The variation in the capsid structure of viral vectors, especially AAV vectors, has a significant impact on immune responses. Different serotypes of AAV have unique capsid structures, which determine their ability to target specific tissues by binding to specific receptors on host cell surfaces. These structural variations affect antigenic properties, viral entry mechanisms, targeting specificity, and transduction efficiency. AAV2 is the earliest and most widely used AAV serotype. It enters cells by binding to heparan sulfate, a type of glycosaminoglycan expressed on the surface of host cells, effectively transducing RPE cells (11). AAV5 and AAV8 serotypes are more effective in transducing retinal photoreceptor cells (12).

After AAV vectors enter the body, the host immune system recognizes specific antigenic epitopes on their capsid proteins such as the major structural proteins VP1, VP2, and VP3 of AAV (13). B cells produce specific antibodies (mainly IgG) against these antigens. These antibodies can bind to AAV vectors, forming antigen-antibody immune complexes, which then trigger the activation of the complement system. Complement activation by AAV IgG immune complexes has been demonstrated in human sera, including with AAV9, generating C3a and C5a and highlighting a potential risk of inflammatory injury at high vector doses (14).The deposition of C3b enhances the clearance of immune complexes and may also form the membrane attack complex (MAC), causing direct damage to the AAV vector and the transduced cells (15). In terms of cellular immunity, variations in the AAV capsid can affect the activation of CD4⁺ T cells and CD8⁺ T cells. The activation of these T cells can trigger more severe inflammatory responses and immune-mediated tissue damage. In murine intravitreal studies, administration of AAV2 provokes clinically evident vitritis within about one week and this largely resolves by about one month, whereas subclinical CD45⁺ infiltration, predominantly T cells, persists; prior exposure to AAV accelerates these adaptive responses (16). In particular, the cytotoxic responses mediated by CD8⁺ T cells may directly damage transduced cells, affecting treatment outcomes (17). Evidence from non-ocular clinical studies shows expansion of capsid specific CD8⁺ T cells after AAV exposure and loss of transduced cells in a hemophilia B trial, directly linking capsid antigen presentation to T cell mediated cytotoxic clearance (18).

The structure and sequence of rAAV genomes critically shape innate sensing and downstream adaptive responses. Across mouse hepatic and pulmonary models, self-complementary genomes expose CpG motifs more efficiently than single stranded genomes and consequently elicit stronger endosomal TLR9 dependent MyD88 signaling, type I interferon (IFN) responses, and capsid specific CD8⁺T cell priming; these effects are markedly blunted in Tlr9 deficient or Myd88 deficient mice (19–21). The GC-rich palindromic inverted terminal repeat (ITR) hairpins, while indispensable for replication and packaging, also harbor CpG motifs recognized by TLR9. Genome engineering strategies, including CpG depletion within the expression cassette and ITRs and the insertion of short TLR9 inhibitory oligomers, reduce inflammatory readouts in mouse liver, muscle, and retina, and after subretinal delivery in pigs. By contrast, in macaques, intravitreal dosing delayed but did not prevent uveitis, underscoring route and species dependent constraints (22, 23).

Beyond TLR9, interspecies differences in cytosolic DNA sensing complicate translation. Murine cells depend mainly on the cGAS-STING pathway, whereas human cells also elicit a strong STING independent response mediated by DNA dependent protein kinase, a kinase that can sense DNA ends and initiate innate signaling (24). Finally, promoter and transgene features, including length, GC content and CpG content, and predicted secondary structure, modulate both expression and engagement of TLR9. CpG depleted cassettes dampen TLR9 dependent cross priming without compromising expression in muscle and liver. CpG depleted ITRs have retained vector activity in vivo (20, 25). Consistent with these model data, human whole blood assays indicate that plasmacytoid dendritic cell TLR9 dependent type I IFN responses to AAV9 require pre-existing anti AAV9 antibodies, highlighting the influence of host humoral background on genome sensing (26).

