Improvements in CRISPR Base Editing: From Molecular Changes to Medical Treatments

The foundation of most base editors is an engineered variant of the CRISPR‐Cas9 protein, typically derived from Streptococcus pyogenes . However, unlike the fully active nuclease employed in conventional CRISPR editing, base editors incorporate Cas9 versions with modified catalytic activity [27]. Early versions of base editors, such as BE1 and BE2, employed a dCas9 that harboured mutations (D10A and H840A) disabling both of its nuclease domains. This modification enabled the dCas9 to bind target DNA, as directed by a gRNA, without introducing DSBs [24].

In contrast, subsequent and contemporary base editors, such as BE3, BE4 and ABEs, predominantly incorporate an nCas9, which cuts only one DNA strand [3]. A widely used nCas9 variant carries the D10A mutation, which inactivates the RuvC nuclease domain tasked with cutting the non‐target DNA strand. Meanwhile, the HNH domain remains active and introduces a single‐strand nick in the target strand that is complementary to the gRNA [28]. Alternatively, the H840A mutation can inactivate the HNH nuclease domain, generating a Cas9 nickase that cleaves only the non‐target DNA strand. However, most base editing systems commonly employ the D10A nickase variant.

Regardless of whether dCas9 or nCas9 is used, the primary role of the Cas9 component in a base editor is to serve as a programmable DNA‐binding scaffold, precisely guiding the fused deaminase enzyme to the genomic locus specified by the gRNA sequence. Furthermore, in base editing systems that employ a nCas9, the resulting single‐strand break (nick) is thought to influence cellular DNA repair pathways, biasing them towards resolving the mismatch in favour of incorporating the intended base change on the opposing strand.

Specificity within the CRISPR‐Cas9 machinery is provided by the sgRNA. This synthetic RNA molecule combines essential features from the natural CRISPR RNA (crRNA) and trans‐activating crRNA (tracrRNA). Its 5′ end contains a variable ‘spacer’ region of about 20 nucleotides, designed to match the target DNA sequence (protospacer). The sgRNA directs the Cas9‐deaminase fusion protein to the target genomic site through base pairing between the spacer and protospacer [33]. Successful targeting also necessitates a short, conserved DNA sequence called the protospacer adjacent motif (PAM), situated immediately next to the protospacer site. While the standard SpCas9 typically requires an NGG PAM sequence (N being any base), engineered Cas9 variants have been developed with different PAM specificities, increasing the number of potential target sites throughout the genome [34].

The deaminase component functions as the catalytic engine of the base editor, performing the chemical transformation on the targeted nucleotide base. As a result, the type of base conversion that occurs is determined by the identity of the fused deaminase. Specifically, CBEs employ cytidine deaminase enzymes to achieve the desired base modification [35]. These enzymes function by catalysing the hydrolytic deamination of a cytosine base, effectively converting it into uracil. While the rat‐derived APOBEC1 is the most frequently utilized deaminase in CBEs, other enzymes such as human APOBEC3A (hA3A) [36], lamprey PmCDA1 [37] and human activation‐induced cytidine deaminase (AID) [15] have also been incorporated or further engineered to potentially improve CBE performance characteristics.

ABEs require a DNA adenine deaminase, an enzyme not found naturally. To overcome this, scientists engineered versions based on the TadA [38]. The initial breakthrough yielded the TadA‐7.10 variant, commonly employed as a heterodimer with wild‐type TadA within the ABE7.10 editor system. Further directed evolution efforts have since produced enhanced versions with improved properties, such as the TadA‐8e variant utilized in the ABE8e editor. These evolved TadA enzymes catalyse the deamination of adenine within the DNA substrate, converting it to inosine [5].

UGI functions as a key repair module in most CBEs, especially those developed from BE2 onwards, significantly boosting their efficiency and specificity [39]. UGI is a small protein from the PBS1 bacteriophage that potently blocks the cell's UDG enzyme [40]. Normally, UDG initiates base excision repair (BER) by identifying and removing uracil from DNA [41]. By inhibiting UDG, the UGI component prevents the excision of the uracil base produced by the CBE's deaminase activity. Protecting this uracil intermediate greatly increases the probability that the resulting U•G mismatch is converted into the desired T•A base pair through cellular DNA replication or repair. Because the inosine (I) generated by ABEs is not efficiently recognized and removed by cellular DNA glycosylases, unlike the uracil produced by CBEs, ABEs typically do not require UGI.

While standard base editors (CBEs and ABEs) perform transition mutations (swapping similar base types), creating transversion mutations (swapping purines for pyrimidines or vice versa) required new strategies. Cytosine‐to‐Guanine Base Editors (CGBEs), such as CGBE1 and Td‐CGBE, achieve C•G to G•C changes [37]. They typically combine a CBE structure (nCas9‐APOBEC1/TadA variant) with engineered DNA glycosylases (like modified UNG or TDG). These glycosylases remove the intermediate uracil base in a way that prompts DNA repair to insert guanine [47]. Td‐CGBE, a TadA‐8e‐derived cytosine‐to‐guanine base editor, achieves C•G‐to‐G•C conversion efficiencies of up to 72.8% and operates within a narrow editing window spanning nucleotides C5 to C6, providing high specificity for targeted genomic loci [46]. Its development expands the scope of base editing beyond transition mutations, enabling precise correction of previously inaccessible cytosine sites in mammalian genomes. Similarly, Adenine Transversion Base Editors (Adenine to Y (Pyrimidine) Base Editor (AYBEs)) use ABEs fused with engineered glycosylases to convert A•T pairs to T•A or C•G [68]. By fusing an evolved adenine deaminase (TadA‐8e) with a nickase Cas9 and N‐methylpurine DNA glycosylase (MPG) or alkyladenine DNA glycosylase (AAG), AYBEs enable precise A•T‐to‐C•G or A•T‐to‐T•A conversions within specific sequence contexts. The engineered mAAG (mouse AAG) excises the adenine‐derived intermediates, achieving A‐to‐C/T conversions of up to 73%, while maintaining a defined editing window and high locus specificity [69]. This development provides a powerful tool for correcting pathogenic point mutations that were previously inaccessible to conventional adenine base editors. Together, these transversion editors significantly broaden the types of point mutations that can be corrected.

Dual‐function base editors enable simultaneous A•T‐to‐G•C and C•G‐to‐T•A conversions within the same target DNA region. These tools are constructed by fusing both an adenine deaminase (like evolved TadA variants) and a cytidine deaminase (such as PmCDA1 or hA3A) to a nCas9 [70, 71]. Tools such as SPACE, A&C‐BEmax, STEMEs and pDuBE1 allow for making multiple different base changes simultaneously (like C•G to T•A and A•T to G•C) or generating intricate patterns of edits within targeted DNA segments [72]. A&C‐BEmax demonstrates robust cytosine editing with substantial adenine activity while reducing RNA off‐target effects, whereas SPACE, STEMEs and pDuBE1 facilitate programmable multi‐base substitutions, providing precise control over intricate DNA modifications.

The emergence of TadA‐derived CBEs (TadCBEs) represents a major advance in precision genome editing. Conventional CBEs such as BE3 and BE4 suffer from broad sequence context dependence, high frequencies of bystander edits and substantial Cas‐independent off‐target DNA and RNA deamination. To overcome these limitations, researchers repurposed TadA as a cytidine deaminase through directed evolution and structure‐guided mutagenesis. TadA variants, including the hyperactive TadA‐8e, were subjected to PACE and iterative site‐directed mutagenesis to shift substrate preference from adenine to cytosine, and the resulting TadCBEs were fused to nCas9 to generate programmable CBEs. Unlike APOBEC1, TadA's compact size (~166 amino acids) makes it favourable for delivery in therapeutic settings, including packaging into size‐constrained vectors such as Adeno‐associated viruses (AAVs) [73]. These engineered TadA mutants retain the structural fold of adenine deaminases but incorporate active site mutations that reconfigure base recognition and catalytic activity towards cytosine substrates. TadCBEs demonstrate comparable or higher editing efficiencies relative to APOBEC1‐CBEs, achieving 56% efficiency at target cytosines with clean editing windows typically spanning positions 4–8 relative to the PAM [73]. Importantly, transcriptome‐wide analyses have shown that TadCBEs exhibit negligible RNA off‐target editing, a significant improvement over APOBEC1‐CBEs. Moreover, their applicability extends to plant systems, where TadCBEa, TadCBEd and TadCBEd_V106W mediated efficient and precise C‐to‐T conversions with editing frequencies ranging from 4.5% to 55% [74].

