Unexpected CEP290 mRNA Splicing in a Humanized Knock-In Mouse Model for Leber Congenital Amaurosis

Leber congenital amaurosis (LCA) is the most severe form of retinal dystrophy with an onset in the first year of life. The most frequent genetic cause of LCA, accounting for up to 15% of all LCA cases in Europe and North-America, is a mutation (c.2991+1655AG) in intron 26 of CEP290. This mutation generates a cryptic splice donor site resulting in the insertion of an aberrant exon (exon X) containing a premature stop codon to CEP290 mRNA. In order to study the pathophysiology of the intronic CEP290 mutation, we generated two humanized knock-in mouse models each carrying ~6.3 kb of the human CEP290 gene, either with or without the intronic mutation. Transcriptional characterization of these mouse models revealed an unexpected splice pattern of CEP290 mRNA, especially in the retina. In both models, a new cryptic exon (coined exon Y) was identified in ~5 to 12% of all Cep290 transcripts. This exon Y was expressed in all murine tissues analyzed but not detected in human retina or fibroblasts of LCA patients. In addition, exon x that is characteristic of LCA in humans, was expressed at only very low levels in the retina of the LCA mouse model. Western blot and immunohistochemical analyses did not reveal any differences between the two transgenic models and wild-type mice. Together, our results show clear differences in the recognition of splice sites between mice and humans, and emphasize that care is warranted when generating animal models for human genetic diseases caused by splice mutations.


Introduction
Leber congenital amaurosis (LCA; OMIM 204000) is a group of rare and severe inherited retinal dystrophies with a prevalence of ~1:50,000 individuals worldwide [1,2]. The clinical characteristics of LCA include severe and early vision loss that appears in the first year of life, amaurotic pupils, sensory nystagmus and the absence of electrical signals on electroretinogram (ERG) [3]. Like other retinal disorders such as retinitis pigmentosa (RP; OMIM 268000), LCA shows a high genetic heterogeneity. Currently, mutations in 19 different genes have been identified (RetNet: https://sph.uth.edu/retnet), mainly segregating in an autosomal recessive manner. The most frequently mutated LCA gene is CEP290 that encodes the centrosomal protein 290 kDa [3,4]. The most prevalent LCA-causing mutation in CEP290, accounting for up to 15% of LCA cases in many Western countries, is a deep-intronic change (c.2991+1655AG; p.C998*) that generates a strong splice donor site which results in the insertion of a cryptic exon with a premature stop codon into ~50% of CEP290 transcripts [4][5][6].
The CEP290 gene was first isolated from human brain cDNA libraries, and encompasses 54 exons that encode a 2479 amino acid protein [12]. CEP290 has been localized to the centrosome and to the transition zone of cilia [13,14]. The exact physiological role of CEP290 remains unclear, although it has been shown that it plays an important role in the regulation of ciliary protein trafficking and cilium assembly [13,15]. In photoreceptor cells, CEP290 localizes to the connecting cilium [16], similar to approximately one third of proteins encoded by retinal dystrophy genes [17]. The connecting cilium is the transition zone of the photoreceptor sensory cilium, connecting the inner and the outer segment of rods and cones, the lightsensitive neurons responsible of the conversion of light stimuli into chemical signals that are transmitted to the brain [18].
Two naturally occurring mutant Cep290 animal models have been described to date. The murine rd16 model carries a genomic deletion encompassing exons 35-39 of Cep290, that results in an in-frame deletion of 897 bp in Cep290 mRNA. These mice show an early, fast and progressive photoreceptor degeneration, where only one row of photoreceptor cell nuclei is detected at postnatal day 30 (P30) [16]. In a subpopulation of Abyssinian cats with retinal degeneration, an intronic mutation (c.6960+9TG) has been identified. This mutation generates a new strong canonical splice donor site that inserts four nucleotides into CEP290 mRNA, causing a frameshift that results in a premature stop codon after two amino acids [19]. The first clinical symptoms appear at the age of seven months when reduced electroretinogram recordings are detected as a result of rod degeneration that is followed by cone death, leading to complete blindness at the age of 3-5 years [20].
In order to study and understand the pathophysiology of the intronic CEP290 mutation c.2991+1655AG, we have generated two humanized mouse models, in which human exon 26, intron 26 (with or without the LCA mutation) and exon 27 were inserted into the murine Cep290 gene via homologous recombination. A detailed characterization of these mice at the transcriptional level revealed unexpected splicing of CEP290 mRNA, that was only partially in line with the aberrant splicing observed in patients with CEP290-associated LCA. Our data suggest species-specific differences in the splicing machinery between mice and humans, indicating that mice may not always represent suitable models to study the pathogenic mechanisms underlying genetic variants that affect splicing.

