Gene replacement therapies for inherited disorders of neurotransmission: current progress in SSADH deficiency

Abstract

Neurodevelopment is a highly organized and complex process involving lasting and often irreversible changes in the central nervous system. Inherited disorders of neurotransmission (IDNT) are a group of genetic disorders where neurotransmission is primarily affected, resulting in abnormal brain development from early life, manifest as neurodevelopmental disorders (NDDs) and other chronic conditions. In principle, IDNT (particularly those of monogenic causes) are amenable to gene replacement therapy via precise genetic correction. However, practical challenges for gene replacement therapy remain major hurdles for its translation from bench to bedside. We discuss key considerations for the development of gene replacement therapies for IDNT. As an example, we describe our ongoing work on gene replacement therapy for succinic semialdehyde dehydrogenase deficiency (SSADHD), a GABA catabolic disorder.

Introduction

Inherited disorders of neurotransmission (IDNT) are a specialized group of genetic disorders characterized by defects in the biosynthesis, recycling, or transport of neurotransmitters or relevant channelopathies¹ (Table 1). IDNT often associate with defects in neurodevelopment, with clinical symptoms manifest in neonatal or infantile stages². and life-long symptoms including intellectual disability^3,4, movement disorders^5,6, and epilepsy^2,7. Traditional treatment options for IDNT are often limited to symptom alleviation rather than amending the underlying genetic cause⁸. Although in certain cases conditions are treatable and controllable⁹, relevant therapies are likely to require life-long administration and provide limited efficacy. Encouragingly, emerging treatment paradigms for IDNT are shifting from symptom-based, supportive, or palliative approaches to disease-modifying, directly targeting the disease-causing mechanisms¹⁰.

Table 1.

List of IDNT, genetic information, and disease-associated phenotypes.

Disease categories	OMIM gene number	HGNC gene symbol	Molecular function	Genetic inheritance trait	Disease phenotype	CDS size (bp) from NCBI	Gene replacement therapy developmental stage
Monoamine synthesis	600225	GCH1	GTP cyclohydrolase I	Autosomal recessive	Hyperphenylalaninemia	753	Clinical trial142,143
	612719	PTS	6-pyruvoyl-tetrahydropterin synthase	Autosomal recessive	Hyperphenylalaninemia	438	Proof-of-concept144
	612676	QDPR	Quinoid dihydropteridine reductase	Autosomal recessive	Hyperphenylalaninemia	735	No known studies
	182125	SPR	Sepiapterin reductase	Autosomal recessive	Dopa-responsive dystonia	786	No known studies
	191290	TH	Tyrosine hydroxylase	Autosomal recessive	Segawa syndrome	1587	Clinical Trial142,143
	107930	DDC	Aromatic L-amino acid decarboxylase	Autosomal recessive	AADC deficiency	1443	On Europe and UK market145
	609312	DBH	Dopamine beta-hydroxylase	Autosomal recessive	Orthostatic hypotension	1056	No known studies
Monoamine catabolism	309850/309860	MAOA	Monoamine oxidase	X-linked recessive	Brunner syndrome	1584	No known studies
Monoamine transport	126455	SLC6A3	Dopamine transporter	Autosomal recessive	Infantile Parkinsonism-dystonia (Dopamine transporter deficiency)	1863	Proof-of-concept146,147
	193001	SLC18A2	Vesicular monoamine transporter	Autosomal recessive	Infantile Parkinsonism-dystonia	1545	No known studies
GABA catabolism	137150	ABAT	GABA-transaminase	Autosomal recessive	GABA-T deficiency	1503	No known studies
	610045	ALDH5A1	Succinic semialdehyde dehydrogenase	Autosomal recessive	SSADHD	1608	Proof-of-concept (Lee et al., in prep)
GABA receptors	137160	GABRA1	GABA(A) receptor, alpha 1	Autosomal dominant	Developmental and epileptic encephalopathy	1371	No known studies
	137190	GABRB1	GABA(A) receptor, beta 1	Autosomal dominant	Developmental and epileptic encephalopathy	1425	No known studies
	600232	GABRB2	GABA(A) receptor, beta 2	Autosomal dominant	Developmental and epileptic encephalopathy	1425	No known studies
	137192	GABRB3	GABA(A) receptor, beta 3	Autosomal dominant	Developmental and epileptic encephalopathy	1422	No known studies
	137164	GABRG2	GABA(A) receptor, gamma 2	Autosomal dominant	Developmental and epileptic encephalopathy	1404	No known studies
GABA transport	137165	SLC6A1	GABA transporter	Autosomal dominant	Myoclonic-atonic epilepsy	1800	Proof-of-concept (Steven Gray, personal communication)
Glycine receptors	138491	GLRA1	Glycine receptor	Autosomal recessive/dominant	Hyperekplexia	1101	No known studies
	138492	GLRB	Glycine receptor	Autosomal recessive	Hyperekplexia	1494	No known studies
Glycine transport	604159	SLC6A5	Glycine transporter	Autosomal recessive/dominant	Hyperekplexia	2474	No known studies
	601019	SLC6A9	Glycine transporter	Autosomal recessive	Glycine encephalopathy	2121	No known studies
Glutamate receptors	138249	GRIN1	Glutamate receptor, NMDA type	Autosomal recessive	Developmental and epileptic encephalopathy	2817	No known studies
	138253	GRIN2A	Glutamate receptor, NMDA type	Autosomal dominant	Epilepsy, focal, with speech disorder	3846	No known studies
	138252	GRIN2B	Glutamate receptor, NMDA type	Autosomal dominant	Developmental and epileptic encephalopathy	4455	No known studies
	602717	GRIN2D	Glutamate receptor, NMDA type	Autosomal dominant	Developmental and epileptic encephalopathy	4011	No known studies
Metabotropic receptors	604473	GRM1	Glutamate receptor, metabotropic	Autosomal recessive/dominant	Spinocerebellar ataxia	3585	No known studies
	604096	GRM6	Glutamate receptor, metabotropic	Autosomal recessive	Night blindness, congenital stationary	2634	No known studies
Channelopathies	182389	SCN1A	Sodium channel, voltage-gated	Autosomal dominant	Dravet syndrome	6000	Proof-of-concept148, clinical trial ongoing
	182390	SCN2A	Sodium channel, voltage-gated	Autosomal dominant	Developmental and epileptic encephalopathy	6018	No known studies
	600702	SCN8A	Sodium channel, voltage-gated	Autosomal dominant	Developmental and epileptic encephalopathy	5199	No known studies
	176258	KCNC1	Potassium channel, voltage-gated	Autosomal recessive	Progressive myoclonic epilepsy	1536	No known studies