Beyond single‐function CBEs, TadA engineering also enabled the development of dual editors such as TadDE, which simultaneously mediates C‐to‐T and A‐to‐G conversions within the same protospacer window, enabling multiplexed mutagenesis from a single editing event [74]. Similarly, CABE‐T (CABEs that use engineered variants of TadA) variants permit programmable cytosine and adenine editing with high precision, expanding both research and therapeutic applications, achieving ~53% average C‐to‐T editing efficiency with < 2% unintended A‐to‐G substitutions [75]. Together, these TadA‐derived CBEs combine high efficiency and precision with broad applicability across mammalian systems, animal models and crops, offering promising potential for therapeutic and agricultural genome engineering.

Recent developments have introduced glycosylase‐based base editors (gGBEs, gTBEs and gCBEs), which utilize engineered DNA glycosylases fused with nCas9 to enable direct base editing that incorporates DNA glycosylase activity to expand editing outcomes beyond conventional deaminase‐based systems [76]. These editors offer precise control over base conversions, expanding the repertoire of editable sites in the genome.

The recently developed deaminase‐free guanine base editor, gGBE, was generated by fusing nCas9 with an engineered human MPG. The gGBE system uses the engineered MPG to specifically recognize and excise guanine, generating an apurinic/apyrimidinic (AP) site at the target locus. This abasic site is subsequently processed by endogenous DNA repair pathways, primarily TLS and BER, which insert a new base, typically T or C, at the damaged position. As a result, the original G•C base pair can be converted into C•G or T•A, enabling G‐to‐Y (Y = C or T) editing [77]. In mammalian cells, gGBE has achieved editing efficiencies of up to ~80% at certain genomic loci. Glycosylase‐Based Thymine Base Editors (gTBEs) employ a similar strategy, utilizing engineered thymine DNA glycosylase (TDG) variants fused to nCas9 to excise thymine bases, enabling T•A to G•C or T•A to C•G or T•A to A•T conversions [78]. In human cells, gTBE has demonstrated T‐to‐G editing efficiencies up to 48.7% and T‐to‐C editing up to 26.2%, with editing primarily occurring within positions 2–6 of the protospacer and maximal activity around position 3. Glycosylase‐Based Cytosine Base Editors (gCBEs) utilize engineered human UNG variants fused to nCas9 to generate abasic sites at targeted cytosines, thereby facilitating precise C•G to G•C transversion via the BER pathway. In mammalian cells gCBEs have achieved C‐to‐G editing efficiencies ranging from ~32% to ~78% across endogenous genomic loci [79]. Additionally, in rice, gCBEs utilizing either engineered hUNG or hUNG2 variants in conjunction with nCas9 have demonstrated C‐to‐G editing efficiencies ranging from 26.1% to 61.1%, with predominant C‐to‐G conversions reaching 58.3% and negligible off‐target activity [80]. These glycosylase‐based editors provide a versatile toolkit for precise base editing across all four nucleotides, offering advantages such as reduced off‐target effects and the ability to edit bases previously inaccessible to traditional deaminase‐based editors. Their development marks a significant advancement in the field of genome engineering, expanding the potential for therapeutic interventions targeting a broader range of genetic mutations.

Genome editing within mitochondria and chloroplasts was historically difficult due to challenges in delivering CRISPR components, particularly gRNAs, into these organelles [81]. A major advancement came with CRISPR‐free base editors. These systems use alternative DNA‐targeting proteins, like TALEs or Zinc Finger Proteins, fused to deaminase enzymes, bypassing the need for gRNAs. For instance, DdCBEs utilize TALE proteins linked to split DddA cytidine deaminase to achieve C•G‐to‐T•A editing in mitochondrial DNA (mtDNA). Subsequently, mitochondrial ABEs (such as dTALED and mTALED) were created by integrating TadA variants into this TALE‐based framework to perform A•T‐to‐G•C edits in mtDNA [82]. These organellar editors open up new research and therapeutic possibilities for mitochondrial diseases and have also been adapted for editing plant chloroplast genomes.

RNA base editing allows for temporary and reversible alterations in gene expression because it targets RNA molecules, in contrast to genome editing techniques that permanently change the DNA [43]. Typically, these editors utilize an RNA‐targeting CRISPR system, like dCas13, fused to an RNA deaminase enzyme. Fusing adenosine deaminases (Adenosine Deaminase Acting on RNA (ADARs)) or engineered versions to dCas13 (e.g., dCas13b) enables A‐to‐I editing (interpreted as G), as demonstrated by the REPAIR (RNA Editing for Programmable A to I Replacement) system [43]. Versions like REPAIRv2 were developed to improve editing specificity and reduce off‐target effects. Alternatively, systems like RESCUE (RNA Editing for Specific C to U Exchange) use cytidine deaminases (e.g., APOBEC) to achieve C‐to‐U editing. The range of possible RNA modifications has recently expanded to include A‐to‐m6A and U‐to‐pseudouridine (Ψ), with newer systems like RtABE (RNA transformer Adenosine Base Editor) enhancing editing precision [83].

Although substantial progress has been made in developing diverse classes of base editors with expanded capabilities, the most mature and extensively optimized platforms remain CBEs and ABEs. These editors have undergone extensive engineering to improve their efficiency, target scope and specificity across multiple biological systems. In contrast, many of the recently developed base editing systems capable of generating transversion mutations or broader nucleotide substitutions are still in relatively early stages of development [76, 84]. While these emerging tools significantly expand the potential scope of genome editing, their efficiency, editing precision and levels of unintended byproducts remain important challenges that require further Optimisation before they can achieve the same level of reliability and therapeutic readiness as CBEs and ABEs.

Viral vectors utilize viruses' inherent capacity to infect cells to deliver genetic material. However, different types of viral vectors offer distinct benefits and drawbacks. AAVs are a top choice for in vivo gene delivery, valued for their low immunogenicity, ability to infect non‐dividing cells and predictable patterns of tissue targeting [26]. However, their significant limitation is a small packaging capacity (around 4.7 kb) [96], which struggles to accommodate large base editor systems based on SpCas9 (over 4 kb). Because of this size constraint, base editor constructs often approach or exceed the packaging capacity of single AAV vectors. To circumvent this size limit, dual‐AAV strategies are commonly used, splitting the base editor components across two vectors [62]. Each fragment is fused to a part of a split intein; upon co‐delivery into a cell, these inteins mediate protein splicing to reassemble the complete, functional editor protein. Despite these advances, dual‐vector systems require efficient co‐transduction and intracellular reconstitution of the editor, which may reduce overall editing efficiency and complicate therapeutic implementation. This dual‐vector approach has successfully enabled in vivo base editing in various tissues (including liver, muscle, brain, eye, inner ear) in mouse models of diseases like progeria, SMA, DMD, ALS, LCA, hereditary deafness and PKU [27]. Despite this success, AAV delivery often results in prolonged expression of the base editor from persistent episomal DNA. This raises concerns about potential cumulative off‐target editing and the risk of immune responses against the Cas9 protein or AAV capsid, which could eliminate edited cells or preclude repeat administrations [61]. Lentiviruses (LVs) provide a substantial cargo capacity (~10 kb), suitable for delivering intact base editors and multiple gRNAs [62]. However, their integration into the host genome poses significant safety risks like insertional mutagenesis and genotoxicity, currently hindering their in vivo clinical application [97]. LVs are still valuable preclinical tools, though, as shown by their use in correcting the Rpe65 gene with an ABE in a mouse model of LCA [27]. Adenoviruses (AdVs) offer a very large cargo capacity, up to 36 kb, making them capable of carrying sizable or intricate base editor systems. However, their tendency to provoke strong immune responses (high immunogenicity) limits their usefulness for applications requiring systemic delivery [98]. Despite this limitation, AdVs have been utilized in some specific preclinical studies exploring base editing technique [27, 99].

