A gene therapy for inherited blindness using dCas9-VPR-mediated transcriptional activation
AbstractCatalytically inactive dCas9 fused to transcriptional activators (dCas9-VPR) enables activation of silent genes. Many disease genes have counterparts, which serve similar functions but are expressed in distinct cell types. One attractive option to compensate for the missing function of a defective gene could be to transcriptionally activate its functionally equivalent counterpart via dCas9-VPR. Key challenges of this approach include the delivery of dCas9-VPR, activation efficiency, long-term expression of the target gene, and adverse effects in vivo. Using dual adeno-associated viral vectors expressing split dCas9-VPR, we show efficient transcriptional activation and long-term expression of cone photoreceptor-specific M-opsin (Opn1mw) in a rhodopsin-deficient mouse model for retinitis pigmentosa. One year after treatment, this approach yields improved retinal function and attenuated retinal degeneration with no apparent adverse effects. Our study demonstrates that dCas9-VPR–mediated transcriptional activation of functionally equivalent genes has great potential for the treatment of genetic disorders.INTRODUCTIONVarious inherited disorders are caused by mutations in genes for which counterparts with similar function but distinct expression pattern exist. CRISPR-Cas–mediated transcriptional activation (transactivation) of such functionally equivalent genes is one attractive therapeutic strategy to compensate for the function of their mutant counterparts. Different transcriptional activators have been fused to catalytically inactive Cas9 (dCas9) proteins and evaluated regarding their transactivation efficiency. Among the tested candidates, the catalytically inactive dCas9 fused to transcriptional activators (dCas9-VPR) transactivating module shows high efficiency across different species and cell types (1). However, because of its size (5.8 kb), dCas9-VPR exceeds the genome packaging capacity of recombinant adeno-associated viral (rAAV) vectors, which are currently the gold standard for gene delivery to native tissues and for gene therapy. A previous study provided a proof of principle for reconstitution of split dCas9-VPR using dual rAAVs in vitro and in vivo in wild-type (WT) mice (2). Nonetheless, the therapeutic potential of this tool has not been evaluated in disease models so far. In particular, the long-term effects, such as efficiency and expression of the transactivated gene as well as the potential adverse effects, remained largely unexplored.Inherited retinal dystrophies (IRDs) affect several million people worldwide. Retinitis pigmentosa (RP) is the most common IRD subtype and primarily affects rod photoreceptors (3). By contrast, achromatopsia (ACHM) is among the most frequent IRDs affecting the cones (4). Many genes associated with RP or ACHM encode members of the phototransduction cascade in rods or cones. The key phototransduction molecules in these cells are encoded by distinct yet functionally equivalent genes, and mutations in many of these genes, such as the visual pigments (opsins) or cyclic nucleotide-gated (CNG) ion channels, are associated with different types of blinding disorders. Mutations in the rhodopsin gene (RHO) are the leading cause for RP, whereas mutations in the cone CNG channel genes (CNGA3 and CNGB3) are the most frequent cause for ACHM. While rods only express rhodopsin, most mammals including mice express two opsin types in cones, the short wavelength–sensitive S-opsin (Opn1sw) and the medium wavelength–sensitive M-opsin (Opn1mw). CNG channels are heterotetrameric complexes composed of the channel function defining CNG A and the modulatory CNG B subunit. The native rod CNG channels contain CNGA1 and CNGB1 and their cone counterparts CNGA3 and CNGB3 subunits, respectively. A previous study has shown that rod and cone CNG A subunits can also form functional units with the CNG B subunits from the other photoreceptor type, i.e., CNGA1 with CNGB3 and CNGA3 with CNGB1 (5). Given the functional similarity between native and chimeric CNG channels, activation of the respective functionally equivalent gene in rods or cones appears an attractive treatment option. Recent work on mouse models has shown that rhodopsin and cone opsins are also functionally equivalent (6–9). This suggests that activation of genes encoding for cone opsins in rods could compensate for the defective rhodopsin in the respective animal models.Here, using dCas9-VPR, we were able to efficiently transactivate the Rho homolog Opn1mw and the rod-specific Cnga3 homolog Cnga1 in vitro. Using a dual rAAV vector system for split dCas9-VPR–mediated Opn1mw activation in a rhodopsin-deficient mouse model for RP, we also demonstrate that this treatment results in safe and efficient long-term expression, gain in retinal function, and delay of retinal degeneration.RESULTSCnga1 and Opn1mw transactivation in transiently transfected 661W and MEF cellsTo test for the feasibility and efficiency of ectopic activation of nonexpressed or poorly expressed genes, we transfected different mouse cell lines with dCas9-VPR in combination with single guide RNAs (sgRNAs) binding to the promoter region of either murine Cnga1 or Opn1mw (Fig. 1A). As the transactivation efficiency can be increased with a growing number of sgRNAs (1), we used a combination of three sgRNAs for each of the genes. Transactivation of Cnga1 was addressed in 661W cells, immortalized derivatives of murine cones lacking Cnga1 expression (10). As 661W cells express Opn1mw endogenously, we used mouse embryonic fibroblast (MEF) cells for transactivation of this gene.
