Long-term Nerve Regeneration in Diabetic Keratopathy Mediated by a Novel NGF Delivery System

Corneal sensory nerves primarily originate from the ophthalmic division of the trigeminal ganglion (TG) and densely innervate the cornea as stromal branch and subepithelial plexus (1,2). These nerves play a crucial role in various physiological functions (3,4), contributing not only to sensory-mediated reflexes but also to the production of trophic factors important for maintaining corneal physiological functions and integrity (5). Factors such as mechanical injuries, infections, chemical injuries, and thermal burns can potentially affect corneal nerves. Additionally, cranial surgeries, trigeminal nerve disorders, and systemic diseases like diabetes can fundamentally affect the integrity of corneal nerves. As of the time of writing, it is estimated that 415 million people worldwide suffer from diabetes, with an annual increase of ∼11 million new diagnoses by 2040 (6). With the increasing prevalence of diabetes, the incidence of concurrent corneal nerve damage is also rising (7). Thirty to fifty percent of patients with a diabetic history exceeding 10 years face a high risk of developing diabetic keratopathy (DK), characterized by a reduction in corneal nerve fiber density and quantity (8,9). In these patients, early ocular symptoms range from eye pain and local irritation to decreased corneal sensitivity. Corneal in vivo confocal microscopy (IVCM), a noninvasive detection method, can detect early-stage corneal nerve degeneration in patients with diabetes (10,11). Owing to its high sensitivity, it is widely employed as an objective tool for assessing the severity of corneal nerve damage and evaluating treatment effectiveness in patients with diabetes (12,13). As corneal nerve damage progresses, it can lead to corneal epithelial punctuates, delayed epithelial repair, persistent corneal epithelial defects, corneal stromal melting, and even perforation.

Early treatments for DK involve artificial tears, bandage lenses, and punctual occlusion. For advanced cases, interventions such as autologous serum eye drops, tarsorrhaphy, amniotic membrane transplantation, conjunctival flap, and even corneal transplantation can be applied (14). Nerve growth factor (NGF) is well known for supporting the growth and differentiation of peripheral nerves, including corneal innervation (15). NGF also maintains the homeostasis of various corneal cells, such as the proliferation and differentiation of epithelial cells and limbal stem cells (16,17), and it has been found to express in corneal endothelial cells (18). In recent years, recombinant human NGF (rhNGF) eye drop (cenegermin) has emerged as a potential medication in managing corneal nerve defects (19–21). Administered for eight consecutive weeks at a frequency of six times per day, it has demonstrated the capacity to facilitate healing of corneal epithelial defects and regeneration of nerve fibers. However, for progressive degenerative conditions such as diabetes, while rhNGF is effective during administration, whether its discontinuation will lead to recurrence remains to be explored (22–24).

Adeno-associated virus (AAV) is widely used in gene therapy (25), and its application in treating eye diseases has been tested (26). The safety and efficacy of AAV-mediated gene therapy have also been preliminarily validated in the treatment of neurodegenerative diseases, such as Parkinson disease, Alzheimer disease, and spinal muscular atrophy (27). However, gene therapy for corneal nerve regeneration and AAV serotype screening targeting corneal nerves are still lacking.

In this study, we screened AAV serotypes to efficiently transfect corneal nerve endings and TG ophthalmic branch and identified AAV.rh10 as a suitable vector. We validated the immune responses of such gene therapy and assessed the expression of the target protein in the cornea and TG. The effectiveness of gene therapy was verified in diabetic models. Comparisons were made between a single injection of AAV.rh10 carrying Ngf and the commercial rhNGF eye drop. Our gene therapy strategy demonstrated a more prolonged protective effect in corneal nerve regeneration, making it a potential novel therapy for clinical application.

Animal ethics approval for this study was obtained from the Animal Investigation Committee of Eye Institute of Shandong First Medical University (Qingdao, China) with approval number SDSYKYJS No.20220812. All animal procedures were performed in accordance with the Association for Research in Vision and Ophthalmology.

C57BL/6 male and female mice aged 6–8 weeks were purchased from the Vital River Laboratory Animal Technology Co., Ltd. (Beijing, China), and maintained in the animal center of Eye Institute of Shandong First Medical University. A streptozotocin diabetic mouse model was constructed using males (28), while the wild-type control mice included both age-matched females and males.

