Retinal gene therapy is increasingly recognized as a novel molecular intervention that has huge potential in treating common causes of blindness, the majority of which have a genetic aetiology1,2,3,4,5. Choroideremia is a chronic X-linked retinal degeneration that was first described in 18726. It leads to progressive blindness due to deficiency of Rab-escort protein 1 (REP1). We designed an adeno-associated viral vector to express REP1 and assessed it in a gene therapy clinical trial by subretinal injection in 14 patients with choroideremia. The primary endpoint was vision change in treated eyes 2 years after surgery compared to unoperated fellow eyes. Despite complications in two patients, visual acuity improved in the 14 treated eyes over controls (median 4.5 letter gain, versus 1.5 letter loss, P = 0.04), with 6 treated eyes gaining more than one line of vision (>5 letters). The results suggest that retinal gene therapy can sustain and improve visual acuity in a cohort of predominantly late-stage choroideremia patients in whom rapid visual acuity loss would ordinarily be predicted.
Choroideremia typically presents with night blindness and progressive visual field restriction in late childhood, leading to profound sight loss in young men beyond the fourth decade7. The choroideremia gene (CHM) encodes Rab-escort protein-1 (REP1), which facilitates intracellular vesicular trafficking8,9. Deficiency of REP1 leads to degeneration of the retinal pigment epithelium and photoreceptors in males10, whereas female carriers generally have a mild disease phenotype due to random X-inactivation11. The choroid degenerates secondary to loss of the pigmented epithelium, leading to exposure of the underlying white sclera and characteristic retinal appearance. The central cone photoreceptors are usually maintained until late stages, due to the centripetal nature of the degeneration12. Hence, long after visual field loss, there is a terminal period during which central visual acuity begins to decline as the underlying central retinal pigment epithelium becomes dysfunctional10. There is therefore a potential window of opportunity for improvement in visual acuity if this dysfunction can be reversed by gene-replacement therapy before these cells are irreversibly lost13.
The adeno-associated virus serotype 2 (AAV2) vector has been used in a number of clinical trials and is particularly effective at targeting outer retinal layers, but only when injected under the retina correctly14,15. Hence, assessment of retinal gene therapy must include consideration of the surgical technique as well as the biological properties of the investigational medicinal product. We previously reported the early safety data of retinal gene therapy for choroideremia and visual acuity changes in the first six patients who received the low dose of an AAV2 vector carrying the human CHM transgene16,17. Here we report the full results of the trial with all 14 participants in both low- and high-dose cohorts having reached the 2-year study endpoint.
A total of 14 patients were recruited (Supplementary Table 1), 13 of whom received either ‘low dose’ 1 × 1010 genome particles (gp; L1–5) or ‘high dose’ 1 × 1011 gp (C2 and H1–7) of AAV2.REP1 vector (Fig. 1), surgically delivered into the subretinal space via an iatrogenic retinal detachment. At the 2-year trial endpoint, the median visual acuity across all 14 treated eyes had improved by 4.5 Early Treatment Diabetic Retinopathy Study (ETDRS) chart letters (interquartile range (IQR): −2.0 to 8.8) and in the 14 untreated eyes had declined by −1.5 letters (IQR: −4.8 to 0.0), hence favoring the treated eyes overall (two-tailed Wilcoxon test: W = 65, P = 0.040) (Supplementary Table 2). The trial thus met its primary endpoint of improving vision following gene therapy compared to untreated fellow eyes, despite any potential adverse effects of retinal detachment.
In 12 out of 14 patients, the retinal gene therapy was performed as per protocol, leading to recovery of visual acuity in all eyes and variable degrees of acuity gains that generally occurred within 6 months of treatment and were sustained up to 5 years (Fig. 2 and Supplementary Table 3). Significant adverse events relating to vector administration occurred in 2 of the 14 patients: C1 and C2 (Fig. 3). In C1, a surgical complication resulted in retinal thinning, and the vector was under-dosed. In C2, there was significant retinal inflammation at 2 weeks postoperatively that was most likely vector-related. The complications in C1 and C2 led to off-protocol treatments, and the ethics committee approved the recruitment of 2 further patients, thereby providing 12 patients treated as per protocol with 2-year follow-up. Considering only the 12 treated eyes that had gene therapy surgery as per protocol without complications, visual acuity improved by a median of 5.5 letters (IQR: 2.5 to 9.0) above baseline levels by 24 months (Wilcoxon test: W = 60, P = 0.016) (Supplementary Fig. 2).
