Adenoviruses have a genome size of 35–40 kb (8), which possess a capsid that accommodates a 26–45 kb linear, double-stranded DNA genome (9). Ads effectively infect dividing cells and does not integrate into the host genome (8), and they bind to various cell surface proteins to facilitate their entry into the target cells (10). The type of tissue infected largely depends on the cell tropism of the virus. It is assumed that Ads infect a broad range of cell and tissue types, including neuronal cells, dendritic cells, and hepatocytes preferentially (8). This is mainly because most human cells express the primary adenovirus receptor and the secondary integrin receptors (11, 12). Different serotypes exhibit distinct cell tropisms. For instance, Ad5 shows strong liver tropism mainly through binding of its hexon protein to coagulation factor X (10). While Ad35, a group B adenovirus, uses CD46 as its primary cellular receptor, which is broadly expressed on dendritic and hematopoietic cells, facilitating preferential infection of immune cell types (13).

Since adenoviral vectors were initially used for brain cell transduction in the early 1990s (14), they have been applied in the treatment of cancer, Parkinson's disease, and Huntington's disease (1). Ads were the first DNA virus to enter rigorous therapeutic development, largely because of its well-defined biology, genetic stability, large transgene capacity and ease of large-scale production (9). However, Ad vectors can induce high-level innate inflammatory responses within the first 24 h of transduction (15). For instance, Schaack (15) reported that adenoviral early region proteins upregulate pro-inflammatory cytokines such as IL-6 and TNF-α, even in the absence of viral replication. In spinal cord injury models, Islam and Tom observed that Ad vector delivery induced notable microglial activation and neutrophil infiltration, highlighting the potential for both therapeutic and detrimental inflammatory effects (16). Besides, they also have limitations like short gene expression of 2 weeks to several months (transient) and high risk of cytopathic effects (8, 16). Additionally, high titters can result in organ damage and even mortality (16). An intravenous administration of Ads may induce acute liver injury, as has been reported in animal models (17). Still, these disadvantages can be corrected by designing and synthesizing other novel serotypes (8). For example, the third-generation adenovirus vectors, also referred to as the helper-dependent or gutless adenovirus, can significantly reduce in vivo immune response compared to the first- and second-generation Ads, although high transduction efficiency and tropism are maintained (18).

Since AAVs were initially discovered as a contaminant in adenovirus preparations (19), their biological characteristics and applications have been widely studied. AAVs have a single-stranded DNA, with genome DNA payloads under 5 kb (8), and have a small (~25 nm) icosahedral capsid composed of three types of structural proteins, namely, VP1, VP2, and VP3 (20). AAVs have the ability to transduce dividing and nondividing cells, but do not integrate into the host genome (8). They are capable of expressing transgenes over several months in nondividing cells (9), and provide a relatively stable expression in dividing cells as well (21). In comparison to Ads, AAVs have a long duration of gene expression in vivo (22). The specificity of adeno-associated virus vector-based cell and tissue targeting is determined by the capsid proteins of AAVs (9). Compared with Ads, a similar cell tropism is shown by AAVs (8). AAVs have at least 12 natural serotypes and many artificial variants, each of which exhibits different cell tropisms in the nervous system (23–26). For instance, AAV2 is one of the most extensively studied serotypes and primarily transduces neurons after direct injection into the brain parenchyma (23). AAV5 shows enhanced transduction in the striatum and cerebellum (23), while AAV8 and AAV9 are capable of crossing the blood–brain barrier (BBB) following systemic administration, thus enabling widespread CNS gene delivery (21). Notably, AAV9 preferentially targets neonatal neurons and adult astrocytes (24). AAV11 has recently been shown to enhance astrocyte-directed transduction and enable efficient retrograde labeling of projection neurons, therefore broadening its potential in circuit-specific interventions (22). These variations among AAV serotypes provide opportunities to tailor gene delivery strategies based on cell-type specificity and disease context.

AAV-mediated gene transfer has great potential as a therapeutic approach (27). Most of the currently developed AAV vectors are directed toward monogenic diseases (28), but they are also versatile and enable one to test for other CNS diseases (9). AAV vectors can lead to severe adverse effects due to off-target expression, such as hepatotoxicity (6), neurotoxicity (29), and even death in critical trials (30–32). Using AAV vectors alone does not elicit a strong immune response similar to that elicited upon using other viruses (9). AAV vectors can also induce unwanted immune responses, while about 50% of humans may have neutralizing antibodies due to previous infections (33, 34). Although certain major obstacles limit the widespread application of AAV vectors, including limited insert size, inefficient transduction, narrowed disease case, and off-target responses, they hold great potential to revolutionize the clinical management of human diseases (9). Overall, AAVs offer a relatively safe and versatile platform for CNS gene therapy, though limitations such as pre-existing neutralizing antibodies and vector immunogenicity must be addressed in clinical applications.

