A novel adeno-associated viral vector derived from human spleen isolate AAV.hu.S17 Authors: Wenyan Guo, Jiawen Sun, Fei Wang, Xiaoyu Nan, Xiaoping Rao, Feng Xiong, Tianlu Zhang, Fei Jin, Qitian Wang, Yujuan Chen, Yafeng Song, Minhua Luo, Yin Shen, Yurong Liu, Yang Wu Published: 2026 ABSTRACT Adeno-associated viruses (AAVs) are widely used vectors for gene therapy owing to their nonpathogenic nature, low immunogenicity, and ability to support long-term transgene expression. However, the tissue tropism and immune response of AAVs are strongly dependent on their capsid serotype. Although hundreds of naturally occurring AAV variants have been isolated, many remain unvectorised and poorly characterized. Therefore, the discovery and development of previously uncharacterised native serotypes retain valuable. Here, we generated a novel recombinant AAV vector rAAV.hu.S17, derived from the human spleen isolate AAV.hu.S17, and systematically evaluated its capsid features, in vitro transduction, and in vivo tissue tropism. Sequence analysis showed that AAV.hu.S17 is closely related to AAV2 and AAV3, and shares 90.9% amino acid identity to AAV2 across the capsid proteins. High-titer vector production was achieved using the triple-plasmid system, with rAAV.hu.S17 yielding approximately 2.68-fold more vector than that of rAAV2. Although rAAV.hu.S17 showed ∼20-fold lower transduction in HEK293T cells than rAAV2 in vitro, it mediated higher neuronal transduction in the mouse primary visual cortex (V1) in vivo. Following intravitreal injection into C57BL/6 mice, rAAV.hu.S17 preferentially transduced photoreceptors in the outer nuclear layer (ONL), with minimal transduction of the inner nuclear layer (INL) and the ganglion cell layer (GCL). Collectively, these results demonstrate that rAAV.hu.S17 exhibits transduction characteristics distinct from those of rAAV2 in both the brain and retina. This study provides the first vectorization and biological characterization of the previously uncharacterized human AAV isolate AAV.hu.S17 and expands the repertoire of naturally occurring AAV capsids available for future biological investigation and vector engineering. Introduction Adeno-associated virus (AAV), first identified in 1965,1 has emerged as a leading viral vector for gene therapy due to its non-pathogenic nature, long-term transgene expression, and broad tissue tropism.2, 3, 4 In neuroscience and ophthalmology, AAVs have shown remarkable success in treating genetic disorders, neurodegenerative diseases, and retinal degenerations.5, 6, 7 AAV can be divided into several serotypes according to different capsid proteins.8 To date, numerous AAV serotypes have been isolated from natural sources and characterized, each exhibiting distinct biological and vector properties.9, 10 For example, AAV2, AAV5, AAV8, and AAV9 are widely used in gene therapy owing to extensive experimental evidence supporting their favorable vector performance.11 AAV2, the earliest and most extensively studied serotype, can support long-term transgene expression in target cells after a single administration. Notably, Luxturna (voretigene neparvovec), an AAV2-based gene therapy approved in 2017, established the safety and efficacy of treating RPE65 mutation–associated retinal dystrophy and opened a new paradigm for inherited retinal disease therapy.12, 13 Upstaza (eladocagene exuparvovec), approved in 2022, likewise employs AAV2 to deliver the aromatic L-amino acid decarboxylase (AADC) gene to the striatum (e.g., the putamen), restoring AADC activity and increasing dopamine and 5-hydroxytryptamine synthesis.14 Although AAV vectors have enabled major advances in gene therapy and neuroscience, limitations in transduction efficiency, tissue tropism, and immunogenicity have become increasingly evident as the field has matured.15 Consequently, developing novel AAV serotypes and engineered capsids with improved tropism and specificity remains a critical challenge for treating neurodegenerative diseases and inherited retinal disorders.16, 17, 18 Most engineered AAV variants derive from natural serotypes. Comprehensive characterization of their tissue tropism and structure–function relationships provide a broader foundation for rational design and directed evolution.19 For example, AAVv66, a multimutant variant closely related to AAV2 and identified by screening human clinical samples, exhibits functional properties distinct from those of AAV2.20 Likewise, the natural serotype AAV11 