Splicing-Correcting Therapeutic Approaches for Retinal Dystrophies: Where Endogenous Gene Regulation and Specificity Matter

Abstract

Splicing is an important and highly regulated step in gene expression. The ability to modulate it can offer a therapeutic option for many genetic disorders. Antisense-mediated splicing-correction approaches have recently been successfully exploited for some genetic diseases, and are currently demonstrating safety and efficacy in different clinical trials. Their application for the treatment of retinal dystrophies could potentially solve a vast panel of cases, as illustrated by the abundance of mutations that could be targeted and the versatility of the technique. In this review, we will give an insight of the different therapeutic strategies, focusing on the current status of their application for retinal dystrophies.

Introduction

Retinal dystrophies are an extremely diversified group of genetic diseases all characterized by visual dysfunctions that can lead to blindness in the worst cases. Today there are more than 200 genes responsible for syndromic and nonsyndromic retinal dystrophies, ¹ each of them carrying several types of mutations leading to very different clinical phenotypes. The development of different gene therapy approaches has given a hope for the implementation of therapies for these otherwise incurable conditions. Messenger RNA splicing is an extremely complex and fundamental cellular process that has been so fare barely considered as a therapeutic target, ^2,3 even if it can be seen as a highly appealing one for its importance in the cell context. The ability to modulate splicing can in fact offer several advantages over other conventional gene replacement approaches, especially in the context of retinal dystrophies. By definition, antisense-based therapeutic approaches act following base-pairing with their mRNA target, thus giving the possibility of obtaining a great specificity of action. Since they act at the mRNA level, the endogenous transcriptional regulation of the target gene is always maintained. This means that the therapeutic effect is obtained only where and when the target pre-mRNA is present. In a highly specialized and organized tissue like the retina, it is particularly important to maintain endogenous gene regulation. Therapeutic interventions for delicate processes like the phototransduction cascade would require the preservation of this control for desirable outcomes. ⁴ Splicing-correction approaches also allow a fine-tuning over the relative abundance of splicing isoforms because, by acting at a pre-mRNA level, it is relatively easy to modulate their ratio. The availability of several different molecular tools that can be used to manipulate splicing renders these approaches a versatile and promising strategy for the multitude of retinal dystrophies known today. Moreover, the retina has some characteristics that make it a perfect target tissue for those therapies. First of all, it is an easy tissue to access, and different drug delivery routes are in use today. ^5,6 Being that the eye is relatively small, enclosed, and separated from the rest of the body by the blood-brain barrier, it minimizes the required dose and systemic dissemination of the therapeutic agent, thus avoiding possible complications due to systemic side effects of the therapy. ⁷ The eye is also an immune-privileged organ, limiting potential immune response to the delivered agent. ⁸ The retina is composed by nondividing cells, thus it is easier to induce prolonged effects or transgene expression, without the need of using integrating vectors. Regarding gene delivery, the presence of different adeno-associated viral (AAV) vector serotypes able to efficiently and stably transduce all retina layers ⁹ is a great advantage for splicing-modulating genetic tools. 

The mRNA Splicing Process

Once transcription of a gene begins in the nucleus, the transcript undergoes a complex series of cotranscriptional processes all devoted to the production of a mature mRNA, collectively dubbed “mRNA processing.” One of these events, called mRNA splicing, consists in the removal of intervening sequences (“introns”) and the joining of the coding portions of the transcript (“exons”). Messenger RNA splicing is a major way by which the cell can induce transcriptional diversity, mainly through alternative splicing, and apply a fine control on this diversity. The proper recognition of introns and exons is mediated by cis-acting sequences and trans-acting factors. The principal cis-acting elements that spatially organize the splicing reaction, consist in the splice donor (DS) site, the polypyrimidine tract (Py), the branch-site (BS), and the splice acceptor (AS) site. There are also other cis-acting sequences that are fundamental for mRNA splicing ¹⁰ : exonic splicing enhancers (ESE) or silencers (ESS) that enhance or inhibit recognition of the exon in which they lay; intronic splicing enhancers (ISE) or silencers (ISS), intronic sequences that promote or suppress recognition of the nearby exons. Trans-acting factors are instead several proteins and ribonucleoproteins able to recognize the different cis-elements. Small nuclear RNA (snRNA) are constitutive components of the small nuclear ribonucleoproteins (snRNP) U1, U2, U4, U5, U6, and allow them to base-pair with different cis-acting sequences mediating the cascade of events leading to the splicing reaction. For example, the U1 snRNP recognizes the DS site, whereas U2 binds to the branch site. The other two groups of trans-acting splicing factors are represented by heterogeneous nuclear ribonucleoproteins (hnRNPs), that have mainly a repressive function, and by serine- and arginine-rich (SR) proteins, that play an important role in splicing regulations by mainly binding to ESE and ISE, thus promoting splicing. ^11–15 All these factors assemble together in a precise temporal sequence in a complex called spliceosome, the cellular machinery devoted to the splicing process. For a more exhaustive description of the splicing process, we refer the reader to more detailed reviews. ^16–18  