LNPs are a leading non‐viral system for delivering nucleic acids, including the components needed for base editing [100]. A typical LNP formulation includes four key elements: an ionizable lipid essential for releasing the cargo from cellular compartments (endosomal escape), helper lipids, cholesterol and a PEG‐lipid to improve stability and reduce detection by the immune system [63]. LNPs offer significant advantages compared to viral vectors [101]. These include generally lower immunogenicity, the flexibility to deliver different types of cargo like mRNA or pre‐assembled RNP complexes (Cas protein plus gRNA), suitability for repeated administrations and the promotion of transient base editor expression [36]. Delivering the editor as an RNP complex, in particular, results in short‐lived activity, which minimizes the risk of off‐target edits and avoids concerns related to integrating genetic material into the host genome. Nevertheless, non‐viral delivery systems still face important challenges related to delivery efficiency, tissue specificity and long‐term safety, which remain areas of active investigation for therapeutic base editing applications.

LNPs efficiently deliver their contents into the cytoplasm of target cells [39]. They possess a natural tendency to accumulate in the liver, largely due to interactions involving apolipoprotein E (ApoE) binding and uptake using the low‐density lipoprotein receptor (LDLR) pathway. This inherent liver tropism makes them particularly effective for hepatic gene editing, a property leveraged by companies like Verve and Beam in programs targeting genes such as PCSK9, ANGPTL3, G6PC and SERPINA1. To enable targeting beyond the liver, current research focuses on engineering LNPs. This includes developing tailored lipid compositions (like SORT LNPs designed for lung and spleen delivery) and adding specific molecules (such as antibodies) to the LNP surface to direct them towards particular cell types, like T cells [102]. Beyond established viral and LNP systems, researchers are actively investigating a range of other strategies to deliver base editing tools, aiming to expand the options for therapeutic application [99].

Virus‐Like Particles (VLPs) are non‐infectious particles derived from viral scaffolds (e.g., retroviruses) engineered to package and deliver protein or RNP cargos, including base editors. VLPs offer the potential for transient delivery of the pre‐assembled base editor RNP complex, minimizing off‐target risks associated with prolonged expression from DNA or mRNA templates [103]. Consequently, directed evolution strategies are being used to improve VLP packaging and delivery efficiency. Furthermore, ENVLPE (Engineered Nucleocytosolic Vehicles for Loading of Programmable Editors) is a cutting‐edge gene‐editing delivery platform that employs non‐infectious VLPs to efficiently deliver gene‐editing tools such as CRISPR‐Cas9, base editors and prime editors into target cells. This advanced VLP‐based system offers programmable, non‐viral delivery and significantly enhances editing efficiency in challenging cell types and in vivo models [104].

Extracellular Vesicles (EVs) are naturally formed nanoscale particles that facilitate communication between cells. Their biocompatibility, low immunogenicity and capacity to cross biological barriers make them attractive candidates for non‐invasive delivery of base editing cargo [105]. Delivering the base editor as an RNP offers the most transient exposure, further enhancing safety by limiting the time window for off‐target activity [106]. Physical methods, such as electroporation (frequently used in ex vivo cell therapy) or nanoneedles, enable the direct introduction of pre‐assembled base editor RNP complexes into cells [107]. This offers highly controlled delivery and transient editor activity. However, difficulties in achieving broad tissue access and efficiency generally limit these techniques to modifying cells outside the body (ex vivo) or for localized treatments in vivo. While electroporation is commonly used for RNP delivery ex vivo, in vivo delivery typically requires packaging within vehicles like LNPs or VLPs. Recent studies suggest packaged RNP delivery (e.g., via enveloped delivery vehicles, EDVs) can be significantly more efficient than electroporation at lower doses [108].

Certain delivery strategies, such as hydrodynamic injection, are primarily used in preclinical settings to establish proof‐of‐concept for in vivo genome editing but are not directly translatable to clinical applications. Hydrodynamic tail‐vein injections have been widely used in preclinical studies as an experimental delivery strategy for liver‐directed genome editing.

A representative example is hereditary tyrosinemia type I (HTI), a recessively inherited metabolic disorder affecting tyrosine metabolism, caused by a deficiency of the enzyme fumarylacetoacetate hydrolase (FAH) due to mutations in the FAH gene [110, 111]. This leads to the buildup of harmful metabolites (fumarylacetoacetate, succinylacetone) causing severe liver damage, renal tubulopathy and increased risk of hepatocellular carcinoma. Mouse models of HTI, often carrying a G•C to A•T point mutation in the last nucleotide of Fah exon 8 causing splicing defects, are widely used for testing liver‐directed therapies.

The potential for in vivo correction of this Fah splice‐site mutation was explored in adult HTI mice employing the ABE6.3 base editor [112]. The editor, encoded on plasmid DNA along with the sgRNA, was delivered systemically via hydrodynamic tail‐vein injection, a method that efficiently targets hepatocytes. Deep sequencing revealed an average A•T to G•C correction efficiency of ~9.5% in liver DNA, with very low indel formation (0.05%). Despite this relatively modest initial editing rate, the treatment led to significant phenotypic rescue. Following the withdrawal of the protective drug nitisinone, ABE‐treated mice maintained body weight, whereas control mice experienced rapid weight loss. Immunohistochemistry showed numerous patches of FAH‐positive hepatocytes throughout the liver parenchyma of treated mice. This occurs because corrected hepatocytes gain a strong selective survival and proliferative advantage in the diseased liver environment, allowing them to clonally expand and repopulate the organ over time. RT‐PCR confirmed partial restoration of correctly spliced Fah mRNA, and some treated mice survived long‐term without nitisinone. Another study demonstrated successful correction and therapeutic benefit using ex vivo ABE treatment of chemically derived hepatic progenitors followed by transplantation into HTI mice [113].

The HTI studies exemplify how even partial in vivo correction can lead to profound therapeutic benefit in the context of liver diseases that favour the proliferation of corrected cells. This suggests that achieving extremely high editing efficiencies may not be necessary for all indications, potentially lowering the threshold for clinical success.

Ex vivo editing of patient‐derived or donor‐derived cells represents one of the most clinically advanced delivery paradigms for base editing therapies. In this approach, target cells are isolated, genetically modified outside the body under controlled conditions and subsequently reinfused into the patient. This strategy allows precise control over editing efficiency, quality control and safety assessment prior to transplantation. Ex vivo approaches are particularly suitable for haematological diseases, where haematopoietic stem and progenitor cells (HSPCs) can be readily isolated, edited and reintroduced following conditioning regimens.

One of the most extensively studied applications of ex vivo base editing is in sickle cell disease (SCD). SCD is an inherited blood condition caused by an A•T to T•A transversion (GAG to GTG) in the β‐globin gene (HBB), causing glutamic acid to be replaced by valine at position six of the β‐globin protein (E6V). This mutation produces sickle haemoglobin (HbS), which polymerizes under hypoxic conditions, deforming red blood cells (RBCs) into a sickle shape. These deformed cells lead to vaso‐occlusion, chronic hemolytic anaemia and progressive organ damage. Base editing presents two main therapeutic strategies for SCD: direct correction of the pathogenic mutation or reactivation of fetal haemoglobin (HbF) expression.

