Gene therapies provide several major advantages compared to all other therapeutics which make them uniquely valuable for the treatment of neurodegenerative diseases (NDDs). First, gene therapies must be delivered intracellularly in order to function (Ingusci et al., 2019). Though this poses a challenge, once successfully delivered, gene therapies are able to directly modify protein production at the source (Ingusci et al., 2019). This is especially advantageous for the treatment of NDDs as the vast majority of pathology in these conditions occur intracellularly (Kovacs, 2019). Second, the vast majority of NDDs are caused by the dysfunction of a single gene, or small number of pathogenic proteins (Kovacs, 2016, 2019), either of which can be directly targeted by gene therapies. Finally, DNA-based gene therapies are adopted by host cells and can reside in the nucleus where they act as a blueprint, allowing cells to produce a therapeutic protein(s) continuously without re-administration (Nathwani et al., 2014). For this reason, DNA-based gene therapies should only require a single dose to give long lasting, or even life-long disease modification. This is particularly valuable for neurologic diseases as the blood–brain barrier (BBB) complicates delivery of large therapeutics to the brain, often requiring more invasive routes of delivery.

Unfortunately, the delivery of gene therapies to the brain faces significant challenges. Delivering genes to the cells in the brain is difficult, as DNA is a large, charged, molecule, and will not penetrate the BBB and cellular membrane (Maeder and Gersbach, 2016). In order to facilitate delivery of gene therapies to the intracellular compartment of target cells, a vector must be employed. Vectors may be non-viral or viral. Non-viral vectors include lipid nanoparticles (Bors and Erdö, 2019), polymer nanoparticles (Liu et al., 2017), cell penetrating peptides (Liu et al., 2017), and cationic microbubbles [used with focused ultrasound (FUS)] (Fan et al., 2016). The major benefit of non-viral vectors is that they lack capsids, which are made up of foreign proteins and therefore are immunogenic, albeit to varying degrees (Bessis et al., 2004). Though non-viral vectors do not possess immunogenic capsids, the risk of immune response to the delivered transgene remains a major problem (Mulia et al., 2020; Ediriweera et al., 2021). Additionally, transgenes, themselves may be toxic to cells through insertion into the host’s chromosomal DNA (insertional mutagenesis) (Cavazzana-Calvo and Fischer, 2007) or undesired alteration of off target protein expression (Balakrishnan et al., 2013). Additionally, even intended manipulation of protein expression could theoretically result in deleterious effects. For example, knocking down the pathogenic protein in a NDD may improve disease phenotype (Alarcón-Arís et al., 2020), however, in many cases it remains unclear if that protein also serves a vital cellular function in humans (Brundin et al., 2017; Wild and Tabrizi, 2018). If that were the case, unintended side effects of gene therapy would be likely. The greater the number of cells transduced, the greater the likelihood of immune or non-immune mediated side effects thus minimization of off target delivery is critical to the safety of both non-viral and viral vector mediated gene therapy. In order to minimize off target delivery, vectors should be targeted to cells involved in disease pathology. Altering both the tissue and cellular tropism of gene therapy vectors will likely be necessary to make the safest gene therapy product possible. Owing to their simplicity of design, precise alteration of tropism is not yet possible with non-viral vectors (Liu et al., 2017).

