In vivo versus ex vivo CRISPR therapies for retinal dystrophy

SUMMARY

Two therapeutic paths have been proposed to treat inherited retinal dystrophy using clustered regularly interspaced short palindromic repeats (CRISPR). One strategy is to genetically correct patient cells ex vivo for autologous transplant, whereas the second is to modify cells in vivo by delivering CRISPR effectors to the retina. The feasibility of both editing strategies has been demonstrated within three years of CRISPR’s adaptation to mammalian systems. However, the functional integration of transplanted cells into host retinae has been a long-standing challenge that currently represents the 2025 moonshot of the National Eye Institute’s Audacious Goals Initiative. The clinical translatability of each path is discussed with regard to current investigations and whether cell replacement can be circumvented by in vivo editing.

KEYWORDS:

• CRISPR
• gene editing
• retinal degeneration
• retinal dystrophy
• retinitis pigmentosa
• translational medicine

1. Ex vivo repaired cells for autologous transplants

The key impact of CRISPR is the RNA-guidance of the CRISPR-associated endonuclease (Cas) to specific sites in the genome for cleavage. The dichotomy in therapeutic strategy was predetermined by distinct endogenous repair mechanisms activated by double strand DNA breaks (DSBs). The process of using a sister chromatid as a template for homology-directed repair (HDR) is restricted to cell division, which likely precludes correction of post-mitotic photoreceptors in vivo (reviewed in Yanik et al. [1]). Bassuk et al. instead suggested autologous transplants and demonstrated ex vivo HDR of a point mutation in induced pluripotent stem cells (iPSCs) from a patient with X-linked RP (XLRP) [2]. The mutation resides in a hotspot of the RPGR locus common to over 60% of XLRP mutations, implying possible applicability to most XLRP patients. Targeting was ostensibly challenged by the high guanine-cytosine-content and repeat-rich sequence, similar to those of polyglutamine disorders such as Huntington’s disease. Nevertheless, the 13% HDR efficiency achieved exceeded the ~1% reported previously in iPSCs [3]. The higher rates of unintended edits (38%) and unedited loci (49%) may be compensated by screening modified clones to expand ex vivo (reviewed in Cox et al. [4]).

The authors demonstrated an important proof-of-principle for autologous cell therapy, but did not test vision rescue in cell-transplanted animals. The success of such ex vivo CRISPR therapies may rely on the functional integration of iPSC-derived photoreceptors in sufficient quantity and longevity to provide meaningful vision rescue, which has yet to be shown. However, cell replacement may not be the predominant mechanism by which transplanted cells rescue vision, as functional improvement in animal models of RP are commonly observed in the absence of integration. The protection mechanisms are poorly defined but include trophic support of photoreceptors and clearance of photoreceptor outer segments to compensate for phagocytosis deficiency in retinal pigment epithelial cells (RPE) (reviewed in Mead et al. [5]). Recent data also suggest that previous reports of photoreceptor replacement may have been conflated with donor-to-host cell transfer of genetic material and/or proteins [6]. On the premise that cell rescue and not replacement preferentially accounts for vision rescue, CRISPR therapies are better utilized for direct in vivo editing, which could eliminate the need for cell transplants.

2. Transferring ex vivo strategies for use in vivo

In vivo HDR can be curative if gene correction imparts a survival advantage to proliferative cells by which their expansion facilitates cell turnover and restored function in organs such as the liver [4]. For post-mitotic neurons, proof-of-concept exists for activating HDR by viral replication mechanisms, but this has not yet been repurposed for CRISPR applications [7]. The inability to expand corrected photoreceptors in vivo implies that the extent of rescue depends on HDR efficiency. Recent improvements of 3–19-fold by various methods used in combination with enhanced editing precision may render HDR strategies amenable to in vivo use (reviewed in Hung et al. [8]). Alternatively, HDR may be circumvented by endowing cytidine deaminase (CD) with the RNA-guidance property of Cas [9]. CD corrected point mutations by deamination of cytidine to uridine in the absence of DSBs. Further engineering of deaminases to broaden editing repertoires and reduce size to allow viral packaging could eliminate the need for HDR of point mutations, and as a result, could replace cell transplants for patients with disease-causing point mutations.

