What Is Prime Editing? From Precise DNA Changes to LNP Delivery and PKU Research
If a genetic disease begins with a single incorrect letter in DNA, could we change that letter back? If two letters are missing, could we replace them while preserving the gene’s native regulation?
These are the kinds of problems prime editing aims to address. It can introduce targeted base substitutions, small insertions, and deletions, expanding the range of changes available to genome editing. But precise editing in cultured cells and effective treatment inside the body require different kinds of evidence.
On June 15, 2026, Allen Y. Jiang, Ana Cristian, and colleagues published Efficient prime editing in vivo and in vitro using lipid nanoparticles in Nature Nanotechnology. The team, which included David R. Liu and collaborators, delivered the RNA components of prime editing to mouse liver using lipid nanoparticles (LNPs). In a phenylketonuria (PKU) mouse model, they corrected a disease-causing variant and lowered circulating phenylalanine.[1]
What draws my attention to this study is the way it treats delivery, RNA design, and the editing reaction as parts of one system. For anyone who has worked on RNA and LNP development, the challenge is familiar: getting RNA into a cell is only the beginning. The components also need to remain intact and active, and work together at the right time.
Three points to keep in mind
The standard prime editing reaction does not require a DNA double-strand break or an external donor DNA template, but unintended editing outcomes still require assessment.
The 49% result is mean bulk-liver DNA editing in the mouse Pcsk9 experiment. The best mean bulk-liver DNA correction in the PKU experiment was approximately 15%; these are different experiments.
The main PKU efficacy assessment extended to day 7. These are preclinical findings, not evidence of human efficacy or a durable cure with LNP-delivered prime editing.[1]
Understanding the CRISPR Family: From Microbial Defense to Genome Engineering
Before discussing prime editing, it helps to put the CRISPR family in context. CRISPR stands for clustered regularly interspaced short palindromic repeats, a term originally describing particular repeated sequences in bacterial and archaeal genomes. The spacers between these repeats can retain information from foreign genetic material. Together with CRISPR-associated (Cas) proteins and RNA, they form defense systems that recognize and act against foreign nucleic acids.
Scientists adapted this RNA-guided recognition into programmable tools. In a commonly used Cas9 editing system, a guide RNA specifies the target sequence, and Cas9 binds and cuts DNA where the targeting requirements are met. In 2012, Jinek and colleagues showed that RNA could program Cas9-mediated DNA cleavage and combined the two natural RNA components into a single guide RNA, establishing an important foundation for subsequent tool development.[14]
Cas protein families and editing strategies are different categories
CRISPR–Cas systems are commonly divided into Class 1 and Class 2 according to the organization of their effector machinery. Class 1 generally uses a complex of multiple proteins. Class 2 mainly uses a single large effector protein; familiar examples include Cas9 in type II, Cas12 in type V, and Cas13 in type VI. This article focuses on tools commonly used in biomedical research, rather than providing a complete catalog of natural CRISPR systems.[15]
Cas9, Cas12, and Cas13 are effector protein families. Base editing, prime editing, CRISPR interference (CRISPRi), and CRISPR activation (CRISPRa) are engineering strategies built around these targeting tools. The prime editing system discussed here uses an engineered Cas9 nickase and a reverse transcriptase; prime editing is not a separate natural Cas family alongside Cas9.[2]
Cas9: DNA cutting, targeting, and regulation
The widely used Streptococcus pyogenes Cas9 (SpCas9) requires a guide RNA and a suitable protospacer adjacent motif (PAM). A PAM is a short DNA sequence beside the target that the Cas protein recognizes. Conventional SpCas9 commonly recognizes NGG, where N can be any nucleotide. Other Cas9 proteins and engineered variants can have different PAM preferences, so NGG is not a universal rule for all Cas9 systems.[14]
After Cas9 nuclease cuts both DNA strands, cellular repair can generate insertions or deletions that disrupt a gene or regulatory sequence. Providing a repair template can also enable precise changes through repair, although the outcome depends on the target, cell type, and repair pathway.[14]
A therapeutic example: Casgevy. Casgevy edits a patient’s own CD34+ hematopoietic stem and progenitor cells outside the body. It targets the erythroid-specific enhancer of BCL11A, reducing BCL11A expression in erythroid-lineage cells and increasing fetal hemoglobin (HbF). This strategy is used for sickle cell disease and transfusion-dependent β-thalassemia.[16]
The interesting point is that it does not directly repair the original disease-causing HBB variant. Instead, it changes another regulatory region to increase a compensating form of hemoglobin. It also uses ex vivo cell editing followed by infusion, a different delivery approach from the in vivo liver-targeted LNP system in this study.
Cas12a: Different DNA targeting requirements and multiplex editing
Cas12a, originally called Cpf1, can also edit DNA, but differs from Cas9 in its RNA requirements, PAM preferences, and cleavage pattern. Commonly used AsCas12a and LbCas12a favor T-rich PAMs, typically TTTV, where V is A, C, or G, and usually produce staggered DNA cuts. Targeting requirements can differ across homologs and engineered variants.[17,26]
A research example: studying several genes at once. Cas12a can process a crRNA array containing multiple guides. In 2017, Zetsche and colleagues used a single customized array to edit up to four genes simultaneously in mammalian cells and three genes in the mouse brain.[18]
This feature simplifies some multiplex experiments and combinatorial screens. Cas9 can also perform multiplex editing, and Cas12a is not automatically the best choice for every multigene task. Target-level efficiency and the possibility of unintended rearrangements from multiple DNA double-strand breaks still need assessment.