Microperimetry is a modified visual field test that assesses retinal sensitivity by determining the minimum threshold of a light stimulus that can be seen at various points across the macula. Hence, in contrast to visual acuity, which measures just one point (usually the fovea), microperimetry provides a mean value of retinal function across a larger area, although with greater test–retest variability15. The mean retinal sensitivity of treated eyes was 4.0 ± 0.7 dB at baseline and 3.3 ± 0.6 at 2 years, representing a small, non-statistically-significant decline of −0.7 dB (paired t-test: n = 12, t = 1.98, d.f. = 11, P = 0.07). In contrast, the untreated eyes fell slightly more from 4.8 ± 0.8 dB at baseline to 3.3 ± 0.7 dB, equivalent to a −1.5 dB loss at 2 years (paired t-test: n = 12, t = 3.62, d.f. = 11, P = 0.004). Although this represented a relative gain favoring the treated eyes over the untreated eyes of 0.8 ± 0.53 dB (95% confidence interval: −0.3 to 1.8 dB) at the 2-year study endpoint, the difference was not statistically significant (paired t-test: n = 12, t = 1.49, d.f. = 11, P = 0.17) (Supplementary Fig. 3).
Microperimetry provides more useful information about fixation, that is, the retinal locus that has maximal sensitivity (usually the fovea). All patients except L1 still retained some degree of foveal or parafoveal fixation, consistent with the centripetal nature of visual field loss in this disease. In L1, however, the fovea had already degenerated at baseline, leaving two peripheral islands of surviving retina, one of which was targeted with gene therapy (Fig. 4). It was previously noted that L1 changed his fixation (or preferred retinal locus) to use this treated island while bypassing the untreated island of retina16. A change in fixation provides good independent evidence of the therapeutic effects of retinal gene therapy and was maintained up to 5 years in this patient, consistent with the sustained improvement in visual acuity (Fig. 2)18.
Anatomical assessments included optical coherence tomography (OCT), which gives a cross-sectional view and measurement of retinal thickness19, and blue-light autofluorescence, which generates a map that can be used to estimate the surviving retinal area20,21. However, it should be noted that the eyes were not selected for anatomical symmetry, and several eyes had residual retinal areas that had degenerated too much to be accurately measured. In choroideremia, the fovea is thickened early in the disease process before the onset of degeneration, and it has been proposed that this is the result of glial cell activation resulting from retinal stress22,23. A small reduction in retinal thickness, if associated with improved retinal function, might therefore be considered a therapeutic effect—as occurs in diabetic maculopathy24. We have previously shown that retinal structural and functional recovery occurred in the five eyes (H3–H7) that received subretinal vector injection using the automated injection system by 1 month25. This would suggest that optimally performed surgery does not damage the retina significantly. Over the 2-year period and across the whole of 12 patients, the mean retinal thickness at the central point of fixation reduced by 17.1 ± 4.0 μm in the treated eyes and by 6.3 ± 2.2 μm in the untreated eyes (paired t-test: t = 2.40, d.f. = 11, P = 0.04). The clinical significance of this marginal difference is unknown. Full plots of retinal thickness changes over time in individual participants are shown in Supplementary Fig. 4.
The area of retinal autofluorescence is correlated to the area of surviving photoreceptors calculated from multiple slices through the ellipsoid zone19,26. Across the whole group of 12 patients who received gene therapy per protocol, similar areas of autofluorescence were preserved in treated and untreated eyes at 2 years (80.7 ± 3.0% and 80.8 ± 2.1%, respectively) (Supplementary Fig. 5 and Supplementary Table 4). It should be noted, however, that shrinkage occurs only from the most peripheral retinal cells located at the leading edge of the degeneration, and 2 years may be an insufficient period of time to assess the long-term effects of retinal gene therapy on the healthier central zones that correspond to the retinal loci responsible for increased visual acuity.
As part of the gene therapy safety assessment, vector shedding through body fluids and anti-AAV2 neutralizing antibody assays were also performed, which did not detect any signs of viral replication or systemic immune response (Supplementary Tables 5 and 6).
Here we report the final outcomes of a 2-year study assessing retinal gene therapy for choroideremia in a phase 1/2 clinical trial. Across the whole cohort of 14 participants, including the 2 patients in whom complications occurred, visual acuity in treated eyes improved relative to untreated eyes over the 2-year trial period. The clinical trial thus met its primary endpoint. Furthermore, three of the study eyes gained three lines or more of vision 12 months after gene therapy. Longer term follow-up with a mean of 3.6 years for the 12 protocol-treated participants confirmed that visual acuity gains were sustained.