Mutations Leading to Retinal Diseases

Retinal dystrophies are caused by mutations in many different genes, leading to a multitude of disease conditions ^1,19 (Table). Today there are 219 genes identified as causative of retinal diseases, and the number is still growing. ¹ The total number of known mutation for 208 of this genes, annotated in the Human Gene Mutation Database (provided in the public domain by HGMD Professional, http://www.hgmd.cf.ac.uk/ac/index.php), is 13.668. ²⁰ The large majority of those are represented by missense mutations, accounting for 34% of the total (Fig. 1). Small deletions follow as the second most abundant type of genetic defect (16%). Bona fide splicing mutations represent an 11% of the total. Generally, mutations residing in introns are categorized as splicing mutations because the amino acid sequence of the protein is not altered, thus the problem most likely concerns proper splicing. These mutations can be located in any of the cis-acting elements present in introns. But splicing mutations can also be found in the exons, altering or not the coding sequence. In this case, their identification as a splicing mutation is much more difficult, as it requires analysis of the splicing pattern. Today it is believed that more than 25% of mutations, normally categorized as missense, nonsense, or silent, actually act by altering the splicing pattern. ^21–23 Their effect can be the disruption of a cis-acting sequence, or the formation of a new one, resulting in exon skipping, intron retention, or use of alternative DS and AS sites. Correct identification of these mutations is of pivotal importance for the development of therapeutic approaches. Aside from splicing mutations, antisense-mediated splicing-correction approaches can potentially be utilized for the correction of missense and nonsense mutations, as well as for small insertions and deletions. In all cases where a mutation causes the introduction of a stop codon or frameshift leading to a premature termination of the transcript, the possibility to interfere with the proper recognition, by the splicing machinery, of the exon carrying the mutation (therapeutic exon skipping) can be the right strategy to follow. The result of this approach is a shorter mature mRNA, missing the portion encoded by the skipped exon, but resulting in a restored ORF. It is then necessary to assess the functionality of the rescued smaller protein. This strategy is more easily applicable to genes where the mutated exon encodes for a repetitive structural element whose loss in the final protein product is less likely to cause structural and functional defects of the protein itself, whereas it is a riskier approach in other cases, where proper function of the skipped protein is less predictable. 

Figure 1

 

Mutation pattern of genes causing retinal diseases.

Table

 