Beyond the capsid and vector DNA, vector encoded transcripts are not immunological bystanders. Engagement of pattern recognition receptors (PRRs) and the magnitude of downstream interferon programs are determined by intrinsic RNA features, including length, sequence composition such as GU richness, chemical modifications, and higher order structure. Within endosomes, TLR3 senses long double stranded RNA, as shown in Tlr3 deficient mice and in polyinosinic polycytidylic acid challenge systems, whereas TLR7 and TLR8 detect GU rich single stranded RNA with pronounced species asymmetry. In mice, responses are dominated by TLR7 and TLR8 activity is weak or non-canonical. In humans, TLR8 is strongly responsive to single stranded RNA in primary monocytes, myeloid dendritic cells, and peripheral blood mononuclear cells (27–29).

In the cytosol, RIG I preferentially recognizes short RNAs bearing a5′-triphosphorylate, including double stranded and single stranded species, whereas MDA5 senses long double stranded RNA, and both receptors converge on MAVS. These assignments are supported by studies in knockout mice, human primary cells, and biochemical mapping of structure and length determinants (30, 31). Relevant to AAV, sustained transduction can generate antisense or minus-strand transcripts from the vector template, enabling sense–antisense pairing and dsRNA biogenesis. Across transformed cell lines (HeLa, HEK293, HepG2, Huh7), primary human hepatocytes (about 10–12 donors), and mice bearing human hepatocyte xenografts, dsRNA engages MDA5, signals through MAVS, and induces IFN-β, thereby curtailing transgene expression. siRNA knockdown of MDA5 or MAVS restores expression in vitro, directly linking transgene derived double stranded RNA to innate activation in human cells and in humanized mice (32). Collectively, these findings establish transgene RNA quality, not capsid alone, as a tunable driver of AAV immunogenicity and durability, and highlight the requirement to factor specie specific TLR7 and 8 biology into translational inference.

Transgenic proteins, as the final expression product of gene therapy, may be recognized as antigens by the host immune system, triggering a specific immune response. The strength and nature of this response depend on several factors related to the transgenic protein, including its structure, function, stability, and folding state (33). In particular, the glycosylation patterns and folding quality of the transgenic protein play an important role in its immunogenicity. For example, abnormal glycosylation patterns may cause the transgenic protein to be identified as a foreign antigen, while a protein that is not fully folded may expose atypical epitopes that can be recognized by the immune system. Detection techniques, such as proteomics (34) and ELISA, can be used in laboratory settings to assess the immunogenicity of proteins.

Viral vector capsids can undergo various PTMs, such as phosphorylation, glycosylation, and ubiquitination (35). These modifications can significantly alter the capsid’s three-dimensional structure, affecting its antigenicity and immunogenicity (36). Different types of PTMs may lead to changes in the immunogenicity of the viral vector in the body, which can impact both the safety and therapeutic effectiveness of gene therapy (37).

The innate immune system is the first line of defense against foreign pathogens, and in gene therapy, both viral vectors and gene editing tools like CRISPR/Cas9 and RNA interference (RNAi) can activate innate immune responses. Components of viral vectors, such as the protein capsid (44), vector DNA/RNA sequences (45), Cas proteins, associated small guide RNA sequences (sgRNA), and non-specific RNA degradation products in RNAi technology can be recognized by the host’s PRRs, triggering innate immune responses (46). PRRs are expressed on the cell membrane and in the cytoplasm of various cells, including innate immune cells (e.g., macrophages, monocytes, neutrophils, natural killer cells, and dendritic cells) (47), adaptive immune cells (such as T cells and B cells), and non-immune cells (such as epithelial cells, endothelial cells, and fibroblasts) (48). In the retina, PRRs are predominantly found in retinal pigment epithelial cells (RPE cells) (49), microglial cells (50), retinal ganglion cells (RGCs) (51), retinal astrocytes (52), and photoreceptor cells (53).