Viral capsids have evolved for over a billion years to be able to effectively allow the deposition of their genetic contents intracellularly, and even to the nucleus of cells. By comparison, non-viral solutions are more simplistic, and thus are less effective at delivering their cargo intracellularly (Liu et al., 2017). Furthermore, viral vector capsids are made of proteins, which, made from DNA blueprints, are modifiable to suit varied purposes (Hudry and Vandenberghe, 2019). Over the past 30 years, many viral vectors have been developed for gene therapy to the central nervous system (CNS), including lentivirus, adenovirus and adeno-associated viruses (Agustín-Pavón and Isalan, 2014; Bedbrook et al., 2018; Hudry and Vandenberghe, 2019; Ingusci et al., 2019). Due to the low immunogenicity, high efficiency, high stability and ease of modification, AAV vectors have become the mainstay delivery tool for gene delivery to the CNS (Choudhury et al., 2016; Lykken et al., 2018). A large variety of serotypes, each with different cellular and tissue tropism have been discovered thus far (Zincarelli et al., 2008; Snyder et al., 2011; Hocquemiller et al., 2016; Hudry and Vandenberghe, 2019; Keeler and Flotte, 2019). Owing to their simple genetic composition (of only three viral proteins named VP1, VP2, and VP3), modifications to the capsid of AAVs is easily implemented (Hocquemiller et al., 2016; Deverman et al., 2018). The only major downside of AAVs, is that their small size limits the size of cargo which can be delivered by a single vector (Bedbrook et al., 2018). Though this poses a problem, there is extensive work being done to develop compact transgenes capable of fitting in an AAV without altering function. For example, one group designed a CRISPR-dCas9 delivery system capable of fitting in a single AAV vector by using a smaller Cas9 variant from Staphylococcus aureus rather than the typical Streptococcus pyogenes (Lau et al., 2019). Ultimately, AAVs provide a valuable tool for gene therapy. It is not surprising that the first approved single dose gene therapy, Onasemnogene abeparvovec-xioi (AVXS-101; or Zolgensma^®) for spinal muscular atrophy (SMA), uses an AAV (specifically AAV9) (Mendell et al., 2017).

Armed with vectors such as AAVs, and with platforms capable of making highly specific alterations to gene expression (Larson et al., 2013; Agustín-Pavón and Isalan, 2014; McMahon and Cleveland, 2016; Savić and Schwank, 2016; Lau and Suh, 2017; Lau et al., 2019; Lim et al., 2020), we are closer to a reality where gene therapy platforms capable of correcting the underlying cause of many NDDs. These diseases include disorders caused by dysfunction of a single gene [e.g., Huntington’s (HD) (Wild and Tabrizi, 2018), Spinocerebellar ataxias (SCAs) (Dong and Cong, 2019; Martier et al., 2019), and lysosomal storage disorders (LSDs) (Ohashi, 2019)]. This approach may also be applicable to treat more common NDDs which are considered multifactorial or polygenic, but result from the toxic accumulation of a single, or small number of, proteins [e.g., Alzheimer’s disease (AD) (Elmer et al., 2019), Parkinson’s disease (PD) (Brundin et al., 2017), and Amyotrophic Lateral Sclerosis (ALS) (Scotter et al., 2015)]. Even if these platforms become available, specific and efficient delivery strategies will likely be necessary for clinical translation.

For a gene therapy to be effective, it must transduce enough target cells to mitigate the disease pathology. Theoretically, one way to achieve this would be to increase the dose of virus until a sufficient level is reached. However, the ability to do this is prevented by dose dependent toxicities which are largely mediated by the immune system (Rabinowitz et al., 2019; Perez et al., 2020; Verdera et al., 2020). The broad translation of gene therapy is contingent upon maximizing gene delivery to target cells, while minimizing the risk of adverse events associated with off-target delivery. A historical example of when this went awry is the death of Jesse Gelsinger in 1999. In this case, Jesse was treated for ornithine transcarbamylase deficiency with an IV adenoviral vector for gene replacement. Tragically, he died 4 days later of a massive immune response and multisystem organ failure (Sibbald, 2001). Though gene therapy has come a long way since this time, similar responses have been seen in non-human primates (NHPs) and AAVs, delivered IV, result in hepatotoxicity (Bessis et al., 2004; Mingozzi and High, 2017; Fitzpatrick et al., 2018), while direct CNS delivery can result in neurotoxicity (Samaranch et al., 2014; Meyer et al., 2015; Hordeaux et al., 2018a; Perez et al., 2020).