Recent advances in genetic medicine, bioengineering, and viral vector technology enable the translational deployment of molecularly-based therapy, such as gene replacement therapy^11–13 Gene replacement therapy involves the delivery of a functioning gene into relevant cells and tissues, replacing or compensating for a disease-causing genetic origin, thereby correcting the underlying genetic defect. Viral vectors are often used for such therapeutic delivery. The commonly used viral vectors are listed in Table 2, where their basic properties, including payload capacity, transgene integration characteristics, and immunogenicity, are compared.

Table 2.

Comparison of commonly used viral vectors for gene replacement therapy.

	Adenovirus	Adeno-associated virus	Lentivirus (LV)
Size	~100 nm	~25 nm	~100 nm
Genome	dsDNA	ssDNA	ssRNA
Packaging capacity	~37 kb29	~4.6 kb (single-stranded) ~2.3 kb (self-complementary)	~8 kb
Transduction	Dividing and non-dividing cells	Dividing and non-dividing cells	Dividing and non-dividing cells
Transduction efficiency in non-human primate neural tissues	Low149	High (>70% for AAV9)52,150,151	High (up to ~70% using a second generation LV vector)152
Expression	Transient	Stable	Stable
Genome integration in brain	Very low chance of integration27,153,154	Confirmed with identified site for wild-type AAV, very low chance of integration for recombinant rAAV40,155	Confirmed integration156
Immunogenicity	High	Moderate	Low

Viral-mediated gene replacement therapy has gained traction in biomedical research in recent years, as the rapid growth of relevant research illustrated by a PubMed search for ‘Gene Replacement Therapy’ between 1985 and 2023 revealing a substantial increase in the number of published articles within the last decade (Figure 1A). Notably, most studies are directed toward neuromuscular disorders, while those related to neurodevelopment or neurotransmitter disorders in the central nervous system remain a minority (Figure 1B). While encouraging recent trends indicate that gene replacement therapy will one day become a prevalent form of IDNT treatment, challenges remain to be overcome for successfully translating gene replacement therapy from bench to bedside.

Figure 1. PubMed search results for “gene replacement therapy.”.

(A) The number of published articles listed in the PubMed database after a keyword search of gene replacement therapy from 1980 to 2023. (B) Published articles in 2023 is classified into different disorder categories.