exhibits retrograde transduction comparable to AAV2-retro, underscoring the continuing value of natural serotypes as both direct therapeutic candidates and templates for further engineering.21 Therefore, ongoing discovery and development of previously uncharacterized natural AAV serotypes are essential for expanding the capsid design space and improving therapeutic performance. This study investigates AAV.hu.S17, isolated from human spleen tissue, whose distinctive capsid surface topology suggests therapeutic potential. Prior sequence analyses indicate that AAV.hu.S17 arose through recombination between AAV2 and AAV322; however, its in vivo performance as a gene-delivery vector has not been defined. Here we generated rAAV.hu.S17-CMV-GFP and, for the first time, we comprehensively evaluated production yield and physical properties as well as its transduction in HEK293T cells and in the brain and retina of C57BL/6 J mice. We further benchmarked rAAV.hu.S17 against the canonical AAV2 serotype to evaluate its potential as a novel targeted delivery vector for neuroscience and ophthalmic disease therapies. Materials and methods Sequence analysis and structural prediction AAV Cap protein sequences for AAV1–AAV13 were obtained from GenBank (AAV1, NP_049542.1; AAV2, YP_680426.1; AAV3, NP_043941.1; AAV4, NP_044927.1; AAV5, YP_068409.1; AAV6, AAB95450.1; AAV7, YP_077178.1; AAV8, YP_077180.1; AAV9, AAS99264.1; AAV10, AAT46337.1; AAV11, AAT46339.1; AAV12, ABI16639.1; AAV13, ABZ10812.1), as well as for AAV.hu.S17 (AAU05370.1) and AAVv66 (QJX58010.1). Multiple sequence alignment was performed using MAFFT. Phylogenetic relationships were inferred in MEGA v12 using the neighbor-joining method with 1000 bootstrap replicates. Furthermore, 42 naturally occurring wild-type human AAV strains belonging to the canonical AAV Clade B sublineage were included as a reference dataset to enable high-resolution comparative phylogenetic positioning of AAV.hu.S17. Together with AAV.hu.S17, these strains comprised a dataset of 43 full-length VP1 capsid sequences for phylogenetic reconstruction, with detailed sequence information provided in Supplementary Table 1. To enable a higher-resolution (“zoom-in”) view of the AAV2-related lineage, a dedicated local phylogenetic subtree was generated.22, 23 Protein structure predictions were generated with AlphaFold 3 (https://alphafold.ebi.ac.uk/), and models were inspected in UCSF ChimeraX with stereo 3D quad-buffering enabled. Plasmid construction To obtain the pAAV-RC.hu.S17 packaging plasmid, AAV.hu.S17 Cap gene sequence was retrieved from GenBank (accession no. AY695376.1), synthesized, and cloned into pUC57 plasmid, which served as the PCR template. The AAV.hu.S17 Cap fragment was amplified with primers S17-F (5′-AATGATTTAAATCAGGTATGGCTGCCGATGGTTATCTTC-3′) and S17-R (5′-ATCAACCGGTTTATTGATTAACACGTAATTACAGATTACGAGTCAG-3′), and then replaced the AAV1 Cap fragment in pAAV2/1 (Addgene plasmid #112862) using SwaI and AgeI restriction sites. For convenient detection of rAAV, a cytomegalovirus (CMV) promoter–driven GFP reporter cassette was used, and the pAAV-CMV-GFP-WPRE-HGHpA core expression plasmid was constructed. Schematic plasmid map and full-length nucleotide sequences are provided in Supplementary Fig. S1 and Supplementary Sequence S1, respectively. All constructs were verified by sequencing. rAAV production and purification HEK293T human embryonic kidney cells were cultured in DMEM supplemented with 10% (v/v) FBS and penicillin–streptomycin (100 U/mL and 100 μg/mL, respectively) at 37 °C in a humidified incubator with 5% CO₂. rAAV.hu.S17-CMV-GFP and rAAV2-CMV-GFP vectors were produced in HEK293T cells by triple-plasmid cotransfection. At 72 h after transfection, cells were harvested and viral particles were purified by iodixanol-gradient ultracentrifugation.24 Viral vectors were aliquoted and stored at −80 °C until further use. Virus titration and silver staining Viral genome titers were determined by quantitative PCR (qPCR) with iQ SYBR Green Supermix (Bio-Rad) according to the manufacturer's instructions. Standard curves were generated from 10-fold serial dilutions of a standard plasmid ranging from 108 to 103 copies/μL. The primers targeting WPRE were Q-WPRE-F (5′-TGCTTCCCGTATGGCTTTCAT-3′) and Q-WPRE-R (5′-GGTTGCGTCAGCAAACACAG-3′). Purified rAAV samples were separated on 10% SDS–PAGE gels and subsequently silver-stained using the Fast Silver Stain Kit (Beyotime) according to the manufacturer's