Immune responses triggered by AAV vectors involve multiple PRR pathways. TLR2 recognizes the protein capsid of AAV, activating the NF-κB pathway and inducing the production of inflammatory cytokines such as TNF-α and IL-1β (39). Unmethylated CpG motifs within AAV genomes are detected by TLR9, particularly in mouse models where genetic ablation of Tlr9 or Myd88 blunts type-I IFN induction and downstream adaptive immunity (19). In the cytoplasm, RIG-I and MDA5 detect double-stranded RNA produced after AAV infection, which activates type I IFN and NF-κB signaling pathways (32). Additionally, the NLRP3 inflammasome responds to AAV-induced cellular stress and damage by activating caspase-1, leading to the release of IL-1β and IL-18. AIM2 detects double-stranded DNA in the cytoplasm, triggering inflammasome activation (54).

Moreover, cGAS recognizes AAV genomic DNA and produces cGAMP, which activates STING. Upon activation, STING induces the production of type I IFNs and inflammatory cytokines. Mouse retina shows cGAS–STING activation in vivo (e.g., diabetic retinopathy), supporting the pathway’s relevance to ocular inflammation and its potential contribution to innate responses triggered by vector (55, 56). Figure 3 shows the AAV transduction process and the innate immune responses triggered in retinal gene therapy. These molecular pathways together form a complex immune response network triggered by various gene therapy methods.

The adaptive immune system (T cells and B cells) is gradually activated within days to weeks after gene therapy, potentially leading to vector clearance, loss of gene expression, and even cytotoxic reactions. In retinal gene therapy, adaptive immune responses mediated by T cells are key to treatment efficacy. CD4⁺ helper T cells activate and regulate immune responses through cytokine secretion, while CD8⁺ cytotoxic T cells directly kill infected cells. Both T cell types are essential for recognizing AAV vector components and transgene products (57).

Antigen-presenting cells (APCs), such as dendritic cells and macrophages, process and present antigenic peptide fragments from viral vectors or transgenes to T cells, triggering T cell activation. After activation, CD8⁺ cytotoxic T lymphocytes (CTLs) proliferate and differentiate, releasing perforins and granzymes to induce apoptosis and kill infected cells (58). During antigen presentation, CD4⁺ T cells recognize viral vector or transgene antigens through MHC II molecules. Upon activation, CD4⁺ T cells differentiate into various subsets, including T helper 1 (Th1), Th2, Th17, and T regulatory cells (Tregs). Each subset plays a distinct role in the immune response (59). T_H1 cells secrete IFN-γ, which promotes CTL activation and enhances macrophage activity, driving cell-mediated immunity (60). In contrast, Th2 cells produce interleukins such as IL-4, IL-5, and IL-13, which are key for promoting B cell-mediated antibody production and humoral immunity (61). Th17 cells secrete IL-17 and IL-22, involved in inflammatory responses and tissue remodeling processes (62). Tregs secrete immunosuppressive cytokines (IL-10 and TGF-β) to suppress excessive immune responses, prevent uncontrollable inflammation, and maintain immune homeostasis (63).

CD8⁺ T cells identify and eliminate target cells expressing viral vector or transgene antigens via MHC I molecules. This process is essential for clearing infected cells and regulating transgene expression (58). Following antigen clearance, some T cells differentiate into memory T cells. Central memory T cells (TCM) and effector memory T cells (TEM) can rapidly respond to subsequent exposures, enhancing immune protection in future gene therapy treatments. B cells play a pivotal role in humoral immunity by recognizing antigens, producing specific antibodies through plasma cell differentiation, and forming memory B cells for long-term immune memory (64). Antibodies neutralize viruses by blocking binding to host cell receptors or by interfering with viral entry into the host cell membrane. This neutralization diminishes the virus’s ability to infect and reduces the transduction efficiency of viral vectors. Furthermore, antibodies can bind to viral antigens on the surface of infected cells, engaging Fc receptors on effector cells such as NK cells, thereby triggering antibody-dependent cellular cytotoxicity (ADCC). Antibodies can also activate the complement system, initiating a cascade that leads to viral lysis and phagocytosis. Figure 4 illustrates the cellular and humoral immune responses during retinal gene therapy.

