Subretinal rAAV2-based VEGF-Trap gene therapy for neovascular age-related macular degeneration: Preclinical assessment and phase 1 trial results Authors: Tong Li, Jingyang Feng, Xiaolu Yang, Shiqi Yang, Jieqiong Chen, Yidong Wu, Xiaosa Li, Zhuoyu Ni, Yang Liu, Minlu Song, Peirong Huang, Kairong Zheng, Hong Wang, Xian Xu, Junran Sun, Fenghua Wang, Huixun Jia, Xiaodong Sun Published: 2026 ABSTRACT SummaryIntraocular vascular endothelial growth factor (VEGF) antagonists for neovascular age-related macular degeneration (nAMD) require frequent injections, leading to treatment burden and suboptimal outcomes. LX102 is an rAAV2-based gene therapy encoding VEGF-Trap via subretinal delivery. In laser-induced choroidal neovascularization (CNV) mouse models, LX102 prevents lesion formation and progression in a dose-dependent manner. In non-human primates, a single LX102 injection reduces grade IV CNV lesions by over 85% and maintains transgene expression for 26 weeks without drug-related ocular abnormalities. In a phase 1 dose-escalation study, 12 participants with nAMD receive a single LX102 administration (2 × 1010 to 1.25 × 1011 vector genomes/eye). LX102 is well tolerated with no clinically significant inflammation. Eleven participants (91.7%) require no supplemental anti-VEGF injections through 1 year, with stable visual acuity and reduced central subfield thickness in most cases. These findings support LX102 as a potential sustained therapeutic approach for nAMD. This study was registered at chinadrugtrials.org.cn (CTR20230194) and ClinicalTrials.gov (NCT06198413). Introduction Neovascular age-related macular degeneration (nAMD) is the leading cause of irreversible blindness and poses a global vision health challenge.1 Vascular endothelial growth factor (VEGF) is pivotal in driving neovascularization in AMD, making VEGF inhibitors the cornerstone of treatment.2,3 However, frequent injections result in fluctuating levels of anti-VEGF protein, leading to inconsistent exposure throughout the dosing cycle. Additionally, the need for regular clinic visits also places a considerable burden on patients, physicians, and healthcare systems,4 often resulting in suboptimal treatment compliance and an increased risk of losing the initial treatment benefits.5,6 This challenge affects both developed and low- and middle-income countries, particularly those with limited medical resources.7 It is further exacerbated by the aging population, turning the management of nAMD into an increasingly urgent public health concern. As a result, there is a strong motivation to develop therapies that provide long-term VEGF suppression. Gene therapy, which aims to deliver sustained levels of anti-VEGF directly to the retina through a “one and done” manner, presents a promising alternative to traditional protein-based therapies.8 Although initial research provided proof of concept, early attempts fell short of expectations; two key studies targeted VEGF by delivering either the endogenous antiangiogenic molecule sFLT-1 or its modified version, sFLT01, using different delivery routes: subretinal and intravitreal injections. While both approaches were found to be safe, they did not significantly improve vision or reduce retinal thickness, possibly due to insufficient transgene expression.9,10 Recently, RGX-314, the first gene therapy product to enter pivotal studies, employs rAAV8 to encode a monoclonal antibody fragment similar to ranibizumab and demonstrated safety in both subretinal and suprachoroidal deliveries. The long-term efficacy of subretinal delivery of RGX-314 remains questionable, as two-thirds of patients required supplemental injection, possibly due to targeting only VEGF-A.11 Suprachoroidal delivery has emerged as a promising, less-invasive approach; however, transduction efficiency may be limited by Bruch’s membrane, and the lack of immune privilege remains a challenge.12,13 Phase 2 study demonstrated that suprachoroidal delivery of RGX314 required higher doses up to 1.0 × 1012 GC/eye to achieve meaningful reductions in anti-VEGF injections, with 50% of patients remaining injection-free at 6 months.14 On the other hand, ADVM-022 uses an engineered adeno-associated virus type 2 (AAV2) capsid for intravitreal administration, carrying a codon-optimized cDNA of aflibercept. Its 2-year phase 1 trial showed significant benefits, with over half of the patients not needing additional treatment.15 Although intravitreal injection offers significant convenience in clinical practice, current data indicate that this procedure inevitably leads to broader exposure of the vitreous and anterior segments, thereby increasing the risks of immunotoxicity and inflammation.15,16,17 This necessitates close follow-up and complex patient management, which ultimately limits its widespread adoption in real-world clinical settings. To address the challenges faced by earlier gene therapy, it is essential to consider the limitations of therapeutic targets, administration routes, and strategies to mitigate immune-mediated responses to viral vectors.17 LX102 utilizes a non-replicating recombinant adeno-associated virus serotype 2 (rAAV2) capsid as a vector to deliver and encode VEGF-Trap, a fusion protein incorporating the second binding domain of the VEGF receptor (VEGFR)-1 and the third domain of the VEGFR-2 and the fragment crystallizable region (Fc) domain of human immunoglobulin (IgG1).10,18 The VEGF-Trap-coding sequence, driven by the ubiquitin C promoter, has been optimized for efficient expression, secretion, and sustained in vivo production. The design includes enhancements to signal peptides and regulatory elements to ensure robust therapeutic protein delivery. LX102 acts as a biofactory, continuously generating therapeutic concentrations of VEGF-Trap, which inactivates VEGF-A, VEGF-B, and placental growth factor (PlGF) via the same binding domains used by aflibercept. Optimized for subretinal administration, LX102 aims to achieve high efficacy with a reduced viral vector load and lower inflammation, potentially overcoming limitations of previous AAV-based therapies. Herein, we report the data on the development of LX102, including preclinical evaluations of safety and potential efficacy in mice and non-human primate (NHP) models, as well as present findings from a phase 1 clinical trial assessing the safety and efficacy of subretinal LX102 injection in patients with nAMD (www.chinadrugtrials.org.cn, CTR20230194; ClinicalTrials.gov, NCT06198413). Results LX102 exhibited comparable binding affinity and inhibition property to aflibercept in vitro The LX102 transgene cassette is flanked by AAV2 inverted terminal repeats, optimized for promoter, coding sequence, and signal peptides for efficient VEGF-Trap expression and secretion (Figure 1A). The potency of LX102 was confirmed by measuring transgene expression in HEK293T cells infected with varying doses of LX102 for 48 h. Aflibercept serves as a positive control. The VEGF-Trap in cell supernatant and lysate was detected by western blot assay, demonstrating a dose-dependent increase in VEGF-Trap transgene protein expression (Figures 1B–1D). Meanwhile, the amount of VEGF Trap in cell supernatant transfected with LX102 is approximately 2.9- to 25.5-fold higher than that in cell lysates normalized by beta-actin. These results demonstrate that LX102 successfully mediates delivery of the VEGF-Trap gene into HEK293T cells, leading to effective expression and secretion of the VEGF-Trap protein. We next compared LX102 transgene protein produced by infected HEK293T cells to aflibercept for their ability to bind and block VEGF in vitro. The binding affinity was measured by equilibrium binding assay, in which different concentrations of LX102 transgene protein or aflibercept were incubated with VEGF-A165 and the amount of free VEGF-A165 was measured (Figures 1E and 1F). The IC50 value of LX102 transgene protein was 55.67 pM, which is similar to that of aflibercept (49.17 pM). Meanwhile, the VEGF-inhibitory capacity of the transgene products of LX102 and aflibercept were evaluated by luciferase assay in a VEGFR2-overexpressing cell line. The luciferase activity induced by VEGF165-VEGFR2 binding, suggested that the IC50 value of LX102 transgene protein was also comparable to that of aflibercept (Figure S1). These results demonstrated that LX102 had comparable VEGF-binding affinity to aflibercept.Figure 1LX102 protein showed comparable binding affinity and inhibition property to aflibercept in vitro(A) Schematic diagram of the LX102 vector (left) and the LX102 transgene cassette (right).(B) Representative western blot showing VEGF-Trap levels in cell supernatant (CS) and cell lysates (CL) after HEK293T cells were infected with varying doses of LX102 for 48 h.(C and D) Quantitative analysis of relative VEGF-Trap expression levels in CS (C) or CL (D) across different dosage groups. Results were from three independent experiments. Statistical analysis was performed using one-way ANOVA with Bonferroni correction for multiple comparisons. ∗∗p < 0.01, ∗∗∗p < 0.001.(E) Binding rates of Eylea to VEGF165 (43.5 pM) across varying Eylea concentrations (n = 4).(F) Binding rates of LX102 transgene product to VEGF165 (43.5 pM) across varying concentrations of LX102 transgene product (n = 4).Data are presented as means ± SEM, and SEM is shown as error bars. LX102 protein showed comparable binding affinity and inhibition property to aflibercept in vitro (A) Schematic diagram of the LX102 vector (left) and the LX102 transgene cassette (right). (B) Representative western blot showing VEGF-Trap levels in cell supernatant (CS) and cell lysates (CL) after HEK293T cells were infected with varying doses of LX102 for 48 h. (C and D) Quantitative analysis of relative VEGF-Trap expression levels in CS (C) or CL (D) across different dosage groups. Results were from three independent experiments. Statistical analysis was performed using one-way ANOVA with Bonferroni correction for multiple comparisons. ∗∗p < 0.01, ∗∗∗p < 0.001. (E) Binding rates of Eylea to VEGF165 (43.5 pM) across varying Eylea concentrations (n = 4). (F) Binding rates of LX102 transgene product to VEGF165 (43.5 pM) across varying concentrations of LX102 transgene product (n = 4). Data are presented as means ± SEM, and SEM is shown as error bars. LX102 effectively inhibited neovascularization in a mouse laser-induced CNV model We first evaluated the transduction efficiency of LX102 following bilateral subretinal injection (2 × 108 vector genomes [VG]/eye) in mice (Figure 2A). Flat mounts of the retina and the retinal pigment epithelium (RPE)-choroid were prepared at 2 and 4 weeks post-administration and evaluated using fluorescence microscopy, respectively. At 2-week time following subretinal injection of LX102, substantial VEGF-Trap expression was detected in the retina and choroid, with expression levels further increased at 4 weeks (Figures 2B and 2C). We then performed an in vivo dose-escalation pharmacodynamic study of LX102, in which 6- to 8-week-old C57BL/6J mice were subretinally injected with LX102 at 1 × 107, 1 × 108, 1 × 109 VG/eye or vehicle. Laser photocoagulation was performed 2 weeks post-injection to induce choroidal neovascularization (CNV) (see Figure S2A for an illustration of AAV treatment timelines). In addition, mice were intravitreally injected with 40 μg/eye aflibercept on the same day of laser induction to serve as a positive control. Vascular leakage in the mouse laser-induced CNV model was evaluated by fundus fluorescein angiography (FFA). Subretinal injections of LX102 at all three doses reduced vascular leakage at 10 days after laser-induced injury, as did aflibercept treatment (Figures S2B and S2C). Subretinal injection of LX102 at 1 × 108 and 1 × 109 VG/eye both resulted in significant reduction in vessel leakiness across all three time points. Consistent with the FFA results, western blot analysis at 10 days post-laser injury showed that VEGF-Trap protein was expressed in the eye tissue of all LX102 groups with an obvious dose-dependent pattern, while very little VEGF-Trap protein was detected in the eye tissue of the aflibercept-treated group (Figures S2D and S2E).Figure 2Subretinal injections of LX102 in mice to evaluate the transduction and efficacy of inhibition of CNV(A) Schematic of subretinal delivery of LX102 vector to mouse eyes.