AVXS-101 is the first approved single dose gene therapy for a neurologic disease (SMA type 1) (Mendell et al., 2017). Though the benefits of this treatment are clear, it is worth noting that significant elevations in the liver enzymes (AST and ALT) have been seen in a large fraction of treated patients (Mendell et al., 2017), signifying subclinical hepatotoxicity. Though the children in the AVXS-101 trial did not develop serious injury from this treatment, the incidence of transaminitis and the necessity co-administration of steroids, suggests that this trial used a near maximum tolerable dose in humans. Further, the sample size in this study was small, Thus, with more subjects, more serious reactions may have been observed. Finally, infants with SMA type 1 are ideal candidates for gene therapies as they have a less mature immune system (Simon et al., 2015), healthy organs, and are unlikely to have developed neutralizing antibodies to AAVs (Fitzpatrick et al., 2018). When it comes to the broad translation of gene therapy for more common neurologic diseases which effect adults (e.g., AD or PD), products with a higher margin of safety are necessary.

There are many gene therapies for NDDs which have shown promise in animal models and are ready for translation to humans (Yang et al., 2017; Arrant et al., 2018; Martier et al., 2019; Sun and Roy, 2021). However, many of these therapies use naturally occurring AAV vectors delivered via inefficient delivery routes and are expressed with non-specific promotors. The utilization of more advanced delivery technologies, which optimize on-target delivery, will likely be necessary to develop safe, effective and broadly translatable therapies.

The process of delivery of AAVs to the brain can be divided into four major stages: (1) the delivery route, (2) BBB crossing, (3) neuronal entry, and (4) transgene expression (Figure 1). Thorough optimization of AAV delivery to the brain should consider each of these stages individually. Methods such intra-arterial delivery (Liguore et al., 2019), physical methods of transient BBB disruption (BBBD) (Foley et al., 2014; Fishman and Frenkel, 2017), AAV capsid engineering (Deverman et al., 2016) and cell-specific promotors (Jackson et al., 2016) have all been shown significantly improve the efficiency of gene therapy delivery, yet none of these methods are being investigated in human studies¹. In our view, the ideal gene therapy for most NDDs of the brain has the following characteristics: (1) global delivery (2) single dose (3) minimally invasive (does not requiring opening skull) and (4) cell specific delivery and gene expression. This manuscript seeks to review a series of recent technological advancements shown to be effective in pre-clinical models, which could facilitate the development of this ideal therapy. It is our hope to provide a guide to methods that could be implemented, in humans, which would allow a translation of gene therapies for the treatment of many NDDs.

Receiving a single dose gene therapy is a life altering event for a patient that poses both significant risk and significant potential benefit. Delivery route has a profound effect on vector biodistribution after injection. Altering this variable can enhance efficiency (off-target divided by on-target delivery) by a factor of tens, hundreds, or even thousands fold (Hinderer et al., 2018a; Marshall et al., 2018; Liguore et al., 2019; Bailey et al., 2020). A major limiting factor preventing the translation of gene therapies to the clinic is dose dependent toxicity which is directly related to high off-target delivery. For this reason, it is critical to use all tools at our disposal to optimize the delivery route for single dose gene therapies to the brain.

The determination of an optimal delivery route is dependent on the tissues involved in disease pathology. Many diseases affect the entire body, as well as the brain [e.g., LSDs (Filocamo and Morrone, 2011)]. As delivery outside the brain may improve disease pathology in these systemic diseases, a greater tolerance for routes with high systemic delivery, such as intravenous (IV), could be considered. Conversely, most NDDs, including as PD (Kovacs, 2019), AD (Kovacs, 2019), HD (Dong and Cong, 2019), and frontotemporal dementia (FTLD) (Svetoni et al., 2016) involve the entire brain, but are not known to pose a significant threat to the rest of the body. In these cases, an optimal route minimizes systemic delivery, while allowing for global delivery throughout the brain. These goals are often at odds with one another, as routes which are capable of delivering vectors evenly throughout the brain are also more likely to deliver vectors to the rest of the body. Finally, for all routes, the invasiveness of delivery must be considered.