This article will discuss key principles and considerations for developing gene replacement therapy for IDNT. First, genetic inheritance traits determine gene replacement strategy. For example, autosomal dominant diseases are more challenging than autosomal recessive diseases because the dominant mutations might require specific silencing in addition to functioning gene supplementation. Mutant proteins might interfere with the ectopically introduced therapeutic functional proteins, often generating their own immunogenic responses. In certain situations, knock-down of mutant proteins might be necessary for functional restoration of protein products. Second, the size of the gene of interest and its expression profile will determine the choice of viral vector and relevant delivery strategies. Due to the limited size of viral vector payload capacity, the gene of interest might be too large to be accommodated in a single viral vector. In such scenarios, a mini-gene alternative approach (e.g., a truncated gene encoding only the functional domains) may be a viable strategy. There are also dual-vector systems that reconstitute a functioning gene from multiple components. Third, the expression profile of the gene of interest will limit the choice of vector delivery route. For example, for a gene expressed in certain brain regions, a local intracortical injection rather than brain-wide coverage might be sufficient and optimal. Fourth, the availability of a viable disease-relevant model will be necessary to investigate target engagement and phenotypic reversibility upon gene replacement. Therapeutic parameters relevant for toxicity and efficacy are vital information that can only be evaluated empirically using an animal or cell-based model exhibiting disease-relevant phenotypes. For bench-to-bedside translational drug discovery, the availability of such experimental disease models is critical in providing valuable information for dose finding, toxicity, and disease reversibility information. As an example of how these principles might apply to the translation of gene replacement therapy for IDNT, we utilize a recently proposed framework for evaluating neurodevelopmental disorders for gene therapy¹⁴, describing our ongoing efforts in developing gene replacement for succinic semialdehyde dehydrogenase deficiency (SSADHD).

Mendelian inheritance traits

The pathogenic mechanisms underlying IDNT include dominant or recessive traits. The inheritance trait of a certain genetic disorder is crucial for the design of corresponding therapeutic approach^15–17. Autosomal recessive inheritance requires variants in both copies of the allele. Mutational content of the gene in autosomal recessive disorders could be identical (homozygous mutants) or different (compound heterozygous), resulting in defective or absence of protein products. For recessive disorders, loss-of-function mutations might be amenable to directly introducing a functional gene in gene replacement therapy (Figure 2A). On the other hand, autosomal dominant inheritance entails a disease-causing variant copy of an allele, in which a genetic trait ‘dominates’ over a normal gene copy. In autosomal dominant disorders, one copy of the gene is pathogenic while the other copy is functional, which renders gene targeted approach particularly challenging¹⁸. An example of gene therapy for autosomal dominant disorder is an ablate-and-supplement strategy employed by Meng and colleagues for Retinitis Pigmentosa (Figure 2B)¹⁹. In the case of mutations leading to early stop codon or other mutations resulting in little or no protein produced, gene replacement therapy or other relevant strategies aimed at augmenting gene products might be sufficient and effective. For example, in Dravet syndrome (DS)-related epileptic encephalopathy where autosomal dominant SCN1A gene mutations result in voltage-gated sodium channel Nav1.1 haploinsufficiency²⁰, an antisense oligonucleotide (ASO) approach called Targeted Augmentation of Nuclear Gene Output (TANGO) rescues seizure and sudden death in a relevant mouse model²¹. Importantly, a Phase1/2a human trial of TANGO-based therapy in patients with DS demonstrated tolerability and favorable therapeutic results (dose-dependent and lasting reductions in convulsive seizure frequency) (Press Release from Stoke Therapeutics, July 2023). However, there are situations where endogenous protein products might interfere with the exogenously supplemented normal-functioning protein products. This is particularly relevant for proteins that require the assembly of multimers, where each component contributes to the protein products’ functional activities. For example, in SLC6A1-related disorders, protein products from a variant allele might interfere with endoplasmic reticulum trafficking²². In such situations, the functional activities of hybrid protein products assembled from mutated endogenous and normal-functioning ectopic proteins will require careful characterization and empirical functional measures. Functional characterization in such heterologous recombinant proteins²³ will offer mechanistic insights into the efficacy of gene supplementation therapy. Furthermore, crystal structures derived from X-ray crystallography²⁴ and molecular dynamics simulations²⁵ of each polypeptide chain in the protein complex resultant from normal and variant peptides might improve our understanding at atomic levels. Overall, additional functional and biophysical approaches may help to understand functional integrity of such protein constitutes or complex formed from functioning ‘wild type’ and mutant proteins. Similar approaches dealing with autosomal dominant disorders, including selective silencing via anti-sense oligonucleotide (ASO)²⁶, might be necessary to dampen the dominating negative effects of endogenous mutant proteins in such situations.