(B and C) Flat mounts of the retina or RPE-choroid complexes were prepared at 2 or 4 weeks after subretinal injection of vehicle (B) or LX102 (C). Immunofluorescence (IF)-stained flat mounts (red) with antibodies against human lgG(H + L) indicate the expression of LX102 transgene product across the retina or RPE-choroid complex (n = 4 eyes per group).(D) Timeline of evaluation of the long-term efficacy of LX102 by using laser-induced CNV model in mice eyes. Laser-induced CNV was generated bilaterally, and the same animals were followed longitudinally.(E) Representative fluorescein angiograms of mice eyes treated with subretinal injection of vehicle or LX102 vector at 3, 5, 10, and 13 days post-laser photocoagulation (1 × 106, 1 × 107, 1 × 108 VG/eye; 2 μL volume), as well as no-treatment group.(F) Percentage of grade III–IV lesions without treatment or with subretinal injection of vehicle (formulation buffer) and varied doses of LX102 vector. The numbers of eyes included were 8 (no treatment), 8 (vehicle), 10 (1 × 106 VG), 9 (1 × 107 VG), and 6 (1 × 108 VG), corresponding to 32, 32, 40, 40, and 24 analyzable lesions, respectively. p values were determined by Fisher’s exact test. Compared with vehicle, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. Compared with no-treatment group, #p < 0.05, ##p < 0.01.(G) Relative VEGF-Trap mRNA expression in eyecups harvested at 13 days post-laser photocoagulation was measured by real-time qPCR and normalized to beta-actin (n = 5 eyes for no-treatment, n = 7 eyes for vehicle, n = 7 eyes for 1 × 106 VG, n = 7 eyes for 1 × 107 VG, and n = 4 eyes for 1 × 108 VG). The mRNA expression of VEGF-Trap was not detectable in no-treatment and vehicle groups. Statistical analysis was performed using one-way ANOVA followed by post-hoc Bonferroni correction for multiple comparisons. ∗∗p < 0.01.Data are expressed as mean ± SD, and SD is shown as error bars. Subretinal injections of LX102 in mice to evaluate the transduction and efficacy of inhibition of CNV (A) Schematic of subretinal delivery of LX102 vector to mouse eyes. (B and C) Flat mounts of the retina or RPE-choroid complexes were prepared at 2 or 4 weeks after subretinal injection of vehicle (B) or LX102 (C). Immunofluorescence (IF)-stained flat mounts (red) with antibodies against human lgG(H + L) indicate the expression of LX102 transgene product across the retina or RPE-choroid complex (n = 4 eyes per group). (D) Timeline of evaluation of the long-term efficacy of LX102 by using laser-induced CNV model in mice eyes. Laser-induced CNV was generated bilaterally, and the same animals were followed longitudinally. (E) Representative fluorescein angiograms of mice eyes treated with subretinal injection of vehicle or LX102 vector at 3, 5, 10, and 13 days post-laser photocoagulation (1 × 106, 1 × 107, 1 × 108 VG/eye; 2 μL volume), as well as no-treatment group. (F) Percentage of grade III–IV lesions without treatment or with subretinal injection of vehicle (formulation buffer) and varied doses of LX102 vector. The numbers of eyes included were 8 (no treatment), 8 (vehicle), 10 (1 × 106 VG), 9 (1 × 107 VG), and 6 (1 × 108 VG), corresponding to 32, 32, 40, 40, and 24 analyzable lesions, respectively. p values were determined by Fisher’s exact test. Compared with vehicle, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001. Compared with no-treatment group, #p < 0.05, ##p < 0.01. (G) Relative VEGF-Trap mRNA expression in eyecups harvested at 13 days post-laser photocoagulation was measured by real-time qPCR and normalized to beta-actin (n = 5 eyes for no-treatment, n = 7 eyes for vehicle, n = 7 eyes for 1 × 106 VG, n = 7 eyes for 1 × 107 VG, and n = 4 eyes for 1 × 108 VG). The mRNA expression of VEGF-Trap was not detectable in no-treatment and vehicle groups. Statistical analysis was performed using one-way ANOVA followed by post-hoc Bonferroni correction for multiple comparisons. ∗∗p < 0.01. Data are expressed as mean ± SD, and SD is shown as error bars. Efficacy of subretinal delivery of LX102 on laser-induced CNV lesion in nonhuman primates We further investigated the therapeutic potential of LX102 in a rhesus monkey laser-induced CNV model, given the advantage of structural similarities between nonhuman primates (NHPs) and human eyes.19 Rhesus monkeys were divided into two groups, including the LX102 group (2.5 × 1011 VG/eye) and a vehicle control group. The experiment design is shown in Figure 3A. The percentage of grade IV lesions and the fluorescein leakage areas of the retina, which reflect the severity of neovascular dysfunction, were determined by FFA (Figure 3B). For the vehicle group, the percentages of grade IV lesions were 70.8% ± 34.4%, 75.0% ± 31.9%, 75.0% ± 31.9%, and 66.7% ± 40.8%, at 2, 3, 4, and 5 weeks after laser photocoagulation, respectively. In addition, the average fluorescence leakage areas of grade IV lesions were 0.40 ± 0.12, 0.50 ± 0.14, 0.54 ± 0.33, and 0.65 ± 0.36 mm2, and the subretinal hyperreflective material (SHRM) thicknesses were 270.0 ± 72.4, 248.7 ± 90.7, 255.0 ± 124.5, and 250.0 ± 130.0 μm, respectively, with all metrics showing minimal fluctuation and no signs of self-healing. In the LX102 treatment group, percentage of grade IV lesions and SHRM thicknesses were significantly lower than those of the vehicle group from 2 to 5 weeks post-laser (p < 0.05 compared with vehicle group, Figures 3C and 3E). The average fluorescence leakage area of grade IV lesions showed a consistent reduction compared to vehicle group, although this difference did not reach statistical significance after multiple-comparison adjustment (Figure 3D). These results suggested that LX102 effectively suppresses the formation of grade IV lesions at doses of 2.5 × 1011 VG/eye.Figure 3Therapeutic potency of LX102 in an NHP laser-induced CNV model and ocular biodistribution following subretinal administration(A) Timeline of studying subretinal administration of LX102 vector at a dose of 2.5 × 1011 VG/eye to evaluate efficacy in an NHP laser-induced CNV model.(B) Representative fluorescein angiograms of rhesus monkey eyes at 2, 3, 4, and 5 weeks post-laser photocoagulation.(C–E) (C) Percentage of grade IV lesions after subretinal injection of LX102 or intravitreal injection of vehicle (n = 4 eyes per group). (D) Mean fluorescence leakage areas of grade IV lesions in LX102 or vehicle group (n = 4 eyes per group). (E) Mean maximum height of subretinal hyperreflective material (SHRM) in LX102 or vehicle group (n = 4 eyes per group). For (C)–(E), group-level summaries (left) are shown alongside individual longitudinal data from the same eyes (right). Group comparisons were performed using the Wilcoxon rank-sum test with Benjamini-Hochberg correction for multiple comparisons. Data are presented as means ± SEM, and SEM is shown as error bars. ∗p < 0.05.(F and G) Ocular biodistribution of LX102 transgene product in a separate NHP study. (F) VEGF-Trap levels in vitreous humor collected at 2, 4, 13, 19, and 26 weeks following unilateral subretinal administration of LX102 (2.5 × 1011 VG/eye). (G) VEGF-Trap levels in retina and choroid tissues collected at 13, 19, and 26 weeks post-injection. Each point represents measurement from independent animals. Data are presented as mean ± SEM calculated on log10-transformed values. Therapeutic potency of LX102 in an NHP laser-induced CNV model and ocular biodistribution following subretinal administration (A) Timeline of studying subretinal administration of LX102 vector at a dose of 2.5 × 1011 VG/eye to evaluate efficacy in an NHP laser-induced CNV model. (B) Representative fluorescein angiograms of rhesus monkey eyes at 2, 3, 4, and 5 weeks post-laser photocoagulation. (C–E) (C) Percentage of grade IV lesions after subretinal injection of LX102 or intravitreal injection of vehicle (n = 4 eyes per group). (D) Mean fluorescence leakage areas of grade IV lesions in LX102 or vehicle group (n = 4 eyes per group). (E) Mean maximum height of subretinal hyperreflective material (SHRM) in LX102 or vehicle group (n = 4 eyes per group). For (C)–(E), group-level summaries (left) are shown alongside individual longitudinal data from the same eyes (right). Group comparisons were performed using the Wilcoxon rank-sum test with Benjamini-Hochberg correction for multiple comparisons. Data are presented as means ± SEM, and SEM is shown as error bars. ∗p < 0.05. (F and G) Ocular biodistribution of LX102 transgene product in a separate NHP study. (F) VEGF-Trap levels in vitreous humor collected at 2, 4, 13, 19, and 26 weeks following unilateral subretinal administration of LX102 (2.5 × 1011 VG/eye). (G) VEGF-Trap levels in retina and choroid tissues collected at 13, 19, and 26 weeks post-injection. Each point represents measurement from independent animals. Data are presented as mean ± SEM calculated on log10-transformed values. Pharmacokinetic and biodistribution studies in NHP Pharmacokinetic and biodistribution studies of LX102 were performed in cynomolgus monkeys subretinally injected with a single dose at 2.5 × 1011 VG/eye in unilateral eyes. As shown in Table S1, the transgene DNA concentrations were concentrated in the posterior segment and extended slightly into the anterior segment, with no significant reduction in target gene concentration observed in ocular tissues up to 26 weeks post-injection. The AUClast values of the transgene DNA across ocular tissues ranked as follows: retina-choroid plexus > iris/ciliary body > vitreous > sclera > cornea > lens > optic nerve > aqueous humor, with the lowest levels found in the conjunctiva. Only minimal levels of the target gene were detected in select tissues of the non-injected eye, at concentrations markedly lower than in the injected eye. In addition, only low transgene DNA concentrations were occasionally detected in the spleen and preauricular lymph nodes, while levels in other systemic organs were generally below the lower limit of quantification, indicating minimal systemic distribution and limited permeability across the blood-brain and blood-reproductive barriers. Transgene DNA concentrations in tear fluid, nasal mucus, and oral mucus were below or near the lower limit of quantification (data not shown), suggesting low environmental shedding. Following administration, transgene product mRNA was predominantly observed in the retina-choroid plexus of the injected eye, with moderate expression in the iris/ciliary body, optic nerve, and cornea and minimal expression in the other ocular tissues (Table S2). The mean levels of LX102 transgene mRNA in the retina/choroid complex was 691.4- to 3,180.2-fold higher than in the aqueous humor and 421.2- to 36,958.0-fold higher than in the vitreous. Transgene product mRNA was not detected in the tissues of the non-injected eye or in systemic organs. The capacity of LX102 to deliver persistent and pharmacologically relevant levels of VEGF-Trap protein following subretinal injection of LX102 was further evaluated in cynomolgus monkeys by ELISA assay. The expression of VEGF-Trap protein was undetectable in the vitreous and retinal/choroidal complex of the vehicle group. In the LX102-treated group (2.5 × 1011 VG/eye), vitreous VEGF-Trap levels were robust and elevated as early as 2 weeks post-injection. The average VEGF-Trap concentrations in vitreous were 215.83 ± 164.41, 845.67 ± 508.53, 831.35 ± 