Gene augmentation for autosomal dominant CRX-associated retinopathies

Abstract

The Cone-rod homeobox (CRX) protein is a key transcription factor essential for photoreceptor function and survival. Mutations in human CRX gene are linked to a wide spectrum of blinding diseases ranging from mild macular dystrophy to severe Leber Congenital Amaurosis (LCA), Cone-Rod Dystrophy (CRD) and Retinitis Pigmentosa (RP). These diseases are still incurable and mostly inherited in an autosomal dominant form. Dysfunctional mutant CRX protein interferes with the function of wildtype CRX protein, demonstrating the dominant negative effect. At present, gene augmentation is the most promising treatment strategy for hereditary diseases. This study aims to review the pathogenic mechanisms of various CRX mutations and propose two therapeutic strategies to rescue sick photoreceptors in CRX-associated retinopathies, namely, Tet-On-hCRX system and Adeno-associated Virus (AAV)-mediated gene augmentation. The outcome of proposed studies will guide future translational research and suggest guidelines for therapy evaluation in terms of treatment safety and efficacy.

1. Pathogenic mechanisms of CRX mutations

The human cone-rod homeobox gene CRX (OMIM #602225) is located on chromosome 19q13.33. There are four exons in CRX, specifically, the first noncoding and three coding exons. CRX codes for a 299 amino acid transcription factor which is predominantly expressed in photoreceptors and the pineal gland, regulating photoreceptor development and maintenance [1, 2]. The protein consists of three major domains: the homeodomain at residues 39-99 facilitates the DNA binding; the activation domain at residues 113-284, including a WSP motif at residues 158-170, contains binding sites for other transcription coregulators; a conserved carboxyl terminus motif, also known as the OTX tail, is found at resides 284-295 [1, 3].

To date, 93 disease-causing CRX mutations have been identified in human patients (www.hgmd.cf.ac.uk). Pathogenic CRX variants are associated with a complex group of inherited retinal diseases (IRDs) such as macular dystrophy [4], cone-rod dystrophy (CRD) [1], retinitis pigmentosa (RP) [5], and Leber congenital amaurosis (LCA) [6, 7]. Mutations within CRX are known to occur de novo or to be inherited mostly in an autosomal dominant pattern, consisting of substitution (nonsense and missense mutations) and frameshift mutations (deletions and insertions) [8]. Our understanding of these mutations has significantly expanded, as a great variety are being analyzed with the advent of cutting-edge tools from the sequencing and analytic technologies. However, a tremendous challenge for the pathogenic analysis of CRX-associated retinopathies, particularly of the autosomal dominant disorders, is to explore the phenotype-genotype correlations between disease severity and CRX mutations.

CRX mutations could cause dominant disorders by two possible mechanisms, namely, the haploinsufficiency of the functional CRX, and/or various dominant negative or gain-of-function effects of the mutant CRX. To the extent of data published, haploinsufficiency may not cause severe phenotypes. The study on Crx^+/− mice corroborates this claim, as they do not develop any detectable functional defects up to 6 months [9]. A rare case of human patients with the heterozygosity of CRX also supports these observations, as only the patients with the nullizygosity of CRX develop LCA [10]. However, in the autosomal dominant disorders, it remains unknown if the mutant CRX allele could partially abrogate the production of a functional CRX from the normal allele. Further studies are needed to address this question in detail. Alternatively, dominant negative activities of various CRX mutant proteins have been demonstrated in animal models [9, 11]. The reported dominant-negative mutations that arise in the homeodomain are mostly missense mutations, and those identified in the activation domain are largely frameshifts [7, 9]. Pathogenic mechanisms thus are constituted by altered DNA binding properties and/or transactivation activities of CRX mutant proteins.

