Base editing and prime editing represent the latest generation of CRISPR-based gene editing technologies, enabling precise single-nucleotide changes in genomic DNA without inducing double-strand breaks (DSBs). For neurodevelopmental epilepsies (NDEs) caused by well-characterized point mutations — including Dravet syndrome (SCN1A), Angelman syndrome (UBE3A), and KCNQ2 encephalopathy — these technologies offer the potential for permanent, precise correction that addresses the root genetic cause.
Unlike traditional CRISPR-Cas9 (which cuts both DNA strands and relies on cellular repair pathways that are error-prone) or ASO approaches (which require repeat dosing and don't permanently alter the genome), base and prime editors make targeted, precise changes with minimal risk of off-target effects.
Base editors consist of three components:
- Catalytically impaired Cas protein (nCas9 or dCas9) — creates a single-strand "bubble" at the target site without cutting both strands
- Deaminase enzyme — chemically converts one nucleotide to another (e.g., C→T or A→G) within the exposed single-strand region
- Guide RNA — directs the complex to the correct genomic location
flowchart TD
A["Guide RNA<br/>+ nCas9"] --> B["Binds target DNA<br/>(PAM required)"]
B --> C["R-loop formation<br/>Single-strand exposed"]
C --> D["Deaminase acts on<br/>exposed strand"]
D --> E["C to U conversion<br/>(CBEs)"]
D --> E2["A to I conversion<br/>(ABEs)"]
E --> F["DNA repair:<br/>U→T,strand copied"]
E2 --> F2["DNA repair:<br/>Inosine→G,strand copied"]
F --> G["Precise C→T (or G→A)<br/>no DSB"]
F2 --> G2["Precise A→G (or T→C)<br/>no DSB"]
style A fill:#e1f5fe
style G fill:#c8e6c9
style G2 fill:#c8e6c9
| Editor |
Conversion |
NDE Applications |
| CBE (Cytidine Base Editor) |
C→T (G→A on opposite strand) |
~60% of disease-causing point mutations |
| ABE (Adenine Base Editor) |
A→G (T→C on opposite strand) |
~20% of disease-causing point mutations |
| CGBE/AGBE |
C→G (G→C) |
Specific applications requiring GC changes |
| **Hypothetical dual base editing |
Simultaneous C+T or A+G |
Expandable targeting scope |
SCN1A (Dravet Syndrome):
- ~40% of pathogenic SCN1A variants are missense mutations — many are C>T or A>G transitions addressable by CBEs or ABEs
- Examples: R1644H, W1204X, G1674E (all C>T candidates for CBE)
- Correction in patient-derived iPSCs has shown restoration of Nav1.1 channel function in neurons
UBE3A (Angelman Syndrome):
- Some Angelman cases are caused by point mutations rather than deletions — these are prime candidates for base editing
- Restoring specific functional residues in UBE3A could reverse the Angelman phenotype
- Important: Must target only the paternal allele (avoiding the maternal deletion) — requires allele-specific design
KCNQ2:
- Gain-of-function mutations (causing self-limited neonatal epilepsy) vs. loss-of-function (causing KCNQ2 encephalopathy) — base editors could upregulate or correct as needed
- Correction of truncating mutations in KCNQ2 could restore functional KCNQ2 potassium channel subunits
Prime editing is the most versatile of the precision editing technologies, capable of making all 12 types of point mutations plus insertions and deletions, without DSBs or donor DNA templates.
The system has two components:
- Prime editor (nCas9 fused to RT) — a Cas9 nickase fused to a reverse transcriptase
- Prime editing guide RNA (pegRNA) — a modified gRNA that contains both the targeting sequence AND the desired edit template
flowchart TD
ApegRNA["ApegRNA + nCas9-RT"] --> B["Binds target DNA<br/>(PAM required)"]
B --> C["Nick the non-edited<br/>strand (3' flap)"]
C --> D["RT polymerization<br/>uses edited strand as template"]
D --> E["3' flap contains<br/>desired edit"]
E --> F["Nicked strand's<br/>3' flap displaced"]
F --> G["DNA repair copies<br/>edit onto nicked strand"]
G --> H["Precise all-12-substitutions<br/>+ insertions + deletions<br/>No DSB, no donor DNA"]
style A fill:#e1f5fe
style H fill:#c8e6c9
SCN1A (Dravet Syndrome):
- Prime editing could correct all missense variants (not just transition mutations)
- Multi-kilobase insertions possible for larger SCN1A corrections
- Could be used to correct the precise variant in each patient, enabling personalized therapy
CDKL5 Deficiency:
- The majority of pathogenic CDKL5 variants are point mutations addressable by prime editing
- Large deletions in CDKL5 could potentially be corrected with insertion editing
STXBP1 Encephalopathy:
- Diverse mutation types (missense, nonsense, frameshift) — prime editing handles all
- Particularly valuable for nonsense mutations (premature stop codons) via insertion of wild-type sequence
| System |
Mechanism |
Efficiency |
Off-target risk |
| PE3 |
Nick target strand, RT fills edit, cellular repair copies |
Moderate |
Lower |
| PE5 |
Nick both strands, more efficient editing |
Higher |
Slightly higher |
Prime editing efficiency in neurons is typically lower than in dividing cells, requiring optimization of delivery, timing, and pegRNA design for post-mitotic neuronal applications.