Lentiviruses have a single-stranded RNA genome and may incorporate constructs up to 9–10 kB in size (35, 36). Lentiviral (LV) vectors can infect both dividing and non-dividing cells, such as neurons (8, 37). While it also can preserve long-term and stable transgene induction (8), which is crucial for adolescents or pediatric patients (9). Genetic engineering makes it possible to produce LVs with specific integration sites for the increased safety of use (38). Unlike Ad or AAV vectors, LV vectors rarely generate neutralizing antibodies (36). Due to their relatively low-immunogenic characteristics (35), It is reported to have low Immunogenicity and elicit no pathogenic effects (39), suggesting that this vector is relatively safe in vivo. LVs are capable of infecting a broad range of host cell types, allowing their use for the treatment of Parkinson's disease (PD) Alzheimer's disease (AD) (1). They have also been demonstrated as efficient gene transfer vehicles for human solid tumor cells. For example, they enabled diphtheria toxin A delivery to suppress prostate cancer xenografts (39), mediated suicide gene therapy using HSV-tk in hepatocellular carcinoma (40), and outperformed retroviral vectors in transducing ovarian cancer cells (41). Although certain problems remain to be addressed, the safe and efficient LV vectors are nonetheless considered promising as a tool for human gene therapy (9). These findings highlight the potential of LV vectors for sustained and low-immunogenic gene delivery in chronic neurodegenerative conditions.

Another commonly used viral vector is retroviruses (RVs), which are single-stranded RNA and integrated into the genome of the host cell (42). They can carry up to 8–9 kb of foreign DNA for transduction (43, 44). RVs can stably integrate into the target cell genome (45), so they are useful for ex vivo delivery of somatic cells. RVs provide a stable and efficient expression of the transgene to patients (4). They have been used in early gene therapy trials for severe combined immunodeficiency (SCID), demonstrating long-term gene correction in hematopoietic stem cells (46). However, they have limitations including immunogenic problems and inability to transduce the nondividing cells (43). Herpes simplex viruses (HSVs) are also one of the recent virus candidates in gene delivery. HSVs are an enveloped virus with a double-stranded DNA (dsDNA) Genome (4). They can usually accommodate up to 40 kb of transgenic DNA (3). Because of its neurotropic features (47), they have strong tropism for neurons. For example, HSV vectors have been studied for the delivery of neurotrophic factors in models of chronic pain and epilepsy (48). HSVs may initiate strong inflammatory response (4), and result in transient gene expression in cells infected with them (49). Vesicular Stomatitis viruses (VSVs) contain a single stranded RNA genome (50). And their cargo capacity is approximately 6 kb (50). VSVs are reported to cause humoral and cellular immune response in previous studies (50, 51). These immunological features, along with their strong cytolytic activity, support their use as oncolytic agents and viral vectors in cancer virotherapy (52). These viral vectors expand the toolkit for gene delivery in the CNS, each offering unique advantages and limitations depending on the therapeutic context.

Intraparenchymal delivery is the most used viral vector delivery method with high delivery efficiency in studies for treating neurological diseases. It's an approach that enables direct delivery to the target structure, bypassing the blood-brain barrier, and usually accomplished via a stereotactic head frame or magnetic resonance imaging guidance (53, 54). While stereotactic guided injection is frequently used for intraparenchymal delivery, risk of surgical intervention and the restricted spread of transgene expression still needs to be concerned. A clinical study on Parkinson's disease patients (NCT00400634) reported severe side effects associated with surgical procedure (55). Intraparenchymal delivery typically affects only a limited range of brain regions, with highest vector distribution and downstream gene product activity around the injection site (56, 57).

A new technology called interventional MRI-guided convection-enhanced delivery (iMRI-CED) has been introduced in recent years, which is the gold standard for verifying accurate vector delivery in real time (58). Convection-enhanced delivery (CED) has been shown to improve specificity and delivery efficiency in a safe, reliable, targeted, and homogeneous manner across the blood-brain barrier (216–219). Therefore, CED holds significant potential for broad application in the treatment of neurological diseases.