CRX R90W mutation presents a hypomorphic missense mutation located in the homeodomain [3, 12], and is associated with a dominant late-onset mild CRD and recessive LCA. The mutant protein has reduced DNA binding activity, and thus malfunctions to transactivate CRX-downstream target genes such as Rhodopsin [3, 12]. The transactivation activity of wild-type (WT) CRX seems normal in the heterozygous model [3], which proves the limited dominant negative effect of the CRX R90W mutation through the process of retinal development. On the other hand, CRX E80A and K88N mutations represent distinct antimorphic missense mutations located in the homeodomain [13–15]. Both mutations manifest LCA in human patients [1, 14]. Interestingly, some LCA-associated CRX homeodomain mutations are located at non-conserved residues in human homeodomain paralogues [4]. These mutant proteins are predicted to bind discrete DNA sequences and show different transactivation activities from the WT protein, although this has not yet been studied in mammalian models. Hence the dominant negative effects of the mutant proteins, if any, remain to be proven.

The frameshift deletion CRX E168d2 presents an antimorphic mutation located in the activation domain [3], and is associated with dominant LCA in human patients [6, 16]. This mutation results in the early truncation of the activation domain, producing a protein that retains the ability to bind target DNA but fails to transactivate CRX-downstream target genes [3]. In addition, CRX E168d2 allele overproduces the mutant protein at about 4 times more than the WT protein in heterozygous mice, which exacerbates the dominant-negative effect on the binding competition [3]. As a result, the heterozygous mice have impaired visual functions at 1 month-old (MO) [3] and limited electroretinography (ERG) response at 3MO (data not shown). Cone degeneration occurs prior to rod degeneration in the heterozygous mice. As compared to the WT samples, the heterozygous mice have only about 30% survived cones at 1 month-old (MO) and no detectable cones at 3MO (data not shown); mutants rods are functional with shorter outer segments (OS) at 1MO but undergo progressive degeneration till complete lost at 6MO [3]. Interestingly, the ratio of mutant to WT CRX proteins directly correlates with the disease phenotype severity and age of onset [3]. In addition, truncation at the last exon by frameshift results in premature terminations [7, 8], hence the production of shortened but stable mutant mRNA that can avoid nonsense-mediated decay [17], which partially contributes to the dominant negative effect. On the other hand, Crx^Rip presents the c.763del1 mutation located in the last exon, which results in a skipping of the OTX tail and a non-homologous extension of 133 residues [18]. The mutant protein does not transactivate CRX-downstream target genes, suggesting its antimorphic activity [18]. More notably, the mutant protein does not bind target DNA [18]. Similar to the retinal manifestation in patients, Crx^Rip/+ mice show a LCA-like phenotype [18]. Photoreceptors in Crx^Rip/+ mice do not form outer segments, due to impaired photoreceptor gene expression and incomplete differentiation at early development [18]. In addition, the mutant protein is not overexpressed in Crx^Rip/+ mice. Taken together, the dominant negative effect of Crx^Rip mutation does not signify a competition between the mutant and WT proteins, but likely arises from the disruption of the photoreceptor transcription factor network.

2. Gene augmentation as a strategy to treat CRX-associated retinopathies

There are currently limited therapeutic options for CRX-associated retinopathies. Fortunately, retinal gene therapy recently shines promise for treatment of various inherited blindness [19]. In particular, the gene augmentation approach has been developed to deliver a healthy gene to a diseased retina to rescue the defective phenotypes and visual functions without replacing the mutant allele. At present, successes of gene augmentation are largely restricted to recessive mutations [20]. Five important questions need to be answered in order to apply gene augmentation to treat CRX-associated autosomal dominant disorders. Firstly, how much of augmented CRX is needed to overcome the dominant negative effect of a given mutation? If a mutant protein competes with WT CRX proteins for binding sites, how much is augmented CRX needed to balance the ratio of mutant to WT CRX proteins in each photoreceptor? Can altering this ratio rescue the mutant phenotypes? Secondly, considering CRX is an early photoreceptor transcription factor that activates the downstream gene regulatory network in the retinal development, could gene augmentation impose overexpression toxicity in a treated retina? Thirdly, when is augmented CRX needed in the treatment? Fourthly, as human patients with CRX-associated retinopathies are often diagnosed at different stages of disease progression, what are the windows of opportunities for gene augmentation to a given disease? Do diseased photoreceptors possess the neuroplasticity for gene augmentation outside the development window? Lastly, can the treated photoreceptors retain their fated cell identity and functional capability upon gene augmentation?

