1 encompasses the entire AIPL1 sequence from exon 1 to 4 amplified from patient gDNA, with the exception of a large central part of intron 2. Minigene 2 encompasses the entire AIPL1 sequence from exon 3 to 5. Minigene 3 encompasses the entire AIPL1 sequence from exon 4 to 6 as well as a small region of the 3′ UTR. The corresponding minigenes from the wild-type allele were used as the control for transcript analysis. Arrows indicate the positions of the AIPL1 variations under investigation. (C) Transcript analysis. AIPL1 cDNA transcripts obtained from the minigene assay were PCR amplified and resolved on 2% to 3% agarose gels. Sanger sequencing was used to confirm the sequence identity and integrity of the resultant amplicons. The molecular weight markers are in base pairs, and the bands analysed are numbered (see Table 3).

Table 3

 

AIPL1 Minigene Splice Products

 

Figure 2

 

Splice site products identified by transcript analysis of AIPL1 variations. The genomic structure of the longest AIPL1 transcript (RefSeq NM_014336) is shown for w/t AIPL1. The transcripts are numbered according to the band number (B#2 to B#24) of the corresponding PCR amplicon in Figure 1C. The AIPL1 variations are demarcated by an arrowhead according to their position in the AIPL1 gene. The resultant predicted AIPL1 protein encoded by each transcript is shown above each transcript. Skipped coding sequence (deletions) and retained intron sequence (insertions) are demarcated by gray and hatched bars, respectively. The sequence data for each transcript is shown in Supplementary Figures S2 through S22. Scale bar: 500 bp.

Figure 3

 

AIPL1 protein isoforms encoded by alternative AIPL1 transcripts. White bars: AIPL1 amino-terminal FKBP-like domain (residues 1–165). The putative farensyl binding motif encompasses residues 89 to 147. Gray bars: carboxy-terminal TPR domain (residues 171–314). The TPR domain encompasses three consecutive TPR motifs, TPR1 (178–211), TPR2 (230–263), and TPR3 (264–297). Each TPR motif consists of a pair of antiparallel α helices (helix A and helix B), and the contiguous series of antiparallel α helices pack against one another to form a chaperone binding channel. Black bars: primate-specific polyproline-rich domain (PRD) (residues 328–384). Hatched bars: inclusion of nonnative AIPL1 sequence induced by a frame-shift variation. Bar filled with small squares: in-frame AIPL1 insertion. The transcripts encoding the splice products p.V33Sfs*57, p.I34Dfs*10, p.F35Lfs*2, p.V33_I92delV156Efs*50, p.H93Afs*66, and p.E215Afs*3 are predicted to be degraded by NMD.

All three w/t constructs yield normally spliced RNA products as their most abundant splice product. Interestingly, w/t minigene 1 produces a second 463 bp splice product that lacks exon 3 and translates to p.H93_Q155del (ΔEx3). The p.H93_Q155del (ΔEx3) product encompasses an in-frame deletion of 62 amino acids from the FKBP-like domain of AIPL1 encompassing the region proposed to be critical for the interaction of AIPL1 with the farnesyl moiety. We did not observe exon skipping with the w/t minigene 2 and 3. 

From Table 3, it can be seen that all the AIPL1 variations investigated (with the exception of c.97_104dupGTGATCTT and c.98_99insTGATCTTG) affect splicing such that the native splice-site is not recognized, with either exon skipping, activation of cryptic splice-sites, or intron retention occurring. Partial recognition of the native splice-site in c.97_104dupGTGATCTT and c.98_99insTGATCTTG leads to premature translation termination resulting in p.F35Lfs*2 and p.I34Dfs*10, respectively. Both mutations also induce a frame-shifted prematurely terminated transcript coding for p.V33Sfs*57. The p.F35Lfs*2, p.I34Dfs*10 and p.V33Sfs*57 transcripts are all predicted to be subject to exon junction complex (EJC) nonsense-mediated mRNA decay (NMD). 

Variations at the 5′ end of intron 2 (c.276+1G>A and c.276+2T>C) induce skipping of exon 2, leading to in-frame deletion of 60 residues from the FKBP-like domain, p.V33_I92del (ΔEx2), and disruption of the putative farnesyl binding motif. Interestingly, skipping of exon 2 induced by c.276+1G>A and c.276+2T>C also couples with the native skipping of exon 3 to create a large 122-residue in-frame deletion of almost the entire FKBP-like domain, p.V33_Q155del (ΔEx2+3). The c.276+1G>A and c.276+2T>C variations also induce aberrant transcripts coding for p.V33_I92delV156Efs*50 and p.H93Afs*66, respectively, both of which are predicted to be degraded by NMD. 

Interestingly, the c.277-2A>G variation at the 3′ end of intron 2 induces only a single aberrant transcript coding for p.H93_Q155del (ΔEx3). Similarly, the most abundant splice product induced by the c.465G>T variation skips exon 3 producing p.H93_Q155del (ΔEx3). The c.465G>T variation additionally induces transcription of a longer minor aberrant splice product leading to an in-frame deletion of 8 residues in the FKBP-like domain at its C-terminus, p.V148_Q155del. 