Statistics analysis was performed using SPSS Statistics and GraphPad Prism 9.2.0. Data were expressed as mean ± SD. One-way ANOVA was used to assess the significance of the differences among groups, and Tukey multiple comparisons were used for intergroup comparisons. Unpaired t test was used to compare differences between two independent groups. Statistically significance was set at *P < 0.05, **P < 0.01, and ***P < 0.001.

The data generated and/or analyzed during the current study are available from the corresponding author upon reasonable request.

In this study, we evaluated six different delivery routes for AAV.rh10-GFP (AAV-Con) to determine the optimal method for delivering gene therapy products into corneal nerves. These routes included subconjunctival injection (route 1), topical eye drops (route 2), hydraulic column on the cornea (route 3), intracameral injection (route 4), traditional intrastromal injection (IS^old) (route 5), and improved intrastromal injection (IS^new) (route 6) (Fig. 1A). Two weeks after AAV-Con injection, the pattern and intensity of green fluorescent protein (GFP) distribution were recorded using corneal whole mounts. Limited GFP close to the limbus was observed via route 1, indicating restricted AAV penetration through the corneal limbus into the stroma. GFP signal was barely observed via routes 2 and 3, suggesting that AAV cannot traverse the corneal epithelial barrier. GFP signals were only concentrated in the peripheral cornea in route 4 (Fig. 1B), suggesting the leakage of AAV from the anterior chamber to the stroma.

Conventional intrastromal injection IS^old exhibited abundant GFP in the corneal stroma, but the GFP distribution was uneven (Fig. 1B), with only about 58.4% of the cornea being GFP positive in week 2 (Fig. 1C). In contrast, the IS^new group displayed homogenous distribution of AAV-Con (Fig. 1B), with 86.2% of the cornea covered in week 2 (Fig. 1C). Moreover, the IS^new group exhibited the strongest fluorescence intensity among all groups (Fig. 1D), indicating IS^new as the most efficient transduction method. The primary distinction between IS^old and IS^new injection methods was that IS^old created a tunnel from the corneal rim to the central cornea, while IS^new made a lamellar incision at the central cornea (Fig. 1A and Supplementary Fig. 1A). The IS^new method restored corneal transparency to its preinjection status by the 1st day after injection, with no noticeable edema of the cornea. In contrast, the longer puncture tunnel in IS^old caused corneal edema and stromal turbidity (Supplementary Fig. 1B), which subsequently diminished 1 week postinjection (Supplementary Fig. 1C). Considering its superior distribution, high transfection efficiency, and minimal side effects, we selected the IS^new method for subsequent experiments.

Different AAV serotypes have different tissue tropism. There is still a lack of screening of AAV serotype with the highest affinity for corneal nerves. In this study, we assessed the corneal nerve transduction efficiency of commonly used AAV serotypes in ocular gene therapies, namely, AAV2, AAV6, AAV8, AAV9, and AAV.rh10. Employing the IS^new method, equal quantities (2 × 10⁹ vector genomes [vg]) and volumes (2 μL) of AAVs were injected into the corneal stroma, and their respective GFP expressions were observed after 4 weeks. AAV6 and AAV.rh10 demonstrated homogenous and robust GFP across the entire corneal area (Fig. 2A and B). In comparison, after intrastromal injection, AAV6 primarily transduced corneal stromal cells, while AAV.rh10 displayed clear neuronal tropism, with GFP significantly accumulating at the basal nerve plexus and within nerve terminals (red arrows) (Fig. 2A and C). Optic clearing of mouse TG confirmed that AAV.rh10 retrogradely transduced to the ophthalmic branch of TG (Supplementary Fig. 2). This transduction capability underscores AAV.rh10’s potentials in promoting nerve regeneration when carrying Ngf.