Intra-thecal (IT) delivery includes any delivery method where AAVs are injected into the subarachnoid space. This delivery can occur anywhere along the neuro-axis including intra-lumbar (IT-IL) via classic lumbar puncture between L4 and L5, intra-cisternal (IT-CM), via direct infusion into the basal cistern via sub-occipital puncture, and intracerebroventricular (IT-ICV), via direct injection through the skull into the lateral ventricle (Piguet et al., 2021). IT delivery is a promising method for the delivery because it naturally bypasses the BBB. In NHPs, IT delivery has been shown to be the most efficient route capable of global delivery to the brain (Yang et al., 2014; Hinderer et al., 2018a; Hordeaux et al., 2018a; Liguore et al., 2019). IT-CM delivery has also been shown to allow highly efficient delivery in mice, even demonstrating disease modification in a mouse model of GM1-Gangliosidosis (Hinderer et al., 2020; Chatterjee et al., 2021). As expected with IT therapy, there is considerably less peripheral uptake to all organs, including the liver (Yang et al., 2014; Hinderer et al., 2018a; Hordeaux et al., 2018a; Liguore et al., 2019). Thus, this route is expected to have considerably lower risk of hepatotoxicity compared to IV delivery. IT-CM delivery is under investigation in humans for several NDDs of the brain including FTLD (NCT04408625) and PD with at least 1 GBA mutation (NCT04127578).

Although IT delivery offers a promising solution to improve delivery throughout the brain, there are challenges with this approach. The first challenge is the concern for CNS toxicity with high doses. It has been demonstrated in NHPs and piglets, respectively that delivery of high doses of AAVs, IT, risks of dorsal root ganglion (DRG) toxicity (Hinderer et al., 2018b; Hordeaux et al., 2020). This CNS toxicity does not appear to be specific to the DRG. One study found cerebellar toxicity was seen after high dose IT-CM administration of AAV9 (Samaranch et al., 2014). The same study found histologic evidence of significant inflammation of the putamen after dIC injection of AAV2 with convection enhanced delivery (Samaranch et al., 2014). These studies together suggest that while injecting AAVs past the BBB may result in significantly less systemic toxicity, at high doses, this may be replaced by CNS toxicity. This suggests that additional measures beyond improving delivery past the BBB may be necessary modulate the inflammatory response to AAV gene therapy.

In addition to CNS toxicity, another major concern with IT delivery is an uneven distribution throughout the CNS. IT-L injection is the least invasive means of IT delivery generally provides good expression in the spinal cord (Bey et al., 2017; Hinderer et al., 2018a; Marshall et al., 2018) but results in little expression in the brain (Bey et al., 2017; Marshall et al., 2018). For this reason, the use of IT-L delivery has primarily been limited to the treatment of disorders involving the peripheral nerves and spinal cord (Rosenberg et al., 2018). One study used real-time MRI monitoring with co-injection with Gadoteridol contrast to estimate AAV delivery after IT-IL, IT-CM, and IT-ICV delivery (Ohno et al., 2019). They found that the IL route resulted in a broad spinal cord but little brain delivery, IT-CM delivery allowed broad distribution throughout the brain and spinal cord, and IT-ICV delivery allowed delivery to the brain only. Consistent with this, other studies also demonstrate broad CNS distribution with the IT-CM route (Bucher et al., 2014; Hinderer et al., 2014, 2018a; Bey et al., 2017; Rosenberg et al., 2018; Ohno et al., 2019).

Although IT-CM does seem capable of broad brain delivery at high enough doses, it may not do so optimally. IT-CM delivery is dependent on the natural flow of CSF through the subarachnoid space (Johanson et al., 2005). The choroid plexus of the lateral ventricles makes CSF (Ohno et al., 2019) which then flows caudally, down the third ventricle, passing through the cisterna magna, then the further to the spinal cord (Johanson et al., 2005). Though a portion of the CSF ascends rostrally, CSF’s net flow remains rostral to caudal, gently pushing intrathecally delivered drugs caudally to the spinal cord (Johanson et al., 2005). The part that ascends to the brain is, in part, drained to the facial lymphatics and into the systemic circulation (Johanson et al., 2005). This flow pattern favors the delivery of vectors to the lower brainstem, cerebellum, upper spinal cord, and caudal/lateral cortex. With IT-CM the brain’s rostral cortex and deep structures (such as the hippocampus, midbrain, and striatum) are exposed to a much lower concentration of drug (Taghian et al., 2020). Rosenberg et al. (2018) reported considerable variability in regional uptake after IT-CM injection, with the highest uptake in the posterior brain (near the injection site) and less in the cerebral cortex.