Figure 2. Gene therapy strategies for autosomal recessive and autosomal dominant disorders.

(A) In autosomal recessive disorders, the underlying etiology is due to the presence of disease-causing DNA mutations in both alleles resulting in non-functional protein products. A viable gene therapy strategy involves functional gene replacement via a viral-mediated approach. (B) In autosomal dominant disorders, disease-causing DNA mutations are found in one allele, resulting in dominant mutant protein products. A corresponding gene therapy strategy requires allele-specific silencing of the dominant mutations at the mRNA messenger level eliminating the dominant mutant proteins, sometimes requiring functional gene supplementation via a viral-mediated approach.

Gene size, expression profile, and routes of administration consideration

Gene replacement therapy utilizes bioengineered viral vector molecules carrying a functioning copy of the disease-relevant gene for delivery producing therapeutic effects. Notable viral vectors gaining recent traction for regulatory approval in clinical trials and eventual clinical use include adenovirus, adeno-associated virus (AAV), and lentivirus²⁷. Table 2 (adapted from Tosolini and Sleigh²⁸) summarizes their properties, including payload capacity, transgene stability, and immunogenicity of these prevalent vectors. Adenovirus infects post-mitotic cells and has large cargo capacity up to 37 kb²⁹, but it triggers strong inflammatory reactions in the brain^30,31. Lentiviral vectors offer long-term and stable gene expression to post-mitotic cells including neurons, but their genome-integrating property presents a realistic risk for safety³². On the other hand, AAV integration has been shown to be mediated by its viral gene Rep³³, which encodes Rep proteins targeting the Adeno-Associated Virus Integration Site 1 (AAVS1) locus in human chromosome 19^34,35. In principle, recombinant AAV (rAAV) lacking viral gene Rep enables transgenes to form circular concatemers that persist as episomes in the nucleus of transduced cells, significantly minimizing the chance of genome integration^36,37. Nevertheless, long-term genome integration by AAV and evidence of subsequent insertional mutagenesis remain an active research area requiring long-term assessment³⁸. Noticeably, recent meta-analysis of rAAV clinical trials revealed 90% of cases involving the CNS achieved transgene durability in follow-up visits³⁹. Nevertheless, AAV is limited by a small packaging capacity of less than 5 kb. Alternative strategies accommodating a larger gene include a micro gene encoding a truncated but functional protein and co-administration of a dual-vector system reconstituting a full-length gene⁴⁰. Besides viral vectors, non-viral-based strategies including biodegradable polymers, lipids, or other inorganic nanoparticles as alternative vehicle materials for gene therapy are under rapid development⁴¹. Non-viral-based vectors have large payload capacity and low immunogenicity but lack cell target specificity in general⁴². Overall, these strategies require a deep understanding of the protein’s structure-function relationship and empirical testing in relevant disease models for therapeutic efficacy.

Another key consideration is the tropism of viral vectors enabling the necessary targeting of cells and tissues relevant for the expression profile of the gene of interest⁴³, in combination with route of administration to achieve physiologically relevant transgene delivery⁴⁴. Certain genes are expressed in a broad range of cell types, or covering multiple brain regions, whereas others are expressed in specific cell types within confined tissue origins. For broadly expressed genes, a delivery strategy enabling wide transgene coverage might be necessary⁴⁵. Viral vectors can be administered via intrathecal (IT) injection, i.e., targeting the subarachnoid space bypassing the blood-brain barrier (BBB) via lumbar puncture^46,47, or intra-cisterna magna (ICM) injection^48–50. Brain-wide coverage might be better achieved through delivery into cerebrospinal fluid (CSF) in the ventricles by intracerebroventricular (ICV) injection^51,52. Intravenous (IV) infusion is another option with clear advantage of less invasiveness compared to the abovementioned delivery methods⁵³, but it requires therapeutic agents with superb BBB penetrance for brain coverage. The development of new AAV capsid with such high BBB penetrating property is an active research area^54–56, including the use of novel molecularly-based screening platforms for high throughput capsid selection^57,58. On the other hand, intraparenchymal delivery might be ideal for local delivery targeting specific groups of neurons or brain structures relevant to the disorders due to underlying etiology^44,59. However, the spread of viral vector is physically limited and is not ideal for brain-wide coverage⁶⁰. Overall, the choice of the administration routes requires simultaneous considerations of viral vector BBB penetrance, cell targeting, and the extent of brain coverage. Their associated risks in terms of dosing, immune response, and corresponding immunosuppressive strategies must be carefully considered^61–64.