Ethics statement
All procedures were performed after obtaining approval from the Radboud University Nijmegen ethics committee for experimental animal research (RU-DEC-2012-023), and according to the regulations of the ARVO statement for the use of animals in ophthalmic and vision research. For human skin biopsies written informed consent was gathered from all individuals in Clinical Genetics Center Nijmegen by signing the declaration of permission for the use of body material (Toestemmingsverklaring gebruik lichaamsmateriaal) of the Radboud University Medical Center and our research followed the tenets of the Declaration of Helsinki. All procedures were carried out in The Netherlands.

Generation of the transgenic models
Two animal models were generated by GenOway (Lyon, France). Briefly, the targeting vector contained the human genomic region between introns 25 and 27, either without (Cep290 hum/hum ) or with the c.2991+1655AG mutation (Cep290 lca/lca ). Neomycine (NEO) and Diphtheria Toxin A (DTA) cassettes were also subcloned for positive and negative selection, respectively. The final vector was electroporated in 129Sv/Pas ES cells. A total of 260 clones were positively selected. After validating the homologous recombination event at the 5'-and the 3'-end by PCR and southern blotting, four ES cell clones for Cep290 hum/hum and one for Cep290 lca/lca were isolated. These clones were injected in blastocysts from C57BL/6J females and injected in pseudo-pregnant females of the same strain. Based on the colour, four highly chimeric males were selected for breeding with Cre deleter females in order to excise the NEO cassette for each model. Excision was validated by PCR and Southern blot. The resulting heterozygous animals were used to obtain homozygous offspring. Wild-type C57BL/6J mice (Cep290 wt/wt ) were used as a reference in all analyses.

Animal handling, tissue dissection and preparation of the samples
Tissue samples were obtained from postnatal day 150 (P150) Cep290 wt/wt , Cep290 hum/hum and Cep290 lca/lca models. Animals were euthanized with CO 2 followed by cervical dislocation. Retina, brain, kidney, liver, spleen, testis and lung were dissected and immediately frozen in liquid nitrogen for RNA and protein isolation, whereas for cryosectioning, eyes were enucleated, embedded in OCT (Optical Cutting Temperature compound, Tissue-Tek, Sakura Finetek, Torrance, CA) and frozen in isopentane and liquid nitrogen. From each of the three models, two animals were used for RNA isolation, two for immunohystochemistry and four for protein analysis.

Cell culture of human fibroblast cell lines
Fibroblast cell lines derived from skin biopsies of individuals with CEP290-associated LCA or healthy controls, were cultured in DMEM, supplemented with 20% fetal bovine serum, 1% penicillin-streptomycin and 1% of sodium pyruvate at 37°C and 5% CO 2 . Informed consent was gathered from all individuals and our research followed the tenets of the Declaration of Helsinki.

RNA isolation
Several tissues from P150 Cep290 wt/wt , Cep290 hum/hum and Cep290 lca/lca mice were obtained for transcriptional analysis. Twenty milligrams of tissue (brain, kidney, liver, spleen, testis and lung) and a pool of two retinas from different animals of the three different mouse models were used for RNA isolation (Nucleospin RNA II, Düren, Germany). For fibroblast cell lines, RNA was isolated from ~1.5 million cells by using the same kit following the manufacturer's protocol. Human adult retina RNA was purchased from Clontech (Mountain View, CA).