The c.642G>C AIPL1 variation generates two aberrant splice products from minigene 2. The smaller 343 bp product skips exon 4 leading to the in-frame deletion of the last nine residues of the FKBP-like domain and the first TPR motif of the TPR domain, p.V156_K214del (ΔEx4). The larger 499 bp splice product results in the in-frame insertion of seven amino acids in the loop region connecting the first and second TPR motifs in the TPR domain, p.K214N_E215insVRGRWPG. 

Two aberrant splice products are observed for the c.784G>A AIPL1 variation in minigene 3, both of which result in a frame-shift and premature translation termination (p.V249Afs*3 and p.E215Afs*3). The less abundant 655 bp transcript coding for p.E215Afs*3 is predicted to be cleared by NMD. The more abundant 757 bp transcript producing p.V249Afs*3 may escape NMD resulting in C-terminal truncation of AIPL1 including half of the TPR domain and the entire polyproline-rich domain (PRD). 

The c.785-10_786del mutation in minigene 3 generates three aberrant splice products that lead to premature termination (p.I263Afs*9), a 14 amino acid deletion in the TPR domain (p.G262_A275del), and a frame-shift read-through into the 3′UTR (p.G262Efs*109), respectively. 

Alternative Transcription of AIPL1 in Normal Retina

Analysis of the RNA-seq dataset publically available through Illumina's Human BodyMap 2.0 project confirmed that AIPL1 is expressed specifically in the retina and pineal gland, and is not expressed in any of the 16 tissue types included in the dataset, including blood, brain, and testes. Annotation of AIPL1 on the Ensembl database (in the public domain, http://www.ensembl.org) shows 11 alternative AIPL1 transcripts, one of which is subject to NMD (AIPL1-007; transcript ID: ENST00000381128). The normally spliced 6 exon wild-type transcript is annotated as AIPL1-001 (transcript ID: ENST00000381129), while a splice variant lacking exon 3 is annotated as AIPL1-002 (transcript ID: ENST00000250087). However, the relative abundance and physiological relevance of these alternative transcripts is unknown. Analyses of RNA-seq data from three retina-derived datasets²³ confirmed that alternative splicing of the AIPL1 gene occurs in normal retina. Quantitative multisample visualization of the three independent mRNA sequencing reads aligned to gene annotations (Sashimi plot) using IGV-2.3.40 (Broad Institute)^24,25 revealed that differential splicing of the AIPL1 gene normally includes skipping of exon 2 (ΔEx2), exon 3 (ΔEx3), and exon 4 (ΔEx4; Supplementary Fig. S1). The transcript abundance from all three datasets was estimated at 87% (95% confidence interval [CI] = ± 0.7) for the full-length transcript (NM_014336.3), 9.7% (95% CI = ± 1.0) for the ΔEx3 transcript (NM_001033054.1), and 3.4% (95% CI = ± 0.4) for the ΔEx2 transcript (NM_001033055.1). Skipping of exon 4 (ΔEx4) appears to be a rare event and was detected in only one of the three independent datasets with the greatest coverage. Of note, the two splice products detected from w/t minigene 1 in the in vitro splicing assay correspond to the two most frequently observed RNA-seq transcripts. 

Discussion

This study describes the detailed investigation of uncharacterized exonic and intronic AIPL1 variations predicted to alter the normal splicing of the AIPL1 gene in LCA patients.^7,16 A caveat of our findings is the limitations inherent in the in vitro approach, however in the absence of patient derived RNA or protein, the in vitro minigene splicing assays have proven vitally important for validating the theoretical predictions and supporting the clinical findings. Our data demonstrates that all nine AIPL1 mutations investigated cause aberrant pre-mRNA splicing to produce transcripts predicted to be degraded by NMD and/or encode functionally deficient protein isoforms. Our findings thus confirm that aberrant alternative transcription of the AIPL1 gene may be an underlying cause of LCA. 

Our study is the first to show that the AIPL1 missense mutations c.465G>T [p.(Q155H)], c.642G>C [p.(K214N)] and c.784G>A [p.(G262S)] alter AIPL1 splicing, a novel finding as alternative transcription from missense mutations is frequently overlooked as a potential disease mechanism. The AIPL1 variation c.784G>A [p.(G262S)] is considered a possible rare benign variant.^7,11 However, our in vitro splice assay revealed that transcription of c.784G>A [p.(G262S)] yields only p.E215Afs*3 and p.V249Afs*3 variants and no p.G262S. While the p.E215Afs*3 transcript is expected to be degraded by NMD, the more abundant transcript producing p.V249Afs*3 may escape NMD, resulting in truncation of the chaperone-interacting TPR domain and the loss of the PRD. Our new findings support the prediction that c.784G>A [p.(G262S)] is a rare loss-of-function disease-associated mutation. Similarly, both c.465G>T [p.(Q155H)] and c.642G>C [p.(K214N)] missense mutations alter AIPL1 transcription in vitro yielding aberrant splice products that disrupt the domain organization of the AIPL1 protein. In all cases, correctly spliced transcripts were not detected. Thus, a complex disease mechanism acting through the disruption of protein function by aberrant splicing rather than a simple amino acid substitution exists for these three missense mutations. 