List of Genes Causing Retinal Diseases

Disease Category	Involved Genes
Cone or cone-rod dystrophy/dysfunctions	ABCA4 ADAM9 AIPL1 BBS12 C2orf71 C8orf37 CA4 CABP4 CACNA1F CACNA2D4 CDHR1 CERKL CNGA3 CNGB3 CNNM4 CRB1 CRX GNAT2 GUCA1A GUCY2D KCNV2 MERTK MKS1 NR2E3 NRL OPN1LW OPN1MW PDE6C PDE6H PITPNM3 PROM1 PRPH2 RAB28 RAX2 RDH12 RIMS1 RLBP1 RPE65 RPGRIP1 TULP1 UNC119
Retinitis pigmentosa	ABCA4 ARL2BP ARL6 BBS1 BEST1 C2orf71 C8orf37 CA4 CEP290 CERKL CLRN1 CNGA1 CNGB1 CRB1 CRX CYP4V2 DHDDS EMC1 EYS FAM161A FSCN2 GPR125 GRK1 GUCA1A GUCA1B GUCY2D IDH3B IMPDH1 IMPG2 KIAA1549 KLHL7 LCA5 LRAT MAK MERTK MFRP MYO7A NR2E3 NRL OFD1 PDE6A PDE6B PDE6G PRCD PROM1 PRPF3 PRPF31 PRPF6 PRPF8 PRPH2 RBP3 RDH12 RGR RHO RLBP1 RP1 RP1L1 RP2 RP9 RPE65 RPGR RPGRIP1 SAG SEMA4A SNRNP200 SPATA7 TOPORS TULP1 USH2A VCAN ZNF513
Leber congenital amaurosis	AIPL1 BBS4 BEST1 CABP4 CEP290 CNGA3 CRB1 CRX DTHD1 GUCY2D IMPDH1 IQCB1 KCNJ13 LCA5 LRAT MERTK MYO7A NMNAT1 NRL RD3 RDH12 RPE65 RPGRIP1 RPGRIP1L SPATA7 TULP1
Macular dystrophy/degeneration	ABCA4 ABCC6 BEST1 CNGB3 CRX EFEMP1 ELOVL4 GUCY2D PAX2 PROM1 PRPH2 RP1L1 TIMP3
Stargardt disease	ABCA4 ELOVL4 PRPH2 CFH HMCN1
Age-related macular degeneration	ABCA4 ARMS2 BEST1 C3 CFH ELOVL4 ERCC6 FBLN5 HMCN1 HTRA1 RAX2 SLC24A1
Stationary night blindness	CACNA1F CABP4 GNAT1 GPR179 GRK1 GRM6 LRIT3 NYX PDE6B RHO SAG SLC24A1 TRPM1
Color blindness	CNGA3 CNGB3 GNAT2 OPN1LW OPN1MW OPN1SW PDE6C PDE6H
Usher syndrome	ABHD12 CACNA1F CDH23 CIB2 CLRN1 DFNB31 GPR98 GUCY2D HARS LRAT MYO7A PCDH15 PDZD7 TRIM32 USH1C USH1G USH2A
Chorioretinal atrophy/degeneration	ABCA4 CRB1 TEAD1
Retinal dystrophies/dysfunctions/degeneration	ABCC6 ABCA4 ADAMTS18 AIPL1 BEST1 C1QTNF5 CAPN5 CDHR1 CERKL CHM CRB1 CYP4V2 FZD4 GUCA1B KCNV2 LRAT LRP5 MERTK NDP NR2E3 NRL OTX2 PANK2 PLA2G5 PROM1 PRPH2 RD3 RDH12 RDH5 RGS9 RGS9BP RLBP1 RPE65 SLC24A1 TSPAN12
Retinopathy of prematurity	LRP5 NDP FZD4
Optic atrophy/aplasia	MFN2 OPA1 OPA3 OTX2 SLC24A1 TMEM126A WFS1
Wagner syndrome	VCAN COL2A1
Bardet-Biedl syndrome	ARL6 BBS1 BBS10 BBS12 BBS2 BBS4 BBS5 BBS7BBS9 CEP290 LZTFL1 MKKS MKS1 RPGRIP1L SDCCAG8 TRIM32 TTC8 WDPCP
Other systemic/syndromic diseases involving the retina	ABCC6 ABHD12 ADAMTS18 AHI1 ALMS1 ATXN7 CC2D2A CDH3 CEP290 CISD2 CLN3 COL11A1 COL2A1 COL9A1 ERCC6 FLVCR1 GNPTG IFT140 INPP5E IQCB1 ITM2B JAG1 KIF11 LRP5 NPHP1 NPHP4 OFD1 OPA3 OTX2 PANK2 PAX2 PEX1 PEX2 PEX7 PHYH RBP4 RPGRIP1L SDCCAG8 TIMM8A TMEM237 TREX1 TTPA TTPA USH1C WFS1

 All RetNet identified genes have been searched and allocated to disease categories according to HGMD Professional entries from December 2013.