Answers to these questions could inform the feasibility and strategies to treat CRX-associated retinopathies. To address these questions at least partially, a new transgenic mouse model, Tet-On-hCRX, has been developed to allow both quantitative and temporal control of augmented human CRX (hCRX) gene expression under a given CRX mutant background (Figure 1). Upon binding of doxycycline, a conformation change of the reverse tetracycline-controlled transcriptional activator (rtTA) is triggered [21], which allows the resulted complex bind to tetracycline-responsive element (TRE) (Tet-On) [22]. FLAG-tagged human CRX protein will hence be produced (Figure 1A). Moreover, the photoreceptor-specific induction can be achieved by the combination of two transgenes, namely, a Cre-dependent pCAG-LSL-rtTA (Figure 1B) and a Crx promoter-driven Cre, pCrx-Cre [23] (Figure 1C). Proof-of-concept validation of gene augmentation by Tet-On-hCRX system shows that CRX expression can be induced and detected in Crx−/− mouse retinae at all tested ages throughout the entire developmental window (postnatal (P) day 5, 14, 21 and 35) (Figure 2), confirming the success of Tet-On-hCRX system. Doxycycline-treated samples also display significant upregulation of CRX-downstream Rho expression. Furthermore, the comparisons between treated mice with different starts of the treatments suggests the ‘early the better’ effect on induced CRX and Rho expression at P35, as samples with P0-35 treatment display significantly higher induced CRX and Rho expression than those with P14-35 or P21-35 treatments (Figure 2A, 2B). However, as compared to the WT samples, treated mice still have significantly lower CRX and Rho expression at P35 under current treatment schemes, implying a partial rescue effect. These results suggest that Crx-null photoreceptors retain the machinery for CRX-mediated gene expression and neuroplasticity for therapeutic intervention.

Figure 1.

Components of Tet-On-hCRX system. (A) TRE-hCRX (B) pCAG-LSL-rtTA (C) pCrx-Cre.

Figure 2.

Quantitative PCR (qPCR) analysis of hCrx and Rho expression in untreated Crx−/−, Crx-BAC-Cre⁺, Tet-On-hCRX⁺; doxycycline-treated Crx−/−, Crx-BAC-Cre⁺, Tet-On-hCRX⁺; WT mice. (A) Treatments started at P0, samples were harvested at P5, P14, P21 or P35. (B) Treatments started at P14 or P21, samples were harvested at P35. Results are plotted as relative Log2 expression to untreated samples (n≥4). Asterisks (**, ***, ****) denote p ≤ 0.01, p ≤ 0.001 and p ≤ 0.0001 respectively by T-test.

As compared to Tet-On-hCRX system, a clinically applicable approach is Adeno-associated Virus (AAV)-mediated gene augmentation. The retina is surgically accessible, which facilitates intravitreal or subretinal administration of the viral vector [24, 25]. Since the retina is a relatively immune privileged tissue [24] and photoreceptors are nondividing and differentiated [26], a small amount of the non-integrating AAV viral vector can be tolerated to generate a therapeutic response [24, 25] with minimal risk of stimulating a severe inflammatory immune response [25]. Yet, two immediate challenges emerge during testing AAV-mediated CRX gene augmentation with animal models, i.e., choice of AAV vectors and cell-type specificity. To date, the AAV2/5 vector yields efficient transduction and good tropism for photoreceptors in P0 mice [27]. And the size of human or mouse CRX cDNA well fits into the AAV packaging capacity. Also, photoreceptor-specific promoters, including CRX or GRK, can be employed to achieve desirable sensitivity and precision.

However, there are still several safety precautions of applying AAV-mediated gene augmentation. Although subretinal administration is currently the most efficient route for targeting photoreceptors [28], it causes a transient detachment of the RPE from photoreceptors, which might damage photoreceptors at the injection site. A possible solution is the use of low-volume and low-concentration multiple blebs to reduce the danger of the RPE detachment and systemic exposure of high concentration by a single bleb, despite of associated minor risks [29, 30]. In addition, immunity and tolerance to AAV is untested in human photoreceptors at early and middle childhood [31], and multiplicity of infection (MOI) of viral vectors per cell may vary at different stages of the disease, all of which require further trials on human patients with early-onset diseases.