RT-PCR and transcriptional analysis
One microgram of RNA (from tissue, fibroblasts or human retina) was used for cDNA synthesis by using the iScript cDNA Synthesis kit (Bio-Rad, Hercules, CA) at a final volume of 20 µl and then diluted by adding 50 µl of RNAse-free H 2 O. For tissue expression analysis, all reaction mixtures (25 µl) contained 10 µM of each primer pair, 2 µM of dNTPs, 1.5 mM MgCl 2 , 10% Qsolution (Qiagen, Venlo, Netherlands), 1 U of Taq polymerase (Roche, Penzberg, Germany) and 5 µl of diluted cDNA. PCR conditions were 94°C for 2 min, followed by 35 cycles of 20 s at 94°C, 30 s at 58°C and 30 s at 72°C, with a final extension step of 2 min at 72°C. Amplicons were analyzed by agarose electrophoresis. Semi-quantification was performed using Image J software [21]. Actin expression was used to compare and normalize the samples, primers for actin amplification were designed to amplify both the human and the mouse gene. Oligonucleotide sequences are listed in Table S1.

Western blot analysis
A pool of two retinas of different animals from each mouse model were homogenized in 200 µl of RIPA buffer (50 mM Tris pH 7.5, 1 mM EDTA, 150 mM NaCl, 0.5% Na-Deoxycholate, 1% NP40 plus protease inhibitors). Protein quantification was performed with a BCA kit (Thermo Fisher Scientific, Waltham, MA). For CEP290 detection, ~170 µg of total protein lysate supplemented with sample buffer was loaded onto a NuPage 3-8% tris-acetate gel (Life technologies, Carlsbad, CA). The electrophoresis was carried out for 4 h at 150 V. For normalization with an anti-α-tubulin antibody, ~30 µg of the same protein lysates were loaded onto a NuPage 4-12% bisacrylamide tris-glycine gel (Life technologies, Carlsbad, CA) and run for 2 h at 150 V. A protein lysate from HEK293-T cells overexpressing FLAG-tagged CEP290 was used as a positive control (40 µg and 10 µg were loaded for CEP290 and tubulin detection, respectively). All lysates were boiled for 5 min at 98°C prior to loading. Proteins were transferred to a PVDF membrane (GE Healthcare, Little Chalfont, UK) overnight at 25 V at 4°C. Blots were blocked in 5% non-fat milk in PBS, incubated overnight at 4°C with rabbit anti-CEP290 (dilution 1:750, Novus Biological, Littleton, CO) or mouse anti-α-tubulin (dilution 1:2000, Abcam, Cambridge, UK) in 0.5% non-fat milk in PBS, washed in PBST (4 x 5 min), incubated with the appropriate secondary antibodies for 1 h at room temperature (RT), washed in PBST (4 x 5 min) and developed using the Odissey Imaging System (Li-Cor Biosciences, Lincoln, NE). Western blot analysis was performed in duplo. Semiquantification was performed using Image J software [21].

Immunohistochemistry
Seven micrometer cryosections were dried for 1 h at RT, washed in PBS to remove the OCT, permeabilized for 20 min in 0.01% Tween in PBS and blocked for 30 min (0.1% ovalbumin and 0.5% fish gelatine in PBS). Primary antibody incubation was performed overnight at 4°C by diluting the antibodies in blocking solution. The next day, sections were washed in PBS (3 x 10 min), incubated for 45 min at RT with the corresponding Alexa fluor-conjugated secondary antibodies and DAPI, washed in PBS (3 x 10 min) and mounted using Prolong Gold antifade kit (Life Technologies, Carlsbad, CA). The antibodies and dilutions used were: 1:300 for rabbit anti-CEP290 (Novus Biological, Littleton, CO), 1:1000 for mouse anti-acetylated tubulin (Sigma-Aldrich, St. Louis, MO) and 1:500 for secondary antibodies Alexa Fluor 488 and 568 raised in goat against rabbit and mouse IgGs, respectively (Molecular Probes, Eugene, OR).