Another limitation of IT-CM delivery is that it involves inserting a needle within millimeters of the medulla (Samaranch et al., 2016; Taghian et al., 2020), posing a risk of brain stem injury resulting in permanent disability or death. Few practitioners are trained in IT-CM delivery, and optimized tools to predictably ensure the safety of patients have yet to be developed.

Interestingly, one study pioneered a new method for IT drug delivery (Taghian et al., 2020) which allows the distribution of IT-CM delivery with invasiveness of IT-L. These authors used an IA catheter, introduced via a typical lumbar puncture, and advanced this catheter to the cisterna-magna (Taghian et al., 2020). This method reduces the risk of directly accessing the cisterna magna and may allow enhanced delivery to the lower spinal cord by performing a second infusion into the lumbar region prior to removing the catheter (Taghian et al., 2020). This method was well tolerated, first in sheep, and then in two human subjects with Tay-Sachs disease (Taghian et al., 2020).

Overall, IT (and especially IT-CM) delivery of AAVs for gene therapy provides a promising alternative to dIC or IV routes. Unlike these traditional routes, IT-CM delivery provides high efficiency, non-localized brain delivery, which is not impeded by the BBB. As discussed, this route is limited by CNS toxicity at high doses, and an uneven CNS delivery of vectors throughout the brain, as well as the potential risks of injecting directly into the cisternal space. These challenges could be overcome with technological advancements. Alternatively, other routes of delivery may eventually be capable of high efficiency delivery past the BBB, and ideally with the benefits as IT-CM but without these limitations.

The properties of an AAV, termed the capsid profile (Wec et al., 2021), includes tissue tropism, cellular tropism, efficiency, immunogenicity, and ability to cross the BBB in the case of certain serotypes (Büning and Srivastava, 2019). The capsid profile is determined by the amino acid sequence of its cap gene which encodes for the viral capsid (Hocquemiller et al., 2016). Manipulating the coding sequence of this gene allows us to change the biologic properties of that capsid. As BBB crossing is a significant barrier of delivery to the brain, most engineered AAVs discovered with tropism for the brain have achieved this by improving the efficiency by which the AAV penetrates the BBB (Deverman et al., 2016; Jackson et al., 2016; Hordeaux et al., 2018b, 2019). Enhancing BBB crossing alters AAV tissue tropism allowing AAVs improved delivery throughout the brain. Similarly, non-target tissues, such as the liver, can be de-targeted (Pulicherla et al., 2011; Choudhury et al., 2016).

Capsid engineering can also be utilized to alter the cellular tropism AAV variants. Engineered AAV variants have been developed with enhanced neuronal tropism or enhanced tropism to brain endothelial cells (Kumar et al., 2020). The potential for capsid engineering extends beyond cellular tropism as well and other strategies for mitigating the immune response including removal of neutralizing antibody binding sites and removal of highly immunogenic antigens has been proposed (Wec et al., 2021). Although we have an idea of what AAV properties are desirable, determining capsid alterations at a sequence level, which would provide novel properties, in humans, poses a challenge. Ideally, one could employ rationale design and make a change that is known to have a targeted effect. Unfortunately this method is challenging using current knowledge and technologies (Bell et al., 2012). In order to overcome this problem, researchers have developed methods based on the same strategy nature has used, namely “evolution” (Pulicherla et al., 2011; Deverman et al., 2016; Chan et al., 2017; Büning and Srivastava, 2019; Kumar et al., 2020; Nonnenmacher et al., 2021).