To further determine the tissue specificity of AAV.rh10, whole eyeball sectioning was conducted 4 weeks after high-dose AAV.rh10 (2 × 10¹⁰ vg) injection. GFP fluorescence was predominantly concentrated in the corneal subepithelial space and the anterior corneal stroma (Fig. 2D). No overlap signal of GFP was observed with K12, which stained the epithelium, and ZO-1, which stained the epithelium and endothelium (Supplementary Fig. 3). Notably, no off-target GFP signal can be detected in other ocular tissues, such as the iris, lens, retina, sclera, and periocular muscles (Fig. 2D). Considering both efficiency and specificity, we selected AAV.rh10 for subsequent experiments.

Screening the AAV dosage is a pivotal step before commencing gene therapy. A gradient of AAV.rh10-Ngf ranging from 2 × 10⁷ vg to 2 × 10¹⁰ vg was injected into the cornea via the IS^new injection method, and the mRNA and protein levels of the target gene were assessed in the cornea and TG 4 weeks after injection. The expression of Ngf is under the control of a CMV promoter fused with a GFP tag (Supplementary Fig. 4). The Ngf mRNA level reached the plateau phase in the cornea after 2 × 10⁸ vg AAV injection (Fig. 3A), while a significant increase in the TG Ngf mRNA level was detected in conjunction with the increasing of AAV concentrations up to 2 × 10¹⁰ vg (Fig. 3A). Interestingly, both cornea and TG displayed a peak in NGF protein level after 2 × 10⁹ vg AAV injection, with a 1.8-fold and 1.6-fold increase in NGF protein in the cornea and TG, respectively (Fig. 3B). These results indicate that 2 × 10⁹ vg AAV.rh10-Ngf is the optimal dosage for efficient NGF protein expression in both cornea and TG.

Subsequently, we monitored the expression timeline of Ngf mRNA in the cornea and TG after a 2 × 10⁹ vg AAV.rh10-Ngf IS^new injection. Significantly elevated Ngf mRNA was observed in both tissues from the 2nd week after injection, reaching a peak at the 4th week in cornea (80.9-fold) and the 6th week in TG (33.9-fold) (Fig. 3C). This observation suggests that AAV was transported directionally from corneal nerve endings to the TG.

Given the crucial roles of NGF in maintaining and recovering corneal nerve functions, we further examined the neurotransmitters in the cornea. Four weeks post–AAV.rh10-Ngf IS^new injection, substance P (SP) and calcitonin gene-related peptide (CGRP), the two important neurotransmitters secreted by corneal sensory nerves were evaluated (29,30). As expected, substantial increases in SP (1.3-fold in the cornea and 1.3-fold in TG) and CGRP (1.1-fold in the cornea and 1.2-fold in TG) were observed (Fig. 3D), indicating additional benefits of AAV-Ngf therapy besides increasing NGF expression.

Despite the eye being a favorable target for gene therapy because of its immune-privileged properties and the ease in local AAV application, intraocular inflammation has been reported in AAV-mediated retinal gene therapies (31–33). Flow cytometry was employed to assess the dynamic immune responses in the cornea post–AAV vector injection. The gating strategy is described in Supplementary Fig. 5. Six immune cell markers were examined, encompassing CD45⁺ for leukocytes, CD11b⁺ for myeloid cells, F4/80⁺ for macrophages, CD4⁺ and CD8⁺ for T cells, and CD19⁺ for B cells. Double staining was avoided because of the GFP signals carried by AAV vectors. Corneas were evaluated at intervals of 1 day (Supplementary Fig. 6), 3 days (Supplementary Fig. 7), 1 week (Fig. 4), 1 month (Supplementary Fig. 8), and 3 months (Supplementary Fig. 9) following the IS^new injection of AAVs. At day 1, there were mainly increases in myeloid cells due to the AAV capsid injection (Supplementary Fig. 6). By day 3, myeloid cells began to decrease (Supplementary Fig. 7), and immune responses gradually diminished by week 1 (Fig. 4). However, a new wave of myeloid infiltration was observed by week 4 (Supplementary Fig. 8), consistent with the peak transcription of the target gene (Fig. 3). Nevertheless, there was only a 0.4% increase in CD11b⁺ myeloid cells in the AAV-Ngf group compared with the PBS group (Supplementary Fig. 8). This mild immune response gradually subsided, with no difference observed by month 3 (Supplementary Fig. 9). To further assess immune response, we performed immunohistology sectioning of corneas following AAV injections and stained with CD45 (Supplementary Fig. 10A and B). Consistent with the flow cytometry results, only a limited number of CD45⁺ cells were observed in the cornea, with no differences among the wild-type, AAV-Con, and AAV-Ngf groups (Supplementary Fig. 10C and D).