606.11, 1,067.27 ± 1,465.45, and 2,811.67 ± 1,394.24 ng/mL at 2, 4, 13, 19, and 26 weeks after subretinal injection, respectively (Figure 3F). Expression of the VEGF-Trap concentrations in the retinal/choroidal complex increased from 13 weeks post-injection, with a mean concentration of 5,601.80 ± 8,459.37 ng/g, to 28,978.33 ± 29,751.72 ng/g at 26 weeks post-injection (Figure 3G). Of note, expression of VEGF-Trap protein in the retina/choroid complex was 6.1- to 10.3-fold higher than in the vitreous. Participants in phase 1 study Between February 2023 and December 2023, 12 patients were enrolled, and all subjects have completed the study with the last patient enrolled completing 1-year follow-up in January 2025. The median age of participants was 70 years (range 60∼82); 4 (33.3%) of 12 participants are female. All patients with a mean of 5.5 years since the first anti-VEGF injection and a mean of 13.9 previous injections (range 4∼45, Table 1). The mean best-corrected visual acuity (BCVA) at baseline was 45.3 ± 14.8 Early Treatment Diabetic Retinopathy Study (ETDRS) letters with a range of 21∼62 letters and the mean baseline central subfield thickness (CST) was 439.9 ± 146.1 μm (range 241∼682; Table 1).Table 1Baseline demographics and clinical characteristics of patients by patient numberPatient numberCohortGenderAge (years)Baseline ETDRSBaseline CSTAnti-VEGF injections before studyTime from last pre-study anti-VEGF injection (months)201002low doseM685963392.7201004low doseM8237346301.3201005low doseF6347682452.8201006medium doseF7146241271.4201007medium doseM7824386111.4206001medium doseM7331Nulla70.5201008high doseM685543143.0201009high doseM6062345101.4201010high doseM672154474.7201011high doseF696260942.3201012high doseF626041782.6201014high doseM754039754.2F, female; M, male.aNull indicates missing data. For subject 206001, the OCT device was damaged at baseline, which prevented the acquisition of central subfield thickness measurements. Table 1Baseline demographics and clinical characteristics of patients by patient numberPatient numberCohortGenderAge (years)Baseline ETDRSBaseline CSTAnti-VEGF injections before studyTime from last pre-study anti-VEGF injection (months)201002low doseM685963392.7201004low doseM8237346301.3201005low doseF6347682452.8201006medium doseF7146241271.4201007medium doseM7824386111.4206001medium doseM7331Nulla70.5201008high doseM685543143.0201009high doseM6062345101.4201010high doseM672154474.7201011high doseF696260942.3201012high doseF626041782.6201014high doseM754039754.2F, female; M, male.aNull indicates missing data. For subject 206001, the OCT device was damaged at baseline, which prevented the acquisition of central subfield thickness measurements. Baseline demographics and clinical characteristics of patients by patient number F, female; M, male. Null indicates missing data. For subject 206001, the OCT device was damaged at baseline, which prevented the acquisition of central subfield thickness measurements. Systemic and ocular safety The primary objective was to evaluate the safety and tolerability of LX102 up to 52 weeks. Subretinal injection of LX102 was successfully performed in all 12 patients and was confirmed by direct visualization and intraoperative optical coherence tomography (OCT). The procedure was well tolerated in all treated patients, and the subretinal bleb was absorbed within 24 h. No dose-limiting toxicity was reported. LX102 was well tolerated at dose concentrations ranging from 2 × 1010 to 1.25 × 1011 vector genomes per eye with one single subretinal injection to participants with nAMD. No LX102-related systemic adverse events were noted. There were no safety signals related to laboratory evaluations or vital signs during the trial. There were no clinically observed immune responses and no post-surgical inflammation beyond what was expected following vitrectomy surgery (Table S3). All participants experienced at least one treatment-emergent adverse event (TEAE), with the majority being of mild to moderate severity. Overall, there were 9 kinds of ocular TEAEs in 12 participants. Most events were mild or moderate in severity, and no serious ocular TEAEs were observed. The most common TEAEs were post-procedure conjunctival hyperemia, which occurred in 11 of 12 participants due to surgical procedure (Table 2). Other ocular TEAEs included conjunctivitis in 5 of 12 study eyes, elevation of intraocular pressure (IOP) in 2 (16.7%) of 12 study eyes, and decreased IOP in 1 (8.3%) of 12 study eyes, which were assessed to be related to the surgical procedure. Aggravation of cataract occurred in 9 (75.0%) of 12 study eyes, and dry eye symptoms occurred in 2 (16.7%) of 12 study eyes. These ocular TEAEs were considered mild or moderate in severity and unrelated to LX102, most of which were resolved within days to weeks without sequalae. The only adverse event related to LX102 was recurrent uveitis in one patient from the high-dose cohort, characterized by 3 episodes of mild keratic precipitates, anterior chamber cells, and anterior chamber flare, with minimal anterior vitreous cells in one of the 3 episodes. The signs of uveitis were detected at weeks 8, 28, and 44 post-LX102 administration, respectively. All of the signs and symptoms were reversible following up to 2 weeks of topical corticosteroid treatment with tobramycin/dexamethasone eye drops (3–4 times daily), while no vision loss compared to baseline through latest visit. Topical steroids were tapered after adequate control of inflammation, with regular IOP monitoring and management of steroid-induced ocular hypertension as needed. Reduced visual acuity, defined as a decrease of ≥15 ETDRS letters from baseline, was observed in 3 (25%) of the 12 study eyes, 2 in medium-dose cohort and 1 in high-dose cohort, and was reported as mild or moderate in severity and considered unrelated to LX102 (cutoff 2024/12/31).Table 2Treatment-emergent ocular adverse events in the study eye by preferred termCohort 1 (n = 3)2 × 1010VGCohort 2 (n = 3)5 × 1010 VGCohort 3 (n = 6)1.25 × 1011 VGTotalReduced or impaired visual acuity02 (66.7%)1 (16.7%)3 (25%)Intraocular pressure increased1 (33.3%)01 (16.7%)2 (16.7%)Intraocular pressure decreased001 (16.7%)1 (8.3%)Conjunctival hyperemia3 (100%)2 (66.7%)6 (100%)11(91.7%)Conjunctivitis2 (66.7%)1 (33.3%)2 (33.3%)5 (41.7%)Dry eye1 (33.3%)1 (33.3%)02 (16.7%)Macular hole01 (33.3%)01 (8.3%)Cataract1 (33.3%)2 (66.7%)6 (100%)9 (75.0%)Intraocular inflammationa001 (16.7%)1 (8.3%)Data are presented as n (%).Participants are counted once for each preferred term regardless of the number of events.Preferred terms are sorted by decreasing frequency in the total group.Adverse event reported terms were coded using Medical Dictionary for Regulatory Activities (MedDRA) 26.1.aInflammation includes anterior chamber cell, anterior chamber flare, anterior chamber inflammation, and vitreous cell. Table 2Treatment-emergent ocular adverse events in the study eye by preferred termCohort 1 (n = 3)2 × 1010VGCohort 2 (n = 3)5 × 1010 VGCohort 3 (n = 6)1.25 × 1011 VGTotalReduced or impaired visual acuity02 (66.7%)1 (16.7%)3 (25%)Intraocular pressure increased1 (33.3%)01 (16.7%)2 (16.7%)Intraocular pressure decreased001 (16.7%)1 (8.3%)Conjunctival hyperemia3 (100%)2 (66.7%)6 (100%)11(91.7%)Conjunctivitis2 (66.7%)1 (33.3%)2 (33.3%)5 (41.7%)Dry eye1 (33.3%)1 (33.3%)02 (16.7%)Macular hole01 (33.3%)01 (8.3%)Cataract1 (33.3%)2 (66.7%)6 (100%)9 (75.0%)Intraocular inflammationa001 (16.7%)1 (8.3%)Data are presented as n (%).Participants are counted once for each preferred term regardless of the number of events.Preferred terms are sorted by decreasing frequency in the total group.Adverse event reported terms were coded using Medical Dictionary for Regulatory Activities (MedDRA) 26.1.aInflammation includes anterior chamber cell, anterior chamber flare, anterior chamber inflammation, and vitreous cell. Treatment-emergent ocular adverse events in the study eye by preferred term Data are presented as n (%). Participants are counted once for each preferred term regardless of the number of events. Preferred terms are sorted by decreasing frequency in the total group. Adverse event reported terms were coded using Medical Dictionary for Regulatory Activities (MedDRA) 26.1. Inflammation includes anterior chamber cell, anterior chamber flare, anterior chamber inflammation, and vitreous cell. Biodistribution For vector shedding analysis, serum and tears samples (from both study and fellow eyes) were assayed using a quantitative PCR method at time points of baseline, day 1, day 2, day 3, week 4, and week 36 after LX102 subretinal administration. The PCR did not reveal any evidence of shedding vector in serum and tear samples from fellow eyes except for a transient detection of AAV vector in 1 fellow eye in the high-dose cohort at day 2 post-LX102 administration, whereas in tear samples from study eyes, the vector was transiently detectable at days 1–3 in 11 (91.7%) of 12 participants (tear samples were not collected due to dry eye in 1 participant) (Figure S3). However, by the time point of week 4, all the positive shedding vector in tears of study eye turned to be negative. The enzyme-linked immunospot assay (ELISpot) results showed that 3 (25%) participants had a cell-mediated immune response to the AAV2 capsid (positive cells were detected at baseline in all 3 participants), and none of the 12 participants showed a positive response to the anti-VEGF protein. No positive sample result corresponded to a clinical finding (i.e., ocular inflammation or other ocular adverse event) (Figure S4). LX102 protein was examined in aqueous samples at baseline and at 36 weeks after LX102 injection in all cohorts. The mean aqueous LX102 protein concentrations in cohorts 1–3 was 155.1 ng/mL (range, 4.9–807.4) at week 36, excluding one patient with supplemental injection. Immune response to AAV2 Neutralizing antibodies to AAV2 and anti-drug antibody (ADA) to AAV2 in the serum were evaluated at time points of day 0, day 14, week 4, week 12, week 36, and week 52 after LX102 administration. One participant from the high-dose cohort showed significantly elevated anti-AAV2 neutralizing antibodies and ADA to AAV2 starting at week 4 and has not recovered to baseline at week 36 visit. This participant developed recurrent uveitis at weeks 8, 28, and 44 post-LX102 administration. However, this patient maintained improved BCVA and retinal structure after LX102 administration, suggesting against a negative effect on therapy. No elevated anti-AAV2 neutralizing antibodies and ADA were obviously detected in the other 11 (91.7%) of 12 participants (Figure S5). Efficacy The BCVA in all participants remained stable, with a mean of 45.3 ± 14.8 ETDRS letters at baseline to 46.9 ± 15.6 ETDRS letters at the latest visit (Figures 4A–4C). BCVA was maintained (±5 ETDRS letters) in 7 (58%) of 12 participants and improved (gain ≥5 ETDRS letters) in 2 (16.7%) of 12 participants treated with LX102 compared with baseline visit. Only one from the low-dose cohort lost 9 letters during the 1-year follow-up. Transient vision reduction in two other participants was considered possibly related to the fluid-air exchange during surgery, with subsequent recovery. Separately, one patient experienced moderate vision loss related to post-vitrectomy cataract at week 20 after LX102 administration, which recovered following cataract surgery. Reductions in CST were graded by an independent reading center (Figures 4B–4D). The mean CST for all patients across the three cohorts was 439.9 ± 146.1 μm at baseline, decreasing to 267.5 ± 53.0 μm at latest visit. FFA assessment showed no recurrence of leakage during the year in 11 of the 12 (91.7%) treated patients; in one patient, increased leakage during follow-up resulted in rescue anti-VEGF injections.Figure 4Longitudinal efficacy measures of LX102(A and B) Longitudinal change of mean BCVA (A) and mean CST (B) following subretinal LX102 administration in all 3 cohorts of 12 patients. (A) Individual measurements of best-corrected visual acuity (BCVA) by Early Treatment Diabetic Retinopathy Study (ETDRS) letters read displayed for all patients from baseline up to 52 weeks after subretinal injection of LX102. (B) Individual measurements of central subfield thickness (CST), as graded by independent graders, are displayed for all patients from baseline up to 52 weeks after subretinal injection of LX102.