One strategy involves employing an ABE to change a pathogenic T•A base pair into a C•G pair by targeting the adenine specifically on the template strand [104]. Although this does not restore the wild‐type sequence (a T>A transversion inaccessible to standard ABEs), it introduces a benign variant known as haemoglobin Makassar. Using an ex vivo approach, haematopoietic stem cells (HSCs) isolated from SCD patients or an SCD mouse model were treated with the ABE, delivered as mRNA or RNP complexes, prior to transplantation into recipient mice. Editing efficiencies reached up to 80% in patient‐derived HSCs. Post‐transplantation, edited cells demonstrated successful engraftment with stable editing levels (~68% in mouse HSCs at 16 weeks). Phenotypic analysis revealed normalization of hematologic parameters, reduced RBC sickling and resolution of hallmark SCD symptoms such as anaemia and splenomegaly. Notably, correcting more than 20% of alleles was sufficient to confer therapeutic benefit in the mouse model.

An alternative therapeutic strategy involves reactivating fetal haemoglobin (HbF, α₂γ₂), which interferes with HbS polymerization and mitigates SCD symptoms. This is achieved by disrupting repressor binding motifs such as those for GATA1 within the erythroid‐specific enhancer of the BCL11A (B cell lymphoma/leukaemia 11A gene) gene, a key modulator of the fetal‐to‐adult haemoglobin switch. Recent studies have utilized base editors (CBEs, ABEs, or dual editors) delivered ex vivo via electroporation of RNPs or RNA transfection to introduce precise nucleotide substitutions within the +58 kb or +55 kb enhancer regions in SCD patient‐derived HSPCs [39]. These approaches achieved efficient single‐site or multiplex editing (up to ~70% for multiplexed edits) with minimal DSB formation or large genomic rearrangement issues commonly associated with Cas9 nuclease‐based editing [114]. Base editing resulted in robust HbF reactivation, with HbF protein levels reaching ~29% in vitro, a level considered therapeutically effective. Edited cells displayed reduced sickling under hypoxia and retained high engraftment capacity in xenotransplantation models, with sustained HbF expression in vivo. Another strategy involves editing the promoter regions of the HBG1 and HBG2 genes to prevent repressor proteins from binding. This approach is designed to mimic naturally occurring HPFH mutations that lead to increased HbF levels [115].

This ex vivo editing paradigm has already progressed to clinical evaluation. BEAM‐101 is an investigational autologous HSC‐based therapy developed by Beam Therapeutics for the treatment of severe SCD. This therapy involves editing a patient's own HSCs ex vivo with an ABE. The edit introduces beneficial mutations, similar to those found in HPFH, into the HBG1/2 promoter to increase protective HbF levels [116]. The ABE enzyme and the specific sgRNA targeting the HBG1/2 promoters are delivered into the HSPCs using electroporation. Although the specific ABE variant used in the final BEAM‐101 product is proprietary, Beam Therapeutics has disclosed the development of advanced ABE versions, such as the ABE8 family, which demonstrate improved editing efficiency and specificity over earlier versions. The goal of BEAM‐101 editing is to trigger a strong, lasting and widespread reactivation of HbF production, leading to significant levels of HbF in the patient's mature red blood cells. Maintaining sufficiently high levels of HbF is predicted to impede the polymerization of HbS. This, in turn, should reduce red blood cell sickling, lessen chronic red blood cell destruction (hemolysis), prevent blood vessel blockages (vaso‐occlusion) and ultimately relieve the symptoms and complications of SCD.

The BEACON Phase 1/2 clinical trial (NCT05456880↗) is an open‐label, single‐arm, multicenter study designed to assess the safety and efficacy of BEAM‐101 in patients with SCD suffering from severe vaso‐occlusive crises (VOC) [117]. The ongoing BEACON Phase 1/2 trial has enrolled over 35 patients (8 dosed by late 2024) and recently expanded to adolescents [118]. Initial results presented in late 2024 indicated successful engraftment, increased HbF production, and improved red blood cell function. Updated results are anticipated by mid‐2025 [95]. These updates will provide further insights into the durability of HbF induction, the impact on clinical outcomes like VOCs and transfusions and the continued safety profile of BEAM‐101. While BEAM‐101's primary indication is SCD, its HbF induction strategy is directly applicable to β‐thalassaemia, as both SCD and β‐thalassaemia are hemoglobinopathies, and elevating HbF levels can ameliorate symptoms in both conditions [119].

A similar ex vivo genome editing strategy is also being applied to β‐thalassaemia, a group of inherited blood disorders manifested by absent or decreased production of beta‐globin chains in haemoglobin, resulting in ineffective erythropoiesis and anaemia of varying severity. Management of transfusion‐dependent β‐thalassaemia (TDT) usually requires lifelong blood transfusions and iron chelation therapy, both of which impose substantial burdens and pose risks of long‐term complications. Allogeneic HSCT provides a potential cure but is constrained by limited donor availability and risks related to immune rejection. Recent progress in genome editing, especially CRISPR‐based technologies, holds promise for transformative, one‐time curative treatments through modification of a patient's own HSCs [120]. Within these technologies, base editing has stood out as an especially promising method [121].

The CS‐101 clinical trial (NCT06024876↗) is a Phase 1, open‐label study assessing the safety and effectiveness of an autologous CD34⁺ haematopoietic stem cell therapy genetically engineered using cytosine base editing. Utilizing a novel transformer Base Editor (tBE), the protocol introduces precise single‐nucleotide edits in the HBG1/2 promoters to disrupt the BCL11A binding motif, thereby derepressing γ‐globin expression and elevating HbF levels in patients with TDT. Interim results reported in 2024 for the first six treated subjects demonstrated robust haematopoietic engraftment, transfusion independence, a marked increase in total haemoglobin, and HbF predominance, without treatment‐related serious adverse events [122]. The trial represents the first clinical application of base editing in β‐hemoglobinopathies and could provide a durable, potentially curative therapy for TDT.

The ex vivo editing framework is also being expanded beyond hemoglobinopathies to engineered cell therapies in oncology. T‐cell acute lymphoblastic leukaemia (T‐ALL) and T‐cell lymphoblastic lymphoma (T‐LL) are aggressive haematological cancers that arise from T‐lymphocyte precursors. Despite significant advances in frontline therapies leading to improved survival rates, outcomes for patients with relapsed or refractory (R/R) T‐ALL/T‐LL remain poor [123]. Although chimeric antigen receptor (CAR) T‐cell therapy has revolutionized the treatment of B‐cell malignancies, applying it to T‐cell cancers presents unique challenges. These hurdles include the difficulty of selecting target antigens found on malignant T‐cells but largely absent from normal immune cells, as well as the problem of CAR‐T cell fratricide, where the therapeutic cells eliminate each other due to mutual expression of the target antigen. Moreover, minimizing the risk of Graft‐versus‐Host Disease (GvHD) remains a key concern when employing allogeneic (donor‐derived) T cells. Allogeneic CAR‐T therapies offer the potential for an ‘off‐the‐shelf’ product, addressing the manufacturing limitations associated with autologous therapies. However, achieving this potential requires sophisticated genetic engineering to ensure both safety and efficacy. BEAM‐201, developed by Beam Therapeutics, is a next‐generation allogeneic CAR‐T therapy that utilizes CBE to disrupt multiple genes in donor‐derived T cells, addressing major challenges in treating R/R T‐ALL and T‐LL. Through precise CBE‐mediated editing of four target genes, it aims to generate a potent, safe and universally applicable “off‐the‐shelf” product targeting CD7‐positive T‐ALL/T‐LL [124]. The multiplex editing strategy enhances both the safety and efficacy of the therapy by targeting key genes in the CAR‐T cells. The CD7 knockout prevents fratricide by removing CD7 expression on the CAR‐T cells, allowing them to effectively target CD7‐positive T‐ALL/T‐LL cells without harming each other. The TRAC (T Cell Receptor Alpha Constant) knockout reduces the risk of GvHD by disrupting the TCR complex, preventing donor T cells from attacking the recipient's tissues. The CD52 knockout resists lymphodepletion therapy, ensuring that BEAM‐201 cells are not depleted by treatments like alemtuzumab. Lastly, knocking out PDCD1 reduces T cell exhaustion by removing the PD‐1 receptor, thereby blocking inhibitory signals and improving CAR‐T cell persistence and activity within the tumour microenvironment. Together, these edits optimize BEAM‐201's safety, persistence, and therapeutic potential in treating CD7‐positive T‐ALL/T‐LL. BEAM‐201 is currently undergoing evaluation in a Phase 1/2, multicenter, open‐label clinical trial (NCT05885464↗) targeting patients with R/R T‐ALL or T‐LL. The trial includes dose‐exploration and dose‐expansion cohorts for adults, as well as a planned cohort for paediatric patients. A Phase 1/2 clinical trial is actively enrolling and treating patients, with early data shared in late 2024 [125].