instructions. Transmission electron microscopy Transmission electron microscopy (TEM) was performed on purified rAAV preparations. Approximately 5 μL of sample was applied to a 400-mesh carbon-coated copper grid and incubated for 2 min. Excess liquid was removed by wicking from the grid edge with Whatman filter paper. The grid was negatively stained with 2% (w/v) phosphotungstic acid (pH 7.4) for 2 min, rinsed with a drop of double-distilled water (ddH₂O), and air-dried. Grids were examined with a Tecnai G2 20 TWIN transmission electron microscope. Negatively stained images were acquired at 100,000× magnification. Heparan sulfate proteoglycan (HSPG) binding assay The HSPG binding assay was performed as described previously.25 Heparin Sepharose 6 Fast Flow resin (Cytiva, USA) was packed into 1.5-mL gravity-flow chromatography columns to a settled bed volume of 500 μL. Before sample loading, the columns were fully equilibrated with 10 column volumes of PBS-MK buffer. AAV samples were individually diluted with PBS-MK to a final volume of 500 μL and then loaded onto separate gravity-flow columns. All procedures were performed by gravity flow alone, and the resin bed was carefully protected from desiccation throughout the assay. Viral genome copies (vg) in the flow-through (FT), wash fractions, all stepwise elution fractions, and input samples were quantified by qPCR. The values were expressed as percentages of the total capsids applied to the column. In vitro transduction assay HEK293T cells were seeded in 96-well plates to reach ∼70% confluence and infected with rAAV.hu.S17-CMV-GFP and rAAV2-CMV-GFP at graded multiplicities of infection (MOI). At 48 h post-infection, GFP expression was assessed using a fluorescence microscope (Olympus X71). The intensity and spatial distribution of GFP fluorescence served as readouts of vector transduction efficiency. In vivo delivery and imaging All animal studies were approved by the Animal Care and Use Committee of the Innovation Academy for Precision Measurement Science and Technology, Chinese Academy of Sciences. Viral transduction was assessed in 8-week-old male C57BL/6J mice. Stereotaxic injections were performed as described previously.21 rAAV (4 × 108 vector genomes in 200 nL) was stereotaxically microinjected into the primary visual cortex (V1). Injection coordinates relative to bregma were AP −3.90 mm, ML ±2.60 mm, and DV −1.30 mm. Four weeks post-injection, mice were euthanized. Brains were removed, fixed in 4% paraformaldehyde (PFA) overnight, cryoprotected in 30% sucrose, and sectioned at 40 μm. Sections were stained with DAPI and subsequently imaged using a virtual slide microscope (Olympus VS120). Intravitreal Injections were performed as described previously.26 Mice were anesthetized with 1% pentobarbital by intraperitoneal injection and given topical oxybuprocaine for corneal anesthesia. After iodophor antisepsis, rAAV (3 × 109 vector genomes in 1.5 μL) was delivered intravitreally into one eye using a 10 μL Hamilton syringe with a 32-gauge needle under a dissecting microscope. Retinal tissue was harvested 4 weeks after injection, frozen-sectioned and stained with DAPI. Images were acquired on a virtual slide microscope (Olympus VS120). Data analysis GraphPad Prism 10.4 (GraphPad Software, La Jolla, CA, USA) was used for data analysis. ImageJ v1.8.0 (National Institutes of Health, Bethesda, MD, USA) was used for the quantification of positive cells. All statistical data were presented as the mean ± standard deviation (SD). Statistical significance was defined as *p ≤ 0.05, **p ≤ 0.01, and ***p ≤ 0.001. Results Sequence analysis and structure predication The AAV.hu.S17 was derived from pediatric spleen tissue which first reported approximately 20 years ago, and not previously developed as a vector. Sequence alignment with AAV2 revealed 90.9% amino acid homology in their capsid proteins, with significant divergence in variable regions (Fig. 1A). The AAV.hu.S17 capsid protein (Cap) gene spans 2205 bp, encoding 735 amino acids. Across VP1–VP3, AAV.hu.S17 largely matches AAV2 in the VP1 N-terminal region (residues 1–180 aa). In contrast, eight amino-acid substitutions occur in the VP2 N-terminal region (139–202 aa) and five in the VP3 N-terminal region (203–235 aa). Most changes in variable regions cluster within five contiguous hypervariable regions (VR-IV to VR-VII), totaling approximately 50 substitutions, whereas only four substitutions are found in other conserved regions. Residues