Simultaneously, we assessed the expression of inflammatory factors, including Il-1β, Il-6, Il-10, Il-12α, Tnf-α, and Ifn-β (Supplementary Fig. 10). In line with the general trend of immune cell changes, there was a significant increase in Il-1β, Il-6, Il-10, and Il-12α mRNA on the 1st day, decreasing on the 3rd day and stabilizing by the 7th day. Unlike the elevation of Il-1β, Il-6, and Il-10 induced by AAV stimulation, the increase in Il-12α appeared to be caused by the injection process. Tnf-α and Ifn-β mRNA, on the other hand, showed no significant change (Supplementary Fig. 11).

Despite increased immune responses following AAV transduction, there were no observable corneal opacities (Supplementary Fig. 12A), alterations in corneal thickness and morphology (Supplementary Fig. 12B), or changes in nerve status (Supplementary Fig. 12C) in wild-type mice receiving AAV-Con or AAV-Ngf injections over 6 months, suggesting sustained safety in our gene therapy.

To evaluate the expression duration of a single injection of AAV.rh10, we monitored GFP from 1 week to 12 months after IS^new injection of 2 × 10⁹ vg AAV.rh10-Con. GFP was readily observed in the cornea at the injection site at week 1 (Fig. 5A) and in the ipsilateral trigeminal nerves (Fig. 5B). Two weeks after injection, GFP signals were distributed homogenously throughout the corneal area, with subepithelial nerve endings visible (red arrow) (Fig. 5C). By 1 month, both corneal nerve endings and the TG ophthalmic branch were abundantly transduced (Fig. 5E and F), and intense GFP signals remained observable until 6 months postinjection (Fig. 5M and N). The GFP signals gradually faded from peripheral to central cornea but were still detectable after 9 (Fig. 5O and P)and 12 months (Fig. 5Q and R). A quantitative analysis of the GFP distribution area and GFP density in the cornea and TG at different time points after AAV injection was also conducted (Supplementary Fig. 13).

To validate the treatment potentials of our gene therapy system, AAV.rh10-Ngf and its control were applied to type 1 diabetic mice (Fig. 6A). Prior to AAV injection, these mice exhibited typical diabetic symptoms and complications, including elevated blood glucose levels above 25 mmol/L, polydipsia, polyuria, weight loss, dull fur, and lethargy (Supplementary Fig. 14A and B). Additionally, we observed corneal nerve fiber degeneration in the diabetic mice, particularly a reduction in the central nerve vortex structure (whorl) (Supplementary Fig. 14C) (11). Nerve whorl is vital for maintaining corneal sensitivity, and it was intact in wild-type mice (Fig. 6B). Fluctuations in body weight and blood glucose were not observed after intrastromal injection of AAV-Ngf (Supplementary Fig. 15). Both the AAV-Ngf and topical rhNGF effectively promoted the regeneration of corneal nerves and the reconstruction of the central whorl in diabetic mice, while no regeneration was observed in the AAV-Con group. The regenerative capacity was comparable between AAV-Ngf and rhNGF groups (Fig. 6B and C). Intriguingly, 1 month after topical rhNGF discontinuation (rhNGF DC), the newly regenerated nerves regressed, with the whorl structure diminished (Fig. 6B), and central nerve area and density decreased back to the level before rhNGF treatment (as referenced in the AAV-Con group) (Fig. 6C). The regression was not observed in the AAV-Ngf group after the single IS^new injection (Fig. 6B), whose nerve density was similar to the wild-type group (Fig. 6C). The long-lasting nerve maintenance ability of AAV-Ngf can be attributed to the high levels of NGF protein in both the cornea and TG, and that there was still a 1.3-fold increase in NGF in both the cornea and TG at 2 months after the injection (Fig. 6D). Consistently, the levels of SP (Fig. 6E) and CGRP (Fig. 6F) also remained at high levels in the AAV-Ngf group 2 months after injection, comparable to the continuous topical rhNGF group in the TG, all at higher levels than in the rhNGF DC group (Fig. 6E and F).