The final step in AAV delivery is the actual expression of transgenes. Modifying the location or context in which the transgene expressed allows a powerful tool which can enhanced the selective of a gene therapy. One method for altering transgene expression is the choice of promoter. Promoters dictate which cells, under what conditions, and with what strength, a transgene will be expressed (Jackson et al., 2016). Currently, most gene therapies use a ubiquitous promoter that is constitutively active in all cells. Commonly used ubiquitous promoters are cytomegalovirus (CMV), chicken β-actin (CBA), human elongation factor 1 alpha (EF1α), and variants of these (Bedbrook et al., 2018). Zolgensma^®, for example, uses a CBA promoter (Hoy, 2019). The benefit of a ubiquitous promoter is that it allows for the highest level of overall transcription (Ingusci et al., 2019). This is especially useful when the vector itself is inefficient as a strong promoter allows for higher gene expression even if relatively few cells are transduced (Jackson et al., 2016).

There are several problems with ubiquitous promoters. The first is that long-term unregulated overexpression of a transgene can result in unintended alterations in cellular signaling (Ingusci et al., 2019). Second, transgenes can have undesirable effects when expressed in a tissue outside of that which it is intended (Ingusci et al., 2019). Finally, the adaptive immune response to gene therapy products can be induced by the transgene product itself in a dose-dependent manner (Samaranch et al., 2014; Perez et al., 2020). Ubiquitous expression in off-target cell types would increase the risk of immune-mediated toxicity.

There are many cell-specific promoters with relevance to gene therapy for NDDs. Promoters can limit expression to a specific brain cell type (Jackson et al., 2016; Bedbrook et al., 2018), brain structure (Chan et al., 2017) or cells producing a specific neurotransmitter (Bedbrook et al., 2018; Ingusci et al., 2019). As the primary site of pathophysiology in NDDs are neurons, neuronal promotors would be optimal for gene therapy in these conditions. The human synapsin (hSyn1) promoter has been demonstrated as an effective neuronal promoter when used with AAVs (Jackson et al., 2016; Bedbrook et al., 2018). One study compared CNS expression of 4 different promotors–hSyn1, human CMV, mouse PGK (mPGK) and short variant of CMV early enhancer/chicken beta actin (sCAG)–each driving expression of GFP delivered via intraparenchymal injection of AAV1 in mice. They found that significantly higher expression with hSyn1 and mPGK compared to sCAG and CMV shortly after the injection (Jackson et al., 2016). Conversely, Jackson et al. (2016) found that expression with synapsin waned after 4–5 months while expression with CMV/Chicken beta actin (CBA) did not. Further investigation into persistence of hSyn expression, ideally in humans, will be important to ascertain its viability.

A novel use of promoters, which, to our knowledge has not yet been described, can be called “environment-specific” promoters. These can be defined as promoters that drive transgene expression only under specific cellular conditions. One strategy would be to use a promoter for proteins specifically activated in the context of proteostatic stress (e.g., in cells nearing neurodegeneration). In this strategy, the transgene would lay dominant in target cells until these cells are under proteostatic stress. As such transgene toxicity could be avoided unless a cell is already going to undergo neurodegeneration. One example of how this could be accomplished uses a response called the unfolded protein response (UPR) (Valenzuela et al., 2018). In this response, proteostatic stress is detected by stress transducers in the ER lumen. These transducer molecules activate receptors, which activate transcription factors such as ATF6f, XBP1s, and ATF4 (Valenzuela et al., 2018). The transcription factors then activate genes whose products can reduce proteostatic stress by a variety of mechanisms (Valenzuela et al., 2018). Because this signaling cascade is capable of identifying when a cell is under proteostatic stress, and is able to use that information to active the expression of specific genes, the promoter which drives these genes could be ideal for gene therapy for NDDs. This would be especially true if the transgene is designed to knock down a protein that, under physiologic conditions, serves a vital cellular function. A similar concept of tumor-specific promoters, has been proposed in the oncology literature (Lu et al., 2005).