Fig. 1 Transactivation of Cnga1 and Opn1mw via dCas9-VPR.(A) Binding position of the three sgRNAs used for targeting dCas9-VPR to the promoter of the Cnga1 or Opn1mw gene, respectively. The relative distance of each sgRNA to the transcription start site (TSS) of the target gene is given in base pairs. (B) Quantitative reverse transcription polymerase chain reaction (qRT-PCR) from 661W cells cotransfected with dCas9-VPR and either Cnga1 or lacZ sgRNAs. Cnga1 expression was normalized to the lacZ control. (C) qRT-PCR from MEF cells cotransfected with dCas9-VPR and either Opn1mw or lacZ sgRNAs. Opn1mw expression was normalized to the lacZ control. A two-tailed unpaired t test with Welch’s correction was used for statistical analysis in (B) and (C). (D to G) qRT-PCR from 661W-pb (D and E) or MEF-pb cells (F and G) cultured at different doxycycline (DOX) concentrations as indicated. Cnga1 expression was normalized to DOX (500 ng/ml), and Opn1mw and dCas9-VPR expression was normalized to DOX (0 ng/ml). In 661W cells transiently cotransfected with the dCas9-VPR cassette and Cnga1-specific sgRNAs, we observed efficient transactivation of Cnga1 on the transcript level, which was absent in control cells expressing the lacZ sgRNA (Fig. 1B and fig. S1A). Nevertheless, no Cnga1 protein signal was detectable in cells labeled with a specific antibody under these conditions.When addressing the Opn1mw transactivation in MEF cells, we detected Opn1mw transcript in both naïve and in lacZ sgRNA–expressing cells, indicating endogenous expression of Opn1mw in this cell line (fig. S1B). The efficiency of dCas9-mediated transactivation is known to correlate negatively with the basal expression level of a given gene (1). Nevertheless, despite the basal Opn1mw expression, a robust transactivation of Opn1mw was detectable in MEF cells coexpressing dCas9-VPR and Opn1mw-specific sgRNAs (Fig. 1C and fig. S1B). A combination of another set of three Opn1mw sgRNAs could not further improve the transactivation efficiency (fig. S1, C and D). These results show that, in transiently transfected cells, both Cnga1 and Opn1mw can be efficiently transactivated using dCas9-VPR.Transactivation in cell lines stably expressing inducible dCas9-VPR sgRNA cassettesNext, we analyzed whether Cnga1 protein can be detected in a transfection-independent system. For this, we created 661W and MEF cell lines with stable integration of expression cassettes for doxycycline (DOX)–inducible dCas9-VPR in combination with Cnga1, Opn1mw, or lacZ-specific sgRNAs [661W–piggyBac (pb) and MEF-pb, respectively]. When analyzing the Cnga1 transcript in the 661W cells, we detected a DOX concentration–dependent transactivation of this gene. However, robust Cnga1 transactivation was also present in the absence of DOX, suggesting a leaky activity of the DOX-dependent promoter driving dCas9-VPR expression. Moreover, an increase in Cnga1 expression was only obtained for the lowest DOX concentration (5 ng/ml), whereas Cnga1 levels were decreasing with further increase in drug concentration (Fig. 1D). Thus, when exceeding the optimal DOX concentration, there was an inverse correlation between dCas9-VPR transcript levels and the efficiency of Cnga1 transactivation (Fig. 1E). Very similar results were obtained in MEF cells stably expressing the dCas9-VPR cassette and Opn1mw sgRNAs, indicating that this effect was gene independent (Fig. 1, F and G). In contrast to Cnga1, there was no apparent Opn1mw transactivation in MEFs stably expressing dCas9-VPR and Opn1mw-specific sgRNAs in the absence of DOX, which is most likely due to the endogenous basal expression of this gene in this cell line.In summary, these results suggest that an optimal window for transactivation exists in which sufficient levels of dCas9-VPR protein support maximal levels of gene expression. Upon exceeding the optimal dCas9-VPR levels, there is a gradual decrease in transactivation efficiency.Analysis of protein expression and function in transactivated cellsNext, we assessed the Cnga1 and M-opsin