R585 and R588 in AAV2, implicated in efficient heparan sulfate proteoglycan (HSPG) binding, were substituted in AAV.hu.S17 (R585N, R588T), thereby likely resulting in a substantial reduction in HSPG binding.Fig. 1Capsid protein analysis and phylogenetic characterization of AAV.hu.S17. A. Sequence alignment of VP1 between AAV2 and AAV.hu.S17. Variable regions (VR-I to VR-IX) are shown in blue, α-helices and β-strands are marked in green, and sequence differences within variable regions are outlined in red boxes. B. Phylogenetic analysis of AAV capsid protein sequences based on amino acid sequences from 15 AAV serotypes, including AAV1 to AAV13, AAVv66, and AAV.hu.S17. The red circle indicates AAV.hu.S17. The phylogenetic tree was constructed using the neighbor-joining method in MEGA version 12 and evaluated with 1000 bootstrap replications. The scale bar denotes the number of amino acid substitutions per site. C. Structural comparison of AAV2 and AAV.hu.S17. The atomic structure of AAV2 was obtained from the Protein Data Bank (PDB ID: 1LP3), and the three-dimensional structure of AAV.hu.S17 was predicted using AlphaFold3. Both structures were visualized and compared using UCSF ChimeraX. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)Fig. 1 Capsid protein analysis and phylogenetic characterization of AAV.hu.S17. A. Sequence alignment of VP1 between AAV2 and AAV.hu.S17. Variable regions (VR-I to VR-IX) are shown in blue, α-helices and β-strands are marked in green, and sequence differences within variable regions are outlined in red boxes. B. Phylogenetic analysis of AAV capsid protein sequences based on amino acid sequences from 15 AAV serotypes, including AAV1 to AAV13, AAVv66, and AAV.hu.S17. The red circle indicates AAV.hu.S17. The phylogenetic tree was constructed using the neighbor-joining method in MEGA version 12 and evaluated with 1000 bootstrap replications. The scale bar denotes the number of amino acid substitutions per site. C. Structural comparison of AAV2 and AAV.hu.S17. The atomic structure of AAV2 was obtained from the Protein Data Bank (PDB ID: 1LP3), and the three-dimensional structure of AAV.hu.S17 was predicted using AlphaFold3. Both structures were visualized and compared using UCSF ChimeraX. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) We performed phylogenetic analysis of AAV.hu.S17 against AAV1–AAV13 and AAVv66, a serotype closely related to AAV2. For comparison, AAVv66 carries only 11 amino-acid substitutions in VR-IV to VR-VII relative to AAV2 and exhibits increased production yield, virion stability, and CNS transduction.20 By contrast, AAV.hu.S17 clusters between AAV2 and AAV3 (Fig. 1B). Subsequent high-resolution phylogenetic analysis focusing on the AAV2-related lineage revealed that AAV.hu.S17 does not cluster within the canonical AAV2 group but instead forms a distinct branch adjacent to the AAV2-related clade within this reference framework (Supplementary Fig. S2). This indicates that AAV.hu.S17 is evolutionarily related to AAV2-associated AAVs but represents a divergent lineage rather than a member of the core AAV2 cluster. Consistently, AAV.hu.S17 shares a high capsid sequence identity with AAV2 (90.9%), further supporting its evolutionary proximity while suggesting functional divergence. Structural comparisons further revealed distinct three-dimensional conformations between AAV.hu.S17 and AAV2 (Fig. 1C). Relative to AAV2, the greater number of substitutions and the observed structural differences in AAV.hu.S17 suggest potential divergence in vector properties. Vectorization of the AAV.hu.S17 serotype In order to further study the characteristics of AAV.hu.S17 as gene delivery vector, we carried out rAAV carrier transformation. Firstly, the Cap.hu.S17 gene fragment was obtained by artificial gene synthesis, and then the Cap1 gene fragment in AAV1 packaging plasmid pAAV-RC1 was replaced by it, and the packaging plasmid of pAAV-RC.hu.S17 serotype was obtained. To assay rAAV vector activity, we constructed the core expression plasmid pAAV-CMV-GFP-WPRE-hGHpA, which contains a CMV promoter-driven green fluorescent protein (GFP) cassette followed by the woodchuck hepatitis virus post-transcriptional regulatory element (WPRE) and the human growth hormone polyA (hGHpA) respectively. Recombinant AAV (rAAV) was produced by triple-plasmid co-transfection of HEK293T cells, and purified vectors were titered by quantitative real-time PCR (qPCR). The yields were 5.60 × 104 vg/cell for