Besides the considerations of transgene packaging into viable vectors and disease-relevant cell or tissue targeting accessible via corresponding delivery procedures, another critical factor is the expression profile of transgene expression. Ideally, therapeutic efficacy and associated risks of gene replacement therapy are related to the transgene expression dosage, i.e., the delivered gene dosage should be within a certain range that produces sustainable therapeutic effects, but not excessive (i.e., overdose) such that overexpression toxicity might overshadow any therapeutic advantage. This may be achieved partly by carefully dosing the delivered vector⁶⁵, but another key factor is the molecular regulatory element or ‘promoter’ incorporated into the vector payload⁶⁶. The choices of ‘promoters’ in gene replacement therapy candidates range from small but constitutive active promoters to self-regulatory elements providing feedback control^67,68. Synthetic promoters might be used to target specific cell types in the CNS⁶⁹. An alternative strategy that has not been commonly adopted is the use of custom design promoters encompassing specific gene regulatory elements in the endogenous promoter⁷⁰. Molecular cloning of certain length DNA sequences (e.g., 2 kb) upstream of the transcriptional start of a given gene is often needed to characterize a given gene’s promoter and associated regulatory elements⁷¹. Truncational analyses of such upstream sequences tethered with a reporter gene assay (e.g., luciferase reporter) are often used strategies to fine-tune the inclusion or exclusion of DNA sequences as the promoter. However, traditional promoter studies often exploit heterologous recombinant cell-based systems such as fibroblasts for excellent transfection efficiency and reporter assay compatibility, but the cellular signaling and genetic contents are often biased or irrelevant for physiologic transgene expression. A fundamentally different approach to designing a promoter would Involve an unbiased assessment of transgene expression in relevant cell types (e.g., single cell RNA-seq), combined with bioinformatics⁷² and machine learning approach^73–75 to ‘mine’ patterns of DNA sequences correlating all genes’ expression profiles across cell types (Figure 3). Ideally, these regulatory elements or promoters should drive a given gene’s functional expression resembling its endogenous profile across developmental stages. These custom promoters are tailored to specific genes, and therefore require empirical testing using relevant disease models, including those with human genomic relevance (e.g., patient derived cells). Remarkably, machine learning approaches are being deployed in the rational design and identification of next-generation viral vectors for higher transduction efficiency, improved tropism, and low toxicity^76–78. The use of novel machine learning algorithms such as large language models⁷⁹ might further facilitate such rational design of viral vector capsid⁸⁰ and therapeutic payload⁸¹.

Figure 3. Proposed machine learning approach optimizing gene-specific promoter sequences.

A novel approach to optimizing the gene-specific promoter sequence of any given gene involves curating single-cell RNA-seq data and applying neural network analysis to seek common patterns of gene expressions across cell types and developmental stages. The optimized promoter sequence will then be used guiding the custom design of vector. The resulting gene replacement therapy candidate can be evaluated in disease-relevant models.

Disease model availability

Successful translational efforts from bench to bedside require empirical testing of any given treatment paradigm or drug candidate at a preclinical stage minimizing risks and maximizing therapeutic efficacy. In principle, disease models containing genetic variant(s) relevant for the disorder should resemble the underlying pathophysiology at a molecular level, displaying a disease-relevant phenotype for preclinical testing. Disease models are broadly classified as in vitro (in a test tube or culture dish) and in vivo (in a living organism), which have their corresponding strengths and limitations. For example, induced pluripotent stem cells (iPSCs) derived from patient samples such as skin biopsy are useful in vitro models for experimental drug testing^82,83. iPSCs have the capacity to differentiate into specific cell types via molecular programming in a culture dish. iPSC-derived cell models are therefore versatile, and they contain genetic content identical to patients. Isogenic controls are often included allowing the gene of interest to be studied specifically in an otherwise identical genetic background. The patients’ genomic and mutational content is particularly important for gene replacement therapy because the functional proteins (from the therapeutic payload delivered by the vector) and the mutant proteins (from patients) might interact and affect each other at a functional level. Mutant proteins might also occupy relevant cellular signaling or trafficking machinery⁸⁴ necessary for the therapeutic proteins to function. Recently, iPSCs are coupled with organoid technology allowing cells to grow in 3 dimensions forming complex structures resembling tissue-like organization and function^85,86. This new approach further enables the modeling of cell-cell interactions, intrinsic/extrinsic cell signaling factors, and cell migration which were not feasible to be studied in a dish^87,88. While in vitro disease models offer unique opportunity to investigate cellular function and properties relevant for disease mechanisms, nevertheless they are limited by the lack of complexity and intrinsic anatomy of an organism, and no disease-relevant behavior at an organismal level can be performed in any cell culture or organoid models.