Antisense Oligonucleotides

A versatile tool to target splicing is represented by antisense oligonucleotides (AONs). These are chemically synthesized molecules, generally around 20 nucleotides long, able to mimic the RNA structure and bind, by reverse complementarity, to specific cellular RNA targets. Even if they are normally used to block mRNA translation or to degrade mRNA by RNase H-mediated cleavage, these effects are unwanted when the goal is to interfere with splicing. By the use of several different chemistries ^24,25 available today for their design, it is in fact possible to direct AONs toward splicing relevant sequences on the pre-mRNA, masking them, and to avoid RNase H activity after binding. The first splice-switch oligonucleotides used a phosphothioate linkage to join nucleosides (DNA-PS). ²⁶ However, these AONs were retaining undesired RNAse H activity. ²⁴ A second generation of oligonucleotides was created from the DNA-PS structure. Inclusion at the 2′ oxygen of a methyl (2′OMe) or a methoxyethyl (2′MOE) protecting group to increase oligonucleotides resistance to degradation and block RNAse H activity originates 2′OMe-PS ²⁷ and 2′MOE-PS ²⁸ ribonucleosides, respectively. The third generation comprised locked nucleic acids (LNA), peptide nucleic acids (PNA), and phosphorodiamidate morpholino oligonucleotides (PMO). Locked nucleic acids derive from the addition of a methylene bridge between the 2′ oxygen and the 5′ carbon on a DNA-PS backbone. ^29,30 Locked nucleic acids have an increased affinity to target RNA and do not activate RNAse H. In PNAs the DNA backbone is instead substituted by a peptide-like mimicry. ³¹ Peptide nucleic acids do not cause RNAse H degradation of their target and show strong affinity to it. Since they are neutrally charged, addition of a lysine residue is commonly used to increase their water solubility and cell-uptake. Phosphorodiamidate morpholino oligonucleotides derive from the substitution of the ribose rings with morpholine ones, and their joining by phosphorodiamidate groups. ³² Phosphorodiamidate morpholino oligonucleotides are neutrally charged oligonucleotides that do not activate RNAse H and are lowly susceptible to degradation. For a comprehensive understanding of these chemistries, we refer the reader to more specific reviews. ^24,25,33,34  

In order to be able to regulate splicing, AONs must gain access to the cell nucleus. There are different splicing regulatory sequences that have been so far targeted with AONs on pre-mRNA to achieve splicing modulation (Fig. 2). For example, by targeting splicing enhancers or silencers, it is possible to induce respectively exon skipping or exon retention by blocking access of splicing factors to their target sites. Another common target of AONs are splice sites that, when bound by an AON, are no longer free to take part in the splicing reaction, thus obliging the spliceosome to use alternative “downstream” sites, again inducing exon skipping. In few cases, AONs have been also engineered to carry an additional tail containing cis-acting sequences that can be bound by splicing factors able to enhance ^35–39 or silence ^40–43 splicing of specific exons. In this way, apart from their antisense activity toward a splicing-relevant sequence, these bifunctional AONs can induce additional effects depending on the sequence they carry on the tail. Since AONs act by base-pairing, they are generally believed to allow a high specificity of action for their desired target. No undesired misspliced products of the target gene or of chosen unrelated genes were in fact detected when investigated after therapeutic application of AONs. ^44,45 Even if these findings are not generalizable and proper design of the antisense molecule should always be considered, they underline the potentiality of AONs in the context of target selectivity. Splicing modulation finds its more advanced application in the cure of Duchenne muscular dystrophy (DMD). Duchenne muscular dystrophy is an X-linked recessive disease caused by mutations in the dystrophin gene. Dystrophin is an important cellular protein whose main role in muscle fibers is the connection of the cellular cytoskeleton with the extracellular matrix. Different mutations in the 79 exons of the gene cause protein truncation due to the loss of the open reading frame. The majority of these mutations can be addressed by exon skipping. ⁴⁶ The commonly mutated exon 51 has been the first target for exon skipping. In the clinical trials completed so far for exon 51 skipping, the different chemistries applied (2′OMe-PS, PMOs) showed overall efficacy and absence of serious adverse effects. ^44,47–49  

Figure 2

 

Possible AONs targets to induce splicing modulation. Schematic representation of cis-acting sequences that are possible target of AONs. Cis-acting sequences that promote exon recognition are reported in green, whereas sequences that suppress it are highlighted in red. Antisense oligonucleotides designed to induce exon skipping are shown in the bottom part. Antisense oligonucleotides designed to promote proper exon inclusion are shown in the upper part.

A phase I clinical trial using a 2′MOE-PS oligonucleotide for splicing modulation has also been recently completed for spinal muscular atrophy (SMA), and a phase II trial has recently started. ⁵⁰ Mutations in the survival motor neuron 1 (SMN1) gene are causative of the disease. In humans, SMN1 has a paralogous, named SMN2. The two genes are identical, apart from a silent mutation in exon 7 of SMN2. This mutation, however, causes exon 7 of SMN2 to be less recognized by the splicing machinery. If exon 7 is not included, the protein is truncated. So a therapeutic strategy is to mask exon 7 ESS using AONs, thus promoting exon 7 inclusion, which results in the production of a functional full length SMN protein from SMN2 that can compensate for mutations on SMN1. 