protein expression in the corresponding stable cell lines treated with the optimal DOX concentration. Under these conditions, no evident increase in M-opsin protein expression was detectable when compared to the control cells (fig. S1E). For Cnga1, however, we could detect a robust signal that was absent in the lacZ sgRNA–expressing 661W cell line (Fig. 2A). To address whether functional channels can be formed from the proteins expressed from the transactivated Cnga1 locus, we conducted electrophysiological recordings in 661W cells cultured under optimal DOX concentration. In contrast to the lacZ control cell line lacking any CNG channel-like responses, several Cnga1 channel–specific characteristics including guanosine 3′,5′-cyclic monophosphate (cGMP) sensitivity, calcium, and magnesium blockage and outward rectification could be measured upon transactivation of this gene (Fig. 2, B to D). This indicates that successful transactivation by dCas9-VPR can lead to fully functional Cnga1 channels.
Fig. 2 dCas9-VPR–mediated transactivation results in functional Cnga1 channels.(A) Immunostainings of 661W-pb cells stably expressing dCas9-VPR and lacZ- (sglacZ, top row) or Cnga1-specific sgRNAs (sgCnga1, bottom row) in the presence of DOX (5 ng/ml) using Cas9- and Cnga1-specific antibodies. Scale bar, 30 μm. (B) Representative current traces recorded from inside out patches of DOX-induced 661W-pb cells in the presence of 300 μM cGMP (left, ICNG) or cGMP and Ca2+/Mg2+ (right, Iblock). (C) Quantification of the cGMP-induced currents in the absence (ICNG) or presence of Ca2+/Mg2+ (Iblock) (unpaired t test with Welch’s correction, two-tailed). (D) Current-voltage plot of cGMP-induced currents from sgCnga1 or sglacZ membrane patches. Cnga1 and Opn1mw transactivation in cells expressing split intein dCas9-VPROwing to the limited genome packaging capacity of rAAVs, the entire dCas9-VPR cassette cannot be packaged into a single rAAV vector for in vivo delivery. To circumvent this limitation, two recent studies took advantage of the split intein technology to reconstitute (d)Cas9-VPR or its derivatives at the protein level upon codelivery of two separate dual rAAVs, each of which expressing one half of the split SpCas9 cassette (2, 11). The split intein-mediated reconstitution efficiency is known to depend on the position of the intein integration within the corresponding protein (12). The aforementioned studies addressed the transactivation of (d)Cas9-VP(R) split either after the amino acid position E573 (11) or V713 (fig. S1F) (2). Nevertheless, to our best knowledge, no quantitative or comparative data with respect to the reconstitution efficiency resulting from these two approaches are available so far. As such data would be very helpful to achieve optimal results in vivo, we first set out to compare the reconstitution efficiency for both approaches side-by-side at the protein level. To this end, we transiently cotransfected human embryonic kidney (HEK) 293 cells with plasmids encoding the SpCas9 halves intersected after the E573 or V713 position (fig. S1G) and quantified the resulting reconstitution efficiency. We found that the reconstitution efficiency of the SpCas9 variant split after V713 (56.9 ± 2.1%) was considerably higher than its counterpart split after the E573 position (33.3 ± 2.1%) (fig. S1H).Next, we examined whether the coexpression of split dCas9-V713-VPR fragments also leads to efficient transactivation of Cnga1 and Opn1mw in the respective cells. While the transactivation efficiency for both target genes originating from the split dCas9-VPR was lower than from full-length dCas9-VPR, it still appeared robust and high enough in relation to the respective controls (fig. S1, I and J). Together, we show that dCas9-VPR split after the V713 position can transactivate both Cnga1 and Opn1mw in cell culture experiments and was therefore used for the subsequent in vivo experiments.Split dCas9-VPR efficiently transactivates cone opsins in WT miceWe next focused on activation of Opn1mw in rod