rAAV.hu.S17 and 2.09 × 104 vg/cell for rAAV2, indicating that rAAV.hu.S17 production was ∼2.68-fold higher (Fig. 2A). The purity of rAAV virus was detected by SDS-PAGE silver staining. Similar to rAAV2, rAAV.hu.S17 displayed three bands corresponding to the capsid proteins VP1 (∼87 kDa), VP2 (∼73 kDa), and VP3 (∼62 kDa) in an approximate 1:1:10 ratio, indicating a highly pure preparation (Fig. 2B). Transmission electron microscopy (TEM) was used to examine rAAV.hu.S17 particles. As shown in Fig. 2D, rAAV.hu.S17 exhibits icosahedral capsids consistent with canonical AAV morphology and appears as relatively uniform spherical particles (∼20–25 nm in diameter), similar to rAAV2. Empty and full particles were differentiated based on the presence of an electron-dense central region. Most of the rAAV.hu.S17 viral particles appeared intact.Fig. 2Characterization of the physical and transduction properties of rAAV.hu.S17. A. Comparison of the yields of rAAV2 and rAAV.hu.S17, as determined by qPCR and expressed as vector genomes per cell (VG/cell). All experiments were conducted in triplicate. Data are presented as mean ± SD. Statistical analysis was performed using an unpaired two-tailed student's t-test (*p ≤ 0.05). B. Silver staining of purified rAAV2 and rAAV.hu.S17. The purified rAAVs were separated by 10% SDS-PAGE and analyzed by silver staining. Approximately 1 × 1010 VG of rAAVs were loaded per lane. C. Heparin binding was assessed by chromatographic separation of rAAVs on heparin-agarose, followed by quantification of the applied, flow-through, and salt-eluted viral fractions using qPCR. The amount of applied virus was defined as 100%. Error bars represent the standard deviations from at least three independent experiments. D. Negative staining transmission electron microscopy (TEM) analysis of purified rAAV2 and rAAV.hu.S17. Representative fields are shown. The original magnification is ×100,000. Scale bars represent 100 nm. E. Transduction activity analysis of rAAV2 and rAAV.hu.S17. HEK293T cells were infected with serial dilutions of rAAV at different MOIs and observed under a fluorescence microscope 2 days post-infection. Representative fields with the optimal MOI for each rAAV serotype are shown.Fig. 2 Characterization of the physical and transduction properties of rAAV.hu.S17. A. Comparison of the yields of rAAV2 and rAAV.hu.S17, as determined by qPCR and expressed as vector genomes per cell (VG/cell). All experiments were conducted in triplicate. Data are presented as mean ± SD. Statistical analysis was performed using an unpaired two-tailed student's t-test (*p ≤ 0.05). B. Silver staining of purified rAAV2 and rAAV.hu.S17. The purified rAAVs were separated by 10% SDS-PAGE and analyzed by silver staining. Approximately 1 × 1010 VG of rAAVs were loaded per lane. C. Heparin binding was assessed by chromatographic separation of rAAVs on heparin-agarose, followed by quantification of the applied, flow-through, and salt-eluted viral fractions using qPCR. The amount of applied virus was defined as 100%. Error bars represent the standard deviations from at least three independent experiments. D. Negative staining transmission electron microscopy (TEM) analysis of purified rAAV2 and rAAV.hu.S17. Representative fields are shown. The original magnification is ×100,000. Scale bars represent 100 nm. E. Transduction activity analysis of rAAV2 and rAAV.hu.S17. HEK293T cells were infected with serial dilutions of rAAV at different MOIs and observed under a fluorescence microscope 2 days post-infection. Representative fields with the optimal MOI for each rAAV serotype are shown. Together, these results demonstrate that the rAAV.hu.S17-CMV-GFP preparation achieved high titer and purity, providing a basis for subsequent in vitro and in vivo transduction studies. In vitro characterization of rAAV.hu.S17 To quantitatively assess the effect of the capsid mutations R585N and R588T in AAV.hu.S17 on HSPG binding, heparin-agarose column binding assays were performed. AAV2 exhibited robust heparin binding, with a binding rate of approximately 86%, whereas rAAV.hu.S17 showed minimal binding, with a binding rate of less than 5% (Fig. 2C). These findings indicate that rAAV.hu.S17 almost completely lost the canonical heparan sulfate proteoglycan (HSPG)-binding phenotype. To evaluate transduction efficiency in HEK293T cells, we seeded 6.0 × 104 cells per well in 96-well plates. Twelve hours later, cells were infected with serial dilutions of each