Efficacy of intravenous delivery largely hinges on the ability of AAV to cross the BBB. Among the naturally existing AAVs, AAV9 is currently the most commonly used AAV serotype for CNS transduction in preclinical and clinical studies (59), which may be attributed to its active-transport mechanisms (60, 61). It has several advantages including non-invasive, widely distributed, and targeting both CNS and peripheral tissues. However, this approach also faces a few challenges. People who have been exposed to AAVs may pre-existing antibodies that can neutralize AAVs (6, 62). As it travels through the bloodstream, immune responses can diminish its efficacy (63). Besides, systemic toxicity such as off-target effects has also been found in previous studies (64, 65).

Another method is intra-CSF injection including intra-cerebroventricular injection (ICV), intrathecal injection (IT), and intracisternal magna injection (ICM). ICV refers to direct administration into ventricles in the brain, allowing the AAV to permeate the barrier and spread widely across different brain regions (66, 67). IT delivers AAV via injection into the spinal canal through the lumbar puncture, resulting in broad coverage of the brain and spinal cord (68). ICM is a technique similar to IT, while its injection site is the cisterna magna (CM). Intra-CSF injection can bypass the BBB and spread widely within the CNS. Compared with intravenous delivery, intra-CSF injection minimizes systemic exposure and avoids neutralization by circulating antibodies. However, it has insufficient coverage, leading to limited access to deeper brain regions and the presence of surgical risks (69–71).

Intranasal delivery is an alternative and non-invasive option which can directly deliver viral vectors through the nose to the brain (72). Intranasal delivery has been used in clinical studies to deliver small-molecule agents to the brain, such as nerve growth factor (NGF) and brain-derived neurotrophic factor (BDNF), thus treating neurodegenerative diseases (73–75). Intranasal delivery is easy to perform and can be repeated, even by patients themselves (72). It is also capable of bypassing the BBB (72). However, intranasal delivery holds limitations. Although it minimizes systemic exposure (76), the cells along the delivery pathways, such as olfactory sensory neurons and trigeminal nerves, may also be transduced and cause off-target effects (72).

Beyond invasive approaches, focused ultrasound with microbubble-mediated BBB opening (FUS-BBBO) is an emerging non-invasive approach for viral vector delivery in treating neurological diseases. It has the potential to overcome existing delivery limitations by providing a means of non-invasive, site-specific gene transfer to the brain, offering advantages such as improved safety and targeted delivery (77–81). Various administration routes have been explored, including intravenous, intranasal, and local injection, which have expanded its applicability for gene therapy across multiple neurodegenerative disorders (81–84). Recent studies have demonstrated the successful application of FUS-BBBO in various neurodegenerative and neurological conditions. For example, AAV has been used in models of Alzheimer's disease and Huntington's disease, while HSV has been applied in glioblastoma (83–85). Importantly, FUS-BBBO may reduce systemic immune activation by enabling lower vector doses, thereby potentially lowering the risk of pre-existing neutralizing antibodies and allowing for safer repeated administration (83). Despite these advantages, FUS-BBBO has its challenges. One primary concern is still the possibility of off-target side effects and toxicity due to the intravenous administration (59). In summary, FUS-BBBO offers a promising delivery strategy that enhances brain-targeted gene therapy while potentially mitigating immune-related complications.

Innate immunity is the first barrier against pathogens as it mounts rapidly and does not require specific adaptation to the pathogens (62). Its response depends on the recognition of pathogen-associated molecular patterns (PAMPs), which are carried by viruses or viral vectors (89). Both the capsid and the DNA components of AAV may contribute to the activation of innate immunity, resulting in adverse reactions and immune toxicity.