Another interesting strategy is synthetic promoters. Synthetic promoters are artificially designed promoters comprised of multiple transcription factor regulatory elements (TFREs) (Brown et al., 2017). These elements may be designed to bind multiple endogenous transcription factors, increasing the sensitivity of the system (Brown et al., 2017). Artificially created promoters can be engineered for enhanced long-term gene expression by removing CpG motifs targeted for methylation-mediated silencing (Brown et al., 2017). Further, targeted mutagenesis and directed evolution can also be used for to improve the efficiency of artificial promoters. One group generated a library of DNA sequences that fused to the 5′ of an artificial promoter and identified a modification that led to a 5.8 fold increase in promoter strength (Jin et al., 2019). Similar to capsid engineering, directed evolution with in vitro and in silico selection methods could allow for the discovery of promoters with optimized properties far beyond those which are naturally occurring.

Finally, manipulations to enhance the specificity of transgene expression is not limited to promoters. Several groups have demonstrated the effectiveness of incorporating cell-specific miRNA binding sites into the transgene. Sequences which bind miRNA produced in off target cells can be chosen. This method allows for specific knock down in cells with undesirable consequences to transgene expression, for example APCs (Piguet et al., 2021).

Re-administration of viral vector gene therapies poses a major risk. After the first treatment with a viral vector, patients will likely develop memory B and T cells to the virus (Rabinowitz et al., 2019; Verdera et al., 2020). If this occurs, a greater adaptive immune in response to any subsequent doses would be expected, greatly increasing the risk of immune-mediated cytotoxicity and/or neutralization by antibodies generated after the initial treatment. While the risk of immune reaction to re-administration is lower with gene therapies which do not require a viral vector, such as antisense-oligonucleotides, the requirement of repeated IT administration (Drahansky et al., 2016) has its own set of risks and challenges.

The ideal gene therapy would be one that requires only a single administration to achieve life-long therapeutic benefit. This should be possible with DNA-based gene therapies, which reside in the nucleus and can be continuously expressed for the life-time of that cell. Though this is theoretically possible, transgene persistence has been a major issue faced by DNA-based gene therapies. Two tong terms follow up studies of patients treated with AAV gene therapy have recently in the hemophilia literature. In one, patients were followed up 3 years after treatment of Hemophilia A with AAV5 delivered human factor VIII gene (Symington et al., 2021). They found that factor VIII activity waned after the first year in most patients. Interestingly, though gene expression had decreased considerably, the clinical benefit of the therapy was sustained through the 3 years of investigation. In the second study, in patients with hemophilia B followed 8 patients after treatment with a “codon-optimized” human factor IX gene delivered by AAV8 (Konkle et al., 2021). Only 1 of 8 patients had sustained factor IX activity at 4 years post-treatment. The authors explain that their codon-optimization program inadvertently added a large quantity of CpG sequences. They hypothesized that their transgene loss was due stimulation of an innate immune response related to these added sequences (Konkle et al., 2021).

In the case of the CNS, Onasemnogene Abeparvovec (Zolgensma^®) is the only DNA-based gene therapy delivered to humans long enough ago to evaluate transgene persistence. Of the patients treated with Zolgensma^®, all had, at minimum, maintained the same motor function (many had developed new milestones) and remained off mechanical ventilation up to 6.2 years post-treatment (Mendell et al., 2021). This is especially noteworthy in SMA type 1 where the natural history of the disease results in 92% of patients requiring mechanical ventilation by 20 months of age (Mendell et al., 2017). Although Mendell et al. do not provide direct evidence supporting persistence of their transgene (i.e., measurement of SMN activity) their clinical results suggest persistence of transgene activity in their patients.