Generation of the humanized models
The murine Cep290 gene contains 53 exons (one less than in humans) and shares a high homology at DNA (86%) and protein (87% homology and 94% similarity) levels with human CEP290 ( Figure 1A). In order to generate a humanized knockin mouse model that would mimic the genotype and phenotype associated with the deep-intronic LCA-causing mutation, a recombination strategy was designed that consisted of the replacement of mouse exons 25 (mE25) and 26 (mE26), as well as intron 25 (mI25), by the human counterparts (hE26, hI26/hI26mut and hE27) ( Figure 1B). All recombination events were validated by PCR and Southern blot analysis (data not shown). The isolated positive ES-cell clones were injected in blastocysts that were implanted in pseudo-pregnant females. Highly chimeric males were bred with Cre deleter females to excise the NEO cassette. The resulting heterozygous mice were used to obtain homozygous animals, which were viable and fertile. Subsequent breedings were done between homozygous animals.

Transcriptional analysis of Cep290 in the humanized models
To determine whether the humanization of the murine Cep290 would compromise the general Cep290 gene expression, we determined by RT-PCR the transcriptional levels of Cep290 in two different regions of Cep290 cDNA, using actin expression as a reference (Figure 2). First, we amplified the region from exon 10 to exon 13 (Figure 2A), which is not affected by the recombination. In addition, we performed a PCR from mouse exons 24 to 27, which encompasses the humanized region (hE26-hE27) instead of mE25 and mE26 ( Figure 2B). In both cases, no differences in the amount of Cep290 transcripts were observed between the two models (Cep290 lca/lca and Cep290 hum/hum ) and the wild-type mice (Cep290 wt/wt ), indicating that the human exons did not alter the expression levels of Cep290 in mouse. Intriguingly however, especially in the retina, additional bands were detected in both the Cep290 hum/hum and the Cep290 lca/lca model but not in wild-type mice. In lymphoblastoid cell lines derived from LCA patients homozygously carrying the intronic CEP290 mutation, two different CEP290 transcripts are detected in equal ratios, the correctly spliced product and an aberrant one, that contains the 128-bp cryptic exon X that introduces a premature stop codon [22]. In the Cep290 lca/lca model, some minor bands were detected, that could represent transcripts which include the 128-bp-exon X. However, additional bands were also observed in the Cep290 hum/hum mice ( Figure 2B), suggesting new unexpected splice variants caused by the humanization of the murine Cep290 locus.
To investigate these aberrant splicing events in more detail, we used primers that were located in the human exons (hE26 and hE27). Due to the high homology of these two exons between both species (E26: 93% and E27: 87%), we designed the primers in the regions with the lowest sequence homology to ensure amplification of only the human exon. As expected, the two humanized models, but not the wild-type mice, showed robust amplification of the human exons. However, also in wildtype mice, some products were detected in the tissues where Cep290 is most highly expressed, such as retina, kidney or testis, likely due to the high similarity between the human and mouse CEP290 exons ( Figure 2C). In line with the results found using primers in mE24 and mE27, again, aberrant splice products were detected in the retinas of both humanized mouse models ( Figure 2C).