On the other hand, in vivo disease modelling can be achieved by precise genetic modifications in a model organism, complementing in vitro models in certain aspects. Ideally, disease models should have a well-established and well-controlled genetic background, and reasonably short life cycle (e.g., within twelve months) allowing practical experimental paradigm and testing duration. A commonly used organism for in vivo disease modelling is Mus musculus (house mouse)⁸⁹. Such mammalian disease models that recapitulate the underlying pathophysiology and disease-relevant phenotypes are particularly valuable for preclinical testing, because key treatment parameters can be investigated and fine-tuned to produce the best therapeutic outcomes in terms of phenotypic reversal. Mice also exhibit behaviors including sensory function, motor function and coordination, cognition, and social communication allowing studies in these complex functions which are often disease relevant^90–92. An in vivo disease model also provides an opportunity to study pharmacokinetics (the study of how an organism affects a drug) and pharmacodynamics (the study of how the drug affects the organism), as well as testing clinically relevant delivery routes and their associated biodistribution^93,94.

Strategies for in vivo disease modelling include the total disruption of a given gene via locus-specific disruption (e.g., knock-out), or precise insertion of disease-causing variant-mimicking sequences (e.g., knock-in)⁹⁵. Alternatively, gene excision can be achieved via Cre-loxP strategy⁹⁶. This strategy typically involves the development of two independent genetic lines (e.g., mouse lines), with one line harboring a pair of ‘loxP’ sites flanking functional domain(s) of a given gene, and the other line expressing Cre recombinase (typically driven by specific promoter for tissue-specific expression)⁹⁷. When these two independent lines are bred together, the resultant offspring expresses Cre recombinase in specific tissues, disrupting a given gene via Cre-lox recombination achieving tissue-specific or cell type-specific deletion of gene function⁹⁸. This strategy is particularly useful for modeling IDNT disease etiology where a given gene’s cell type-specific role can be investigated in detail. Relevant for gene replacement therapy in IDNT, a variation of the Cre-loxP strategy called ‘lox-STOP’ can be employed. Here, a genetic cassette harboring a STOP codon flanked by a pair of loxP sites will be inserted into a given gene, disrupting its baseline function⁹⁹. This results in essentially a gene ‘knock-out’ situation, but the STOP codon can be readily removed upon the introduction of Cre (e.g. via genetic breeding with a Cre-expressing line, or the injection of a Cre-expressing viral vector)¹⁰⁰. This system allows precise reactivation of a given gene, mimicking a gene restoration scenario. Clinically relevant parameters for gene replacement therapy including rate, time, and cell-specificity can be investigated using such animal models.

Importantly, in vivo disease modeling is subject to limitations. For example, disease-causing variants of a given gene in patients might not cause the same phenotype in the model organism due to species-specific genetic compensation or other variance requiring careful design and experimental planning to circumvent potential pitfalls¹⁰¹. The disease relevant phenotype in the model organism might be too severe (e.g., embryonic or neonatal lethal) or not manifested at all. In such situations, a combination of knock-out and hypomorphic variant (partial reduction of gene product) might be useful to fine-tune gene dosage recapitulating disease phenotype^102,103.