vector at defined multiplicities of infection (MOI). Forty-eight hours post-infection, GFP expression was assessed by fluorescence microscopy. As shown in Fig. 2E, both rAAV.hu.S17-CMV-GFP and rAAV2-CMV-GFP produced robust GFP signals, which decreased with increasing dilution. Notably, rAAV.hu.S17-CMV-GFP at MOI of 5.0 × 105 yielded approximately the same GFP expression as rAAV2-CMV-GFP at MOI of 2.5 × 10,4 indicating that rAAV.hu.S17 exhibits ∼20-fold lower transduction efficiency than rAAV2 in HEK293T cells in vitro. The altered HSPG-binding capacity of rAAV.hu.S17 may have indirectly contributed to its reduced infectivity in HEK293T cells in vitro. In vivo characterization of rAAV.hu.S17 Accordingly, we investigated the in vivo tropism of the new serotype. We first performed stereotaxic intracranial injections targeting the primary visual cortex (V1) of wild-type C57BL/6J mice with rAAV.hu.S17-CMV-GFP to assess neuronal transduction (Fig. 3A). Four weeks after injection, fluorescence imaging revealed numerous GFP-positive neurons in the V1 region, with well-resolved morphology. rAAV.hu.S17 demonstrated a larger transduction range than rAAV2, resulting in 3.1-fold more EGFP-positive neurons in the V1 region (Fig. 3B). The mean numbers of GFP-positive cells were approximately 81 for rAAV.hu.S17 and 26 for rAAV2. These findings indicate that rAAV.hu.S17 mediated broader transgene expression and a greater number of GFP-positive cells than rAAV2 in the V1 region under the experimental conditions used in this study.Fig. 3Comparison of brain transduction profiles between AAV2 and rAAV.hu.S17 following stereotaxic injection into the primary visual cortex. A. Schematic illustration of stereotaxic injection targeting the primary visual cortex (V1) region. B. Fluorescence images showing rAAV-mediated transduction in the brains of C57BL/6J mice four weeks after injection. Each vector (rAAV2-CMV-GFP or rAAV.hu.S17-CMV-GFP; 4 × 108 vg in 200 nL) was delivered into the V1 region. Green fluorescence represents GFP expression, and blue fluorescence (DAPI) labels cell nuclei. Scale bar, 100 μm. C. Quantitative analysis of the number of GFP+ cells transduced with the AAV vectors in V1. Statistical analysis was performed using an unpaired two-tailed student's t-test (***p ≤ 0.001). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)Fig. 3 Comparison of brain transduction profiles between AAV2 and rAAV.hu.S17 following stereotaxic injection into the primary visual cortex. A. Schematic illustration of stereotaxic injection targeting the primary visual cortex (V1) region. B. Fluorescence images showing rAAV-mediated transduction in the brains of C57BL/6J mice four weeks after injection. Each vector (rAAV2-CMV-GFP or rAAV.hu.S17-CMV-GFP; 4 × 108 vg in 200 nL) was delivered into the V1 region. Green fluorescence represents GFP expression, and blue fluorescence (DAPI) labels cell nuclei. Scale bar, 100 μm. C. Quantitative analysis of the number of GFP+ cells transduced with the AAV vectors in V1. Statistical analysis was performed using an unpaired two-tailed student's t-test (***p ≤ 0.001). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) We next evaluated transduction efficiency and cellular tropism of C57 mouse retina by intravitreal injection (Fig. 4A). The retinas of C57 mice were collected 4 weeks after administration and processed for staining. In eyes injected with rAAV2-CMV-GFP, GFP fluorescence was detected broadly across retinal layers, including the ganglion cell layer (GCL), sparsely in the inner nuclear layer (INL), and with occasional signal near the retinal outer nuclear layer (ONL). In contrast, rAAV.hu.S17-CMV-GFP predominantly transduced retinal photoreceptors in the ONL, with minimal labeling of bipolar cells or retinal ganglion cells in the INL and GCL. rAAV.hu.S17 exhibited a transduction pattern distinct from that of rAAV2, resulting in 2.5-fold more EGFP-positive neurons in the ONL. The mean numbers of GFP-positive cells were approximately 94 for rAAV.hu.S17 and 37 for rAAV2. These results indicate that rAAV.hu.S17 displays retinal transduction characteristics distinct from those of rAAV2.Fig. 4Comparison of retinal transduction profiles between AAV2 and rAAV.hu.S17 following intravitreal injection. A. Schematic illustration of rAAV intravitreal injection into the eyes of C57BL/6J mice. B. Fluorescence images showing retinal transduction by rAAVs four weeks after injection. Each virus (rAAV2-CMV-GFP or rAAV.hu.S17-CMV-GFP; 3 × 109 vg in 1.5 μL) was delivered into the vitreous cavity of C57BL/6J mouse eyes. Green fluorescence represents GFP expression, and blue fluorescence (DAPI) marks cell nuclei. ONL, outer nuclear layer; INL, inner nuclear layer; GCL, ganglion cell layer. Scale bar, 100 μm. C. Quantitative analysis of the number of GFP1 cells transduced with the AAV vectors in ONL layer. Statistical analysis was performed using an unpaired two-tailed student's t-test (***p ≤ 0.001). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)Fig. 4 Comparison of retinal transduction profiles between AAV2 and rAAV.hu.S17 following intravitreal injection. A. Schematic illustration of rAAV intravitreal injection into the eyes of C57BL/6J mice. B. Fluorescence images showing retinal transduction by rAAVs four weeks after injection. Each virus (rAAV2-CMV-GFP or rAAV.hu.S17-CMV-GFP; 3 × 109 vg in 1.5 μL) was delivered into the vitreous cavity of C57BL/6J mouse eyes. Green fluorescence represents GFP expression, and blue fluorescence (DAPI) marks cell nuclei. ONL, outer nuclear layer; INL, inner nuclear layer; GCL, ganglion cell layer. Scale bar, 100 μm. C. Quantitative analysis of the number of GFP1 cells transduced with the AAV vectors in ONL layer. Statistical analysis was performed using an unpaired two-tailed student's t-test (***p ≤ 0.001). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) In summary, rAAV.hu.S17 exhibited biological properties distinct from those of rAAV2, including markedly reduced HSPG binding, lower transduction efficiency in HEK293T cells, broader transgene expression in the mouse visual cortex, and preferential transduction of photoreceptor-rich retinal layers following intravitreal administration. Discussion Adeno-associated virus (AAV), as a non-pathogenic and low-immunogenicity gene delivery tool, has been widely adopted in gene therapy. Its unique capsid architecture and genomic properties enable efficient and sustained transgene expression across diverse tissues.27, 28 In recent years, AAV-based therapies have achieved clinical success in treating genetic disorders such as retinitis pigmentosa, spinal muscular atrophy, and hemophilia.29, 30, 31 However, existing AAV serotypes remain constrained by limitations in tissue specificity and immune evasion capabilities, restricting their application to a broader spectrum of diseases.32, 33 Various AAV serotypes have been isolated from tissue-culture stocks, human samples, and non-human primates.9 However, numerous AAV serotypes have yet to be thoroughly investigated. In 2005, Chen et al. isolated AAV.hu.S17 from pediatric spleen tissue and reported that it was phylogenetically intermediate between AAV2 and AAV3.34 Over 50 capsid amino-acid substitutions are clustered within five contiguous hypervariable regions (VR-IV to VR-VIII) of AAV.hu.S17. To date, there have been no reports on vectorizing AAV.hu.S17 or characterizing its properties. The extensive sequence divergence within the capsid variable regions suggested that AAV.hu.S17 may possess biological properties distinct from those of previously characterized serotypes. In this study, we successfully constructed and produced rAAV.hu.S17, achieving a yield ∼2.68-fold higher than that of rAAV2. These findings indicate that AAV.hu.S17 can be efficiently produced using standard triple-transfection methods and may provide useful insights into determinants of AAV production efficiency. However, compared with rAAV2, rAAV.hu.S17 exhibits ∼20-fold lower in vitro transduction efficiency in HEK293T cells (Fig. 2D). This reduction is consistent with substitutions at the heparan sulfate proteoglycan (HSPG)–binding residues (R585N, R588T) in AAV.hu.S17 Cap protein, which reduce HSPG affinity (Fig. 2C). Similar findings have been reported for AAVv66 identified in a human clinical sample by high-throughput.20 In vitro performance of AAV vector does not reliably predict infectivity or its transduction in vivo, because of the complex microenvironment, anatomical and cellular barriers, and the host immune response, among other factors.35 For example, rAAV9 shows lower in vitro transduction in HEK293T cells but markedly higher in vivo transduction of neural cells than rAAV2.11 Given the widespread use of AAV vectors in studies of the central nervous system and retina, we evaluated the in vivo transduction characteristics of rAAV.hu.S17 in these tissues.33 In stereotaxic injections targeting the primary visual cortex (V1) of C57 