The visual acuity gains appeared similar in both low- and high-dose cohorts, but this was a small study group with only five patients treated at the lower (1 × 1010 gp) dose and generally the patients had only a small fraction of the macula remaining, for which a lower dose of vector might be adequate. Both doses are, however, within the ranges shown previously to be therapeutic in RPE65-related Leber congenital amaurosis. Moreover, the inclusion of the translational enhancer Woodchuck hepatitis virus post-transcriptional regulatory element (WPRE) could potentially increase transgene expression when compared with other AAV constructs27,28. Since the retinal pigment cells primarily affected in choroideremia have a key role in supporting the visual cycle of the overlying photoreceptors, their dysfunction before cell death is likely to have an adverse effect on visual acuity. Hence, some improvement in visual acuity is a likely consequence of improving surviving cone function at the fovea following AAV2.REP1 gene therapy. Alongside visual acuity gains in some participants within 6 months, subjective improvements in color perception were also described; however, color vision assessment using the Farnsworth–Munsell 100 hue test was unreliable owing to the constricted visual field (Supplementary Fig. 6).
With regard to anatomical changes, it should be noted that advancement of this very slow degeneration measured over the 2-year timeframe was only at the peripheral retina, whereas visual acuity gains arose mainly centrally. The retina in patients with choroideremia is difficult to detach peripherally, and the biconvex shape of the subretinal space following detachment means that the height of the bleb is greatest centrally. The centripetal nature of fluid reabsorption may provide the central retina with several additional hours of exposure to the vector compared with more peripheral areas of the detachment.
While visual acuity was maintained or increased in all protocol-treated eyes throughout the duration of the study, this was not seen with retinal sensitivity. Although microperimetry is generally more variable15, the tests do have subtle differences in what they measure. Microperimetry assesses retinal function from many points averaged over the central retina, whereas visual acuity is a measurement taken from a single point (usually the fovea) with maximal sensitivity29. Maintaining retinal sensitivity therefore requires successful transduction of the entire area being measured, right up to the edges of the surviving tissue, and may also be reduced owing to the development of cataract in treated eyes, which is known to impact on microperimetry more than visual acuity30.
Among the adverse events encountered in the trial (Supplementary Table 7), one incident of retinal stretch (C1) was clearly related to surgery. The other significant adverse event of inflammation (C2) might equally be related to surgery if significant vector reflux into the vitreous cavity occurred, which is known to trigger inflammation (Supplementary Fig. 7). Following the protocol change midway through the trial, we developed an automated system for subretinal injection, which was further facilitated by intraoperative retinal scanning using OCT, which helped to identify the correct plane of the subretinal space in some of the more advanced patients.
A recent Canadian phase 1 gene therapy trial also showed a significant visual acuity gain in one of six patients using the same batch of vector. This was corroborated by an improvement in cone thresholds together with preservation of outer retinal structures measured with OCT in the same patient31. Conversely, one of their six patients had a surgical complication of subretinal air and hemorrhage during vector injection and developed inflammation at a later stage. Although their results were mixed, it should be noted that they did not use intraoperative OCT, which should improve the precision of subretinal vector delivery in future studies. Robot-assisted infusion of vector is also being developed to improve surgical consistency and safety32.
Nevertheless, the results of this phase 1/2 clinical trial show that gene therapy for choroideremia is generally safe. Small but sustained visual acuity gains were seen over a period of several years in end-stage eyes, in which rapid visual acuity loss would ordinarily be expected, with several patients experiencing gains of three lines or more, an improvement widely accepted to be clinically significant.
Gene therapy vector design
The AAV serotype 2 vector comprised a chicken β-actin promoter with a cytomegalovirus enhancer flanking a rabbit β-globulin intron/exon splice site—collectively termed CAG promoter—driving the cDNA of the human CHM gene, which encodes the REP1 protein (Fig. 1a)9. The vector also included WPRE to enhance gene expression and a bovine polyA signal28.
Validation of the AAV2.REP1 vector
During the trial period, the clinical-grade gene therapy vector was tested annually to confirm REP1 protein expression and prenylation activity using in vitro assays33. Briefly, HEK293 cells were transduced with clinical-grade AAV2.REP1 vector at a multiplicity of infection of 10,000 gp per cell (in triplicates). Untransduced control cells (in triplicate) were processed in parallel. Cells were collected at 5 days post-transduction and prenylation reactions were prepared with 20 μg of total protein extract. The positive control consisted of untransduced cell lysate supplemented with recombinant fish REP1 protein (25 nM). Western blot was used to detect any increase in incorporation of biotinylated prenyl groups into a RAB6A substrate, which would be proportional to the amount of vector-derived REP1 prenylation activity. The immunostaining of mouse retina shown was performed 5 weeks after subretinal injection of the AAV2.REP1 vector at 1 × 109 gp to confirm correct localization of the REP1 protein. The animal work under the UK Home Office approved project licence (33/3363) complied with local and national regulations on the use of animals in scientific research (see Life Sciences Reporting Summary). After fixation in 4% paraformaldehyde, retinal sections were blocked and incubated overnight at 4 °C with rabbit anti-human REP1 primary antibody (HPA003231, Sigma-Aldrich) 1:1,000, and then for 1 h at room temperature with donkey anti-rabbit Alexa Fluor 568 secondary antibody (A10042, Thermo-Fisher Scientific) 1:500. All sections were counterstained with Hoechst 33342 and mounted with ProLong Gold for imaging.