(C and D) Mean BCVA and CST over 52 weeks for each cohort. Longitudinal efficacy measures of LX102 (A and B) Longitudinal change of mean BCVA (A) and mean CST (B) following subretinal LX102 administration in all 3 cohorts of 12 patients. (A) Individual measurements of best-corrected visual acuity (BCVA) by Early Treatment Diabetic Retinopathy Study (ETDRS) letters read displayed for all patients from baseline up to 52 weeks after subretinal injection of LX102. (B) Individual measurements of central subfield thickness (CST), as graded by independent graders, are displayed for all patients from baseline up to 52 weeks after subretinal injection of LX102. (C and D) Mean BCVA and CST over 52 weeks for each cohort. Before enrollment, most patients required extensive treatment with anti-VEGF (average 13.9 previous injections). During the 52-week follow-up period, each patient was evaluated every 2 months to assess the need for supplemental aflibercept injections (starting after week 12). Rescue criteria were prespecified based on BCVA, OCT, and fundus photography (FP) assessment. Of the 12 patients, 11 (91.7%) required no supplemental injections. Only one from low-dose cohort required 7 aflibercept supplemental injections. Figure 5 shows longitudinal OCTs, BCVA, and supplemental injections needed for study eyes in cohorts and illustrates the range of responses to subretinal injection of LX102.Figure 5Longitudinal fundus photographs, OCT scans, and BCVA (ETDRS letters) of three patients treated with LX102(Left) Male aged 82 years in low-dose cohort (dose 2 × 1010 VG/eye). This participant received 30 anti-VEGF-A injections since neovascular AMD diagnosis, six within the 12 months before LX102 administration; did not require any supplemental anti-VEGF-A intravitreal injections; and maintained stable BCVA letters. (Middle) Female aged 71 years in medium-dose cohort (dose 5 × 1010 VG/eye). This participant received 27 anti-VEGF-A injections since neovascular AMD diagnosis, three within the 12 months before LX102 administration; did not require any supplemental anti-VEGF-A intravitreal injections; and had a change in BCVA of +12 letters. (Right) Female aged 69 years in high-dose cohort (dose 1.25× 1011 VG/eye). This participant received five anti-VEGF-A injections within the 12 months before LX102 administration, did not require any supplemental anti-VEGF-A intravitreal injections, and had a change in BCVA of −3 letters. Longitudinal fundus photographs, OCT scans, and BCVA (ETDRS letters) of three patients treated with LX102 (Left) Male aged 82 years in low-dose cohort (dose 2 × 1010 VG/eye). This participant received 30 anti-VEGF-A injections since neovascular AMD diagnosis, six within the 12 months before LX102 administration; did not require any supplemental anti-VEGF-A intravitreal injections; and maintained stable BCVA letters. (Middle) Female aged 71 years in medium-dose cohort (dose 5 × 1010 VG/eye). This participant received 27 anti-VEGF-A injections since neovascular AMD diagnosis, three within the 12 months before LX102 administration; did not require any supplemental anti-VEGF-A intravitreal injections; and had a change in BCVA of +12 letters. (Right) Female aged 69 years in high-dose cohort (dose 1.25× 1011 VG/eye). This participant received five anti-VEGF-A injections within the 12 months before LX102 administration, did not require any supplemental anti-VEGF-A intravitreal injections, and had a change in BCVA of −3 letters. Discussion Our studies demonstrate that the LX102 gene therapy has a favorable safety profile, with minimal or no intraocular inflammation or clinically evident immune response, as shown in preclinical and phase 1 clinical data. Furthermore, this rAAV2-based gene therapy LX102, which encodes a VEGF-Trap, offers preliminary evidence of improvements in anatomic structures with maintained visual outcomes. Notably, subretinal injections of LX102 effectively controlled clinically relevant exudation in most patients without requiring supplementary treatments. These data are encouraging and indicate that subretinal AAV2-LX102 injection could effectively address compliance challenges by providing sustained anti-VEGF protein with a single administration, preserving the initial treatment benefits. Collectively, the findings highlight LX102 as a promising and durable therapy option for wet macular degeneration. Our preclinical studies demonstrated that LX102 was efficiently transduced into the retina and RPE-choroid tissues by subretinal injection. AAV2-mediated expression of VEGF-Trap was maintained for at least 6 months by subretinal delivery in C57BL/6J mice. In a mouse laser-induced CNV model, subretinal administration of LX102 at a low dose of 1.0 × 107 GC/eye significantly prevented the formation and progression of grade IV CNV lesions and exhibited a significantly increased reduction in the leakage area of IV CNV lesions compared with vehicle. It is worth noting that species-dependent differences in VEGF ligand structure and receptor activation constitute an inherent limitation of murine models for evaluating human VEGF-targeted therapies.20,21 While receptor-based VEGF-Trap binds VEGF-A and PlGF isoforms across multiple mammalian species with similar high affinity, humanized monoclonal antibodies such as ranibizumab and bevacizumab were found to fail to efficiently neutralize murine VEGF-A and, therefore, show limited activity in mouse models.10,22,23,24 Within this context, the CNV model serves an appropriate proof-of-concept platform for evaluating LX102, while NHP studies are essential for translational relevance. Notably, through pharmacokinetic and biodistribution studies in NHPs, we found that the majority of LX102 transgene product was retained in the retina, with only a small amount spreading outward. This biodistribution differs from intravitreal injections of anti-VEGF antibodies or fusion proteins, which generate transient high vitreous peaks but typically achieve no more than 10% of that concentration in the retina and choroid.25 The level of the LX102 transgene product (VEGF-Trap) was 6.1- to 10.3-fold higher in the retinal/choroidal complex than in the vitreous, suggesting higher accumulation of transgene product in the posterior segments. As a result of this remarkable retinal retention, LX102 at a dose of 2.5 × 1011 VG/eye exhibited significant inhibition of the grade IV lesions in the laser-induced CNV NHP model. Furthermore, in a long-term transgene expression assay in NHP, LX102 protein was observed to be persistently expressed in the vitreous and retina/choroid tissue for 26 weeks. The VEGF-Trap levels measured in vitreous of LX102-treated eyes at 2.5 × 1011 VG/eye at 26 weeks post-injection (2.9 ± 1.2 μg/mL; mean ± SEM) were comparable to levels described by Grishanin et al.26 from the vitreous of African green monkey eyes measured 28 weeks after intravitreal administration of ADVM-022 (AAV2.7m8-C11.CO.aflibercept) at 2.0 × 1012 VG/eye (3.4 ± 0.5 μg/mL; mean ± SEM). Corresponding aqueous humor levels in ADVM-022-treated NHPs (0.7 ± 0.2 μg/mL) were similar to those measured in NHPs 2 weeks after intravitreal aflibercept (IVA), and in nAMD patients 1 month after IVA,27,28 indirectly suggesting that LX102 achieves therapeutically relevant VEGF-Trap concentrations. While vitreous or aqueous concentrations provide a pharmacokinetic reference, they may not be considered a direct surrogate for therapeutic efficacy, which depends primarily on sustained expression in the posterior segment. Despite the enormous advantages of rAAV-based gene therapy for ocular diseases, rAAV-mediated gene delivery exhibits potential dose-dependent toxicity, thus raising concerns about high-dose rAAV administration and extraocular vector distribution.29,30 The LX102 transgene expression cassette is optimized with coding sequences, signal peptides, as well as optimal combination of transcription and translation regulatory elements to enhance sustained expression of VEGF-Trap protein in vivo at reasonably lower dose range. The vector design uses a UbC promoter instead of the commonly used CMV/CAG promoter, which is highly CpG rich and optimizes the coding sequence for humanized codons with reduced CpG content. Since CpG hypermethylation of promoters is a well-known hallmark of gene silencing, while CpG island DNA sequences can serve as a stimulatory signal for TLR9-mediated innate immune responses,31,32,33,34 this strategy is designed to promote efficient expression in human cells while minimizing gene silencing and nucleic acid-induced immune responses. Our biodistribution study in NHPs indicated that the transgene product following subretinal injection of LX102 was predominantly localized to ocular tissues, with minimal detection in non-ocular tissues and none in reproductive organs. These results support LX102 as a potential low-dose, one-time subretinal treatment for nAMD. All participants experienced at least one TEAE, with the majority graded as either mild or moderate in severity, consistent with previously reported subretinal gene therapy studies.11,35 Importantly, none of the serious AEs were reported during at least 9 months’ follow-up, including intraocular inflammation, retinal detachment, endophthalmitis, and vision loss. In our study, the most common ocular TEAEs were conjunctival congestion, observed in 11 participants (91.7%) and attributed to the surgical procedure. AAV-induced inflammation is one of the most critical safety concerns in gene therapy; the only LX102-related adverse event was recurrent uveitis in one high-dose patient, with all symptoms resolving within 2 weeks of topical steroid treatment and no vision loss observed during follow-up. Recent advances in anti-VEGF drugs have enhanced safety and durability.36,37 While gene therapy holds the promise of “delivering one-time, long-lasting benefits,” unlike rare diseases with no treatments, its safety is crucial for AMD, where existing therapies are available, making it essential for approval and acceptance. The current trial offers important insights into the safety profile of AAV injections into the subretinal space, which seem to reduce immune system exposure compared to injections into the vitreous cavity. Although intravitreal injections are easier to perform, they typically require a higher vector dose and carry a greater risk of off-target effects, which can lead to increased inflammation.38 While most conditions are manageable, the necessity for close patient monitoring diminishes the convenience advantage offered by gene therapy. Reduced visual acuity occurred in 3 of the 12 study eyes, with 2 cases attributed to gas tamponade during the surgery, both of which self-recovered within 2 weeks, while the remaining case was due to post-vitrectomy cataract. Vision in the latter case recovered 9 months post-vitrectomy, with an improvement of 7 letters compared to baseline after cataract surgery. It is worth noting that a macular hole developed in one subject from the medium-dose cohort 2 weeks post-surgery. We speculate that the macular hole was most likely caused by strong adhesion between the posterior hyaloid membrane and the fovea during the posterior vitreous detachment procedure, which may have led to increased postoperative traction from fibrous