3. Unanswered questions and challenges

One-step gene augmentation may not rescue all CRX-associated autosomal dominant disorders. For examples, gain-of-function activities of a mutant protein with altered DNA binding specificity, such as CRX K88N, may generate the irreversible but detrimental effects on the developmental program. In such cases, expression from the mutant allele needs to be silenced to prevent the disease progression. Similar approaches of treating RHO-associated autosomal dominant RP in the mouse and canine models have shown encouraging outcomes [32, 33]. In addition, the CRISPR-Cas9/gRNA technique [34–37] can also be used to knockout the mutant allele.

Since gene augmentation utilizes the transcription machinery of the host cells to express the functional protein, cellular conditions of the host cells require close monitoring. In a developing retina, CRX gene augmentation may not fully rescue mutations that block early differentiation in photoreceptors, due to an insufficient but early window of opportunity. On the contrary, when the host cells are at the late stage of degeneration or the degeneration tends to progress too fast, gene augmentation may not succeed. Prerequisite anti-apoptotic or neuroprotective therapies can be combined to target the defective photoreceptors in order to gain a required time window for gene augmentation [38, 39].

Maculopathy has been reported in human patients of different CRX mutations with large variability in disease progression and symptoms [4, 40–42]. In addition, CRX may be involved in foveal development, as the haploinsufficiency of CRX results in subclinical foveal abnormalities in human patients [10]. None of these observations can be validated and studied in non-primate models.

The phenotypic and onset variability between patients sharing the same CRX mutant allele has constantly been reported [4, 7, 10, 43, 44]. It may be due to, firstly, possible polymorphisms in the CRX promoter region; secondly, the impacts of possible epigenetic modifications in coding regions; thirdly, any differential expression of other transcription factors such as NRL and NR2E3; fourthly, variable expression levels of the WT and/or mutant alleles; lastly, sexual dimorphisms [42], suggesting CRX may interact with specific targets on the sex chromosomes. None of these hypotheses has systematically been tested in animal models. In addition, the stochastic effects of CRX mutations and overexpression toxicity on the retinal degeneration are currently unknown, which appends uncertainty on the application of gene augmentation.

4. Methods

4.1. Transgenic mouse lines.

All male and female mice in this study were on the genetic background of C57BL/6J. TRE-hCRX line was generated for this study. The transgene had a TRE promoter upstream of full-length human CRX cDNA sequence, all inserted into the H11 locus. pCAG-LSL-rtTA (Rosa26:pCAG-LSL-rtTA) line was obtained from the Jackson Laboratory (Stock No: 029617). pCrx-Cre (BAC-Tg Crx-Cre) line was obtained from a published colony [23]. Doxycycline diets (200mg/kg, Bio-Serv, NJ) were provided to mice (> P14) and breeding dams. Intraperitoneal (IP) injection of 100μL of 20mg/mL doxycycline solution was provided to mice (> P21) once a week to maximize the treatment effect. All animal procedures were conducted according to the Guide for the Care and Use of Laboratory Animals of the National Institute of Health, and were approved by the Washington University in St. Louis Institutional Animal Care and Use Committee.

4.2. Quantitative PCR (qPCR).

Each RNA sample was extracted from 2 retinae of a mouse using the NucleoSpin RNA Plus kit (Macherey-Nagel, PA). 1μg of RNA was used to produce cDNA using First Strand cDNA Synthesis kit (Roche, IN). Technical triplicates were run for each gene. Primers (5’ to 3’) used in this study were Crx (F) TGTCCCATACTCAAGTGCCC, (R) TGCTGTTTCTGCTGCTGTCG; Rho (F) GCTTCCCTACGCCAGTGTG, (R) CAGTGGATTCTTGCCGCAG; Ubb (F) CAACATCCAGAAAGAGTCAACC, (R) ATGTTGTAATCAGAGAGGGTGC. The reaction master mix consisted of EvaGreen polymerase (Bio-Rad Laboratories, CA), 1μM primer mix, and diluted cDNA samples. Samples were run using a two-step 40-cycle protocol on a Bio-Rad CFX96 Thermal Cycler (Bio-Rad Laboratories, CA). The statistical analysis is done by Student’s t-test with p<0.05, CI:95%.