photoreceptors to provide an in vivo proof of concept for transactivation of functionally equivalent genes using split dCas9-VPR dual AAVs. For this purpose, we coinjected titer-matched dual rAAVs expressing split dCas9-VPR under control of a human rhodopsin promoter and Opn1mw-specific sgRNAs (Fig. 3A). Given the high percentage of rods, an activation of Opn1mw in this cell type is expected to superimpose the endogenous Opn1mw mRNA and protein signal originating from the more sparse cones, which make up only 3% of the total photoreceptor population in mice. For initial experiments, we used WT (+/+) animals subretinally injected on postnatal day 28 (P28) with split dCas9-VPR dual rAAVs. Four weeks after injection, a robust increase in Opn1mw transcript levels was observed in eyes coinjected with dual rAAVs compared to saline-injected eyes (Fig. 3B). This up-regulation of Opn1mw transcript was also confirmed by RNA sequencing (RNA-seq) (fig. S2A). Moreover, we could detect a strong M-opsin signal using an M-opsin–specific antibody in rod photoreceptor outer segments (Fig. 3C). This rod photoreceptor–specific expression of M-opsin was absent in untreated retinas, which only show M-opsin signal in cone photoreceptors (Fig. 3D). Encouraged by this finding, we addressed whether S-opsin (Opn1sw), another Rho homolog and potential candidate for transactivation in rhodopsin-deficient mice, can also be transactivated in rods using the same approach. Akin to Opn1mw, in retinas expressing split dCas9-VPR dual rAAVs and Opn1sw-specific sgRNAs, we could also detect S-Opsin protein in rod outer segments (Fig. 3, E and F). Notably, there was an evident variability in protein expression of transactivated M-Opsin and S-Opsin from experiment to experiment (fig. S2, B to E).
Fig. 3 Split dCas9-VPR–mediated transactivation of Opn1mw and Opn1sw in WT mice.(A) Scheme of split dCas9-VPR and sgRNA encoding dual rAAV vectors used for transactivation of Opn1mw (B to D) and Opn1sw (E and F) in C57Bl/6J WT (+/+) mice. (B) qRT-PCR for Opn1mw expression upon transactivation 4 weeks after injection (Opn1mw-ta). The expression was normalized to control eyes injected with NaCl solution (saline, n=6 eyes, paired t test, two-tailed). (C to F) Immunolabeling of two +/+ mice injected with Opn1mw- (C) or Opn1sw-transactivating dual AAVs (E). The contralateral eyes (D and F) of both mice served as control (untreated) (D and F). Peanut agglutinin (PNA; magenta) was used as marker for cones. Scale bars, 30 μm. Opn1mw transactivation delays retinal degeneration and improves retinal function in Rho+/− miceNext, we also tested whether Opn1mw transactivation is sufficient to ameliorate the RP phenotype in the heterozygous rhodopsin-deficient (Rho+/−) RP mouse model (13). In contrast to Rho−/− mice, which completely lack rod outer segments from birth, heterozygous mice are capable of developing shortened but functional outer segments (13, 14), which is expected to be an important prerequisite for treatments requiring injections at later time points. Rho+/− mice were subretinally injected on P14 with titer-matched dual rAAV vectors expressing the split dCas9-VPR and Opn1mw sgRNAs. The contralateral control eye was injected with a NaCl (saline) solution. As Rho+/− mice show a slow course of retinal degeneration (13), we assessed the effects of our treatment 1 year after injection. Retinal degeneration is accompanied by a reduction of photoreceptors, a condition that can be addressed noninvasively by optical coherence tomography (OCT) measuring the thickness of the outer nuclear layer (ONL). OCT recordings from eyes expressing split dCas9-VPR and Opn1mw sgRNAs revealed an increase in the ONL thickness compared to the contralateral saline-injected eye, suggesting that our treatment is capable of delaying the degeneration (Fig. 4A and fig. S2F). Similar results were obtained when comparing the ONL thickness between treated and saline-injected eyes using an independent method based on histological analysis of retinal cryosections (fig. S2G).