tissue, ultimately resulting in the formation of the macular hole. This finding highlights the importance of assessing posterior hyaloid membrane adhesion with enhanced vitreous imaging OCT scans before surgery, particularly for patients with thin macula, to reduce the risk of macular hole formation. Moreover, RPE atrophy near the injection site has emerged as a critical safety concern in the field of retinal gene therapies39; for instance, pigmentary changes were observed in 69% of participants (29/42) in the RGX314 trial, potentially due to surgical mechanical damage, AAV vector toxicity, or individual immune responses.11 In our study, however, no concerning cases of RPE atrophy following subretinal injections of LX102 have been observed so far, except at the touchdown site of the subretinal cannula. This may be at least partially attributed to the optimized vector design, which limits potential immune responses while maintaining efficient recombinant protein secretion. Additionally, the injection protocol, which includes careful selection of the injection site, appropriate injection pressure, lower infusion pressure, pneumatic delivery, and microscope-integrated OCT monitoring, helps to minimize direct injury to the RPE. Specifically, to mitigate the risk of RPE atrophy, injections should be placed 3–4 optic disc diameters away from the fovea within relatively normal RPE regions, preferably between the retinal artery and retinal vein in the supratemporal or inferior temporal vascular arcades. Maintaining an injection pressure no more than 20–30 psi and halting the injection as the bleb approaches the macular region may also help reduce the risk of RPE atrophy.40 Additionally, anatomical differences in the Bruch’s membrane and lipofuscin pigments of RPE between Asian and Caucasian populations, may also influence susceptibility to such effects.41,42 Overall, our preliminary findings suggest that LX102 is safe and well tolerated, supporting its continued use in future trials to further investigate its efficacy outcomes. During the 52-week follow-up period, visual acuity was well maintained in 10 study eyes, while 11 patients showed a favorable and sustained reduction of the intraretinal or subretinal fluid. These efficacy findings are encouraging, especially considering that 11 of 12 patients did not require additional anti-VEGF-neutralizing protein therapy during at least 1 year of follow-up. These data suggest that the optimization of the plasmid’s promoter and other structures has enhanced the transfection efficiency. The single patient who required five supplementary injections was diagnosed with polypoidal choroidal vasculopathy (PCV) based on preoperative indocyanine green angiography,43 and baseline OCT revealed RPE tear within the macular region. Although refractory PCV cases often experience persistent subretinal or intraretinal fluid despite monthly anti-VEGF injections, it is noteworthy that, among the 7 PCV patients in our study, only this one required supplemental treatment. This suggests that RPE damage, rather than the PCV itself, is likely a key factor contributing to the need for frequent injections, highlighting the importance of preoperative fluorescein angiography assessments of RPE integrity. Patients with RPE tears may not be optimal candidates for similar gene therapies. Interestingly, our data indicated that there was no strong correlation between LX102 concentrations in the aqueous humor and therapeutic response. Unlike the RGX-314 study, where a relatively higher protein concentration in the aqueous humor was suggested as a potential efficacy indicator,11 our patients achieved fluid resolution despite relatively low aqueous concentrations. This difference may be attributed to an additional lavage of intravitreal fluid following by fluid-air exchange used during our surgical procedure, aimed at minimizing reflux of the therapeutic solution into the vitreous cavity and reducing off-target expression in the anterior segment. Additionally, the relatively higher aqueous protein concentration reported by RGX314 might be influenced by the pseudophakic status of the enrolled patients, as zonular relaxation following cataract surgery may facilitate protein diffusion into the anterior chamber. Along with the high retinal retention observed in our preclinical NHP models, these findings reinforce that aqueous concentrations are not a reliable surrogate for efficacy of subretinal gene therapies. The sustained fluid reduction and stable visual acuity in our patients were likely driven by localized transgene expression within the target tissues. In conclusion, the translational work presented here suggests that LX102 may offer therapeutic benefits through the robust expression and efficacy of a VEGF-Trap delivered via subretinal administration of a gene therapy vector. Preclinical evidence from studies in mice and NHPs, along with early clinical data from patients with nAMD, showed that LX102 was generally well tolerated, and control of exudation was observed in all dose cohorts for up to 1 year, informing the approach for the pivotal program. Ongoing randomized, controlled phase 2 trial of LX102 in patients with nAMD will provide further insight into the clinical effects of this unique agent (ClinicalTrials.gov, NCT06196840). Limitations of the study This study has limitations inherent in an early-phase study with a small sample size and no control group. Additionally, it is worth noting that all 12 subjects in this study were phakic, as vitrectomy can significantly increase the risk of postoperative cataract formation, which could impact the assessment of post-treatment vision. Therefore, the actual visual benefits for the subjects in this study may be greater than those observed in the study. In the subsequent phase 2 study, we included pseudophakic status as a stratification factor to further evaluate the necessity of prophylactic cataract removal surgery in subretinal gene therapy. Resource availability Lead contact Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Xiaodong Sun (xdsun@sjtu.edu.cn). Materials availability Biotin-anti-hVEGFR1 generated in this study will be made available on request from the lead contact with a completed materials transfer agreement. Data and code availability •Participant-level clinical data are deposited to Electronic Data Capture Database (managed by the Happy Life Tech. Inc.) and can be accessed via the lead contact by reasonable request.•This paper does not report any original code.•Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request. •Participant-level clinical data are deposited to Electronic Data Capture Database (managed by the Happy Life Tech. Inc.) and can be accessed via the lead contact by reasonable request. Participant-level clinical data are deposited to Electronic Data Capture Database (managed by the Happy Life Tech. Inc.) and can be accessed via the lead contact by reasonable request. •This paper does not report any original code. This paper does not report any original code. •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request. Acknowledgments We thank all the patients participating in this study. This work was supported by the 10.13039/100014718National Natural Science Foundation of China (82388101, 82571246, and 82471130), 10.13039/501100003399Science and Technology Commission of Shanghai Municipality (23J41900200), and 10.13039/100017950Shanghai Municipal Health Commission (GWVI-11.2-XD18). Innostellar Biotherapeutics provided funding sources for the phase I trial. Author contributions X.S., F.W., H.J., T.L., and Z.N. conceptualized the preclinical and phase 1 clinical projects. X.S., T.L., J.F., H.W., J.S., X.Y., S.Y., J.C., Y.W., P.H., and K.Z. enrolled participants in the clinical study and supervised the clinical components of the trial (clinical review, safety, and dose monitoring). H.J., T.L., Y.L., and M.S. performed statistical planning and analysis of the clinical study, interpreted the results, and critically revised the manuscript. F.W., Z.N., Y.L., and T.L. oversaw the cynomolgus monkey study and data analysis and interpreted results. F.W., Z.N., Y.L., M.S., T.L., and J.S. were involved in the design and interpretation of the preclinical studies and prepared the preclinical portion of the manuscript. The original manuscript was written by X.S., H.J., T.L., and J.F. and edited by J.S., F.W., Z.N., X.Y., S.Y., Y.L., M.S., P.H., and Y.W. All authors reviewed the manuscript, approved the final version, and agreed to be accountable for all aspects of the work. Declaration of interests X.S. reported being consultant of Roche, Alcon, Allergan, Bayer Healthcare, Boehringer Ingelheim, Innovent Biologics Inc., and Kanghong Biotech Inc. F.W. reported being the founder of Innostellar Biotherapeutics. Z.N., Y.L., and M.S. are employees of Innostellar Biotherapeutics. T.L. reported being consultant of Innostellar Biotherapeutics. Innostellar Biotherapeutics provided funding sources for the phase 1 trial. F.W. and Z.N. were named as inventors in the following patent application related to LX102: China National Intellectual Property Administration, 202311513893.5. STAR★Methods Key resources table REAGENT or RESOURCESOURCEIDENTIFIERAntibodiesHRP-conjugated goat anti-human IgG (H + L) antibodyYeasenCat#33501ES60: RRID:AB_3751338β-Actin Mouse Monoclonal AntibodyBeyotimeCat#AF0003: RRID:AB_2893353HRP-conjugated goat anti-mouse IgGBeyotimeCat#A0216; RRID:AB_2860575Alexa Fluor 594 AffiniPure Goat Anti- Human IgG(H + L)YeasenCat#33222ES60; RRID:AB_3751339Goat Anti-Human IgG, Monkey ads-UNLBSouthernBiotechCat#2049-01; RRID:AB_2795694Biotin-anti-hVEGFR1This paperN/ABiological samplesBlood samplesPatients in this studyN/ATear samplesPatients in this studyN/AChemicals, peptides, and recombinant proteinsEylea® (aflibercept)Bayer PharmaceuticalsN/ARecombinant Human VEGF165(His Tag)East-Mab Biomedical TechnologyCat#Y04702Tropicamide Phenylephrine Eye DropsSanten PharmaceuticalN/A0.4% oxybuprocaine hydrochloride eye dropsSanten PharmaceuticalN/AOxyfloxacin eye ointmentSINQIN/AFluorescein sodiumAlconCat#110LEFetal bovine serumSigmaCat#F8687Penicillin/streptomycinGibcoCat#15140-122DMEMBasalmediaCat#L110KJNativePAGE Sample Buffer (4X)InvitrogenCat#BN2003Spectra Multicolor Broad Range Protein LadderThermo ScientificCat#266238%GLASS Gel Tris-Glycine gelShareBioCat#SB-TG1000815PVDF Blotting MembraneMilliporeCat#ISEQ00010Triton X-100SigmaCat#T9284TRIzolInvitrogenCat#15596026Bovine serum albuminBeyotimeCat#ST023DAPIYeasenCat#40728ES03Fluoromount-GSouthernbiotechCat#0100-01Zoletil®50Virbac LabN/ACritical commercial assaysBio-GloTM Luciferase Assay SystemPromegaCat#G7940Human quantitative VEGF ELISA kitSino BiologicalCat#KIT11066Super ECL Detection ReagentYeasenCat#36208ES76PrimeScript RT reagent Kit with gDNA eraserTakaraCat#RR047ATB Green Premix Ex Taq IITakaraCat#RR820APremix Ex Taq(Probe qPCR)TakaraCat#RR390AMagNA Pure 96 DNA and Viral NA Large Volume KitRocheCat#06374891001MagNA Pure 96 DNA and Viral NA Small Volume KitRocheCat#06543588001Experimental models: Cell linesHEK293T cellsATCCCRL3216H-VEGF reporter 293 cell lineGenomeditechGM-C09057Experimental models: Organisms/strainsC57BL/6J miceGemPharmatechN/ARhesus monkeysGuangxi Xiongsen Primate Laboratory Animal Breeding Development Co., Ltd.N/ACynomolgus monkeyHainan Jingang Biotech Co., Ltd.N/AOligonucleotidesPrimers: VEGF-Trap Forward: GACTTCAACTGGGAGTACThis paperN/APrimers: VEGF-Trap Reverse:AGTGCTCAGGAATTTCTT-3This paperN/APrimers: beta-actin Forward: CTGGAACGGTGAAGGTGACAThis paperN/APrimers: beta-actin Reverse:GGGATGTTTGCTCCAACCAAThis paperN/ARecombinant DNApAAV-Ubc-VEGF trap-pAThis paperN/ASoftware and algorithmsImageJNIHhttps://imagej.nih.gov/ij/GraphPad Prism 10.0GraphPad Softwarehttps://www.graphpad-prism.cn/SAS® Version 9.4.SAS Institute Inc.https://support.sas.com/software/94/HEYEX