BEAM‐201 demonstrates the promise of base editing technology in developing complex, multiplex‐engineered cell therapies. By precisely introducing four distinct gene edits, BEAM‐201 aims to create a more effective and potentially safer allogeneic CAR‐T therapy for T‐ALL/T‐LL, addressing the significant unmet need in relapsed/refractory cases. Early data from the ongoing Phase 1/2 trial show a manageable safety profile and promising signs of efficacy, supporting further investigation. Additional clinical evaluation is essential to fully determine BEAM‐201's safety, efficacy and durability and to define the future role of multiplex base editing in CAR‐T and other cell‐based therapies.

Together, these examples illustrate how ex vivo base editing approaches enable precise genome engineering in clinically accessible cell types such as haematopoietic stem cells and T cells. By allowing editing, quality control and expansion outside the body, this delivery paradigm provides a controlled and scalable pathway towards therapeutic genome editing.

For diseases requiring in vivo correction, AAVs remain one of the most widely used delivery platforms due to their ability to achieve efficient and tissue‐specific gene transfer. AAV‐mediated delivery has been widely explored for base editing applications targeting tissues such as the liver, muscle, retina and the central nervous system. However, the therapeutic use of AAVs is strongly shaped by tissue‐specific delivery demands and by the restricted cargo capacity of the vector, which is approximately 4.7 kb, whereas many base editors exceed 5–6 kb. As a result, AAV‐based base editing strategies generally fall into two categories: single‐AAV systems enabled by compact editor architectures, and dual‐AAV systems that rely on split‐intein or related reconstitution strategies to accommodate larger constructs. These complementary approaches have provided valuable platforms for evaluating base editing across diverse disease models.

Single‐AAV delivery strategies are enabled by the use of compact Cas variants, optimized base editor architectures or favourable construct design that allows the entire editing system to be packaged within a single vector. This approach simplifies vector design, reduces manufacturing complexity and facilitates more efficient in vivo delivery.

This strategy is exemplified in lysosomal storage diseases (LSDs), where early and efficient systemic delivery is critical for therapeutic efficacy. LSDs are inherited metabolic disorders marked by the buildup of undigested substrates inside lysosomes, caused by deficiencies in particular enzymes, resulting in cellular dysfunction and damage in multiple organs. A preclinical study demonstrated the potential of in utero adenine base editing to correct genetic mutations responsible for LSDs. In particular, the researchers focused on correcting the G to A (W392X) mutation in the Idua gene of a mouse model for Mucopolysaccharidosis type I (MPS‐IH), commonly known as Hurler syndrome [126]. This mutation causes a deficiency of α‐L‐iduronidase, leading to the buildup of glycosaminoglycans (GAGs) in multiple organs and resulting in severe systemic disease. Utilizing AAV9 vectors, the team delivered an ABE to the fetal mice at embryonic day 15.5. The ABE effectively corrected the W392X mutation in hepatocytes and cardiomyocytes, while a lower level of editing was detected in the brain. This correction correlated with marked improvements in survival rates and alleviation of metabolic, musculoskeletal and cardiac defects. Importantly, prenatal treatment reduced GAG buildup and improved cardiac function, highlighting the therapeutic promise of early genetic correction. This study demonstrates the potential of in utero base editing as a viable approach for treating genetic diseases prior to the appearance of clinical symptoms. Although this proof‐of‐concept study provides encouraging insights, additional research is needed to evaluate the safety, effectiveness and long‐term effects of these interventions in humans.

Similarly, single‐AAV delivery has been successfully applied in neuromuscular disorders such as Duchenne muscular dystrophy (DMD). DMD is a serious and advancing genetic condition linked to the X chromosome, resulting from mutations in the DMD gene responsible for producing dystrophin, a critical protein for muscle function. Mutations such as deletions, duplications, and point mutations cause the loss or malfunction of dystrophin, which leads to fragile muscle fibres, persistent inflammation fibrosis, and progressive muscle degeneration. Approximately 15%–20% of DMD occurrences are due to nonsense point mutations that introduce premature termination codons (PTCs), halting dystrophin translation [127]. ABE technology is well‐suited to correct many of these nonsense mutations (e.g., TAA>CAA, TAG>CAG, TGA>CGA) [128].

Several studies have explored ABE‐mediated correction of DMD nonsense mutations. An ABE fused to the SpG Cas9 variant (SpG‐ABE) which recognizes NGN PAMs was utilized to expand targeting possibilities [129]. Specific nonsense mutations (c.4174C>T creating a TAA stop codon; c.2977C>T creating a TGA stop codon) were targeted in DMD patient‐derived iPSC‐cardiomyocytes and in humanized mouse models carrying these mutations. AAV vectors were used for delivery, either via local intramuscular injection or systemic administration in neonatal mice. Efficient A•T to G•C editing was achieved in vitro (~58% in iPSCs) and in vivo (e.g., ~13% genomic editing in muscle after local delivery, ~37% in heart after systemic delivery). This correction led to substantial recovery of dystrophin expression in cardiomyocytes in vitro and in heart and skeletal muscle fibres in vivo (e.g., > 90% positive fibres in the heart after systemic delivery). Importantly, systemic ABE treatment in neonatal mice resulted in significant improvements in motor performance.

The nonsense mutation (CAA>TAA) in exon 53 of the commonly used mdx4cv mouse model was targeted for correction in a separate research using an improved ABE coupled to an NG‐PAM recognizing Cas9 (iABE‐NG) [127]. Systemic delivery via AAV9 in adult mice led to high levels of T•A to C•G editing at the target site, particularly in the heart, where editing reached ~85% at the mRNA level after 10 months. Consequently, dystrophin protein expression was restored to near‐complete levels in the heart, accompanied by significant enhancements in muscle contractile function.

Modification of splice donor sequences using base editing to induce exon skipping was demonstrated as an approach to recover the DMD gene reading frame in cases of exon deletions [130]. The researchers employed ABEs at key positions within splice donor sites, leading to the skipping of exon 50 or 45, depending on the mutation. This targeted editing successfully restored an in‐frame dystrophin transcript in patient‐derived iPSC‐cardiomyocytes and in a ΔEx51 mouse model. Using AAV9 vectors for in vivo delivery, the team achieved dystrophin restoration and phenotypic rescue in muscle tissue. The study demonstrated that base editing can precisely reprogram splicing patterns to correct frame‐disrupting mutations in DMD.

Systemic single‐AAV delivery has also demonstrated strong therapeutic potential in multisystem disorders such as Hutchinson–Gilford progeria syndrome (HGPS). HGPS is a rare and severe genetic disorder characterized by premature aging, mainly resulting from a single C•G to T•A point mutation (c.1824C>T, G608G) in the LMNA gene encoding nuclear lamin A. The mutation in LMNA exon 11 creates an abnormal splice site, leading to the production of a toxic truncated protein called progerin. Progerin accumulates in cells and drives disease pathology, with particularly severe effects on the vasculature. A landmark study demonstrated the clinical potential of in vivo base editing using a mouse model of HGPS that was homozygous for the human LMNA c.1824C>T mutation [109]. The researchers employed an ABE variant (ABE7.10max‐VRQR), designed to recognize an NGA PAM sequence adjacent to the mutation, to revert the pathogenic T•A base pair back to the wild‐type C•G. A single systemic administration of AAV9 encoding the ABE and a gRNA was delivered at postnatal day 14. Six months after treatment, substantial correction of the mutation was observed across multiple tissues, with editing efficiencies ranging from approximately 20%–60% and reaching ~23% in the critical aortic tissue. This genomic correction corresponded with reduced levels of aberrantly spliced LMNA mRNA and diminished progerin protein expression. Phenotypically, ABE‐treated mice showed striking improvements. The therapy preserved vascular smooth muscle cells (VSMCs) in the aorta and prevented adventitial fibrosis—two key pathological features of HGPS [131]. Most notably, a single ABE injection improved the median survival time of progeria mice from 215 days in untreated controls to 510 days—a 2.4‐fold increase approaching the age associated with the onset of old age in healthy mice. This study provided compelling evidence that in vivo base editing can directly repair the causative genetic defect in HGPS and deliver unprecedented therapeutic benefits in a preclinical model.