Fig. 4 Opn1mw activation improves retinal phenotype in Rho+/− mice.(A) OCT measurements from Rho+/− mice injected with a NaCl solution (+/− saline, n=10 eyes) or dual rAAVs expressing split dCas9-VPR and Opn1mw-specific sgRNAs (+/− treated, n=10 eyes). Age-matched C57Bl/6J WT mice (+/+, n=14 eyes) were used as controls. Statistical analysis was performed using one-way analysis of variance (ANOVA) with Tukey’s post hoc test. (B and C) Averaged photopic and scotopic traces of the same Rho+/− and WT mice at 10 cd.s/m2. (D and E) Quantification of light-adapted photopic a- and b-wave amplitudes of the same groups. (F and G) Dark-adapted scotopic a- and b-wave amplitudes of the same groups plotted against different light intensities. All P values for each comparison (+/− treated versus +/− saline, +/− treated versus +/+, and +/− saline versus +/+) are shown in fig. S5E. To assess beneficial effects of our approach on rod-mediated (scotopic) and cone-mediated (photopic) retinal function, electroretinography (ERG) measurements were performed in dark- and light-adapted Rho+/− mice, respectively (Fig. 4, B to G, and figs. S3 and S4). The associated statistical analysis has been conducted in two different ways: (i) We performed a multiple comparison test to compare the ERG amplitudes at the different light intensities of all three groups (WT and treated or saline-injected Rho+/− mice; Fig. 4), and (ii) we made a paired comparison of ERG amplitudes between treated and saline-injected eyes only (fig. S5).In light-adapted animals, an increase in the a-wave amplitude of dual rAAV–injected eyes at the highest light flash intensity was achieved compared to saline-injected eyes and untreated WT animals (Fig. 4D). Moreover, when performing a pairwise comparison of the treated eyes to the corresponding saline-injected counterparts only, an increase in photopic a-wave amplitudes could be observed for the two highest flash intensities (fig. S5A). This suggests that under these conditions, M-opsin expressing rods might respond to cone-activating intensities. A slight tendential improvement was also detectable in the photopic b-wave amplitudes in treated eyes at the highest flash intensity (Fig. 4E and fig. S5B). When addressing the rod-mediated function in dark-adapted animals, we could measure a slight tendency for improvement of the scotopic a-wave toward WT-like responses (Fig. 4F and fig. S5C). In comparison, a robust trend toward an improvement of the scotopic b-wave was observed when comparing the treated eyes to their saline-injected counterparts (Fig. 4G and fig. S5D). This trend was more pronounced with increasing light intensities, further supporting the assumption that treated rods expressing M-opsin are capable of responding to cone-activating stimuli. The individual P values for all ERG measurements shown in Fig. 4 are summarized in fig. S5E. Conclusively, these data suggest that Opn1mw transactivation can ameliorate retinal degeneration and results in partially improved retinal function in the Rho+/− RP mouse model.Opn1mw transactivation reduces apoptosis without inducing gliosis or invasion of immune responsive cells in Rho+/− miceIn another set of experiments, we also analyzed the expression of M-opsin protein and markers for potential gliosis, apoptosis, or immune response in the retinas of the mice used for OCT and ERG measurements. Analogous to the results obtained in WT mice (cf. Fig. 3C and fig. S2), we found a considerable expression of transactivated M-opsin in the rod outer segments of injected animals, which, however, varied between animals (Fig. 5, A and B, and figs. S6 and S7). Furthermore, to assess the translational potential of this approach, we examined whether our treatment induced persistent gliosis or immune responses, which would be accompanied by proliferation of glial fibrillary acidic protein (GFAP)–positive Müller glia or ionized calcium binding adaptor molecule 1 (Iba-1)–positive microglial or mononuclear cells in the retina. Immune labeling of the retinas with these markers revealed no obvious increase in the number of glial, microglial, or mononuclear cells between the different groups in contrast to retinas of rd1 (retinal degeneration 1) mice exhibiting a fast retinal degeneration peaking on P13 (Fig. 5, C to H, and fig. S8) (15). To investigate whether photoreceptor degeneration is caused by apoptosis in the Rho+/− mouse model, we conducted a terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick end labeling (TUNEL) assay on retinal sections from the treated Rho+/− mice (Fig. 6, A and B). In this assay, we could detect a low but considerable number of TUNEL-positive cells, indicating that apoptosis is the underlying mechanism for the photoreceptor loss in this mouse model. Moreover, by comparing the number of TUNEL-positive cells per area in the transduced versus untransduced part of the treated retinas, we show that Opn1mw transactivation reduces apoptosis (Fig. 6C). These data further emphasize the beneficial effects of our treatment on photoreceptor survival.