Version 3.2cHeidelberg Engineering GmbHhttps://www.heidelbergengineering.com/en-US/products/heidelberg-eye-explorerOther10-μL syringeWorld Precision InstrumentsNanofil36-gauge blunt needleWorld Precision InstrumentsCat#NF36BL38-gauge subretinal cannulaMedOne SurgicalCat#3255MicroDose™ Injection KitMedOne SurgicalCat#3275 REAGENT or RESOURCESOURCEIDENTIFIERAntibodiesHRP-conjugated goat anti-human IgG (H + L) antibodyYeasenCat#33501ES60: RRID:AB_3751338β-Actin Mouse Monoclonal AntibodyBeyotimeCat#AF0003: RRID:AB_2893353HRP-conjugated goat anti-mouse IgGBeyotimeCat#A0216; RRID:AB_2860575Alexa Fluor 594 AffiniPure Goat Anti- Human IgG(H + L)YeasenCat#33222ES60; RRID:AB_3751339Goat Anti-Human IgG, Monkey ads-UNLBSouthernBiotechCat#2049-01; RRID:AB_2795694Biotin-anti-hVEGFR1This paperN/ABiological samplesBlood samplesPatients in this studyN/ATear samplesPatients in this studyN/AChemicals, peptides, and recombinant proteinsEylea® (aflibercept)Bayer PharmaceuticalsN/ARecombinant Human VEGF165(His Tag)East-Mab Biomedical TechnologyCat#Y04702Tropicamide Phenylephrine Eye DropsSanten PharmaceuticalN/A0.4% oxybuprocaine hydrochloride eye dropsSanten PharmaceuticalN/AOxyfloxacin eye ointmentSINQIN/AFluorescein sodiumAlconCat#110LEFetal bovine serumSigmaCat#F8687Penicillin/streptomycinGibcoCat#15140-122DMEMBasalmediaCat#L110KJNativePAGE Sample Buffer (4X)InvitrogenCat#BN2003Spectra Multicolor Broad Range Protein LadderThermo ScientificCat#266238%GLASS Gel Tris-Glycine gelShareBioCat#SB-TG1000815PVDF Blotting MembraneMilliporeCat#ISEQ00010Triton X-100SigmaCat#T9284TRIzolInvitrogenCat#15596026Bovine serum albuminBeyotimeCat#ST023DAPIYeasenCat#40728ES03Fluoromount-GSouthernbiotechCat#0100-01Zoletil®50Virbac LabN/ACritical commercial assaysBio-GloTM Luciferase Assay SystemPromegaCat#G7940Human quantitative VEGF ELISA kitSino BiologicalCat#KIT11066Super ECL Detection ReagentYeasenCat#36208ES76PrimeScript RT reagent Kit with gDNA eraserTakaraCat#RR047ATB Green Premix Ex Taq IITakaraCat#RR820APremix Ex Taq(Probe qPCR)TakaraCat#RR390AMagNA Pure 96 DNA and Viral NA Large Volume KitRocheCat#06374891001MagNA Pure 96 DNA and Viral NA Small Volume KitRocheCat#06543588001Experimental models: Cell linesHEK293T cellsATCCCRL3216H-VEGF reporter 293 cell lineGenomeditechGM-C09057Experimental models: Organisms/strainsC57BL/6J miceGemPharmatechN/ARhesus monkeysGuangxi Xiongsen Primate Laboratory Animal Breeding Development Co., Ltd.N/ACynomolgus monkeyHainan Jingang Biotech Co., Ltd.N/AOligonucleotidesPrimers: VEGF-Trap Forward: GACTTCAACTGGGAGTACThis paperN/APrimers: VEGF-Trap Reverse:AGTGCTCAGGAATTTCTT-3This paperN/APrimers: beta-actin Forward: CTGGAACGGTGAAGGTGACAThis paperN/APrimers: beta-actin Reverse:GGGATGTTTGCTCCAACCAAThis paperN/ARecombinant DNApAAV-Ubc-VEGF trap-pAThis paperN/ASoftware and algorithmsImageJNIHhttps://imagej.nih.gov/ij/GraphPad Prism 10.0GraphPad Softwarehttps://www.graphpad-prism.cn/SAS® Version 9.4.SAS Institute Inc.https://support.sas.com/software/94/HEYEX Version 3.2cHeidelberg Engineering GmbHhttps://www.heidelbergengineering.com/en-US/products/heidelberg-eye-explorerOther10-μL syringeWorld Precision InstrumentsNanofil36-gauge blunt needleWorld Precision InstrumentsCat#NF36BL38-gauge subretinal cannulaMedOne SurgicalCat#3255MicroDose™ Injection KitMedOne SurgicalCat#3275 Experimental model and study participant details Cell culture and cell lines The HEK293T cell line was purchased from ATCC (#CRL3216), and the H-VEGF reporter 293 cell line was purchased from Genomeditech (#GM-C09057). All these cells were cultured in specialized growth media according to their product information. Additionally, all cells were incubated in humidified incubators maintained at 37°C with 5% CO2. The HEK293T cell line was tested negatively for mycoplasma contamination, while the H-VEGF reporter 293 cell line was not tested for mycoplasma. Neither cell line underwent additional short tandem repeat authentication for this study. Animals Six-to eight-week-old male (2-week post-SR administration efficacy study) or 6-month-old female (6-month post-SR administration study) C57BL/6J mice were obtained from GemPharmatech Co., Ltd. The C57BL/6J mice were housed under specific pathogen-free (SPF) conditions (22°C–26°C with humidity at 45%–55%) at PharmaLegacy Laboratories (Shanghai) Co., Ltd., with a 12-h light/12-h darkness cycle. Mice were fed normal chow. At the end of the study, the animals were deeply anesthetized and euthanized by carbon dioxide exposure. The animal experimental procedures were approved by the Animal Care and Use Committee of Innostellar Biotherapeutics (approval numbers: LA202207013 and LA202409011). Male rhesus monkeys of 3.5–6.5-year-old with weight ranging from 4.16 to 6.26 kg were obtained from Guangxi Xiongsen Primate Laboratory Animal Breeding Development Co., Ltd., and used for pharmacology studies. Cynomolgus monkeys of three to six-year-old with weight ranging from 2.35 to 6.35 kg were purchased from Hainan Jingang Biotech Co., Ltd., and used for comprehensive pharmacokinetic and biodistribution studies (n = 30, 15 males and 15 females). The rhesus monkeys were housed in Joinn Laboratories (China) Co., Ltd., while cynomolgus monkeys were housed in Tianfu Drug Research Center of West China-Frontier PharmaTech Co., Ltd. (WCFP). Both of the institutions are accredited by AAALAC International (Association for Assessment and Accreditation of Laboratory Animal Care International). The experiments conducted with rhesus monkeys and cynomolgus monkeys followed the regulations of the Institutional Animal Care and Use Committee (IACUC) and the Guide for the Care and Use of Laboratory Animals.44 The animal experiments were designed in accordance with the Association for Research in Vision and Ophthalmology Statement for the Use of Animals in Ophthalmic and Vision Research. These experiments have been approved by the Institutional IACUC under the approval numbers: ACU21-275 (Joinn), and IACUC-SW-A2021040-K065-01 (WCFP), respectively. Human subjects The phase 1 trial was an open-label, dose-escalation study of LX102 gene therapy and enrolled 12 patients (8 male and 4 female) with previously treated nAMD. Subjects were allocated to three dose-escalation cohorts in a non-randomized manner according to the study protocol. The study was conducted at three sites in China, and the protocol was approved by the institutional review board at each participating study site (Approval ID: [2023]-009, [2023]-772, [2023]-368). Participants aged 50 or older with macular neovascularization secondary to nAMD were eligible if they had received recent anti-VEGF therapy, defined as at least 2 anti-VEGF treatments within 6 months in the study eye prior to screening. With evidence of an anatomical response on spectral domain–optical coherence tomography (SD-OCT), while still having active disease indicated by the presence of macular fluid at screening, confirmed by Optical Coherence Tomography (OCT) or Fundus Fluorescein Angiography (FFA). Additionally, participants needed to have a best-corrected visual acuity (BCVA) in the study eye between 20/800–20/63 (i.e., between 5 and 63 Early Treatment Diabetic Retinopathy Study [ETDRS] letters). At study entry, investigators discussed the standard of care, alternative options, and investigational therapies with each patient adhered to the International Ethical Guidelines for Health-related Research Involving Humans.45 Subjects were enrolled only if they declined continued standard therapy and voluntarily agreed to participate in this Phase I trial, with full understanding of risks, limitations, and the possibility of supplementary anti-VEGF therapy. All participants provided written informed consent before enrollment and were informed of their right to withdraw at any time. The study was conducted in accordance with the principles of the Declaration of Helsinki and the International Conference on Harmonization Good Clinical Practice guidelines. The complete information of inclusion and exclusion criteria is presented in the clinical protocol (Methods S1). Method details Viral vector production and purification LX102 was produced in the HEK293 cells via the triple plasmid transfection system, including the transfer gene plasmid, one encoding the AAV2 Rep and Cap proteins, and the helper plasmid. LX102 was purified by ultracentrifugation or chromatography to remove cellular contaminants and empty capsids. Properly formulated capsids were achieved by filtration before utilization. The amino acid sequence of the secreted VEGF trap protein from LX102 is provided in Data S1. Preclinical study In vitro potency assay The in vitro expression of the LX102 transgene product expressed in cell supernatant and lysate from transfected HEK293T cells (ATCC, #CRL3216) were measured by western blotting. Briefly, HEK293T cells were seeded into 24-well plates at 4×105 cells per well and cultured overnight before infection with varying doses of LX102 [multiplicity of infection (MOI) = 0, 40, 200, 1000, 5000 VG/cell]. Samples were diluted with NativePAGE Sample Buffer (4X) (Invitrogen, #BN2003), following by probing in 8% Tris-Glycine (ShareBio, #SB-TG1000815), and then transferred to PVDF membranes (Millipore, #ISEQ00010). After washing, the membranes were blocked and incubated with HRP-conjugated anti-IgG (H + L) antibody (YEASEN, #33501ES60) and anti-beta actin antibody (Beyotime, #AF0003). The LX102 transgene protein bands were directly visualized with chemiluminescence using a Super ECL Detection Reagent kit (YEASEN, #36208ES76), while the beta-actin protein bands were visualized after washing and incubation with secondary antibody. Quantification of the protein bands was performed using ImageJ software (National Institutes of Health). Different amounts of aflibercept (Eylea, Bayer Pharmaceuticals) were diluted with NativePAGE Sample Buffer (1X) to reach concentrations of 0.1, 0.3, and 1 ng/μL as positive controls. The in vitro VEGF inhibition of the LX102 transgene product was measured by enzyme-linked immunosorbent assay (ELISA) and a VEGF bioassay. For ELISA, the supernatants from transfected HEK293T cells were collected after 48-h post-infection with LX102. Different concentrations of aflibercept and LX102 transgene product ranging from 12.5 pM to 200 pM were treated with 43.5 pM purified recombinant human VEGF165 protein (East-Mab Biomedical Technology, #Y04702) for competition binding assay. A human quantitative VEGF ELISA kit (Sino Biological, #KIT11066) was used to detect unbound VEGF according to the manufacturer’s instructions. The inhibition rates were determined by the difference between the OD values of sample wells with the regarding control wells and dividing by the OD values of the control wells. The binding rate was calculated by 100% minus the inhibition rate. The concentration of transgene products and the corresponding binding rates were fitted to the four-parameter curve to calculate the half maximal inhibitory concentration (IC50) values. For the VEGF bioassay, the H-VEGF reporter 293 cell line (Genomeditech, #GM-C09057) was used as a bioluminescent assay. These cells are engineered to express VEGFR2 response element upstream of luciferase based on transcription factor activity. Briefly, a range of concentrations of transgene products or aflibercept (0.008–2.10 nM) were incubated with 0.453 nM human VEGF165 (East-Mab Biomedical Technology, #Y04702) and cultured with H-VEGF reporter 293 cells for 7 h, followed by quantification using a Bio-Glo luciferase assay system (Promega, G7940). The binding rate was calculated by 100% minus the inhibition rate based on the relative light unit (RLU) values, and the concentration of transgene products and their corresponding binding rates were fitted to the four-parameter curve to calculate IC50. In vivo pharmacological study Mice