Antisense oligonucleotides are known to be able to target all retinal layers following intravitreal, subretinal, or topical administration. ^51–56 They have long been used to elicit RNAse H degradation or to block transcription in the eye for several different diseases, having been applied, for example, against cytomegalovirus (CMV), herpes simplex virus (HSV), vascular endothelial growth factor (VEGF), transforming growth factor-β (TGF-G), fibroblast growth factor (FGF). ^5,57  

An example of the use of splice-switch oligonucleotides in the eye is that of vascular endothelial growth factor receptor 2 (KDR). The KDR gene has two distinct products: membrane-bound KDR (mbKDR), that is prohemangiogenic, and soluble KDR (sKDR), antilymphangiogenic. Soluble KDR needs the recognition of an alternative polyadenilation site on KDR intron 13 to be translated. By intravitreal administration of PMO directed against murine Kdr exon 13 DS site, it was possible to increase the sKDR/mbKDR ratio at mRNA and protein level in the retina and vitreous, interfering with the spliceosome ability of mediating intron 13 splicing ⁵⁸ (Fig. 3). This resulted in a block of hemangiogenesis and lymphangiogenesis in a model of choroidal neovascularization and corneal injury. ⁵⁸ Another recent example of the use of AONs as splicing regulators to treat a retinal dystrophy is the one of the gene centrosomal protein 290kDa, (CEP290). Centrosomal protein 290kDa mutations are responsible for ∼15% ^59,60 of Leber congenital amaurosis cases, as well as for other genetic diseases such as Joubert syndrome, Senior-Løken syndrome, Meckel-Gruber syndrome, and Bardet-Biedl syndrome. A transition on intron 26 (c.2991+1655A>G) is among the most common mutations of CEP290. ⁶¹ The mutation introduces a new DS site on intron 26, causing an aberrant exon to be included in the mature messenger RNA between exon 26 and 27. This aberrant exon carries a stop codon, resulting in a premature truncation of the protein. By the design of 2′OMe-PS directed toward predicted ESE sequences at the 3′ of the aberrant exon it was possible to demonstrate, on patient fibroblast, its skipping from the mature mRNA, so to efficiently restore proper splicing between exon 26 and 27 ⁶² (Fig. 4). Another study by Gerard X and colleagues, ⁶³ by using 2′OMe-PS targeting a different predicted ESE sequence, came to a similar result. Moreover, they were able to show an increase of full length protein levels in patient fibroblast following AON administration, as well as a faster ciliation. 

Figure 3

 

Antisense oligonucleotides approach for KDR. Scheme of action of AON against KDR exon 13 DS site: normally mbKDR, originating from intron 23 splicing, is more abundant than sKDR—that is, instead generated by the use of an alternative polyadenylation site in the retained intron 13. By interfering with the E13 DS site it was possible to increase the sKDR form, and decrease the mbKDR one.

Figure 4

 

Antisense oligonucleotides approach for CEP290. (A) Proper joining of CEP290 exon 26 and 27 (green lines and arrow) is impaired by a mutation in intron 26 (red star). The mutation causes the aberrant inclusion of a cryptic exon in a portion of the mature mRNA (red arrows). (B) Using different AONs (black lines) it was possible to increase the fraction of correctly spliced mRNA (black arrows).

Chimeric and Adapted snRNAs

The use of AONs as a therapeutic approach for genetic diseases poses one major problem. Their effect is time-limited, so to have a durable effect, repeated administration is required. In this view, the use of engineered snRNAs offer a major advantage, as they can be delivered in expression cassettes in the same way as it is done in conventional gene replacement therapies. By using viral or nonviral delivery systems, it is in fact possible to transduce or transfect target cells, and then produce the snRNA exploiting endogenous transcription. Like for AONs, one of the advantages of this class of RNA molecules is their specificity of action, as undesired activity of snRNAs has not been reported so far. ^64,65 Today there are two classes of snRNAs that have been successfully modified to be able to modulate splicing: U1 and U7. The first step of spliceosome assembly is mediated by U1 recognition of the DS site. ⁶⁶ U7 snRNA is instead not involved in splicing, but in the processing of the 3′ end of histone mRNA. ⁶⁷ They both can be used to manipulate splicing exactly as AONs. Antisense U1 and U7 snRNA have been applied for masking cis-acting sequences, thus inducing therapeutic exon skipping, for Duchenne muscular dystrophy. ^68–70 Bifunctional U7 snRNA, acting in a similar way as bifunctional AONs, have also been designed for DMD ⁷¹ and SMA. ^72,73  