mice, rAAV.hu.S17 resulted in a greater number of GFP-positive cells and a broader transduction range than rAAV2 (Fig. 3B). rAAV.hu.S17 displayed a retinal transduction pattern distinct from that of rAAV2 following intravitreal injection, with predominant transduction in the photoreceptor layer and minimal transduction in the ganglion cell and other inner retinal layers (Fig. 4B). By contrast, rAAV2 transduced the ganglion cell and inner retinal layers and also reached photoreceptors, consistent with prior report.26 These findings demonstrate that rAAV.hu.S17 exhibits transduction characteristics distinct from those of rAAV2 in both the brain and retina. The observed differences warrant further investigation across additional tissues and experimental settings. Further studies should evaluate rAAV.hu.S17 tropism across additional neural tissues and the durability of transgene expression during long-term use. A limitation of the present study is that comparisons were restricted to rAAV2, which was selected because of its close phylogenetic relationship and high capsid sequence identity to AAV.hu.S17. Consequently, the relative performance of rAAV.hu.S17 compared with other commonly used CNS- and retina-directed serotypes, including AAV8 and AAV9, remains unknown. Therefore, the present study should be regarded as an initial characterization of AAV.hu.S17 rather than a comprehensive benchmarking analysis.11, 26 Future studies incorporating direct side-by-side comparisons with established serotypes, together with broader tropism and immunogenicity assessments, will be necessary to more fully define the biological characteristics and potential applications of rAAV.hu.S17. Thus, evaluating rAAV.hu.S17 alongside these serotypes in future studies would be of highly meaningful. Moreover, comprehensive immunogenicity profiling will be essential to support clinical translation. In conclusion, this study reports the successful vectorization and initial characterization of the previously uncharacterized human AAV isolate AAV.hu.S17. Relative to rAAV2, rAAV.hu.S17 displayed distinct production characteristics, receptor-binding properties, and transduction profiles in vitro and in vivo. These findings expand the repertoire of naturally occurring AAV capsids available for future biological investigation and vector engineering and provide a foundation for future studies aimed at defining the biological characteristics and potential applications of AAV.hu.S17 relative to established AAV serotypes. Author’s contribution J.S., F.W. and X.N. performed the experiments and acquired the data and drafted the paper. F.J., Q.W and Y.C. carried out viral vector production; X.R., T.Z. and Y.S. guided the animal experiments; F.X., M.L. and Y.S. guided sequence analysis and structural prediction; and W.G., Y.L. and Y.W. supervised the project and edited the manuscript. Disclaimer The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of the journal and/or the editor(s). The journal and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. CRediT authorship contribution statement Wenyan Guo: Writing – review & editing, Visualization, Supervision, Resources, Project administration. Jiawen Sun: Writing – original draft, Visualization, Validation, Project administration, Investigation, Formal analysis, Data curation. Fei Wang: Writing – original draft, Visualization, Validation, Project administration, Investigation, Formal analysis, Data curation, Conceptualization. Xiaoyu Nan: Writing – original draft, Validation, Investigation, Conceptualization. Xiaoping Rao: Validation. Feng Xiong: Software, Formal analysis. Tianlu Zhang: Validation. Fei Jin: Validation. Qitian Wang: Validation. Yujuan Chen: Validation. Yafeng Song: Software, Formal analysis. Minhua Luo: Software, Formal analysis. Yin Shen: Validation. Yurong Liu: Writing – review & editing, Visualization, Supervision. Yang Wu: Writing – review & editing, Visualization, Supervision, Resources, Project administration. Funding This research was funded by the National Natural Science Foundation of China (82271438, 82071413, 82301383); the National Natural Science Foundation of China for Original Exploration Project (82151309); the General Program of the Department of Education of Hunan Province (24C0273); the Hunan Provincial Natural Science Foundation Regional Joint Fund (2025JJ70068) and Wuhan Key Research and Development Program (2025020602030101). Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.