CRISPR-based strategies to inactivate genes utilizing the non-homologous end joining (NHEJ) pathway are better poised to impact ophthalmic therapies. Unlike HDR, repair through NHEJ is not restricted by cell-cycle and is therefore permissive of photoreceptor editing in vivo. Random insertions and deletions (indels) of nucleotides during DSB repair frequently shift open reading frames (ORFs) and introduce early stop codons that result in functional gene ablation. This was demonstrated in a dominant RP model by selective-ablation of the pathogenic rhodopsin (Rho) S334ter transgene [10]. A single base-pair (bp) variation at the target site established the protospacer adjacent motif (PAM) required for Cas activity, and was sufficient to discriminate the mutant from wild type (WT) Rho for ablation. Editing rates of 33% and 36% in photoreceptors conferred survival and rescued visual function. Native RHO mutations characterized as novel PAMs may be similarly targeted, and viral delivery of a single construct can be accomplished using the smaller Cas from Staphylococcus aureus to maintain consistency with current clinical practice for gene therapy. Mutations refractory to cleavage may become targetable through the use of Cas orthologs or rational-design engineering to tailor PAM specificities [8].

RHO P23H is the most common dominant RP mutation, and a target for mutation-specific ablation strategies using CRISPR [11,12]. Current investigations of RHO ablation therapy stand on the shoulders of those who demonstrated vision rescue by silencing the Rho P23H transcript almost two decades ago [13]. These investigators also showed the window of opportunity for photoreceptor rescue to not diminish with disease progression [14]. This counters the ‘point of no return’ hypothesis, which posits that rescue becomes impossible beyond a threshold of degeneration [15]. This hypothesis was further weakened by an elegant demonstration in an inducible RP model, whereby the simultaneous genotype correction in all photoreceptors halted retinal degeneration when only 30% of photoreceptors remained [16]. Photoreceptors persisted for the remaining 10 months of the study, suggesting long-term vision rescue is possible regardless of disease progression, and perhaps inheritance pattern.

3. Mosaicism of genetic correction in the retina

The extent to which these principles apply to in vivo editing is unclear as not all edited photoreceptors will be corrected. The stochastic nature of indel formation reduces the likelihood of ORF disruption based on the principle of the genetic code. Indels of 3n bps induce frameshifts with a one-third probability of failed gene ablation, and instead add or delete amino acids to the peptide sequence. Pathogenicity is likely retained, and the target site may be degenerate to subsequent editing. Photoreceptor survivability will be conferred in mosaic distribution with corrected photoreceptors neighboring those in which apoptosis remains subject to cell-autonomous processes. Precise DSB repair without indel formation in vivo reduced the rescued photoreceptor number closer to one-third [10].

Insight to whether retinal mosaicism reduces disease penetrance long-term or provides only transient rescue comes from CRISPR repair of recessive RP mice at the blastocyst stage. HDR of mutated Pde6b generated mice with 19% and 36% mosaicism, which disproportionately translated to photoreceptor rescue of 50% and 76%, respectively [17]. Visual function was also improved disproportionately and did not diminish in the time between functional tests. This is consistent with CRISPR-induced mosaicism in mice used to verify mutational pathogenicity in Leber congenital amaurosis (LCA) 16. Founder mice mosaic for a Kcnj13 mutation affecting RPE function showed mild but long-term photoreceptor preservation adjacent to WT RPE [18]. Furthermore, a unique clinical finding in which 13% mosaicism of a RHO mutation was identified in an unaffected individual suggests feasibility to dominant RP patients [19]. These findings support the idea that mosaic distribution of genetically corrected photoreceptors provides meaningful benefit.

CRISPR repair of frameshift mutations may utilize both HDR and NHEJ. In a CRISPR-repaired mouse model of dominant nuclear cataracts, a 1bp deletion in Crygc was corrected by indels of +1bp, −2bp, and −5bp in the coding sequence to restore the correct ORF [20]. Inverse to the one-third failure probability for ablation, two-third of indels will fail to restore the ORF and in vivo efficiency may be expected at half the 33–36% observed for ablation [10]. Interestingly, CRISPR-corrected mice were repaired by NHEJ at rates comparable to HDR, indicating an additive therapeutic advantage if HDR becomes amenable to in vivo use.