Small nuclear RNAs can also be applied to a specific set of mutations not targetable by AONs. When a mutation disrupts a DS site, it leads to complete or partial loss of the splicing machinery's ability to recognize it. The design of mutation-adapted U1 snRNA able to interact by base-pairing with the mutated splice site can reestablish spliceosome recognition. ^64,74 This is possible by exploiting U1 natural function in splicing. Unfortunately, the same strategy is so far not applicable in a similar way to mutations of the AS site, It is also possible to engineer any viral vector with even a combination of different snRNAs, by taking advantage of their limited size. ⁷⁰ Tanner and colleagues ⁷⁵ applied mutation-adapted U1 snRNAs to rhodopsin (RHO), one of the genes responsible for autosomal dominant retinitis pigmentosa. An exonic point mutation interfering with the DS site was found responsible for exon 4 missplicing. By using minigenes as reporter systems in COS 7 cells, they were able to show rescue of exon 4 proper recognition with an efficiency of around 90% after treatment with mutation-adapted U1 snRNAs ⁷⁵ (Fig. 5). The same strategy was used for a splice donor mutation of RPGR inducing exon 10 skipping. Proper inclusion of exon 10 was achieved in patient fibroblast using mutation-adapted U1snRNAs ⁷⁶ (Fig. 5). Mutations in Bardet-Biedl syndrome (BBS) 1 result in more than 20% of cases of BBS, a ciliopathy characterized by retinal dystrophy, cognitive impairment, obesity, polydactyly, hypogonadism, and renal disease. ⁷⁷ Bardet-Biedl syndrome 1 is a member of a protein complex called BBSome, involved in trafficking of vesicles to the cilia. ⁷⁸ Schmidt and collaborators ⁷⁹ identified in a family affected by BBS a splice donor mutation on exon 5 of BBS1 causing missplicing. They were able to show, after administration of mutation-adapted U1 snRNAs, restoration of proper splicing in COS-7 cells, using minigenes as splicing reporter. Similar results were obtained in patient-derived fibroblast transduced with lentiviral vectors encoding for the modified U1-snRNAs (Fig. 6). A recent innovative approach has been developed for the treatment of mutations occurring at position +5 of DS sites. ⁶⁵ The synergic use of both mutation-adapted U1 and U6 snRNAs was sufficient to achieve efficient correction of aberrant splicing caused by BBS1 mutations, whereas the only use of mutation-adapted U1 snRNAs resulted in low levels of splicing correction. 

Figure 5

 

Mutation-adapted U1 snRNA for RHO and RPGR DS site mutations. The altered splicing pattern caused by the different mutations is reported in red. The correct splicing pattern is reported in green. The antisense sequence of the best U1 snRNA used to correct the effect of each mutation is reported. (A) Mutation at the last base of exon 4 of RHO causes skipping of the exon or missplicing due to the use of an alternative DS site. (B) An intronic mutation affecting the DS site of exon 10 of RPGR leads to exon 10 skipping.

Figure 6

 

Mutation-adapted U1 snRNA for BBS1 DS site mutations. (A) Normal splicing of BBS1 (green) is altered (red) by a mutation at the end of exon 5. The mutation causes exon 6 skipping (bottom left) or intron 5 retention (bottom right). (B) Black arrows represent the correcting effect of the best mutation-adapted U1 snRNA, able to partially restore proper splicing. Arrows dimension represent the amount of the different splicing products.

Trans-Splicing

Another correction approach that acts at the splicing level is spliceosome-mediated RNA trans-splicing (SMaRT). This technology is based on a cellular process, called trans-splicing. First discovered in trypanosome, ^80,81 trans-splicing has been also described in mammals. ^82,83 It consists in the ability of two different pre-mRNAs to originate a chimeric mature mRNA following a recombination event during splicing. Trans-splicing can be exploited to correct aberrant mRNA by using an artificial RNA sequence, called pre–trans-splicing molecule (PTM). The PTM consists of a correct portion of the cDNA of the gene of interest, flanked by a region containing all important element for splicing and the binding domain (BD), important for specific binding of the PTM to the target endogenous pre mRNA, mainly on an intronic sequence. There are three types of PTMs that can be exploited to achieve 5′ trans-splicing, 3′ trans-splicing or internal exon replacement, correcting respectively the 5′, the 3′, or a central region of a transcript (Fig. 7). Pre–trans-splicing molecules are delivered in expressing vectors as in a normal gene transfer approach. They have been administered in vivo using different viral vectors, ^84–86 or nonviral delivery sistems. ^87–89 The peculiarity of this technique, compared with other splicing-correction approaches, is the fact that it is mutation-independent, thus the same PTM can be used to treat different mutations located in the same region of the transcript. 