Expression of CEP290 cryptic exons in humans and mice
The fact that the Cep290 hum/hum model also showed evidence of aberrant Cep290 splicing, whereas only one transcript was expected ( Figure 3A), prompted us to study the nature of these aberrant splice products in depth. Sanger sequencing analysis revealed the presence of a new exon of 116 bp (coined exon Y), which was found to be also located in human intron 26, downstream of exon X (starting position exon Y: c.2992-1603). RT-PCR analysis using a primer located in exon Y allowed us to observe that Cep290 transcripts encoding exon Y were expressed in almost all tissues, but most prominent in the retina ( Figure 3B), both in the Cep290 hum/hum and in the Cep290 lca/lca model. When using a primer located in exon X, two bands were observed derived from the retina of Cep290 lca/lca mice ( Figure 3C). Sequence analysis revealed that the most prominent band corresponded to a Cep290 transcript containing both exons X and Y. Cep290 transcripts that contained only exon X were also found, but expressed at very low levels that were hardly detectable. In total, we were able to identify one transcript in the Cep290 wt/wt animals, two in the Cep290 hum/hum and four in the Cep290 lca/lca model ( Figure 3D). Semi-quantification of all aberrant splicing events revealed that in the Cep290 hum/hum model, approximately 5% of Cep290  Table S1). doi: 10.1371/journal.pone.0079369.g001 transcripts contained exon Y, whereas in the Cep290 lca/lca model, ~14% of all Cep290 transcripts contained aberrant exons, i.e. exon Y, exon X or both ( Figure 3D). Following the identification of these unexpected splicing events, we investigated whether exon Y is a naturally occurring exon in human tissues. For that purpose, RNA from human retina as well as from fibroblast cells from healthy and LCA (homozygously carrying the CEP290 intronic mutation) individuals were used, together with retinal RNA from the Cep290 wt/wt and Cep290 lca/lca mice. RT-PCR analysis revealed that transcripts containing exon Y were not detected in healthy retinas nor in fibroblasts derived from individuals with CEP290associated LCA (Figure 4). Moreover, we observed again that the transcripts containing exon X are less abundant than those containing exon Y in Cep290 lca/lca retinas (Figure 4).

Western blot and immunohistochemical analysis of mouse retinas
The insertion of exon X, exon Y, or both, to the regular Cep290 transcript, in all cases is predicted to cause a frameshift and premature termination of the Cep290 protein.
To determine the effect of these various splicing events on the amount of CEP290 protein levels, Western blot analysis was performed using retinal protein lysates from our mouse models. Large amounts (>150 µg) of total protein were required to robustly detect CEP290. Western blot analysis revealed no significant differences among the three models ( Figure 5), indicating that the inclusion of any or both cryptic exons to only a small percentage of Cep290 transcripts does not significantly alter Cep290 protein levels.
Finally, retinal sections from Cep290 wt/wt , Cep290 hum/hum and Cep290 lca/lca mice at P150 were analyzed by immunohistochemistry. The overall morphology of all retinas was similar, as were the amounts of photoreceptor cell nuclei in the different models. In addition, CEP290 was correctly localized at the connecting cilium of photoreceptors in all three models (Figure 6), illustrating that the low abundance of aberrant Cep290 splicing also does not appear to affect the structural integrity of the retina.

