Human Gene Therapy

The global burden of chronic and genetic kidney diseases poses a significant challenge to healthcare systems. Current therapies, including dialysis, transplantation, and supportive pharmacotherapies, cannot halt disease progression or address root causes, especially in monogenic disorders like Alport syndrome and Fabry disease. Adeno-associated virus (AAV)-based gene therapy is promising, enabling targeted correction of underlying genetic defects. However, renal delivery faces challenges, including cellular heterogeneity, anatomical barriers, and pre-existing immunity. This review evaluates advances in AAV capsid engineering to overcome these obstacles, focusing on strategies to enhance kidney-specific tropism, transduction efficiency, and immune evasion. We outline the evolution from conventional serotype selection to precision engineering via rational design, directed evolution, and in silico approaches. Artificial intelligence (AI) has emerged as a pivotal accelerator, with machine learning models and generative frameworks enabling data-efficient capsid optimization despite limited datasets. Multimodal AI, reinforcement learning, and agentic systems can refine renal targeting by balancing glomerular penetration, cell specificity, and safety. Future progress relies on scaling high-quality datasets through collaborative consortia, lab-in-the-loop validation, and explainable AI. By combining capsid engineering with renal pathophysiology insights, this roadmap paves the way for curative AAV therapies that move beyond current suboptimal treatments to correct underlying pathogenic mechanisms.

Rheumatoid arthritis is a chronic autoimmune disease characterized by persistent synovial inflammation and progressive joint destruction. Although biological disease-modifying anti-rheumatic drugs (DMARDs) have transformed treatment, their systemic immunosuppression, high cost, and incomplete efficacy in certain joints remain significant challenges. To address these limitations, we developed a localized gene therapy using the human indoleamine 2,3-dioxygenase (hIDO) gene driven by the E-selectin (ELAM-1) promoter, which is responsive to inflammatory cytokines. We assessed ELAM-1 promoter responsiveness to interleukin-1β (IL-1β) stimulation in vitro mainly using the HeLa cell line, measuring hIDO expression and enzymatic activity by enzyme-linked immunosorbent assay, quantitative reverse transcription-polymerase chain reaction (qRT-PCR), Western blotting and kynurenine assays. In vivo, we established a robust collagen-induced arthritis (CIA) model in BALB/c mice, which are normally resistant to CIA, and delivered pELAM-1pro/hIDO plasmids intra-articularly via liposomes. Therapeutic efficacy was evaluated by clinical scoring, ankle circumference, histopathology, immunohistochemical quantification of synovial CD4⁺ T cells and CD68⁺ macrophages, and endpoint immunohistochemical assessment of IDO expression. In vitro, the ELAM-1 promoter allowed rapid, high-level expression of hIDO within 2 h of IL-1β stimulation, without IDO-induced suppression of the promoter. In vivo, pELAM-1pro/hIDO treatment significantly reduced ankle swelling and clinical arthritis scores more rapidly than constitutive CMV-driven hIDO, achieving marked histological improvement on day 7. The treated joints exhibited substantial reductions in pathogenic CD4⁺ T cells and CD68⁺ macrophages, consistent with the IDO immunoregulatory mechanism. This study supports the feasibility of a localized gene therapy approach using an inflammation-responsive promoter and suggests localized immunomodulation at disease sites in a manner consistent with inflammation-associated promoter activity.

Emerging evidence suggests CAR-NK cell therapy shows great promise in cancer treatment. ROBO1 is highly expressed in various cancer types, including glioblastoma, hepatocellular carcinoma, lung cancer, breast cancer, and uterine cancer. Our and other laboratories’ studies have shown that ROBO1 CAR-NK cells exhibit promising tumor therapeutic effects. However, the results still have some limitations. Cbl-b, an E3 ubiquitin ligase, has been reported to negatively regulate NK cell activation, homeostasis, and antitumor immunity. ¹ Therefore, we attempted to further enhance the antitumor activity of ROBO1 CAR-NK92 cells by knocking out Cbl-b using CRISPR/Cas9 gene-editing technology. In this study, we conjugated Cbl-b sgRNA with Cas9 protein to form ribonucleoprotein complexes, which were then delivered into ROBO1 CAR-NK92 and NK-92 cells (control cells) via electroporation. Through fluorescence-activated cell sorting, limiting dilution, and sequencing, we obtained monoclonal Cbl-b-knock-out (KO) cell lines. Both in vitro cytotoxicity assays and in vivo tumor xenograft experiments were conducted to examine whether Cbl-b knockout enhances the target cell killing and tumor suppression capacities of ROBO1 CAR-NK92 cells. In this study, monoclonal cell lines of ROBO1 CAR-NK92-Cbl-b-KO and NK92-Cbl-b-KO were successfully established. In vitro, at an effector-to-target (E:T) ratio of 0.1:1, ROBO1 CAR-NK92-Cbl-b-KO (50.55%) cells exhibited significantly higher cytolytic activity against ROBO1-positive T47D target cells after 3 h of coculture than ROBO1 CAR-NK92 (34.10%), NK92-Cbl-b-KO (22.22%), and parental NK-92 cells (3.28%). In vivo, tumor volume and weight measurements demonstrated that mice treated with ROBO1 CAR-NK92-Cbl-b-KO cells developed significantly smaller tumors than all control groups, achieving a tumor growth inhibition (TGI) rate of 32.45%, indicating enhanced antitumor efficacy conferred by Cbl-b knockout. In vitro and in vivo data confirmed that Cbl-b knockout potentiates the antitumor efficacy of ROBO1 CAR-NK92 cells. The overall cytotoxic capability ranked as follows: ROBO1 CAR-NK92-Cbl-b-KO > ROBO1 CAR-NK92 > NK92-Cbl-b-KO > NK-92.