Although the results from Mendell et al. are encouraging, waning transgene expression remains a major concern which should be optimized in any way possible. Consideration of optimizing transgene persistence should start with identification of potential mechanisms by which transgene loss can occur. Proposed mechanisms include (1) loss of transgene expression due to immune response (Mulia et al., 2020; Konkle et al., 2021; Symington et al., 2021), (2) episomal loss due to cellular turnover without transgene replication (Lufino et al., 2008; Mulia et al., 2020), (3) epigenetic modifications resulting in gene silencing (Mulia et al., 2020). Loss of transgene expression due to the immune system has been proposed to occur either by adaptive or innate immune responses. CpG sequence activation of TLR-9 has been implicated in the innate response while, transgene loss due to an adaptive response is thought to be due to T-cell mediated cytotoxicity (Konkle et al., 2021). In either case, “transgene loss” would be the result of destruction of the cells harboring the transgene (rather than isolated loss of a transgene in an otherwise intact cell). In the liver, cell loss due to immune destruction may be clinically silent as hepatocytes are capable of replicating to replace lost cells. However, as neurons are largely incapable of replication, immune-mediated transgene loss therefore causes permanent neuronal loss synonymous with immune-mediated neurotoxicity (Wang et al., 2019). This process has been demonstrated to irreversible and mediated by cytotoxic T-Cells, IFN-γ, TNF-α, and perforin (Wang et al., 2019). Strategies to counter immune-mediated toxicity (and therefore transgene loss) are a major focus of earlier sections in this manuscript and will not repeated here.

With AAV vector delivery, the majority of delivered transgenes will exist as an episome, rather than integrating into the host genome. This can be beneficial, as it reduces the risk of insertional mutagenesis, however, unless extra measures are taken, the transgene will not replicate with rest of the genome, overtime, transgene expression will naturally be lost (Mulia et al., 2020). Though this is a major challenge in most other organs, the highly limited neuronal replication is beneficial, in this case, and this problem may not need to be addressed when considering gene therapy for diseases predominantly involving neurons.

Epigenetic silencing poses a threat to delivered transgenes (Lufino et al., 2008; Mulia et al., 2020). There are several means to combat this. First, it has been demonstrated that viral promoters, such as the CMV promoter, are subject to greater rates of silencing (Mulia et al., 2020) thus mammalian promoter (ideally human) should be preferentially considered. Second, specific sequences (e.g., CpG sequences) are known to be sites of gene silencing (Mulia et al., 2020; Konkle et al., 2021), minimization of these, and other sequences known to be targeted for methylation should be considered to improve transgene persistence. Finally, extrachromosomal DNA (i.e., episomes) are less subject to silencing compared integrated DNA (Lufino et al., 2008). For this reason, vectors such as AAVs which rarely integrate into the host genome would be preferred to avoid epigenetic silencing.

Overall, the length of persistence of episomal transgenes after AAV gene therapy is not known, especially in the case of treating neurologic diseases. While there is evidence of persistent clinical benefit for at least 6 years, it is unclear how long this might last. As there was no quantitative measurement of SMN it is unclear if the clinical Zolgensma^® has reached its full potential, or if it is being blunted by waning gene expression. Future studies investigating transgene expression over time after CNS gene therapies would be valuable to address these concerns, however it is noteworthy that this would likely require repeated spinal taps to obtain CSF.

Gene therapy provides a logical means for treating many neurologic diseases. NDDs are especially amenable to gene therapy as the pathogenesis of most is mediated either by dysfunction of a single gene or by a single, or small number of proteins (Kovacs, 2019). Gene therapy delivery to the brain can be accomplished using AAV vectors. However, these vectors, and the transgenes they carry, pose significant risks to patients, especially when delivered at high doses. Optimizing the efficiency with which AAVs are delivered is critical to the success of these therapies. This manuscript discusses a subset of AAV delivery technologies thought to contribute to our vision of the ideal gene therapy for the treatment of NDDs of the brain. The vision is to have a single dose gene therapy delivered globally to the brain, with minimal systemic delivery. This therapy would preferentially target neurons, and avoid immune cells such as APCs. Finally, and importantly, the delivery of the therapy does not require opening of the skull. To accomplish this vision, we consider each step of AAV delivery individually, and highlight methods which can improve delivery at that step. We envision a multifaceted approach for optimizing AAV delivery to the brain, which, if successful, would allow for the translation of gene therapies for the treatment of many diseases of the brain.

The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author/s.

JF did most writing and all figures. PF provided the regular guidance, education, suggestions and edited the multiple revisions. Both authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.