Figure 7

 

Schematic representation of the three possible trans-splicing approaches. The different PTMs are constituted by a region harboring: splicing cis-acting sequences, shown in green; the coding sequence in blue; and the BD. The mature mRNA resulting from the three trans-splicing approaches is shown with the endogenous sequence in orange, and in blue the sequence introduced by the PTMs.

Even if SMaRT has never been applied so far for the correction of a genetic disease of the retina, it has been successfully tested in several in vivo models for spinal muscular atrophy, ⁸⁸ hemophilia A, ⁹⁰ hyper-IgM X-linked immunodeficiency, ⁸⁶ and tauopathy. ⁸⁴ As there are already exhaustive reviews ^91,92 about the first three in vivo approaches, to give an example of possible application of SMaRT, we spend a few words on the last and more novel one regarding tauopathy caused by mutations in MAPT, the gene encoding tau protein. Tau, a protein important for microtubule stabilization in the CNS, is subject to active alternative splicing as it is present in humans with six different isoforms. ⁹³ Tau exon 10 encodes for a tandem repeat and by alternative splicing originates two sets of different tau isoforms: with 4 (4R; +E10) or 3 (3R; −E10) tandem repeats. Splicing mutations that lead to a change in the levels of E10 containing transcripts cause an imbalance in the 4R/3R isoform ratio, and lead to frontotemporal dementia with parkinsonism linked to chromosome 17 (FTDP-17). Avale and colleagues ⁸⁴ designed a PTM for 3′ trans-splicing with the BD binding to intron 9, and carrying the cDNA of the last three exons of tau (E10-E11-E12). They applied the PTMs to Htau mice that express only human MAPT, resulting in equal amounts of 4R and 3R. ⁹⁴ Since normal adult mice express only the 4R isoform, the model recapitulates the effect of a splicing mutation abolishing E10 inclusion. Following delivery of the PTM in the prefrontal cortex of Htau mice by stereotaxic injection using lentiviral vectors, they were able to show effective trans-splicing at RNA and protein level. 

RNA Interference

RNA interference (RNAi) is a regulatory mechanism used by the cell to silence specific transcripts at the posttranscriptional level. The endogenous effectors of this mechanism are micro RNAs (miRNAs). They are transcribed into primary precursors (pri-miRNA) that are then cleaved first by Drosha into a ∼70-bp precursor hairpin (pre-miRNA), and then by Dicer into a ∼22-bp RNA/RNA duplex. ⁹⁵ One of the two strands (the guide strand) is then loaded into the RNA-induced silencing complex (RISC) as a mature miRNA, where it can recognize complementary mRNAs. Silencing by RISC is caused by translational repression if the complementarity between the miRNA and its target mRNA is not perfect. When the miRNA perfectly matches its target, silencing is instead mediated by cleavage and subsequent degradation of the mRNA. Small interfering RNAs (siRNAs) are double-stranded RNA molecules that result from processing by Dicer of exogenous double-stranded RNAs. Alternatively, they can be chemically synthetized and delivered as such for therapeutic purposes. Small interfering RNAs are directly loaded into RISC. Another class of interfering RNA is composed by short hairpin RNA (shRNAs), stem-loop structures that enter the miRNA processing pathway as substrates for Dicer. Short hairpin RNAs are often obtained from the transcription of a delivered DNA transgene hereby guaranteeing stable expression. On the contrary, the effect of siRNA is transient, even if different chemistries are today available to improve their stability. Small interfering RNAs have already been used to treat retinal diseases for a few years. The first RNAi therapeutic applications to enter the clinical phase have in fact been two naked siRNAs developed for the treatment of AMD: Bevasiranib, directed against VEGF; and AGN211745, targeting VEGF receptor (VEGFR1). Clinical development of these two drugs has been discontinued during the last years, mainly because of their failure in meeting efficacy endpoints. ⁹⁶ Even if other clinical trials are in progress for other naked siRNA, ⁹⁷ the lessons that we can learn from the two described trials is that it is a challenge to deliver naked siRNA into cells, even in an easy system such as the eye, and that their efficacy is not mediated by RNAi, but via an off-target sequence independent effect on cell surface Toll-like receptor-3 (TLR3). ^98,99 New chemistries seem to overcome these limitations, ¹⁰⁰ and the availability of different types of formulations (polymer- or lipid-coated nanoparticles, oligonucleotide nanoparticles, and conjugates delivery systems ⁹⁷ ) can potentially help in improving cellular uptake in the retina. However, the current trend for in vivo applications is to utilize viral vectors to deliver shRNAs to different retinal layers. ^101–106  