Current progress in developing gene replacement therapy for SSADHD

Clinical background and preclinical considerations for gene replacement therapy

SSADHD is a rare autosomal recessive disorder due to loss-of-function mutations of the gene ALDH5A1 essential for GABA catabolism¹⁰⁴. The genetic cause of SSADHD is monogenic, although multiple variants have been reported among patients¹⁰⁵. The size of the human ALDH5A1 gene coding region is 1.6 kb¹⁰⁶, a size compatible with commonly used viral vectors (Table 2). The underlying pathophysiology of SSADHD is due to the pathologic accumulation of GABA and its metabolite γ-hydroxybutyrate (GHB), the biochemical hallmarks of this disorder¹⁰⁷. SSADHD symptoms include developmental delay, intellectual disability, and autistic behaviors typically detectable at an early stage (2–3 years of age)¹⁰⁸. A natural history SSADHD study is ongoing, and thus far revealed a 15% incidence of sudden unexpected death in epilepsy (SUDEP) in this patient population among other developmental and behavioral morbidities¹⁰⁹. Current treatment options for SSADHD are symptomatic with limited efficacy¹¹⁰. Gene replacement therapy is a potential disease-modifying treatment for SSADHD but is currently unavailable^111,112. The broad expression profile of ALDH5A1 in the brain suggests that a brain-wide targeted approach is likely necessary.

To develop gene replacement therapy for SSADHD, we evaluated the current preclinical and clinical landscape of this disorder following our previously proposed framework¹⁴. Given our current research progress, combined with a highly engaged family community and readily accessible patient population, we estimated a composite GTS (gene target suitability) score of 34 for SSADHD (out of a maximum score of 40), indicating overall very positive favorability among some of the prime examples of neurodevelopmental disorders currently under gene therapy development (Table 3). Ongoing work on gene replacement therapy for SSADHD focuses on disease-relevant modeling and clinical biomarkers, including electroencephalography (EEG) and transcranial magnetic stimulation (TMS)^113,114.

Table 3.

The GENE TARGET framework evaluation for SSADHD.

Category	Description	Current status for SSADHD	Score
G	Genetic mechanism understood	Autosomal recessive mutation causing ALDH5A1 loss of function	6/6
E	Early diagnosis typical	Mean diagnosis age ~6.6 years	1/2
N	Natural history documented	Natural History Study ongoing	2/3
E	Endpoints validated	EEG and TMS-based under development	1/3
T	Tools targeting relevant tissues available	Brain and liver targeting viral vector tools available	5/6
A	Availability of alternative effective treatment	No disease-modifying treatment available	2/2
R	Reversibility demonstrated	Phenotypic reversibility demonstrated in a mouse model	4/4
G	Gene is tolerant to dosage changes	Effective gene dosage threshold and range exist	3/4
E	Ethical principles considered	Moderate to severe level of disability often requiring life-long personal care, risk of epilepsy, and SUDEP	6/6
T	Target population is engaged and accessible	Well-organized clinics and engaged individual/family communities worldwide	4/4
Total score			34/40

At the disease modeling front, we utilized the ‘lox-STOP’ strategy as described above, creating a novel SSADH mouse model allowing gene restoration ‘on demand’¹¹². This new aldh5a1^lox-STOP mouse model enables systematic investigation of three key parameters (i.e., rate, age, and cell-specificity) for successful phenotypic rescue. As a step toward clinical translation, current efforts also include the development of a custom-designed AAV vector encompassing an ALDH5A1-specific promoter tethered with the human ALDH5A1 coding sequence. In addition, human iPSC models are also under parallel development, which will provide a powerful testing platform with human genomic relevance. Furthermore, functional SSADH enzyme exists as tetramers¹¹⁵. The intracellular fate of endogenous mutant proteins (i.e., intracellular trafficking and recycling) and the molecular interaction between patients’ endogenous mutant proteins and the functioning proteins provided by the gene replacement therapy payload might play a key role in determining treatment efficacy, which can be tested empirically by a range of functional assays and biophysical methods¹¹⁶.

While ongoing experiments assessing the safety and efficacy of a candidate gene replacement therapy yield preliminary favorable results^117,118, compensatory changes at a molecular level might have taken place during early brain development in SSADHD (Figure 4). Broadly, compensatory changes that occur during neurodevelopment are common among IDNT and are realistic concerns because unwanted adverse effects might result from suboptimal gene restoration. For example, restoring the GABA-degrading enzyme in a setting of down-regulated GABA receptors in SSADHD might inadvertently lead to a hypo-GABAergic condition, increasing the chance of seizures and brain injury¹¹¹. Here, we describe two potential mechanisms relevant to compensatory changes that might fine-tune GABAergic inhibition.

Figure 4. Potential GABA-related compensatory mechanisms underlying SSADH deficiency.

A schematic diagram showing the conceptual molecular changes in SSADH deficiency where GABA catabolism is impaired. At the presynaptic level, pathologic accumulation of GABA and its metabolite GHB might be accompanied by reduced GABA synthesis. At the postsynaptic site, synaptic GABA_A receptors (particularly of those γ-subunit containing receptors) might be removed from cell surface due to enhanced internalization.

