The study, published in the Journal of Translational Medicine, describes how the researchers retooled the AAV5 capsid using rapid directed evolution. Their strategy centered on variable region VIII of the viral capsid protein, a surface-exposed loop that governs how the virus engages cellular receptors. At residue Q574 within this region, they inserted libraries of random heptapeptides, seven-amino-acid sequences that collectively presented an enormous diversity of receptor-binding profiles. Each randomized capsid was packaged with its own encoding genome, so that the viral particles effectively carried the blueprint of their own success. The library was then subjected to iterative selection on 661W cells, a well-established mouse photoreceptor-derived cell line, allowing only those variants that entered these cells efficiently to amplify through successive rounds.

This selection scheme is a classic example of artificial Darwinism applied to a gene therapy vector. Rather than rationally designing a single capsid modification and hoping it improves tropism, the team let the biology choose. Millions of capsid variants competed for entry into photoreceptor-like cells, and the fittest clones were recovered, re-amplified, and challenged again. After several rounds, one variant emerged as the clear winner and was designated AAV5-AlyT. The name reflects its tropism for the outer retinal layers, and the variant carries a heptapeptide insertion at the Q574 position that presumably alters the capsid’s interaction with receptors on the inner limiting membrane and on photoreceptor cells themselves, the two principal barriers that natural AAV5 must overcome after intravitreal delivery.

In vitro comparisons confirmed that AAV5-AlyT transduced 661W cells far more efficiently than its parental AAV5 scaffold. The critical question, however, was whether that cellular advantage would translate into living eyes, where the vitreous, the inner limiting membrane, and the layered architecture of the retina impose constraints that cell culture cannot reproduce. In mice, a single intravitreal injection of AAV5-AlyT carrying a reporter transgene produced robust gene expression concentrated in the outer retina, the compartment containing photoreceptor inner and outer segments. The expression was not a transient burst; it persisted for more than six months after the single injection, an essential property for any therapy aimed at chronic degenerative diseases in which the treated cells must be supported for years.

The vector’s performance extended beyond mice. In rabbits, whose eyes are larger and anatomically closer to the human eye in several respects, AAV5-AlyT achieved efficient transduction across retinal layers after intravitreal delivery. Cross-layer transduction is a notable feature, because it suggests the capsid can penetrate the inner limiting membrane and then distribute genetic cargo to cells in multiple retinal strata rather than being confined to a narrow band. For translational purposes, this matters enormously: rabbit and non-human primate eyes are the standard large-animal models used to de-risk retinal gene therapy before human trials, and a vector that fails to scale from mouse to larger eyes rarely survives the development pipeline.

Having established delivery efficiency, the researchers tested whether AAV5-AlyT could actually treat disease. They chose two complementary models of retinal degeneration. The first was the ABCA4 knockout mouse, a model of Stargardt disease, an inherited macular dystrophy caused by mutations in the ABCA4 transporter gene that lead to toxic lipofuscin accumulation in photoreceptors and the retinal pigment epithelium. The second was light-induced retinal degeneration, a widely used model that recapitulates aspects of dry age-related macular degeneration, in which excessive light exposure triggers photoreceptor death and inflammatory degeneration. In both models, the therapeutic cargo was not a replacement gene but the complement regulator Crry, a mouse analogue of human complement regulatory proteins that dampens the cascade of complement activation driving tissue damage.

The results were striking. AAV5-AlyT-mediated delivery of Crry markedly rescued visual function in both disease models, as measured by electroretinography, the standard electrophysiological test of retinal responsiveness to light. Immunofluorescence and histological analyses showed that the treatment suppressed pathological complement activation and reduced damage to the retinal pigment epithelium, the support layer whose dysfunction is central to both Stargardt disease and dry AMD. These findings position the vector not merely as a delivery tool but as a validated platform for anti-inflammatory and neuroprotective gene therapy in the outer retina, where complement dysregulation is increasingly recognized as a key driver of degeneration.

One of the most clinically consequential findings concerns pre-existing immunity. A significant fraction of the human population carries neutralizing antibodies against common AAV serotypes, generated by prior natural infection, and these antibodies can neutralize therapeutic vectors before they reach their target cells, rendering treatment ineffective or forcing exclusion of patients from trials. The team screened 300 human serum samples for neutralizing activity against AAV5 and against the engineered AAV5-AlyT. The engineered variant showed a 34.9 percent lower rate of pre-existing neutralizing antibody response than parental AAV5. The heptapeptide insertion, while enhancing retinal tropism, apparently also perturbed an antibody epitope on the capsid surface, a serendipitous but valuable outcome that could broaden the eligible patient population for future therapies built on this capsid.

The technical achievement rests on a few principles worth underscoring. Directed evolution of AAV capsids has produced several notable vectors in recent years, but the field has struggled to find variants that combine outer retinal tropism with the practicality of intravitreal delivery. The Q574 locus in variable region VIII proved to be a productive engineering site, likely because it sits within a capsid region that influences both receptor binding and antibody recognition. The iterative selection on a photoreceptor cell line provided a clean, high-throughput fitness landscape, and the two-animal-species validation, followed by therapeutic proof of concept in two distinct degeneration models, gives the work an unusually complete translational arc for a single study.

Caveats remain, as they always do at this stage. The efficacy data come from mouse and rabbit models, and the human eye presents additional barriers, including a thicker inner limiting membrane and a much larger vitreous volume, that will require testing in larger animals and eventually in clinical trials. The Crry protein used as therapeutic cargo is mouse-specific, so human applications would require an equivalent human complement regulator. The authors also note a patent application related to the capsid, reflecting the commercial interest that engineered AAV vectors typically attract. Even so, AAV5-AlyT addresses the central logistical obstacle that has limited retinal gene therapy to specialized surgical centers: the need for subretinal injection. A vector that reaches photoreceptors through an intravitreal shot, sustains expression for over six months, evades a meaningful fraction of pre-existing neutralizing antibodies, and has already demonstrated functional rescue in models of both inherited and age-related retinal degeneration represents a substantial step toward making outer retinal gene therapy as routine as the injections ophthalmologists deliver every day.

Subject of Research: Directed evolution of an adeno-associated virus capsid for outer retinal gene delivery

Article Title: An engineered AAV variant with high-efficiency outer retinal tropism for gene delivery

Article References: Li, Y., Zhao, Z., Chen, Y., Zai, X., Chen, Z., Zhang, J., Zhang, Z., Wang, B., Song, X., Long, J., Wu, S., & Hou, L. (2026). An engineered AAV variant with high-efficiency outer retinal tropism for gene delivery. Journal of Translational Medicine. https://doi.org/10.1186/s12967-026-09022-z

Image Credits: AI Generated

DOI: 10.1186/s12967-026-09022-z

Keywords: adeno-associated virus, directed evolution, gene therapy, retinal degeneration, intravitreal injection, photoreceptors, Stargardt disease, age-related macular degeneration, viral vectors, complement cascade, neutralizing antibodies, AAV5-AlyT