study Mice studies included a 2-week post-subretinal (SR) administration efficacy evaluation, and a long-term expression and efficacy study of 6 months post-SR administration of LX102. Subretinal injection and intravitreal injection in mice For subretinal injection of mice, procedures were performed by highly experienced technicians following a standard protocol based on established microsurgical techniques.46 Briefly, mice were anesthetized with tiletamine–zolazepam (50mg/kg; Zoletil50, Virbac Lab, France) and xylazine hydrochloride (6.65mg/kg) intraperitonealy. After pupil dilation with 0.5% tropicamide and 0.5% phenylephrine mixed eye drops (Santen, Japan), mice received local anesthesia in the eyes using 0.4% oxybuprocaine hydrochloride eye drops (Benoxil; Santen, Japan), and placed on the operating table after anesthesia. A limbal hole was made with a 34G needle by insertion into the vitreous at 1mm from the posterior edge of the cornea under a stereomicroscope (Shanghai Yide, DX-2000). Then, a Nanofil 36-gauge blunt needle (World Precision Instruments, #NF36BL) connected to a 10 μL syringe (World Precision Instruments, Nanofil) was inserted through the initial hole and directed toward the contralateral subretinal space, avoiding touching the lens. For the 2-week pharmacodynamic study, each eye was given 2 μL of rAAV vector solution with a titer of 5×109, 5×1010, 5×1011 vg/ml, or vehicle. For the 6-month study, each eye was given 2 μL of rAAV vector solution with a titer of 5×108, 5×109, 5×1010 vg/ml, or vehicle. Immediately after injection, subretinal blebs were confirmed under the stereomicroscope. Injections creating subretinal blebs without massive vitreous or subretinal hemorrhage were considered to be successful. For the aflibercept (Eylea , Bayer Pharmaceuticals, 40 mg/mL, 1μL) treated group, the mice were bilaterally intravitreally injected immediately after laser photocoagulation. Ofloxacin eye ointment was applied to the eyes after the procedure. Immunofluorescence of retinal and RPE-choroid flat mounts Flat-mounts of the retina and RPE-choroid were obtained two weeks post-SR administration of LX102 and prepared as previously described protocols.47,48 Briefly, the eyes were enucleated and fixed with 4% paraformaldehyde for 2 h. The retina and RPE-choroid complex were obtained separately by carefully removing the anterior segments and optic nerve. Approximately four relaxing radial incision were made to flatten the tissue. After three times washing with phosphate-buffered saline (PBS), the flat mounts were incubated with 0.2% Triton X-100 (Sigma, #T9284) in PBS for 1 h at RT. After washing, the flat mounts were then incubated with blocking buffer [5% bovine serum albumin (Beyotime, #ST023) in PBS] for 2 h at RT. Flat mounts were incubated overnight at 4°C in blocking buffer containing antibodies against Alexa Fluor 594–conjugated goat anti-human lgG(H + L) (1:200; YEASEN, #33222ES60). Nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI). The flat mounts were carefully washed with PBS three times before they were embedded and cover-slipped. The flat mounts were observed using a fluorescence microscope (Thermo Fisher, EVOS M7000). Laser-induced CNV model in mice Mice were anesthetized with intraperitoneal injection of tiletamine–zolazepam (50mg/kg; Zoletil50, Virbac Lab, France) and xylazine hydrochloride (6.65mg/kg). Their pupils were dilated with 0.5% tropicamide and 0.5% phenylephrine mixed eye drops (Santen, Japan), following by local anesthesia with 0.4% oxybuprocaine hydrochloride eye drops (Benoxil; Santen, Japan). A small amount of oxyfloxacin eye ointment (SINQI, China) was applied to maintain corneal moisture. CNV lesions were induced in mice using a 532-nm laser system (Visuals 532S, Carl Zeiss Meditec) based on previously described protocols,(3,4) with modifications. Briefly, laser radiation was delivered between the major retinal vessels and at equal distances from the optic disc through gently pressing a coverslip to fully expose the visual axis. C57BL/6J mice were subjected to laser-induced rupture of Bruch’s membrane (160 mW in intensity, 50 μm in size, and 100 ms in duration) at three or four locations per eye. Only laser spots in which a visible bubble was produced without hemorrhage, indicating a perforated Bruch’s membrane, were considered effective and included in the study. After the procedure, mice were resuscitated in a warm environment and returned to the cage after awakening. Fundus fluorescein angiography in mice In the mouse studies, CNV lesions were evaluated by fundus fluorescein angiography (FFA) with a confocal scanning laser ophthalmoscope (Heidelberg retinal angiograph, Heidelberg Engineering GmbH) at 3, 5, 10, or 13 days after laser photocoagulation. Infrared photographs were obtained before. Late-phase angiograms were captured at 5 min after intraperitoneally injected with fluorescein sodium (10%, 0.1 mL/mouse; Alcon, Tokyo, Japan). The fluorescein angiograms were graded by two masked and experienced examiners. Lesion grades were graded following the grading scale, with grade I representing no hyperfluorescence, grade II representing hyperfluorescence without leakage, grade III representing early hyperfluorescence and late mild leakage within the spot area, and grade IV representing early hyperfluorescence and late severe dye leakage, transiting beyond the borders of the spot area.48 Real-time qPCR Mice were sacrificed and eyecups were harvested by removing the anterior segment. Total RNA was extracted from eyecup tissues using TRIzol reagent (Thermo, #15596026). Then, RNA was reverse transcribed to cDNA by using the PrimeScript RT reagent Kit with gDNA eraser (Takara, #RR047A) according to the manufacturer’s instructions. DNase was used in RNA extraction and transcription to ensure elimination of DNA contamination. Quantitative PCR was performed with TB Green Premix Ex Taq II (Takara, # RR820A). The VEGF-Trap protein primers are as follows: Forward, 5′-GACTTCAACTGGGAGTAC-3′, Reverse, 5′-AGTGCTCAGGAATTTCTT-3’; beta-actin primers are as follows: Forward, 5′-CTGGAACGGTGAAGGTGACA-3′, Reverse, 5′-GGGATGTTTGCTCCAACCAA-3’. The specificity of the qPCR was confirmed by detection of a single distinct peak on examination of the dissociation curve profile of the reaction product. Relative gene expression to endogenous control was calculated using the formula 2−ΔΔCt, where ΔCt represented the magnitude of the difference between cycle threshold (Ct) values of the target and endogenous control (beta-actin), and the result expressed as a percentage of the mean value of the low-dose group. Nonhuman primate study NHP studies included a pharmacology evaluation in rhesus monkeys using a laser-induced CNV model established 21 days post-subretinal (SR) administration of LX102, and a comprehensive pharmacokinetic and biodistribution study in cynomolgus monkeys. For the pharmacology study, clinical signs and ocular evaluations (via slit-lamp, fundoscopy, and multimodal imaging) were longitudinally monitored at scheduled intervals through day 56 post-injection of LX102. Subretinal injection in NHPs Rhesus monkeys were anesthetized with tiletamine–zolazepam (8 mg/kg, 50 mg/mL; Zoletil50, Virbac Lab, France) intramuscularly, while cynomolgus monkeys were anesthetized with 2.5% pentobarbital (25 mg/kg) intravenously. Pupils were dilated using 0.5% tropicamide and 0.5% phenylephrine mixed eye drops (Santen, Japan). Oxybuprocaine hydrochloride eye drops were used for local anesthesia before administration. The eyes to be dosed were disinfected with povidone iodine solution. The animal’s eyelids were opened with the eyelid opener so as to exposed the operation site. The scleral puncture needle was injected at 2 mm from the corneoscleral limbus toward the center of the eyeball. Then, a blunt 36-gauge needle (RPE-KIT, WPI) was inserted through the scleral tunnel and reached the subretinal space at proper site of the retina under a surgical microscope combined with a retinoscope. For the pharmacology studies, each eye of the rhesus monkey was given 50 μL of rAAV vector solution with a titer of 5×1012 VG/ml. For the biodistribution study, the left eye of the cynomolgus monkey received rAAV vector solution at a dose of 2.5×1011 VG/eye (50 μL, titer: 5×1012 VG/ml). After the injection, ofloxacin eye ointment were applied to each eye to protect the cornea and prevent infection. Laser-induced CNV model in NHPs Rhesus monkeys were anesthetized with tiletamine–zolazepam (12 mg/kg, 50 mg/mL; Zoletil50, Virbac Lab, France) intramuscularly. After the NHPs were adequately anesthetized, the pupils were dilated using tropicamide drops. CNV was induced in rhesus monkeys with a different laser system (VISULAS Trion multi-wavelength laser, ZEISS). Six to eight lesions were symmetrically placed in the perimacular area of both eyes, approximately 1.5–2 optic disc diameters away from the macular fovea while avoiding blood vessels for photocoagulation. The approximate laser parameters were as follows: spot size of 50 μm, laser power of 400∼500 mW, and exposure time of 100 ms. Appearance of a small blister and a popping sound were considered to indicate a puncture of Bruch’s membrane. Slit lamp Biomicroscopy and fundoscopy Anterior segments were examined by slit lamp microscope for evaluation of eyelids, conjunctiva, cornea, anterior chamber, iris, pupils and lens. Evaluation of the posterior fundus and vitreous cavity was performed by posterior segment slit lamp exam employing a 90-diopter lens. Retinal infiltrates and hemorrhage, vascular dilation, tortuosity, and sheathing, and optic disc edema were also evaluated during the fundoscopy. Fundus photography and fundus fluorescein angiography (FFA) Bilateral color fundus photographs were obtained using a fundus camera (Digital fundus angiography camera, APS-BER, Kanghua) under pupil dilation. FFA was obtained by a confocal scanning laser ophthalmoscope (Digital fundus angiography camera, APS-BER, Kanghua) with intravenous injection of 100 mg/mL fluorescein sodium (10 mg/kg) at 14, 21, 28 and 35 days after laser photocoagulation. The angiograms were obtained at several time points, including the arterial phase, early arteriovenous phase, and several late arteriovenous phases, to monitor leakage of fluorescein associated with CNV lesions. Grading of fluorescein angiograms were performed by two masked and experienced examiners following the grading scale (I to IV).47 The leakage area of the grade IV lesions was measured by a masked evaluator using ImageJ to delineate the leakage boundary and calculate in square microns (mm2). Optical coherence tomography (OCT) Under anesthesia, the animals were fixed to keep an eye open and maintain the position of the head. OCT (Spectralis HRA+OCT, Heidelberg Engineering GmbH) scans were performed by experienced physicians to evaluate the structural integrity and subretinal hyperreflective material (SHRM) by computer-assisted manual segmentation using the proprietary Heidelberg Eye Explorer (HEYEX) software. A multiple-linear scanning protocol was introduced to cover the entire CNV lesions. The cross-sectional scans through the grade IV lesions were selected to measure the maximum height of SHRM within each corresponding lesion. Specifically, under the condition that one specific lesion was evaluated as a grade IV lesion at any time points during the follow-up examinations after laser photocoagulation, the SHRM within the corresponding lesion was measured at all time points after laser coagulation. Pharmacokinetic study in NHPs Thirty cynomolgus monkeys received unilateral SR administration of LX102 solution at a dose of 2.5×1011 VG/eye (50 μL, titer: 5×1012 VG/ml). The animals were euthanized at 2, 4, 13, 19, and 26 weeks post-injection. At each time point, blood samples, ocular tissues (including the conjunctiva, cornea, aqueous