Compensatory GABAergic synapse changes

GABAergic synapses in the brain consist of presynaptic and postsynaptic domains, which are highly regulated during development. At the pre-synaptic site, neurotransmitter release is mediated by a complex system involving syntaxin¹¹⁹, synuclein¹²⁰, synaptotagmin¹²¹, synaptobrevin¹²², and synaptophysin¹²³. Distinct developmental events involving relevant protein components are highly coordinated and often arranged sequentially. For example, vesicle pools are first formed and refined during synapse formation events^124,125. Vesicle release undergoes transitional changes from spontaneous events to evoked release upon activity-dependent mechanisms¹²⁶. Calcium-mediated vesicle fusion is then further coupled in an activity-dependent manner, strengthening synapse integrity¹²⁷. At the postsynaptic site, GABA receptors are found in hetero-pentameric structures consisting of a typical 2α/2β/1γ composition¹²⁸. In rodent early postnatal life, GABA receptors are predominantly α2/3-containing^129,130. This expression pattern is replaced with α1-containing GABA receptors as the brain matures^131,132. It is known that α2/3 subunits contribute to a slow kinetics of GABA current, while α1 subunits mediate fast decay. Strikingly, this subunit developmental switch precisely coordinates GABA current dynamics throughout development¹³³. These findings have broad indications related to the function and computation of local neural circuits affecting higher-level functions. Interestingly, in pathologic conditions such as epilepsy, this GABA receptor subunit compositional switch is often delayed¹³⁴. Patients with SSADHD have reduced GABA receptor expression¹³⁵, highlighting an altered GABA developmental trajectory in these patients. Beyond the synaptic regions, the extrasynaptic GABAergic system mediating tonic inhibition is predominantly mediated by δ-containing GABA receptors¹³⁶. Overall, the dynamic changes of presynaptic, postsynaptic, and extrasynaptic regions involved in GABAergic neurotransmission in SSADHD remain an active research topic that likely provides insights into de-risking and fine-tuning gene replacement strategy.

Chloride homeostasis disruptions

Fast GABAergic inhibition depends on the neuronal transmembrane chloride gradient regulated by chloride transporters¹³⁷. The neuronal-specific potassium chloride co-transporter KCC2 is a key player in chloride homeostasis in mature neurons, and thus plays a crucial role in maintaining the efficacy of GABA-mediated inhibition¹³⁸. In the early developmental stage, however, KCC2 expression is relatively low compared to the chloride intruder sodium-potassium chloride co-transporter NKCC1, rendering GABA less inhibitory or potentially depolarizing in premature neurons¹³⁹. Interestingly, GABA itself plays a role in promoting KCC2 expression, implying that this developmental switch from depolarizing to hyperpolarizing GABA current is an autonomous process¹⁴⁰. KCC2 activity also plays a key role in inhibitory neuron migration, a fundamental process in functional brain circuit development¹⁴¹. In a condition such as SSADHD, where pathologic GABA accumulation begins in early life, chloride homeostasis is likely significantly affected, resulting in a profound and sustained pathologic phenotype. The reversibility of these changes will be necessary for successful gene replacement therapy in SSADHD.

Concluding remarks: a generalizing principle beyond SSADHD

Given the recent rapid advancement of genetic medicine and viral technology, gene replacement therapy for inherited disorders of neurotransmission (IDNT) (especially those of monogenic origin) quickly emerged as a realistic treatment option. The translational research of gene replacement therapy is multi-disciplinary, requiring combined efforts from various areas of expertise, including bioengineering, molecular biology, in vitro and in vivo disease modeling, pharmacology, clinical care, and regulatory experience. Here, we discuss the practical aspects of developing gene replacement therapy for IDNT, focusing on the involved genetic content (i.e., inheritance, gene size, expression profile), as well as the availability of disease models allowing the explicit testing of key parameters for maximum therapeutic efficacy and minimizing risk. We use our ongoing program of gene replacement therapy in SSADHD as an example, demonstrating our thought process and describing experimental efforts for agile bench-to-bedside translation. While current work toward SSADHD treatment is in progress, we believe our experience provides necessary insights and guiding principles generalizable for initial efforts to develop gene replacement therapy for other indications within the IDNT family.

Data availability statement

Data presented in this review is accessible from public databases PubMed (https://pubmed.ncbi.nlm.nih.gov/) and OMIM (https://www.omim.org/).