In the context of AAV-mediated gene transfer, one mechanism is considered to trigger inflammatory responses is a PAMP within the capsid proteins binding to Toll-like receptor 2 (TLR2) (90), positively expressed on myeloid dendritic cells. Another mechanism involves the AAV double-stranded DNA (dsDNA) genome, in particular its unmethylated CpG sequences, recognized by Toll-like receptor 9 (TLR9) (91–94). TLR9 engagement increased antigen presentation on class I major histocompatibility complex (MHC-I) and further associated with enhanced activation of AAV-specific CD8+ T cell (94–96). Further CpG Depletion of Vector Genome in animal models has shown that a reduction of the vectors' CpG motifs blunts CD8+ T cell responses without modifying B cell responses (97, 98). In addition to the vector DNA genome, double stranded RNA (dsRNA) may also participate in the induction of innate immunity from long-term AAV transduction, which could interact with melanoma differentiation-associated protein 5 (MDA5) (99). It is well known that MDA5 is cytoplasmic viral RNA sensor serving as a cytoplasmic pattern recognition receptor (PRR) and ubiquitously expressed. MDA5 is capable of activating type I interferons (type I IFNs) signaling pathways after virus infection, leading to the expression of type I IFNs (99). Besides, it is also found that AAV vectors activate type I IFNs expression in plasmacytoid dendritic cells (pDCs) of human and murine (93). TLR9 can interact with myeloid differentiation primary response gene 88 (MyD88) for downstream signaling, leading to type I IFNs production in pDCs and further promoting activation of CD8+T cells (93, 96). pDCs recognize AAV genomes via TLR9 and release type I interferons, which in turn activate conventional dendritic cells to prime capsid-specific CD8⁺ T cells—highlighting the interplay between innate sensing and adaptive immune activation (62). These studies underscore the important role of type I IFNs in bridging innate and adaptive immunity following AAV administration.

An adaptive immune response requires a longer time than innate immunity and is considered the second line of defense against pathogens. It is characterized by antigen-specific and the ability to eliminate pathogens while generating an immunological memory (62). AAV-specific adaptive responses can be induced by previous natural infection or they can be stimulated or recalled by AAV gene transfer.

After AAV vector administration, T and B lymphocytes are activated following molecular recognition of an antigen presented by antigen-presenting cells (APCs) (62). Cytotoxic CD8+T cell responding via MHC class I to AAV capsid protein can affect therapeutic efficacy (94, 95, 100), possibly by clearing AAV-transduced cells thus inducing inflammation in the target tissue (101, 102). It is also reported that increased AAV-antigen presented on capsid-derived MHC class I together with higher CD8+T cells activation in dose-dependent in vitro experiments (103, 104). The concurrent presentation of CD4+ T helper cells activated by capsid-derived MHC class II facilitate both humoral and cell-mediated immune responses (105). AAV-specific B cell responses induced by previous natural infections can neutralize the viral vectors before they can deliver their payload, thus precluding a successful AAV-mediated gene transfer (6). Efforts have made to avoid its neutralization by modification to the vector's capsid (106). However, it showed limited success as many of the antibodies bind to domains that are crucial for transduction (106, 107). We can see that T cells play an important part in adaptive Immune responses against viral vectors. Liver toxicity caused by T cells directed to epitopes, which bind with low affinity to MHC class I molecules, is more worthy of our attention. Therefore, cellular responses, in particular the T cells, are needed to further study and make a breakthrough.

Anti-transgene immune responses were documented in the clinical trials, mostly after intramuscular delivery of AAV vectors (108, 109). For example, evidence of T cell-mediated anti-transgene cytotoxic T cell responses was documented in a phase I/II trial of intramuscular gene transfer in Duchenne muscular dystrophy patients (108). The risk of antibody responses to the transgene product is influenced by many factors, including the route of vector administration, specific vector design, types of viral vectors, and vector dose. Additional host factors may include specific aspects such as tissue inflammation, target tissue and the host genetic background (105). Clinical experience indicates that immune responses directed against AAV vectors are dose dependent (110, 111). Compared to Ad and LV vectors, AAV vectors have relatively weak and transient innate response and least efficient inducer of CD8+ T cells (105). The clinical data presented suggest that disease-specific conditions, such as the ongoing inflammation in muscle dystrophies (108), are likely to increase transgene immunogenicity after gene transfer.

As a progressive neurodegenerative disorder, the earliest phase of Alzheimer's disease happens in parallel with accumulating amyloid β, inducing the spread of tau pathology. More than 40 Alzheimer's disease-associated genetic risk loci already have been identified, of which the APOE alleles have the strongest association with the disease (134). Preclinical and clinical studies have explored various strategies through viral vector delivery for the treatment of AD, including risky genes and related pathways.

Parkinson's disease (PD) is one of the common neurodegenerative diseases in middle-aged and elderly people, characterized by loss of dopaminergic neurons in the substantia nigra (SN). The focal nature of SN pathology in PD has been long considered amenable to gene therapy. 3–5% of Parkinson's disease is explained by genetic causes linked to known Parkinson's disease genes, whereas 90 genetic risk variants collectively explain 16–36% of the heritable risk of polygenic Parkinson's disease (47), leading to approaches that can target genes directly linked to the disease.