The key challenge for base/prime editing in NDEs is delivering the editing components to the right cells in the brain:
| Delivery Approach |
Advantages |
Disadvantages |
| AAV |
Neuronal tropism (some serotypes), long-term expression |
Cargo limit (~4.7kb) limits full editor + gRNA; immune response |
| LNP |
Large cargo, scalable, redosable |
Limited BBB crossing; transient expression |
| Exosome |
Natural BBB crossing, low immunogenicity |
Manufacturing challenges, variable yield |
| High-capacity adenovirus (HdAd) |
Large cargo, long-term expression |
Immune response, less neuronal tropism |
AAV delivery constraint: Most base editors (CBEs: ~5.3kb including SpCas9nick, ABEs: ~5.7kb) barely fit in AAV with a single gRNA. Strategies:
- Split-base editor systems (separate AAVs for each component)
- Use of smaller Cas9 orthologs (e.g., Cas12a, Cas9 from S. aureus)
- Minimized linker architectures
For SCN1A/Dravet, editing must be confined to GABAergic interneurons — editing in excitatory pyramidal neurons could worsen disease:
- Promoter-specific editors: Use GABAergic neuron-specific promoters (e.g., GAD1, GAD2) to drive editor expression
- Cell-type specific gRNA design: For allele-specific editing, exploit differences between mutant and wild-type alleles
- Transient delivery: Non-integrating delivery (mRNA + protein/RNP) to minimize off-targets
Both base and prime editors can cause unintended edits at genomic sites with partial homology to the target sequence:
- Whole-genome sequencing (WGS) is the gold standard for off-target detection
- DISCOVER-seq and CHANGE-seq provide unbiased off-target profiling
- Engineered deaminases (e.g., evoAPOBEC, TadA variants) show improved specificity
Cytidine base editors can cause widespread RNA editing (C-to-U conversions in mRNA), which was a significant concern with first-generation CBEs. New-generation CBEs (e.g.,evoAPOBEC, BE4max) have substantially reduced RNA off-targets through deaminase engineering.
- High-fidelity editor variants: Use latest-generation editors with optimized deaminases and Cas9 variants
- Bioinformatics gRNA design: Choose gRNA targets with minimal off-target homologies
- RNP delivery: Protein/gRNA complexes (rather than plasmid/mRNA) minimize exposure time
- Partial editing: Incomplete correction may be sufficient for haploinsufficient conditions (SCN1A, KCNQ2)
¶ Preclinical Data and Timeline
Patient-derived iPSCs have been edited to model and test base/prime editing approaches:
- SCN1A iPSC models show that Nav1.1 dysfunction can be corrected by restoring wild-type sequence
- KCNQ2 iPSC models demonstrate that loss-of-function can be rescued by precise correction
- CDKL5 iPSC models show that editing can restore normal neuronal electrophysiology
- PE3 in mouse neurons: Prime editing in post-mitotic neurons has been demonstrated with ~10-20% efficiency using RNP electroporation
- CBE in mouse brain: AAV-delivered CBE has edited neurons in vivo with low but detectable efficiency
- NHP studies: Prime editing in non-human primates is ongoing, with results expected 2026-2027
| Milestone |
Expected |
| Proof-of-concept in NHP models |
2026 |
| IND-enabling studies for NDE indication |
2027-2028 |
| First-in-human clinical trial |
2029-2030 |
| Factor |
Base Editing |
Prime Editing |
Traditional CRISPR-Cas9 |
| Precision |
Single nucleotide (C>T or A>G) |
All 12 substitutions + indels |
Requires DSB + homology |
| Efficiency |
High |
Moderate |
High |
| Off-targets |
Low (CBE: RNA concern) |
Low-moderate |
Higher (DSB-induced) |
| Cargo size |
~5kb |
~6kb+ |
~4kb |
| Indel formation |
Minimal |
Very low |
Common |
| In neurons (in vivo) |
Demonstrated |
Demonstrated |
Demonstrated |
| BBB crossing |
Requires delivery optimization |
Requires delivery optimization |
Requires delivery optimization |
| NDE readiness |
Earlier (simpler, higher efficiency) |
Later (more versatile) |
Mid-term |
- Can base/prime editing achieve sufficient efficiency in human GABAergic neurons to provide therapeutic benefit in Dravet syndrome?
- What delivery platform (AAV, LNP, exosome) will best balance cargo size, BBB crossing, neuronal tropism, and durability for NDE applications?
- How many cells need to be edited to achieve meaningful clinical benefit in haploinsufficient NDEs (SCN1A, KCNQ2)?
- What is the safety profile of in vivo base/prime editing in the developing pediatric brain?
- Can allele-specific editing be achieved for patients with deletions vs. point mutations, particularly in Angelman syndrome?