Fig. 5 Transactivation of Opn1mw in Rho+/− mice does not evoke any microglial activation or reactive gliosis.(A and B) Representative immunostainings of retinas from Rho+/− mouse #1 injected with either split dCas9-VPR and Opn1mw-specific sgRNAs (A) (treated) or NaCl (B) (saline, contralateral eye). Peripherin-2 antibody (PRPH2, cyan) was used as rod and cone outer segment marker and PNA (magenta) as marker for cones. (C to F) Immunolabeling of the same retinas with Iba1 or GFAP (cyan) to visualize microglial cells or reactive gliosis in the treated (C and E) and saline-injected contralateral eye (D and F). (G and H) Immunolabeling of retinas from Pde6b-deficient (rd1) mice on P13 with Iba1 (G, cyan) or GFAP (H, cyan) served as a positive control. Scale bars, 30 μm.
Fig. 6 Transactivation of Opn1mw in Rho+/− mice reduces apoptosis.(A) Representative sections of the immunolabeled retina from Rho+/− mouse #1 injected with split dCas9-VPR and Opn1mw-specific sgRNAs showing a transduced (left) or untransduced (right) area of the same retina 1 year after injection. (B) Immunolabeling of the rd1 mouse retina on P13 served as a positive control. TUNEL staining (magenta, top) was used to visualize apoptosis, PRPH2 (cyan) was used as rod and cone outer segment marker (bottom). Scale bar, 30 μm. (C) Quantification of TUNEL+ cells in transduced versus untransduced areas of retinas from eight Rho+/− mice injected with split dCas9-VPR and Opn1mw-specific sgRNAs. A paired t test (two-tailed) was used for statistical analysis. Characterization of untreated Rho+/− miceThe interpretation of the results obtained so far was largely based on the comparison of the treated Rho+/− mice to the saline-injected contralateral eye. To address whether the injection of saline might induce some effects on its own, we characterized the retinal degeneration and function, reactive gliosis, recruitment of immune reactive cells, and apoptosis in age-matched (1-year-old) untreated Rho+/− mice and compared these parameters side by side to those obtained from saline-injected control eyes. Except for a very slight decrease in the photopic b-wave amplitude at one single light flash intensity (0.1 cd.s/m2) upon saline injection, there were no noticeable differences in any of these parameters between the two groups (fig. S9). The injection of the saline solution thus most likely does not affect the interpretation of our results and the treatment success.DISCUSSIONIn this study, we provide the first evidence that ectopic transcriptional activation of functionally equivalent genes using split dCas9-VPR can ameliorate a disease phenotype, further expanding the spectrum of possibilities for treatment of genetic disorders. Given that many IRD-linked genes expressed in rods or cones are encoded by functionally equivalent genes, the transactivation approach provides an attractive option for therapy of these diseases for several reasons: (i) Transactivation of functionally equivalent genes can be used for the treatment of large and frequent IRD genes, such as ABCA4 or MYO7A, which cannot be efficiently reconstituted using classical dual AAV vectors. ABCA4 and MYO7A belong each to a large protein family with many functionally equivalent partners expressed across different cell types. Some of these partners, e.g., ABCA1 or MYO7B, show a very high structural conservation to ABCA4 and MYO7A, respectively, with at least 50% amino acid identity and 65% amino acid similarity and therefore represent highly attractive candidates for transactivation. (ii) A transactivation approach can also be used for therapy of genes with autosomal dominant inheritance, for which gene delivery has to be combined with simultaneous knockout of the diseased allele. Recent studies demonstrated that sgRNAs with shortened spacer lengths (