Gene therapy clinical trial design and summary
In this unmasked, non-randomized, prospective interventional gene therapy clinical trial, 14 participants were recruited with informed consent and underwent gene therapy treatment to one eye using the AAV2.REP1 vector (ClinicalTrials.gov ref. NCT01461213)16. The clinical trial protocols were approved by the UK National Research Ethics Committee (London – West London; ref. GTAC171) and the study adhered to the Declaration of Helsinki 2013. Since choroideremia affects both eyes fairly symmetrically over the longer term15, the primary endpoint was defined in terms of vision change in the treated eyes compared to the untreated eyes in each patient. Eight of the 14 treated eyes had visual acuities greater than 70 letters (6/12) at baseline. In these eyes, significant visual acuity gains would not be expected since a 3-line (15 letter) gain would require them to surpass 85 letters (6/6) after surgery. All participants were male, ranging from 25 to 73 years of age, with confirmed null mutations in the CHM gene (Supplementary Table 1). The primary objective of the trial was to assess safety in relation to maintaining vision by two years after surgery. Initially, 12 patients were to be recruited into 2 dose cohorts of 6 patients, each of whom would be monitored for 24 months. However, complications in 2 patients led to a 24-month delay midway through the trial and a change in protocol relating to improved surgical technique and immune suppression regimen. The ethics committee approved an extension of the trial together with the recruitment of 2 further patients so that 12 patients in total received the gene therapy treatment as per the protocol without complications. Consequently, all 14 patients have now reached the 2-year follow-up point that signifies the formal end of the trial. In addition, longer-term data up to five years are also available for the patients recruited before the mid-way protocol amendment.
Subretinal administration of AAV vector
Surgical delivery of the AAV2.REP1 vector into the subretinal space has previously been described in detail16,34,35. In the first cohort of 6 patients, a subretinal injection of up to 1 × 1010 gp assayed using a supercoiled plasmid reference was performed as a two-step procedure. This comprised an initial detachment of the retina with balanced salt solution delivered through a 41 gauge Teflon cannula (DORC BV) and secondary injection of the AAV2.REP1 vector into the newly created subretinal space. In patient C1, difficulties in detaching the retina and stretching of the papillomacular bundle resulted in a reduced gene therapy dose of ≤6 × 109 gp and subsequent retinal thinning, but all other patients received either the full low dose of 1 × 1010 gp (L1–5) or high dose of 1 × 1011 gp (C2 and H1–7), as per protocol. Initially oral prednisolone was administered at 1 mg kg−1 for 3 days before and 7 days after gene therapy, but following the development of visually significant vitritis and retinitis in C2 two weeks post-operatively, the protocol was amended so that H1–7 received an extension of the prednisolone regime: 0.5 mg kg−1 (days 8–14), 0.25 mg kg−1 (days 15–16), then 0.125 mg kg−1 (days 17–18). Further details of the surgery and visual function tests can be found in the Supplementary Methods.
Due to the ceiling effects of including eyes with near-maximal visual acuity in the study, the letter scores were found to be skewed (Shapiro–Wilk normality test at 0.05 alpha) and are therefore presented as median values with IQRs36. Changes between treated and control eyes were compared using a two-tailed Wilcoxon signed-rank test. Microperimetry data and anatomical assessments were found to be normally distributed. These data are therefore presented as mean ± standard error of mean and compared using a two-tailed paired t-test.
Further information on research design is available in the Nature Research Reporting Summary linked to this article.
The authors declare that all of the data supporting the findings of this study are available within the paper and the Supplementary Information and are available from the corresponding author upon reasonable request.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
We thank all trial participants for their commitment to attending extensive follow-up visits, K. M. Jasani for helping with collection of the OCT and adverse events data, M. Hassall for helping with analysis of the AF data, and staff members of the Eye Research Group Oxford (ERGO) for their support throughout the study. This work was supported primarily by a grant (HICF-1009-006) from the Health Innovation Challenge Fund, a funding partnership between the UK Department of Health and the Wellcome Trust. Additional funding support was from the Health Foundation; Fight for Sight; the Lanvern Foundation; the Special Trustees of Moorfields Eye Hospital; the Royal College of Surgeons of Edinburgh and the National Institute for Health Research (NIHR) Biomedical Research Centres (BRC) at the Oxford University Hospitals NHS Foundation Trust (which includes the University of Oxford) and Moorfields Eye Hospital NHS Foundation Trust (which includes the University College London Institute of Ophthalmology).