Beyond monogenic disorders, single AAV‐delivered base editing has also been explored in oncology models. In the context of hepatocellular carcinoma (HCC), a prevalent mutation in the TERT promoter, specifically the −124 C>T alteration, leads to increased telomerase activity and contributes to tumorigenesis. Researchers have developed a sgRNA‐guided Campylobacter jejuni Cas9 (CjCas9)‐fused ABE system capable of efficiently converting this mutation back to the wild‐type sequence in HCC cells [132]. This correction results in decreased TERT expression, reduced telomerase activity and inhibited tumour cell proliferation. In vivo experiments have shown the therapeutic promise of this base editing method. Delivery of the ABE system via AAV vectors to mouse models harbouring the TERT promoter mutation led to significant inhibition of liver tumour growth. The treated mice exhibited reduced tumour volumes and improved survival rates, emphasizing the efficacy of base editing in correcting oncogenic mutations and halting tumour progression. The therapeutic effect is driven by the targeted deamination of the mutated adenine to inosine leading to a permanent A‐to‐G base substitution. This targeted correction restores the normal transcription factor binding sites in the TERT promoter, thereby normalizing TERT expression levels. Likewise, local injection of AAVs delivering sgRNA‐guided CjABE into mouse models with the TERT promoter mutation led to significant suppression of brain tumour growth [133]. By leveraging the specificity and efficiency of ABEs, this approach provides a promising strategy for treating cancers driven by specific genetic mutations, opening new avenues for tailored and more efficient treatments.

In contrast to single‐AAV systems, many base editor constructs exceed the packaging capacity of AAV vectors, necessitating the use of dual‐AAV delivery strategies. In these approaches, the base editor is divided into two separate components, each packaged into an individual AAV vector, which reconstitute into a functional enzyme within the target cell through split‐intein or related protein reassembly mechanisms. While this strategy overcomes cargo size limitations and expands the range of targetable genomic loci, it introduces additional complexity, including the requirement for efficient co‐transduction of both vectors and potential variability in reconstitution efficiency.

This approach is exemplified in metabolic liver disorders such as phenylketonuria (PKU), where large base editor constructs require dual‐AAV delivery for effective in vivo correction.

PKU is a genetic condition resulting from mutations in the phenylalanine hydroxylase (PAH) gene, which causes increased phenylalanine levels and related neurological complications. A CBE approach was employed to correct a disease‐causing point mutation in the Pah gene associated with PKU [134]. Specifically, a split‐intein CBE system was utilized as a method to enable the delivery of the rather large base editor components through a dual AAV vector approach. When the AAV‐base editor system was delivered intravenously to adult Pah^enu2 mice, a model of human PKU, the treatment successfully corrected the mutation, lowering blood phenylalanine levels to within the normal physiological range below 120 μmol/L. Significantly, mRNA correction rates reached as high as 63%, resulting in restored PAH enzyme function and the reversal of the light fur phenotype typical of PKU in these mice. These results highlight the practicality of employing AAV‐mediated base editing for in vivo correction of genetic mutations, presenting a promising therapeutic strategy for metabolic liver disorders such as PKU.

Similarly, dual‐AAV delivery has been applied in a gene therapy strategy used CBEs to silence the activity of the mutant SOD1 gene by introducing a premature stop codon [135]. Because of the restricted loading capacity of AAV vectors, the researchers implemented an intein‐based split method to deliver the base editor in two parts using dual AAVs, which reassemble in vivo. Following intrathecal injection of these dual AAVs into G93A‐SOD1 ALS mice, the split‐intein CBE successfully introduced a nonsense mutation in the target gene, causing a significant delay in disease progression. Treated mice showed prolonged survival, slower muscle atrophy, better neuromuscular function, reduced muscle denervation, and up to 40% fewer toxic SOD1 protein aggregates compared to untreated controls. This investigation emphasizes the capabilities of base editing specifically, split‐intein‐mediated delivery of CBEs as a viable therapeutic strategy for ALS and possibly other genetic diseases, overcoming vector size limitations while achieving precise, permanent gene silencing.

Dual‐AAV systems have also been successfully utilized in inherited retinal diseases such as Leber congenital amaurosis (LCA), where localized delivery enables efficient reconstitution of split base editors. LCA is a serious inherited retinal disorder caused by mutations in genes such as RPE65, resulting in early‐onset vision impairment. Two pivotal studies have examined the implementation of ABEs to correct mutations in the Rpe65 gene, a typical reason for LCA, using the rd12 mouse model. While both studies employed ABEs delivered via AAV vectors, they employed distinct methodologies and targeted different aspects of visual function restoration, leading to varying outcomes.

The initial research centered on correcting the p.R44X nonsense mutation within the Rpe65 gene [136]. In vivo base editing was performed using dual AAV9 vectors carrying split ABEs linked by a trans‐splicing intein. Subretinal delivery of these vectors into rd12 mice induced an A‐to‐G conversion in RPE cells, leading to the restoration of wild‐type RPE65 mRNA and protein. This correction led to the recovery of light‐induced electrical responses in the retina, as measured by electroretinography (ERG), with sustained functional improvements observed for at least 3 months post‐treatment.

The objective of the second study was to protect cone photoreceptors, cells essential for colour perception and sharp vision [137]. The study involved laboratory screening (in vitro) to pinpoint the most effective ABE/sgRNA combinations for fixing the Rpe65 mutation. Delivery of the optimized ABE and sgRNA through split AAV2 vectors via subretinal injection corrected up to 40% of Rpe65 transcripts, restored cone‐mediated vision and preserved cone cells in rd12 mice. Additional analysis with single‐cell RNA sequencing showed increased expression of genes linked to cone phototransduction and survival pathways in the treated mice. These findings suggest that base editing not only corrects genetic mutations but also promotes the expression of genes critical for cone function and longevity. The study showed that base editing therapy can provide long‐term restoration of cone photoreceptor function and survival.

Together, these studies demonstrate that dual‐AAV strategies provide a practical solution to AAV cargo limitations, enabling the use of full‐length or less compact base editors across diverse tissues. However, compared to single‐AAV systems, they introduce additional delivery complexity and depend on efficient co‐transduction, which may affect overall editing efficiency and translational scalability.

LNPs have emerged as one of the most promising delivery platforms for in vivo base editing, particularly for liver‐targeted therapies. LNPs can efficiently encapsulate mRNA encoding base editors together with gRNAs and deliver them to hepatocytes following systemic administration. This transient delivery approach avoids long‐term nuclease expression and reduces the risk of integration‐related complications. The primary trade‐off in this paradigm involves delivery safety versus durable efficacy.

One of the most advanced applications of LNP‐mediated base editing targets is the PCSK9 gene (Proprotein Convertase Subtilisin/Kexin type 9), a well‐validated regulator of LDL‐C levels. The primary goal is to achieve durable LDL‐C reduction through permanent gene inactivation in the liver, offering a potential single‐treatment alternative to chronic therapies like antibodies or siRNA. ABEs are particularly suited for this application. These editors are often delivered via LNPs that efficiently target hepatocytes, aiming to introduce a specific base change at a critical PCSK9 splice site, thereby disrupting gene function and preventing the production of the PCSK9 protein.