Our analysis also reveals that alternative transcription of AIPL1 occurs normally in the retina, producing transcripts that lack exon 2, 3, or 4 as a result of deletion of these 0-0 phase exons. Interestingly, several of the disease-associated AIPL1 variations investigated in this study also produced transcripts lacking exon 2, 3, or 4. Both c.276+1G>A and c.276+2T>C induced in-frame skipping of exon 2 to produce p.V33_I92del (ΔEx2), c.277-2A>G and c.465G>T [p.(Q155H)] induced in-frame skipping of exon 3 to produce p.H93_Q155del (ΔEx3), and c.642G>C [p.(K214N)] induced skipping of exon 4 to produce p.V156_K214del (ΔEx4). Notably, in-frame skipping of exon 3 was also detected as a minor transcript from wild-type AIPL1 in the in vitro splice assay, and a low abundance of the transcript was also detected in vivo in normal retina. Therefore the splice variants c.277-2A>G and c.465G>T [p.(Q155H)] shift the relative abundance of this transcript in retina. Interestingly, the c.465G>T [p.(Q155H)] mutation was recently identified as a homozygous mutation in all affected members of a consanguineous family diagnosed with autosomal recessive retinal degeneration, with heterozygous carriers being unaffected.¹⁵ It is therefore possible that there is a tolerance threshold for expression of this naturally occurring alternatively spliced AIPL1 transcript, and that the expression of this transcript in the homozygous state may be associated with a less severe disease phenotype. Similarly, there may normally be a tolerance threshold for expression of the p.V33_I92del (ΔEx2) and p.V156_K214del (ΔEx4) alternative transcripts. 

Analyses of the predicted impact of aberrant splicing on the AIPL1 protein suggest that all of the AIPL1 variations investigated in this study are invariably loss-of-function mutations. To summarize (Fig. 3), alternative aberrant transcripts of AIPL1 that shift the open reading frame and lead to early premature termination are predicted to be degraded by NMD and are loss-of-function mutations. However, alternate aberrant transcripts of AIPL1 that involve in-frame exon skipping, and therefore in-frame deletion of specific domains within the AIPL1 protein escape NMD, and are likely to exhibit functional deficits associated with the domain-specific deletion. In-frame skipping of exon 2 or exon 3 disrupts the FKBP-like domain of AIPL1 leaving the TPR domain intact, whilst skipping of exon 4 primarily disrupts the TPR domain leaving the FKBP-like domain intact. The observation that skipping of exons 2, 3, and 4 normally occurs in the retina, albeit a rare occurrence, suggests that low levels of expression of alternative AIPL1 transcripts lacking exons 2, 3, or 4 are normally tolerated in the retina. However, our data suggest that the exclusive or abundant expression of alternative AIPL1 transcripts lacking exons 2, 3, or 4 as a result of splice site variations lead to pathogenesis. Variants affecting the splicing of exons 5 and 6 of AIPL1, which are out-of-phase, result in truncation or disruption of the TPR domain and PRD leaving the FKBP-like domain intact. Because the PRD and a significant proportion of the TPR domain are encoded by exon 6, many of the resultant transcripts likely escape NMD to express a faulty protein unable to interact with the molecular chaperones. 

In conclusion, our data has identified aberrant transcription of AIPL1 as a potential underlying cause of LCA. The analysis has been important in solving the molecular mechanism of disease in LCA patients harboring these variations and has increased our understanding of the role of aberrant RNA processing as a cause of LCA associated with variations in AIPL1. This opens up the possibility that a number of previously uncharacterized AIPL1 variations of unknown pathogenicity may cause aberrant splicing of AIPL1. Moreover, unidentified variations in the promoter, untranslated regions, cis-acting elements or regulatory elements in the introns or exons, such as splicing enhancers and silencers, may affect gene expression, RNA stability, or splicing and contribute to disease. It is therefore possible that the prevalence of LCA attributed to AIPL1 mutations may be higher than previously estimated. Confirmation of this proposal would require further investigation and screening of AIPL1 in patients with retinal degeneration and LCA, in combination with in-depth investigations of the effect of newly identified AIPL1 variations at the transcript and protein level. 

Acknowledgments

The authors thank Vincent Plagnol (UCL Genetics Institute, London, UK) for assistance with the RNA-seq data analysis. They also thank Alison Hardcastle (UCL Institute of Ophthalmology, London, UK) for critical review of the manuscript. 

Supported by grants from Moorfields Eye Hospital Special Trustees (London, UK), Rosetrees Trust (Edgware, Middlesex, UK), Fight for Sight (London, UK), Retinitis Pigmentosa Fighting Blindness (Buckingham, Buckinghamshire, UK). 

Disclosure: J. Bellingham, None; A.E. Davidson, None; J. Aboshiha, None; F. Simonelli, None; J.W. Bainbridge, None; M. Michaelides, None; J. van der Spuy, None