By exploiting the mechanism of RNAi-mediated gene silencing, it is possible to solve different situations in which correct splicing is compromised. Mutations that cause impairment in the splicing process can lead to misspliced products that in some cases can show a dominant negative or a gain of function effect. If this is the case, the simple decrease of the misspliced mRNA, without affecting the correctly spliced counterpart, can revert the pathological condition. This is possible using RNAi. ¹⁰⁷ A recent example of the application of this strategy in the retina has been shown for retinitis pigmentosa. ¹⁰⁸ Mutations in rhodopsin (RHO) are the most prevalent cause of autosomal dominant retinitis pigmentosa (adRP), accounting for 25% of the cases. ¹⁰⁹ Two mutations causing adRP (c.531-2A>G and c.937-1G>T) generate missplicing products. They abolish the proper recognition of the DS site of exon 3 and 5, respectively. Mutation c.531-2A>G originates two misspliced mRNAs with a partial intron 2 retention or a partial deletion of exon 3. Mutation c.937-1G>T results in a partial deletion of exon 5. Using RNAi against the misspliced mRNA, it was possible to successfully decrease their level without affecting the properly spliced isoform, with the exception of the second product of c.531-2A>G mutation, where the RNAi was not able to distinguish between the wild type (WT) and the aberrant mRNA (Fig. 8). 

Figure 8

 

RNA interference for adRP splicing mutations. Left: the effect on splicing of the two RHO mutations c.531-2A>G and c.937-1G>T is shown. Mutation c.531-2A>G originates two misspliced products (A, B). Right: The different misspliced products can be selectively recognized by siRNA (black lines) targeting sequences or exon-exon junctions absent in the WT transcript. Small interfering RNA directed toward misspliced product b of c.531-2A>G mutation was not able to discriminate between the WT mature mRNA and the aberrant one. Adapted from Hernan et al. ¹⁰⁸

Conclusions and Future Perspectives

Antisense-mediated, splicing-correction approaches, after their first and successful application for DMD, acquired a momentum that prompted the efforts for their application to several disease conditions. Citing only some of the main applications in the CNS, splicing modulation therapies have been implemented for ataxia telangiectasia (AT), frontotemporal dementia, and parkinsonism linked to chromosome 17 (FTDP-17), Alzheimer disease (AD), spinocerebellar ataxia (SCA), neurofibromatosis type 1 (NFT1), and many others. ¹¹⁰  

So far their application in the eye has just begun, showing great promise. However, there are some limitations that need to be faced for a successful implementation of such therapies. The first drawback is that, apart from trans-splicing, all other approaches are patient-specific. Even if the shift toward personalized medicine offers great advantages in a long-term frame, its development poses many challenges in the immediate future. The limited number of patients treatable with each antisense molecule raises questions on the feasibility of normal clinical trials. Moreover, the current need to consider antisense therapies, even of the same class and for the same disease, as different drugs if they are targeting different mutations, creates a great economic obstacle to their development. Luckily a debate between researchers and regulatory agencies has already started on these issues. ¹¹¹ Another problem regarding the retina is the difficulties of having relevant animal models that would allow splicing manipulation of the gene of interest. The implementation of such models will definitely foster clinical application of splicing-correction approaches for retinal dystrophies. 

Acknowledgments

The authors thank Erik Dassi for his helpful assistance in retrieving and analyzing data from HGMD. 

Supported by the Italian Ministry of Health, Young Italian Researchers Grant 2008, Project GR-2008-1136933. 

Disclosure: N. Bacchi, None; S. Casarosa, None; M.A. Denti, None 

Footnotes

 SC and MAD contributed equally to the work presented here and should therefore be regarded as equivalent authors.