Pivotal preclinical validation came from studies in non‐human primates (NHPs), primarily cynomolgus monkeys [138]. Researchers employed ABEs to precisely edit the PCSK9 gene in the livers of mice and cynomolgus macaques. Delivered via LNPs, the ABE system achieved up to 67% editing efficiency in mice and 34% in macaques. This resulted in a 95% decrease in plasma PCSK9 levels and a 58% reduction in LDL cholesterol in mice, and a 32% reduction in PCSK9 and 14% in LDL cholesterol in macaques. Importantly, the ABE mRNA was rapidly cleared from the system, and no unintended mutations were observed. However, an antibody‐mediated immune response against the ABE was observed upon re‐dosing in macaques. These results highlight the promise of ABE‐mediated gene editing as a therapeutic approach for treating monogenic liver diseases, such as familial hypercholesterolemia.

The use of in vivo CRISPR base editing aiming at the PCSK9 gene was also explored in primates as a strategy to reduce cholesterol levels [139]. Following a single intravenous infusion of LNPs carrying the base editor, an almost complete suppression of PCSK9 expression was achieved in the liver. This resulted in roughly a 90% decrease in circulating PCSK9 protein and about a 60% reduction in LDL cholesterol, with these effects remaining stable for at least 8 months. The findings support a promising one‐time intervention for lowering LDL cholesterol and managing atherosclerotic cardiovascular disease (ASCVD). Furthermore, this study offers strong proof‐of‐concept for using CRISPR base editors to perform precise single‐nucleotide modifications in disease‐related genes within the liver, and potentially in other tissues.

ASCVD encompasses conditions resulting from plaque accumulation in arterial walls, which narrows or blocks arteries and raises the risk of heart attack, peripheral artery disease and stroke [140]. Base editing, a technology advanced by companies such as Verve Therapeutics, presents a promising therapeutic strategy for ASCVD due to its capacity for precise and permanent correction of genetic variants associated with increased cardiovascular risk. The initial programs focus on three key genetic targets associated with elevated cholesterol and ASCVD: the PCSK9 gene to reduce LDL cholesterol, the ANGPTL3 (angiopoietin‐like protein 3) gene to lower triglycerides and remnant cholesterol and the LPA gene to decrease lipoprotein(a) levels [141, 142, 143].

The key strategy involves using insights from human genetics to mimic the protective effects of natural gene variants, like PCSK9 loss‐of‐function mutations known to cause lifelong low LDL‐C levels and lower ASCVD risk [144]. VERVE‐101, the first clinical candidate, employs an ABE delivered via LNPs to inactivate the PCSK9 gene in the liver by introducing a targeted A•T to G•C base pair substitution [145]. The therapeutic goal was to achieve permanent reductions in LDL‐cholesterol for patients suffering from Heterozygous Familial Hypercholesterolemia (HeFH) and established ASCVD. VERVE‐101 consists of two components: a gRNA that targets the splice donor site between exon 1 and intron 1 of the PCSK9 gene to introduce a premature stop codon, thereby halting the production of functional PCSK9 protein, and an mRNA encoding the ABE8.8 m enzyme [146]. This base editor comprises a modified SpCas9 protein fused to an engineered TadA adenosine deaminase domain.

The Phase 1b heart‐1 trial (NCT05398029↗), launched in 2022, provided the first proof‐of‐concept for in vivo base editing in humans and showed encouraging early outcomes in reducing LDL‐C levels [147]. Interim data from the human trial, released in late 2023, showed dose‐dependent decreases in blood PCSK9 protein and LDL‐C levels [148]. Notably, the single participant in the highest dose group (0.6 mg/kg) exhibited a 55% reduction in LDL‐C that was sustained for at least 18 months. However, safety issues emerged in this high‐risk patient group during the trial. Findings encompassed temporary elevations in liver enzymes, along with two serious cardiovascular events: a fatal cardiac arrest (concluded unrelated to treatment due to the patient's severe pre‐existing heart condition) and a heart attack with a possible connection to the treatment. Additionally, later laboratory tests identified low platelet counts (thrombocytopenia) and elevated liver function tests (LFTs), which were considered likely side effects of the particular LNP delivery system used in VERVE‐101 [149]. Due to these findings, enrollment in the heart‐1 trial was halted.

The occurrence of serious cardiovascular events illustrates the inherent complexity of conducting first‐in‐human gene editing trials in patients with severe, advanced ASCVD. In this high‐risk population, the frequent incidence of events such as myocardial infarction and cardiac arrest complicates the assessment of whether these outcomes are linked to the investigational therapy. Although the myocardial infarction soon after dosing was tentatively linked to the treatment, the fatal cardiac arrest occurring weeks later emphasizes the challenge of separating therapy‐induced events from the expected progression of severe cardiovascular disease. Investigations, which included nonclinical studies, determined that the observed laboratory abnormalities specifically elevated liver enzymes and thrombocytopenia were likely driven mainly by the LNP delivery system used in VERVE‐101, rather than the base editing components themselves.

Although regulatory filings for VERVE‐101 are still ongoing, greater emphasis is now being placed on VERVE‐102. This drug candidate employs the same PCSK9‐targeting base editor and gRNA as VERVE‐101 but is delivered through a different LNP system [150]. The ionizable lipid included in this alternative LNP formulation has a history of previous use and good tolerability in other clinical trials. Unlike LNPs relying solely on ApoE for uptake via the LDLR, the VERVE‐102 LNP incorporates the targeting ligand N‐acetylgalactosamine (GalNAc) to achieve more specific and efficient delivery to the liver [151]. This improved targeting could potentially allow for lower effective doses or enhance the therapeutic index by minimizing exposure to non‐target cells. VERVE‐102 is set to be evaluated in patients with HeFH and ASCVD in the Phase 1b Heart‐2 trial (NCT06164730↗).

Additionally, VERVE‐201, an investigational therapy which also uses LNP‐delivered base editing but targets the distinct ANGPTL3 gene, is being advanced through preclinical development [152]. The ANGPTL3 gene plays a central role in regulating cholesterol and triglyceride levels, contributing to the risk of ASCVD via a mechanism distinct from the PCSK9 pathway. ANGPTL3, a liver‐derived protein, inhibits two essential enzymes involved in lipid metabolism: lipoprotein lipase (LPL) and endothelial lipase (EL). By inhibiting LPL and EL, ANGPTL3 increases levels of triglycerides (TG), cholesterol‐rich remnant lipoproteins, as well as LDL‐C and HDL‐C. Suppressing ANGPTL3 protein production in the liver has been shown to lower both LDL‐C and TG concentrations [153]. The fact that individuals naturally lacking a working ANGPTL3 gene have much lower LDL‐C and triglycerides without obvious health problems indicates that targeting this gene for therapy could be both effective and safe [154]. VERVE‐201 employs the same GalNAc‐LNP delivery system utilized for VERVE‐102, which incorporates the clinically validated ionizable lipid and the GalNAc targeting ligand for efficient and specific hepatocyte delivery. Clinical trials for VERVE‐201 are anticipated to begin soon. Success in these programs could mark a new era in cardiovascular medicine, moving from chronic disease management to long‐term disease modification through precision genetic medicine.

Similarly, GSD1a is a rare inherited metabolic disorder with autosomal recessive inheritance, caused by mutations in the G6PC gene that result in deficient activity of the glucose‐6‐phosphatase (G6Pase) enzyme [155]. This enzyme deficiency disrupts the final stages of glycogenolysis and gluconeogenesis, impairing the body's ability to regulate glucose levels. GSD1a typically presents in early infancy with severe fasting hypoglycemia, lactic acidosis, hyperlipidemia, hyperuricemia and notable hepatomegaly and nephromegaly due to glycogen and fat buildup. Long‐term complications can include hepatic adenomas with a risk of malignant transformation, progressive renal disease, osteoporosis and gout. The R83C mutation in the G6PC gene, common in Caucasian populations including those of Ashkenazi Jewish descent, causes a complete loss of G6Pase enzyme activity, contributing to the severe clinical phenotype of GSD1a [115]. The GSD1a‐causing R83C mutation arises from a G>A transition on the coding strand of the G6PC gene. Consequently, the complementary non‐coding strand carries a C>T change at the corresponding position. To restore the wild‐type G6PC sequence, this aberrant T on the non‐coding strand must be converted back to a C. This requires changing the A•T base pair at the mutation site back to the original G•C base pair.