We continued following the AAV-injected diabetic mice for up to 5 months after injection (Fig. 7A). No corneal opacity or edema was observed during the observed period in both AAV-Con and AAV-Ngf injection groups (Fig. 7B). The whorl structure was well maintained through the 5 months in the AAV-Ngf group, compared with the dramatic degeneration of corneal nerve endings in the AAV-Con group (Fig. 7C).

The corneal nerve network is essential for maintaining corneal epithelial integrity, where nerve damage would potentially lead to epithelial defects. Therefore, we applied a corneal epithelial debridement assay to examine the functions of regenerated nerves (Fig. 8A). The AAV-Ngf group showed a comparable epithelial healing capability to the topical rhNGF group (Fig. 8B and C). However, discontinuation of topical rhNGF for only 1 month compromised this protective effect, with both the repair speed and defect area reverting back to the levels observed in untreated control mice (Fig. 8B and C). There was also a loss of nerve regeneration ability in the rhNGF DC group, with nerve area, nerve density, and corneal sensitivity dramatically decreased compared with the continuous topical rhNGF group (Fig. 8D and E), whereas a single injection of AAV-Ngf was still functional in corneal epithelial healing during repetitive corneal epithelial injuries (Fig. 8D and E). The levels of SP and CGRP were the highest in the AAV-Ngf group compared with the other groups in both the cornea and TG (Fig. 8F and G).

Overall, our results demonstrate that a single injection of AAV-Ngf shows comparable nerve regeneration ability to rhNGF eye drops and provides protective effects on the corneal nerves for at least 5 months.

In clinical practice, patients with diabetic corneal lesions may not exhibit obvious ocular symptoms in the early stages. However, additional irritations such as ophthalmic examinations, medications, or surgeries can disrupt the fragile ocular surface homeostasis in patients with diabetes. This disruption may lead to recurrent corneal epithelial roughness and defects, accompanied by severe symptoms such as decreased vision, eye pain, and photophobia. Prolonged and chronic local irritating symptoms not only lead to severe sensational discomfort but also give rise to a series of psychological problems such as insomnia and anxiety, creating a vicious cycle. The rhNGF eye drop was developed to treat mild to severe corneal neuropathy, primarily focusing on restoring epithelial homeostasis. It is typically prescribed for 8 weeks, with most patients achieving complete corneal epithelial healing within this period (19–21,34). However, our results showed a regression of regenerated nerves and defects in corneal epithelial repair once topical rhNGF was discontinued. In comparison, gene therapy provided a long-lasting protective effect against corneal nerve degeneration and epithelial defects.

Intrastromal injection is frequently employed in ophthalmic clinical practice, with its safety well recognized (35). In this study, we developed an innovative intrastromal injection method, with transient corneal edema observed only within the 1st day postinjection. This novel injection method facilitated the rapid and uniform dispersion of the injected AAV across the corneal area. In contrast to the traditional intrastromal injection, which involved the creation of an intralamellar long tunnel to prevent drug leakage during injection, our new method used a 0.5-mm-long lamellar incision with an 11-0 suture needle for microinjection needle insertion. This modification aimed to reduce the invasiveness of the surgical procedure, preventing the formation of corneal scars. This alternative approach is not only useful for animal experiments but also can be applied for intrastromal drug delivery in clinical practice.

There have been approximately 47 clinical trials investigating AAV-mediated gene therapy for ocular diseases, as reported on the website ClinicalTrials.gov using the search terms “(AAV OR adeno associated virus OR adeno associated vector) AND (ocular OR eye).” However, almost all of these trials aim at treating retinal diseases, such as Leber hereditary optic neuropathy, Leber congenital amaurosis, X-linked retinitis pigmentosa, achromatopsia, and wet age-related macular degeneration. The AAV vectors used in these clinical trials are recombinant AAV2/2 and AAV2/5. To date, there are no ocular gene therapy clinical trials reporting using the AAV.rh10 vector. AAV.rh10 has been used to treat hemophilia B in adults (https://clinicaltrials.gov; search term “[AAV.rh10]”), but no treatment outcomes or safety results have been posted. Our study not only determined the nerve tropism of AAV.rh10 after intrastromal injection but also assessed the safety and efficacy of its usage in eyes.