humor, iris/ciliary body, lens, vitreous, retina-choroid complex, sclera, and optic nerve), and various organ tissues (forebrain, brainstem, thalamus, heart, liver, spleen, lung, kidney, adrenal gland, bone marrow, skeletal muscle, ovary, uterus, testis, epididymis, stomach, pancreas, jejunum, ear, and anterior lymph nodes) were collected. In addition, secretion samples (tears, nasal mucus, oral mucus, feces, and urine) were harvested at 4-, 13-, and 26-week post administration. For transgene detection, DNA was extracted from the tissues of interest. Solid tissue samples were thawed on ice. All ocular tissues were homogenized. For organ tissues, approximately 50 ± 10 mg of tissues samples (at least four sites from different parts) were transferred to a 2.0-mL grinding tube and mixed with three to five ceramic beads (3 mm diameter) and appropriate volume of lysis buffer based on the tissues weight (25 mg:200 μL; for less than 25 mg of tissues, 200 μL of lysis buffer was added). Samples were homogenized at 6500 rpm for 20 s using a homogenizer for three repeated cycles with a 20-s interval between each homogenization step (repeated once or twice in case of incomplete homogenization) and then immediately centrifuged. The harvested samples were collected at the bottom of the tubes. Liquid tissues samples did not require homogenization. Homogenate or liquid tissue samples (200 μL) were added to the bottom of the wells of a MagNA Pure 96 Nucleic Acid Purification System (Roche). The actual volume added was accurately recorded and 1×PBS was added to reach a total volume of 200 μL if the sample volume was insufficient. DNA was extracted using a kit (MagNA Pure 96 DNA and Viral DNA Large Volume Kit, Roche, #06374891001) and stored below −66°C. The transgene was detected by quantitative PCR (qPCR) using TaqMan probes (forward primer, 5′-GACTTCAACTGGGAGTAC-3’; reverse primer, 5′-AGTGCTCAGGAATTTCTT-3’; Probe, CAGCAAGCACCAACACAAGAAG). The method was validated, the validation results are shown in Table S4. The transgene mRNA expression level after subretinal injection in NHPs was assessed by RT-qPCR. Briefly, NHPs where sacrificed at 2-, 4-, 13-, 19-, 26-week post injection, and retinal tissues were harvested. Total RNA was extracted from retinal tissues using a Macherey-Nagel RNA extraction kit (Roche, #6374891001) according to the manufacturer’s protocol. RNA quality and quantity were assessed using a DeNovix DS- 11 FX Series of Spectrophotometers/Fluorometers (DeNovix) and 7 μL of RNA was then reverse transcribed to cDNA by using High Capacity cDNA reverse transcriptase kit (Takara, #RR047A). And TaqMan gene assays was used to quantify the mRNA expression level on a Real-time PCR System (Thermo QuantStudio5) using a validated RT-qPCR method (validation results shown in Table S5). For transgene product detection, the solid tissues were supplemented with PBS containing protease inhibitor at a ratio of 5 mL PBS per 1 g tissue; vitreous body was homogenized directly. All homogenates were centrifuged at 5000 ×g and 4°C for 5 min. The obtained supernatant was aliquoted at approximately 100 μL/tube, and the remaining supernatant was stored in another tube. The liquid tissues were directly aliquoted (approximately 50 μL in one tube and the rest in another one). The above homogenates were stored below −65°C. Briefly, 100 μL of Goat Anti-Human IgG, Monkey ads-UNLB (1 μg/mL in 0.05 M carbonate buffer solution, Southern Biotech) was added to a 96-well microplate. The microplate was sealed and incubated overnight at 2°C–8°C. The coating solution was decanted and the plate was washed three times with 300 μL wash buffer. The microplate was blocked for 2 h at 10°C–30°C. Next, 50 μL of sample (including reference standards, quality control samples, and study samples) was added to the microplate. The microplate was covered and incubated for 1 h at 10°C–30°C. After three washes, 100 μL of biotin-labeled anti-VEGFR1 was added and plates were incubated for a further 1 h. After three washes, 100 μL of Streptavidin-HRP was added and plates were incubated for a further 0.5 h. After three washes, 100 μL of TMB chromogenic solution was added for color development. The reaction was stopped with the stop solution and the absorbance at 450 nm was recorded (Thermo, Varioskan LUX). Phase 1 study-design and surgical procedures The primary objective was to evaluate the safety and tolerability of a single dose of subretinal LX102 up to 52 weeks. Participants received a single injection sequentially enrolled in three dose-ranging cohorts: cohort 1, 2×1010 vector genomes per eye, n = 3; cohort 2, 5×1010 vector genomes per eye, n = 3; and one maximum tolerated dose cohort 3, 1.25×1011 vector genomes per eye, n = 6.The study eye received an intravitreal injection of aflibercept (2 mg/0.05 mL) on day −14. Approximately 14 days later, a subretinal injection of LX102 was administered in one eye by a trained vitreoretinal surgeon (Day 0). Dose-limiting toxicity was assessed at week 4, and rescue injections were provided from week 4 onwards based on disease activity. Participants were followed up at week 2, week 4, week 8, week 12 and every 8 weeks afterward for up to week 52. A standard subretinal injection protocol has been established based on prior experience with retinal gene therapy injections in collaboration with experts in the field.40,49,50 Specifically, under local anesthesia in an operating room, a small-gauge three-port standard vitrectomy was performed with removal of posterior hyaloid facilitated by diluted triamcinolone acetonide to visualize the hyaloid. A trained vitreoretinal surgeon, under the supervision of a surgical proctor, performed the subretinal injection of LX102 using a one-step technique without creation of a pre-bleb. Following core vitrectomy and hyaloid removal, the vector was injected into the subretinal space using a 38-gauge subretinal cannula (#3255, MedOne Surgical, Inc., FL, USA) which was attached to a 1-mL syringe driven setup (#3275, MedOne Surgical, Inc., FL, USA), connected to the Viscous Fluid Injection (VFI) module of the vitrectomy system, controlled via foot pedal (Figure S6). The flow rate was initially increased gradually until a steady continuous stream appeared, then reduced to create a discontinuous stream of individual droplets. The 38-gauge needle was positioned just above the retina, superior to the supratemporal or inferior temporal vascular arcades, and the foot petal was activated to deliver the vector stream, penetrating the retina. Infusion pressure was reduced to approximately 10–15 mmHg prior to injection. Although care was taken to try to inject full prescribed volume of vector, the final volume in the subretinal space may have varied somewhat due to inaccuracies in the volume markings on the syringe and leakage of a small amount of vector through the puncture site in the retina after the injection. Intraoperative microscope-integrated optical coherence tomography (MI-OCT) was used to monitor bleb formation and confirm subretinal placement while avoiding excessive retinal stretching or foveal detachment (Figure S7). An additional lavage of intravitreal fluid was performed if visible reflux was observed. A partial fluid-air exchange was then performed, the cannulas were removed, and any non-self-sealing sclerotomies were sutured. Supplemental anti-VEGF injections criteria: At each study visit after week 4, participants could receive an intravitreal anti-VEGF injection at the investigator’s discretion if any of the following criteria were met: a loss of more than 10 letters in BCVA compared to baseline, attributable to intraretinal or subretinal fluid on OCT; an increase in CST of more than 100 μm compared to baseline; or the presence of vision-threatening hemorrhage due to nAMD. Phase 1 study-outcomes The primary outcomes of the trial were safety and tolerability through 52 weeks after LX102 administration as assessed by the incidence of ocular and non-ocular adverse events. Secondary outcomes were to evaluate the efficacy of LX102, specifically focusing on changes in best-corrected visual acuity (BCVA) and central subfield thickness (CST) in the study eye from baseline to 52 weeks. BCVA was measured by ETDRS protocol, and CST was defined as the circular area 1 mm in diameter centered around the center point of the fovea, with its thickness provided as a quantifiable value on the HEYEX software. Other secondary outcomes included the need for supplemental anti-VEGF injections up to week 52. Ophthalmic assessments included BCVA, slit-lamp and indirect ophthalmoscopy examinations, and ocular imaging (SD-OCT using the SPECTRALIS system [Heidelberg Engineering; Franklin, MA, USA], fundus photography using a Topcon TRC50-DX Retinal Camera [Tokyo, Japan] or ultra-wide fundus photography [Optos; Marlborough, MA, USA], and fluorescein fundus angiography [Heidelberg Engineering; Franklin, MA, USA]). Exploratory outcomes in this study encompassed several key clinical laboratory tests designed to assess both the safety and efficacy of the treatment. These included the analysis of vector biodistribution and shedding in peripheral blood and tears, detection of total AAV2 antibodies and AAV2 neutralizing antibodies in serum, detection of anti-VEGF trap antibodies in serum, measurement of VEGF trap concentration in aqueous humor, and detection of changes in interferon gamma (IFN-γ) enzyme-linked immunospot (ELISpot) responses to AAV2 antigen and VEGF-trap peptide in whole blood. AAV capsid biodistribution and shedding were assessed using a validated TaqMan-PCR method with a lower limit of quantification (LLOQ) of 5 copies/μL and a detection limit of 0.8 copies/μL; samples with a Ct value greater than 40 were considered negative. Total AAV2 antibodies were measured using an electrochemiluminescence bridging assay with a detection sensitivity of 1.31 ng/mL, and antibody titers were calculated by linear extrapolation. AAV2 neutralizing antibodies were quantified using a cell-based assay, with a 50% inhibition rate used as the threshold for titer determination. Anti-VEGF trap antibodies were detected using an ELISA-based bridging assay, with positive samples confirmed by a more sensitive confirmatory assay (19.1 ng/mL). Finally, the VEGF trap concentration in aqueous humor was measured by ELISA, with an LLOQ of 3 ng/mL. Quantification and statistical analysis For preclinical assessments, two-sided Student’s t test, one-way ANOVA and chi-square test were used for data comparison. Statistical analyses were performed by using GraphPad Prism 10.0 (Graphpad). Statistical significance was defined as ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, and ∗∗∗∗p < 0.0001. For phase I study, data from all participants enrolled in the study were included to assess safety and efficacy of treatment. No formal calculations were performed to determine the sample size. AEs were coded using MedDRA classification. Oversight for the safe conduct of the study was provided by an independent data monitoring committee. Descriptive statistics were used to assess efficacy by cohort, dose received, as well as for the study overall. BCVA data was presented as the ETDRS letter score at each time point, and CST was quantified at each time point. Categorical data were summarised by the number and percentage of participants within each category. Continuous data were summarised by descriptive statistics, including number of observations, mean (SD), and median (range). The observed value and change from Baseline for both BCVA and CST were summarized using continuous descriptive statistics. Statistical programming and analyses were performed using SAS Version 9.4. Additional resources The Phase 1 study is registered with www.chinadrugtrials.org.cn (Trial Identifying Number: CTR20230194) and ClinicalTrials.gov NCT06198413.