In retinal gene therapy, immunosuppressive drugs play a pivotal role in managing post-treatment inflammation and enhancing safety. Corticosteroids such as dexamethasone and prednisone are commonly used to suppress the release of inflammatory mediators, significantly reducing retinal tissue inflammation. Dexamethasone, well-known for its strong anti-inflammatory effects, inhibits cytokine and chemokine release, thus controlling immune responses to gene therapy (74). Additionally, it induces dendritic cells to exhibit tolerogenic properties, reducing the expression of co-stimulatory molecules and limiting T cell activation (75). Prednisone reduces inflammation by blocking the NF-κB pathway and decreasing inflammatory gene expression (76), offering protection to the retinal microenvironment. Clinical trials also indicate that prednisone helps mitigate chronic inflammation and retinal degeneration over prolonged treatment periods (6). Despite their therapeutic benefits, corticosteroids are not a panacea and present significant limitations. Local administration carries substantial risks, with 20-30% of long-term users developing corticosteroid-induced glaucoma and cataract formation. Systemic administration (oral/intravenous routes) may lead to more extensive complications, including immunosuppression and metabolic disturbances with prolonged use (77).

Beyond steroids, calcineurin inhibitors modulate T cell activation. Cyclosporine A mitigates disease in rat experimental autoimmune uveoretinitis (EAU), supporting a T cell targeted mechanism in vivo (78). Tacrolimus (FK506) similarly inhibits calcineurin and reduces ocular inflammation in rabbit and rat EAU after intravitreal delivery, with preservation of retinal structure and electroretinography (ERG) function (79, 80). Although ocular AAV specific tacrolimus data are limited, systemic tacrolimus prolonged rAAV8 and rAAV9 transgene expression in NHPs after skeletal and muscle delivery, indicating translatable T cell control across species and tissues (81, 82). Other agents are being explored. Mycophenolate mofetil inhibits lymphocyte proliferation. However, in mouse liver directed models, it lessens single stranded AAV transduction and may thereby limit overall efficiency (83). These challenges mirror broader findings in AAV gene therapy, where a systematic review of 73 clinical and real-world studies identified immunosuppression optimization as a critical determinant of treatment safety and efficacy (84).

Immune modulation is essential in controlling immune responses during retinal gene therapy. Immunoglobulin (IVIG), a key immunomodulator, neutralizes pathogenic antibodies and regulates immune functions, significantly reducing immune-mediated adverse reactions. IVIG is commonly used in clinical settings to manage acute immune reactions triggered by AAV vectors by binding to immune molecules, reducing inflammation, and improving treatment outcomes.

Plasmapheresis, though rarely used, can effectively reduce inflammation by removing pathogenic antibodies and immune complexes from the bloodstream, alleviating retinal inflammation (70, 85). In NHPs with pre-existing AAVrh74 immunity, two to three exchanges lowered neutralizing titers and enabled redosing, capsid based immunoadsorption columns also deplete anti-AAV IgG ex vivo and in animal models (70).