Our study also provided insights into the dynamic immune reaction profile following intrastromal injection of AAVs. We employed flow cytometry to analyze corneal immune cells from day 1 to month 3. Because of the green fluorescence of the AAV vectors, we opted for single antibody staining to prevent fluorescence leakage. Although double labeling of immune cells was lacking, the identity of immune cells could be speculated. CD11b⁺ cells constituted the most dramatically increased immune cell population. F4/80⁺ macrophages partly contributed to this elevation, while neutrophiles appeared to be another major participant. T cells are reported to be closely related to capsid-specific responses (32). The adaptive immune responses mediated by AAV typically initiate at around 1 week after AAV injection (36,37). In our study, neither CD4⁺ nor CD8⁺ T cells exhibited a dramatic increase after 1 week of AAV injection, possibly because of the blood vessel–free status in cornea compared with other retinal gene therapy clinical trials.

As the dosing of AAV-Ngf increased from 10⁷ vg to 10¹⁰ vg, Ngf mRNA expression continued to rise, with the NGF protein level peaking at 10⁹ vg of injected AAV-Ngf. There might be a saturation point in the NGF production and secretion machinery in corneal and TG cells when 10⁹ vg of AAV-Ngf is introduced. Although NGF is the endogenous neurotransmitter produced by corneal nerves and corneal constituent cells, the consequences of overwhelming NGF supplementation remain unclear. Throughout the entire follow-up period of our AAV-treated diabetic or wild-type mice, no visible corneal or ocular abnormities were observed, implying that the intrastromal injection of 10⁹ vg of AAV-Ngf was well tolerated in this study.

In clinical practice, DK patients often present with persistent corneal epithelial defects. It has been reported that both the corneal epithelium and corneal nerves can be influenced by NGF. The observed corneal epithelial repair is frequently a result of corneal nerve recovery, which gradually normalizes the ocular surface structure through a positive feedback loop. Additionally, the corneal whorl diminishment in DK serves as an early indicator of corneal nerve degeneration, and IVCM can noninvasively diagnose early nerve changes in DK. Therefore, our gene therapy combined with fluorescein sodium staining can assess epithelial repair, while its integration with IVCM can evaluate early nerve repair, allowing for comprehensive monitoring and evaluation of the long-term efficacy of DK treatment (38). Furthermore, IVCM can screen DK patients who are suitable for gene therapy, even in the absence of corneal sensation impairment, enabling early detection and intervention.

However, our study has some limitations. Experiments on nonhuman primates were not conducted, which is necessary before proceeding to clinical trials. In our study, we employed the corneal sensitivity assay to monitor the functional recovery of corneal nerves. Healthy corneal sensory nerve fibers consist of polymodal nociceptor neurons, cold thermoreceptor neurons, and selective mechanonociceptor neurons (39). For a comprehensive evaluation of the functions of AAV-Ngf, additional electrophysiological recordings of all three types of neurons can be considered. Finally, although our improved intrastromal injection minimized the side effects, this method is still invasive compared with the eye drops. Future efforts can be employed to develop ocular surface topical delivery of gene therapy vectors.

Acknowledgments. The authors thank Wai Kit Chu, Department of Ophthalmology and Visual Sciences, The Chinese University of Hong Kong, for his constructive suggestions.

Funding. This study was supported by the National Natural Science Foundation of China (82101091 to B.N.Z., 82271059 to S.L.), the Shandong Provincial Key Research and Development Program (2021ZDSYS14 to L.X.), the Joint Innovation Team for Clinical & Basic Research (202405 to L.X.).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. Q.Z., Y.C., B.N.Z., and L.X. contributed to the conception and design of the study. L.C., B.Q., and S.C. carried out experiments. R.L. instructed the flow cytometry experiments. S.L. provided rhNGF. L.C. and B.N.Z. carried out data analysis. L.C. and B.N.Z. wrote the manuscript. All authors read and commented on the manuscript. B.N.Z. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.