4. Circumventing cell transplants and gene replacement

NHEJ-mediated excision of a genomic segment can also restore ORFs. The intronic mutation in LCA10 patients causes CEP290 mis-splicing, ORF disruption, and loss of protein expression, which accounts for 21% of all LCA cases [21]. Dual cleavage at sites flanking the mutation restored the ORF and protein expression in patient fibroblasts [22]. This strategy was tested ex vivo with its intended use for in vivo editing. Thus, the approach for CEP290 circumvents cell transplants. It also averts technical limitations for viral delivery of large genes, and is safer as CEP290 expression remains regulated by its endogenous promoter.

The promise of ORF restoration by excising a mutation hotspot in the dystrophin gene generated substantial attention for the treatment of Duchene’s muscular dystrophy. The similar approach for CEP290 may have even greater merit as exons are not removed, and the retina’s compartmentalization simplifies delivery. An analogous strategy is also being explored to excise a splicing mutation in USH2A for Usher syndrome, and PAX6 is a target of mutation-specific ablation for dominant Aniridia, similar to the approach for RHO P23H [12]. Nevertheless, autologous transplants are proposed using iPSCs corrected ex vivo for mutations in MAK [12] and TRNT1 [23] for recessive RP, ABCA4 for Stargardt’s cone-rod dystrophy [24], and CLN3 for Batten disease [25]. For genes in which in vivo correction is impossible and/or impractical due to extensive mutational heterogeneity, alternative-editing strategies may still offer alternatives to cell transplant. For example, conversion of rods to cones by Nrl in vivo ablation can effectively nullifies mutations in rod-specific genes such as Rho and Pde6b, whereas the opposite conversion could mitigate cone-specific dystrophies [26].

5. Necessary precautions

Collectively, current investigations for CRISPR-based treatment of retinal dystrophies are represented for all major inheritance patterns, and apply to some spontaneous monogenetic mutations.

Allele-specific ablation is only applicable to dominant conditions not subject to haploinsufficiency, and the risk of permanent ablation of the remaining WT allele necessitates stringent off-target validation. The non-fatal nature of RP makes it difficult to rationalize any threshold for off-target tolerance, even if compensated by overall reduced disease penetrance because iatrogenic vision loss risks a devastating setback for clinical development, similar to the decade-long translational pause in the gene therapy field. Several properties of the eye make it particularly suited for evaluating experimental therapies, making ophthalmologists stewards of the gene editing frontier. As such, translational precautions to prevent irreversible harm will likely be integrated into CRISPR designs. Cas expression can be regulated by self-cleavage mechanisms [27], photoreceptor-specific promoters [28] or viruses [29], or transient Cas activity can be ensured by delivering preformed Cas-protein/RNA complexes [30].

6. Future outlook

The ability to modify the genome at specific sites has generated much hope for treating diseases and for good reason; logic dictates that monogenetic diseases are curable by eliminating the mutations that cause them. As CRISPR systems advance, technical caveats will be overcome or circumvented and novel applications realized. Rational-design engineering has endowed Cas enzymes with advantageous features that allow non-permanent transcriptional repression or activation, which may be useful in conjunction with optogenetic regulation (reviewed in Hung et al. [8]). The converse is also true; endowing RNA-guidance to enzymes capable of DSB-independent base-editing has equal therapeutic potential [9]. Such advances will initially produce therapies intended to halt retinal degeneration to preserve existing vision. This fundamentally differs from the aim of cell replacement, which is to restore lost vision. As attenuating dystrophy likely renders the retinal environment more permissive for transplanted photoreceptor survival and de novo synaptogenesis, a combination of the two is likely inevitable.

Declaration of interest

The author is listed as co-inventor by Cedars-Sinai Medical Center on Non-Provisional Patent Application 15/130,846 846 filed in the United States Patent and Trademark Office on April 15, 2016 for the use of CRISPR/Cas9 as in vivo gene therapy to generate targeted genomic disruptions in genes bearing dominant mutations for retinitis pigmentosa. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

Acknowledgments

The author thanks Dr. Ritchie Ho and Dr. Alex Laperle for their critical review of the manuscript.

Additional information

Funding

This paper was supported by the Cedars-Sinai Medical Center’s Board of Governors Regenerative Medicine Institute.