Rituximab (anti-CD20) depletes B cells, thereby reducing antibody mediated responses. In mouse and NHP models, B cell depletion, frequently paired with sirolimus, mitigated anti-capsid immunity and permitted AAV redosing. In clinical practice, rituximab is also employed for refractory ocular inflammation (86, 87). mTOR inhibition with rapamycin or sirolimus suppresses T and B cell activity. Phase III trials demonstrated that intravitreal sirolimus improves non-infectious uveitis. In mice and NHPs, rapamycin loaded tolerogenic nanoparticles (ImmTOR) attenuated anti-AAV immunity and enabled vector redosing (88, 89). JAK inhibition similarly reduces inflammatory signaling. Tofacitinib suppresses experimental autoimmune uveitis in mice, and ruxolitinib alleviates endotoxin-induced uveitis in rats, supporting pathway relevance to ocular inflammation even if direct AAV data are preclinical (90, 91). Moreover, immunoenzymes that transiently debulk antibodies are particularly promising in AAV settings. IgG-degrading endopeptidases (IdeS/IdeZ) restore AAV transduction in mice passively immunized with human Ig and permit in vivo gene transfer in NHPs in the face of NAbs. IceMG, a dual protease engineered to cleave IgM and IgG, rapidly clears both isotypes in rhesus macaques, reinstates AAV transduction in mice carrying human antisera, and reduces complement activation (67, 92, 93).

In retinal gene therapy, cytokine modulation plays a key role in managing inflammation. AAV vector injections activate Müller glial cells and microglia, triggering the release of pro-inflammatory cytokines including TNF-α and IL-1β (94). These glia–cytokine dynamics are consistent with broader microglia and Müller biology in mouse retina.

Although not specific to gene therapy, clinical experience in humans with non-infectious uveitis supports targeting these cytokines when intraocular inflammation threatens structure or function. Anti-TNF-α agents (e.g., infliximab) achieve corticosteroid sparing control in refractory uveitis (adult and pediatric cohorts), while anti-IL-6R (tocilizumab) is effective for uveitic cystoid macular edema and in juvenile idiopathic arthritis associated uveitis refractory to anti-TNF therapy. In ophthalmic practice they are generally systemic and used off label to suppress ocular inflammation (95–98).

Targeting the NF-κB and PI3K/Akt signaling pathways can also reduce immune overactivation, providing an additional strategy for immune response modulation. In mouse models of retinal and choroidal inflammation and neovascularization, microglial NF-κB activation is a driver of lesion development. Pharmacologic or genetic dampening of this axis reduces inflammatory infiltration and pathology. In parallel, PI3K/Akt signaling in retinal microglia governs activation and cytokine release, and blocking PI3K (e.g., LY294002) reduces retinal inflammatory programs in the oxygen-induced retinopathy model while dampening BV2 activation. Microglia-focused PI3K regulation (e.g., PIK3IP1) further links this pathway to retinal inflammatory angiogenesis in mouse models. Together, these data support NF-κB and PI3K/Akt as tractable nodes for adjunctive immunomodulation around ocular gene transfer, while acknowledging translation to AAV specific settings remains to be fully defined (99–101).

Modifications to AAV vector capsids are critical in reducing immunogenicity and improving therapeutic efficacy. Glycosylation of the capsid helps mask antigenic epitopes, making the vector less recognizable by the immune system (102). Techniques such as directed evolution and synthetic biology enable the development of AAV variants with lower immunogenicity. Mutations in capsid proteins, sequence recombination, and structural optimization can reduce immune responses while maintaining high transduction efficiency (103, 104). CRISPR/Cas9 technology also facilitates precise modifications to existing AAV vectors, further enhancing their properties (105). Non-viral vectors, such as liposomes and nanoparticles, also hold potential in retinal gene therapy, with modifications in size, surface charge, and composition improving their ability to evade the immune system and deliver genes more efficiently (106, 107). The use of AAV serotypes with low immunogenicity, such as AAV8 and AAV9, is an established strategy to enhance treatment outcomes. These serotypes exhibit high transduction efficiency and reduced immunogenicity in certain tissues (108).

Polyethylene glycol (PEG) modification of AAV vectors can reduce immune recognition, extending their circulation time in the body (109). Immunomodulatory coatings, such as the regulatory protein CTLA-4, can also be applied to the surface of AAV vectors to reduce immune responses (110). Decoy capsid technology, where immune cells bind to non-therapeutic capsids, can further reduce immune activation against the therapeutic vector (111). Epitope masking, which conceals antigenic epitopes, is another promising approach to reducing immune recognition (112).