Discussion
Rodents, in particular mice, have been widely used in the last decades to shed light on the pathophysiology of several diseases as well as for the development of therapeutic trials. In the field of vision science, they are particularly useful because of their relatively short lifespan, which allows a quick read-out of the progression of retinal degeneration. In addition, there are several naturally occurring retinal degeneration models [23,24], the interactions between retinal cells are similar in humans and mice [25], and more importantly, the genetic manipulation and subsequent phenotypic characterization are very well established. Here, we generated a transgenic CEP290 knock-in model by introducing the deep intronic mutation that causes LCA in humans [5] and its flanking exons to elucidate the mechanisms that cause photoreceptor cell death in these patients. In addition, mimicking the splice abnormalities that are observed in LCA patients in a mice model would allow to determine the efficacy of therapeutic interventions, like for instance antisense oligonucleotide-based rescue of aberrant CEP290 splicing that was shown to be effective in patient cell lines by us and others [22,26]. The high homology between human and mouse exons allowed the generation of the animal model without changing the open reading frame of Cep290 and only slightly altering the amino acid composition of CEP290. In addition, the naturally occurring Cep290 mouse model (rd16) shows a clear earlyonset retinal phenotype, indicating an essential role for this gene in the mouse retina [16]. As expected, the expression levels of the Cep290 gene were not altered in our humanized  mouse models, with highest expression shown in retina and testis. However, when the composition of Cep290 transcripts was analyzed in detail, we did not detect the same ratio of correctly and aberrantly spliced transcripts (approximately 1:1) as detected in cell lines of LCA patients carrying the intronic CEP290 mutation [22]. Instead, unexpected aberrant splicing events were detected, that included the recognition of a novel exon Y, in addition to exon X. In total, in the Cep290 lca/lca model, four different transcripts were observed, including two that contained the aberrant exon X associated with LCA ( Figure  3D). Nevertheless, the amount of all aberrantly spliced transcripts collectively was less than 15% of the total Cep290 expression, which is most probably not enough to compromise retinal function. This is further supported by the absence of significant differences in CEP290 protein levels between the models assessed by Western blot analysis, and of any abnormalities in retinal structure following (immuno)histochemical analysis of retinal sections.
In the past, others have successfully used humanized knockin mouse models to mimic the pathophysiological mechanisms underlying splice site mutations for several inherited diseases [27][28][29][30]. In this study, even though the humanization of mouse Cep290 was not compromising the gene expression in general, the splicing machinery of the mouse appeared not to be able to recognize the human splice acceptor site or the splice donor site that is generated by the intronic CEP290 mutation, suggesting differential recognition of splice sites in humans vs. mice. In addition, the substitution of the murine intron by the human one generated another cryptic exon, robustly expressed in the mouse retina, but not found in human retinas nor in fibroblasts. Again, these results point to differences in the splicing machineries in mice vs. humans, as it has already been suggested for other retinal dystrophy genes [31]. One example is CERKL, which shows a high transcriptional complexity in human and mouse retina, showing different alternative splice events between species in addition to the use of alternative promoters [31]. In addition, it has been showed that the recognition of splice sites are species-, regulator region-and RNA-structure-dependent [32], suggesting that the human exon X might have an inaccessible structure for the murine spliceosome or that some enhancers favor the insertion of exon Y in mice.
In the last decade, the use of other non-murine animal models has been regularly preferred to study retinal dystrophies, for a number of reasons. It is well known that like in humans, primate and raptor retinas contain a cone-enriched zone called fovea, whereas in mice the cones are homogeneously distributed throughout the retina [18], thus less reminiscent of the human situation. In addition, several genes in which mutations are causative for retinal dystrophy in humans do not have an orthologue in rodent species, like for instance EYS [33,34]. Other mutant mouse models do not show any phenotype, potentially because of functional redundancy. For example, the GCAP1 (GUCA1A) and GCAP2 (GUCA1B) proteins are guanylate cyclases activators involved in phototransduction. Although in humans, mutations in either of these genes can cause retinal dystrophy, mice only show a phenotype when both genes are knocked out at the same time [35,36]. Occasionally, the protein might not be localized in the expected cell type, as is the case with the aforementioned CERKL gene, which in mouse is mainly expressed in ganglion cells rather than in photoreceptors [37]. Other genes that, when mutated in mice, do not resemble the human phenotypes are USH2A and RDH12. In Ush2a -/-mice retinal degeneration only appears at very late stages [38], whereas in Rdh12 knock-out mice, no retinal phenotype is observed at all [39]. Taken together, these examples suggest that proteins involved in visual function might not play the same roles or are not involved in the same critical functions in humans vs. mice. The question that now arises is which is the best model to study retinal dystrophies. Although there are models with anatomically more similar eyes (i.e. dogs or primates), in these medium-and large-sized animal models, genetic manipulation is practically impossible due to experimental, financial and ethical restrictions. In the past, other naturally occurring animal models have enabled the study of the retinal degeneration and the start of the development of new therapies, such as RPE65 gene augmentation therapy in dogs [40]. However, compared to mice, the disease is only manifested after several years and also the progression is slower.
In summary, we have generated and characterized a humanized transgenic mouse model that carries the prevalent intronic Cep290 mutation causative for LCA. Despite the recognition of expected and unexpected cryptic exons, the total amount of aberrantly spliced Cep290 transcripts was less than 15%, and thereby does not mimic the aberrant CEP290 splicing observed in individuals with CEP290-associated LCA.
Our results indicate that the mouse may not be a suitable model to study the pathophysiology of the intronic CEP290 c. 2991+1655AG mutation, due to differential recognition of splice sites between humans and mice. Together, our results emphasize that caution is warranted when generating animal