Site-specific integration of large genes in human primary stem cells remains a significant challenge in gene therapy, particularly for treating multiallelic diseases. Gene editing efficiency in primary stem cells is heavily influenced by the delivery strategy, which often faces issues with programmability, efficiency, and specificity. Here, we developed a dual-viral delivery system, targeted integration via virus-like particles and integrase-deficient lentivirus (TIVID). This system combines virus-like Cas9 edit particles for delivering Cas9/sgRNA ribonucleoprotein complexes and integrase-deficient lentiviral vectors for delivering HDR donor templates. The TIVID system achieves a knock-in efficiency of 65% ± 5% in human induced pluripotent stem cells (iPSCs). In erythroid progenitor HUDEP2 cells, TIVID mediates precise integration of a 7.1 kb HBB-GFP cassette (from cut site to cut site) at the AAVS1 locus with 20% efficiency and stable expression. Crucially, we demonstrate that TIVID overcomes stringent packaging constraints to deliver an approximately 6 kb full-length HBB therapeutic cassette into primary human CD34⁺ hematopoietic stem and progenitor cells. This platform achieved 5–10% targeted integration efficiency and preserved robust lineage-specific differentiation capacity, demonstrating its potential for treating β-thalassemia and other multiallelic disorders. In head-to-head comparisons, TIVID outperformed lentivirus-derived nanoparticles (∼50% vs. <10% at AAVS1 in K562 with M3814) and plasmid-based eePASSIGE in iPSCs (∼20% vs. ∼1.5%). Compared with traditional electroporation delivery, TIVID offers lower early cytotoxicity, promotes predominantly mono-allelic integration, and exhibits enhanced compatibility with primary stem cells. By decoupling nuclease and donor delivery, TIVID circumvents the payload constraints of single-vector systems and the toxicity of physical transfection, providing a robust ex vivo engineering platform for complex gene replacement therapies.

To treat patients affected with bestrophinopathies caused by mutations in the BEST1 gene, a recombinant adeno-associated virus vector (OPGx-BEST1 or rAAV2/2-VMD2-BEST1) is being developed. The vector construct includes the human BEST1 cDNA under control of the VMD2/BEST1 promoter and is packaged in an AAV2 capsid. This study evaluated the efficacy and toxicity of OPGx-BEST1 administered by subretinal injection in dogs that were homozygous mutant or compound heterozygotes for 3 naturally occurring mutations in BEST1. Twelve BEST1-mutant dogs were divided into 4 groups of 3 animals each, and they received in their left eye a subretinal injection of 0.15 mL of OPGx-BEST1 at 1 of 3 concentrations (9.5 × 10⁹ vector genome [vg]/mL; 3.0 × 10¹⁰ vg/mL; or 3.0 × 10¹¹ vg/mL) of OPGx-BEST1, resulting, respectively, in a low (1.4 × 10⁹ vg), high (4.5 × 10⁹ vg), and highest (4.5 × 10¹⁰ vg) dose or vehicle control. The right eyes were not injected. Subretinal injections were well tolerated and were not associated with any systemic or ocular toxicity. Electroretinography showed improved rod- and cone-mediated responses in eyes treated with OPGx-BEST1. Noninvasive retinal imaging by optical coherence tomography showed improved structural integrity with a reduction or prevention of appearance of vitelliform lesions and reversal of microdetachments in the retinal areas treated with OPGx-BEST1. These results support the use of OPGx-BEST1 in clinical studies with patients affected with bestrophinopathies and define the no-observed-adverse-effect level at 4.5 × 10¹⁰ vg/eye (0.15 mL, 3.0 × 10¹¹ vg/mL).

Comprehensive recombinant adeno-associated virus characterization is essential for establishing the knowledge base required to ensure clinical safety and efficacy, yet current long-read methods suffer from library preparation biases that obscure genome integrity. We present AviNP-seq, a blindspot-free nanopore sequencing framework utilizing one-end-sufficient ligation and Cas9–ribonucleoprotein (RNP) linearization to minimize terminal selection. Applied to a 1.5–6.5 kb panel, AviNP-seq delineates a sharp packaging cliff at 5.0–5.2 kb and reveals that sequence structure modulates integrity by 2–5× at fixed lengths. It unmasks covalent head-to-tail tandems in sub-3 kb vectors, detecting them with significantly higher sensitivity than PacBio HiFi. The Cas9–RNP step boosts ligation yield ∼7-fold, providing an unbiased assessment of genome integrity (≥95% inverted terminal repeat [ITR]-to-ITR). In addition, the assay quantifies plasmid impurities down to 0.05% with linear response. By integrating integrity mapping, tandem detection, and impurity profiling into a rapid (<36 h), low-input workflow, AviNP-seq provides a robust analytical tool to guide vector design and de-risk early-stage process development.