ABE technology is well‐suited for correcting the R83C mutation in GSD1a. BEAM‐301, developed by Beam Therapeutics and delivered via LNPs, uses ABE to revert the A•T base pair to the wild‐type G•C pair. This precise correction restores G6Pase enzyme activity and addresses the metabolic imbalances associated with GSD1a [100]. A Phase 1/2 clinical trial evaluating the dose, safety, and efficacy of BEAM‐301 (NCT06735755↗) is currently underway. This follows the FDA's clearance of the Investigational New Drug (IND) application for the therapy in the second quarter of 2024. BEAM‐301 holds significant potential as an advancement in treating GSD1a, shifting from palliative dietary management to a possible one‐time genetic cure. If successful, this in vivo base editing therapy could fundamentally alter the disease's progression, relieving patients from lifelong dietary restrictions and reducing the risk of severe hypoglycemia and long‐term organ complications.

Another liver‐targeted base editing therapy currently under development addresses alpha‐1 antitrypsin deficiency (AATD). AATD is a genetic metabolic disorder characterized by reduced circulating levels of alpha‐1 antitrypsin (AAT) protein, leading to an increased risk of serious lung and liver diseases [116]. AAT is a serine protease inhibitor (serpin) predominantly produced and secreted by liver cells. Its main role is to inhibit neutrophil elastase in the lungs, protecting the delicate alveolar tissue from proteolytic damage, particularly during inflammatory responses. Mutations in the SERPINA1 gene on chromosome 14q32 cause AATD, with the PiZ allele identified as the most clinically significant variant [117]. The PiZ allele results from a single point mutation (c.1096G>A) that substitutes glutamic acid with lysine at codon 342, causing the AAT protein (Z‐AAT) to misfold and accumulate in the liver, leading to hepatocyte stress, inflammation and an increased risk of cirrhosis and liver cancer. This misfolding also impairs AAT secretion, drastically reducing circulating AAT levels, which leaves the lungs vulnerable to neutrophil elastase, resulting in progressive alveolar damage, early‐onset emphysema and potentially bronchiectasis.

The PiZ mutation in AATD is a G>A transition. ABE technology is ideally suited for correcting this specific type of mutation [118]. Using in vivo base editing, BEAM‐302 is an investigational therapy from Beam Therapeutics designed to target the genetic origin of AATD. The therapeutic strategy centers on correcting the specific PiZ point mutation (G>A) in the SERPINA1 gene directly within the patient's liver cells [100]. The base editing components, mRNA encoding the ABE and gRNA targeting the PiZ mutation, are delivered via LNPs. After intravenous administration, the LNPs are designed to preferentially accumulate in the liver. Hepatocyte uptake is facilitated by ApoE adsorbed on the LNP surface, which interacts with LDLR on hepatocytes. Once internalized through endocytosis, the LNPs release their mRNA and gRNA cargo into the cytoplasm, where the mRNA is translated into the ABE protein. The ABE then binds to the gRNA to form a complex capable of editing the target DNA. Correction of the PiZ mutation reduces toxic Z‐AAT accumulation in hepatocytes and restores normal M‐AAT production. By targeting the gene's native locus, this approach enables physiologically regulated expression, boosts protective AAT levels and may provide a durable, one‐time treatment.

Building on promising preclinical results, a Phase 1/2 open‐label, multicenter clinical trial (NCT06389877↗) was initiated to assess BEAM‐302 in adults with AATD. The study comprises dose‐escalation (Phase 1) and dose‐expansion (Phase 2) phases, aimed at assessing the safety, tolerability, pharmacokinetics and therapeutic efficacy of single ascending doses, while determining the optimal biological dose (OBD) for further development. Dosing in the Phase 1/2 trial commenced in mid‐2024 [119]. Initial data from the first cohorts (released March 2025) showed dose‐dependent increases in functional AAT (reaching therapeutic levels), significant reductions in the mutant protein and a favourable safety profile (mild/moderate transient events) [156]. Ongoing studies focusing on dose Optimisation, long‐term outcomes, and patients with liver involvement will be key to defining its full clinical potential. If successful, BEAM‐302 could shift AATD management from ongoing symptomatic care to a durable, gene therapy solution.

The ultimate synthesis of base editing's modularity and delivery efficiency is seen in ultra‐rare, personalized cases. For a neonate with carbamoyl‐phosphate synthetase 1 (CPS1) deficiency, a personalized in vivo therapy named kayjayguran abengcemeran (k‐abe) was designed to correct the patient's paternal CPS1 variant (Q335X) [157]. The therapy, consisting of mRNA encoding the ABE and a custom gRNA delivered via LNPs, progressed from genetic diagnosis through preclinical validation in patient‐derived cells and animal models (including mouse efficacy and nonhuman primate toxicology studies) to IND application approval and clinical administration within approximately 7 months.

The patient received two IV infusions of k‐abe at approximately 7 and 8 months of age, starting with a 0.1 mg/kg dose followed by 0.3 mg/kg. Following treatment, the patient demonstrated notable clinical improvements, including the ability to tolerate increased dietary protein and a 50% reduction in the required dose of nitrogen‐scavenger medication. A significant outcome was the infant's ability to endure subsequent viral illnesses without developing hyperammonemic crises, which had previously been a substantial risk. While transient elevations in liver transaminases were observed, no serious adverse events occurred. Although long‐term safety and efficacy require further monitoring, this case establishes a new paradigm for the rapid development and deployment of customized gene‐editing therapies for individuals with rare or unique genetic diseases.

This study's primary significance lies in demonstrating the feasibility of developing and deploying a personalized, in vivo gene‐editing therapy with unprecedented speed within just 6 months of a patient's birth. This work validates a crucial proof‐of‐concept workflow, demonstrating a platform approach that enables the rapid creation of customized treatments for ultrarare genetic diseases by modifying only specific components, such as the gRNA. This success paves the way for a future where rapid deployment of patient‐specific gene‐editing therapies could become a routine strategy for managing a wide array of genetic disorders [158].

Taken together, these delivery platforms highlight fundamental trade‐offs that shape the clinical translation of base editing technologies. Preclinical approaches such as hydrodynamic tail‐vein injection provide valuable proof‐of‐concept for efficient in vivo editing but lack direct clinical applicability. Ex vivo strategies offer the highest level of control over editing outcomes, enabling precise quality assessment and selection of modified cells prior to reinfusion. However, they are limited to clinically accessible cell types and require complex manufacturing workflows. In contrast, in vivo delivery approaches expand therapeutic reach. AAV‐mediated systems enable efficient gene transfer across diverse tissues, including muscle, retina and the central nervous system, and have already demonstrated clinical feasibility through multiple approved gene therapies. However, they are constrained by limited cargo capacity and prolonged transgene expression, often necessitating dual‐vector strategies. LNP‐based delivery, particularly for liver‐targeted applications, provides a clinically scalable and non‐viral alternative, enabling transient expression of base editing components and broader patient applicability. Notably, LNP‐mediated base editing has already entered human clinical trials targeting genes such as PCSK9 and has also enabled rapid, personalized therapeutic approaches, as exemplified in CPS1 deficiency. However, this approach offers reduced control over cell‐specific editing outcomes and may introduce delivery‐associated toxicities. Collectively, these platforms emphasize the importance of aligning delivery modality with disease biology, tissue accessibility and therapeutic objectives to maximize the clinical impact of base editing interventions.