Huntington's disease is a progressive, fatal, neurodegenerative disorder caused by an expanded CAG repeat in the huntingtin gene(HTT), which encodes an abnormally long polyglutamine (polyQ) repeat in the huntingtin protein (180). The mutant HTT protein causes selective neuronal degeneration, especially in the striatum and cortex (180). Therefore, directly targeting the HTT gene using viral vectors offers a rational and potentially disease-modifying therapeutic approach.

GRN mutations cause frontotemporal dementia (GRN-FTD) due to deficiency in progranulin (PGRN), a lysosomal and secreted protein with unclear function, which plays roles in lysosomal homeostasis, neuronal survival, and inflammation regulation (192). Deficiency of the gene GRN results in gangliosidosis in frontotemporal dementia, causes neurodegeneration through lysosomal dysfunction, defects in autophagy, and neuroinflammation (193, 194).

Preclinical studies performed AAV-driven expression of progranulin to animal model of frontotemporal dementia due to GRN mutations (195, 196). Results have shown that it not only increased progranulin levels in the cerebrospinal fluid, but also normalized histological and biochemical markers of progranulin deficiency, including reduced lipofuscin accumulation and improved lysosomal enzyme activity (195, 196). These studies support the translation of GRN gene therapy for FTD from preclinical studies into the clinic. Interim results of a phase 1/2 trial showed preliminary insights into the safety and bioactivity of PR006 (AAV90-GRN), demonstrating increased CSF progranulin and reduced neurofilament light chain (NfL), a biomarker of neurodegeneration, while its long-term safety and potential efficacy remain to be confirmed in ongoing studies (NCT04408625, NCT04747431) (195).

In other neurodegenerative diseases, GDNF can also be used in patients with Multiple System Atrophy (MSA), a rare and rapidly progressive disease involving multiple brain systems, as it has a trophic effect on the Purkinje cells (197). A Phase 1 clinical trial delivering AAV2-GDNF to the putamen in patients with either a possible or probable diagnosis of MSA has been ongoing since 2023 to evaluate its safety and potential clinical effect (NCT04680065). Amyotrophic lateral sclerosis (ALS), a progressive neurodegenerative disease affecting motor neurons, has also been a target of experimental GDNF-based therapies. Delivering a retroviral vector encoding GDNF into hind limb muscles of rodent models of ALS increased motor neuron survival and delayed the “ALS-related” decline in performance on motor tests, demonstrating functional benefits in preclinical models (198). Injection of AAV-GDNF showed evidence of retrograde transport of GDNF to motor neurons and impeded neurodegeneration in transgenic models of ALS (199). In addition to GDNF, preclinical studies using AAV-IGF1 in rodent models of ALS have also shown a significant reduction in motor neuron loss and improvement in behavior, particularly in motor function and survival extension in some models (200–202). However, further clinical translational studies are needed to assess the safety and efficacy of these approaches in humans.

In addition to adult-onset neurodegenerative disorders, recent clinical advancements in CNS-targeted gene therapy have been largely driven by rare monogenic pediatric diseases that share progressive neurodegenerative characteristics. Among them, spinal muscular atrophy (SMA) has become a flagship example of CNS-targeted gene therapy, with the FDA-approved AAV9-based Zolgensma demonstrating significant clinical efficacy. In the pivotal trial NCT02122952, systemic administration of Zolgensma, an AAV9–mediated gene therapy targeting the SMN1 gene, led to marked improvements in motor function and survival in SMA1 patients (203), findings further supported by a comparative study with a natural history cohort (204). The success of Zolgensma was further confirmed in the phase III SPR1NT trials, where presymptomatic infants with two or three copies of SMN2 achieved age-appropriate motor milestones, such as sitting and walking, underscoring the benefits of early intervention (205–207). However, the application of AAV9-based gene therapy also presents immunological challenges. In clinical trials including STR1VE and SPR1NT, immune-mediated adverse events such as transient hepatotoxicity and thrombocytopenia were commonly reported. These responses are likely due to host reactions against the AAV9 capsid or transgene product. To manage this, all patients received prophylactic corticosteroids (typically prednisolone), with liver function and hematologic parameters closely monitored and treatment adjusted accordingly (205, 207). Beyond SMA, Gene therapies for mucopolysaccharidoses (MPS, e.g., NCT03580083) and Rett syndrome (e.g., NCT06856759) are actively being explored. While most treatments remain in early or mid-stage trials with limited clinical efficacy so far, these efforts provide important proofs of concept for CNS-targeted gene therapy. Their progress offers valuable lessons on delivery strategies and immune response management, paving the way for future applications in adult neurodegenerative diseases.