Cas13: Targeting RNA
The Cas13 family primarily targets RNA. Abudayyeh and colleagues demonstrated Cas13a-mediated RNA knockdown in mammalian cells and developed RNA-binding applications.[19]
These tools offer another option when the aim is to reduce a transcript or study RNA localization and function. Acting on RNA generally does not directly alter genomic DNA, but this does not establish an absence of adverse effects. RNA selectivity and possible collateral cleavage depend on the protein variant, target RNA abundance, and cellular context. Collateral cleavage refers to cleavage of other RNAs after target recognition; it is mechanistically different from off-target activity caused by partial guide matching.[23]
A research example: REPAIR RNA editing. Cox and colleagues fused catalytically inactive dCas13b to the deaminase domain of ADAR2, enabling targeted conversion of adenosine (A) to inosine (I) in RNA. Inosine is generally read as G during translation, allowing sequence information to be changed at the RNA level.[20]
This strategy has cellular proof-of-concept evidence. It differs from DNA prime editing and should not be presented as established clinical efficacy. Because the substrate is RNA, the duration of an effect also depends on transcript turnover and how long the system remains active.
CRISPRi and CRISPRa: Adjusting gene expression
Inactivating both Cas9 nuclease activities produces dCas9. A guide RNA can still direct it to DNA, but it no longer cuts DNA through Cas9 nuclease activity. Researchers can use it to interfere with transcription or recruit repressive or activating factors, creating CRISPRi and CRISPRa systems.[21,22]
An experimental example: to ask whether reducing expression of a gene affects cell survival, a researcher could design a CRISPRi experiment. To test whether increasing expression changes differentiation or cell function, CRISPRa may be useful. These are examples of experimental design, rather than claims of demonstrated efficacy for a particular disease.
These approaches primarily regulate expression rather than changing the DNA base sequence. Reversibility and persistence depend on the regulatory machinery, delivery method, and chromatin state; they cannot be assumed to be identical across systems.
Where base editing and prime editing fit
Conventional DNA base editors commonly use Cas9 nickase or dCas9 as a targeting scaffold, coupled to an enzyme that changes a base. Prime editors use Cas9 nickase and a reverse transcriptase, with a pegRNA specifying the intended DNA change.[2,24,25]
Disease-related examples: the PAH R408W variant has been studied using both adenine base editing and prime editing in mice. A two-nucleotide deletion in NCF1 has also been addressed in an early human study using ex vivo prime editing of hematopoietic cells followed by infusion.[8–10] The first example shows that a single variant may have more than one editing option; the second illustrates the use of prime editing to restore missing sequence.
This context helps explain the LNP study. Its task is to introduce a specified DNA change using an RNA template. Cas9 targeting alone is therefore insufficient: reverse transcriptase activity, intact epegRNA, and appropriate cellular repair must work together.
Prime Editing Basics: How an RNA Template Specifies a DNA Change
In 2019, Andrew V. Anzalone and colleagues introduced prime editing by linking a Cas9 nickase, which cuts one DNA strand, to an engineered reverse transcriptase (RT). They paired this protein with a prime editing guide RNA (pegRNA) that supplies both targeting information and instructions for the intended change.[2]
Think of it as a targeting tool carrying its own revision instructions. Cas9 finds the site, the pegRNA provides the new sequence, and the reverse transcriptase uses the RNA template to synthesize DNA containing the change. Cellular processing and repair then determine whether that change becomes established in the genome.
What does a pegRNA add to a conventional guide RNA?
A conventional guide RNA supplies targeting information and a scaffold that binds Cas9. A pegRNA additionally includes a primer-binding site (PBS) and a reverse-transcription template (RTT). The PBS pairs with the DNA 3′ end generated by nicking, providing a starting point for reverse transcription. The RTT specifies the sequence to be written. The integrity, structure, and design of the pegRNA therefore affect whether the reaction can be completed.[2,3]
There is more to the reaction than changing a letter
The prime editor forms a complex with the pegRNA, recognizes a target with an appropriate PAM, and nicks one DNA strand. The exposed DNA 3′ end pairs with the PBS, and the RT extends the DNA using the RTT, generating an edited 3′ flap. Flap processing, ligation, and repair then influence which sequence is retained.[2]
This sequence of events helps explain why the same editor can perform very differently across variants, cells, and delivery methods. Prime editing supports a broad range of changes, but that does not mean every variant is easy to correct.
Choosing Between CRISPR Nucleases, Base Editing, and Prime Editing
CRISPR describes a broad family of systems, and prime editing also uses CRISPR-based targeting. The useful comparison is between editing strategies.[2]
Cas9 nuclease
Main reaction and use: Creates a DNA double-strand break. Repair can disrupt a gene; a repair template can support precise changes.
Key considerations: Unintended deletions, rearrangements, and other repair products are possible. Precise repair depends on cell state.
Conventional CBE or ABE
Main reaction and use: Uses deamination to enable particular substitutions, commonly C•G→T•A or A•T→G•C.
Key considerations: Requires suitable targeting and an editing window. Other bases within that window may undergo bystander editing.
Prime editing
Main reaction and use: Uses an RNA template and reverse transcription to specify any of the twelve base substitutions, as well as small insertions or deletions.
Key considerations: More components and reaction steps require target-specific optimization. Indels, unintended insertions, and off-target outcomes still require assessment.
If the aim is simply to disrupt a gene, a nuclease may be sufficient. For a point mutation compatible with an ABE or CBE, with controllable bystander editing, base editing may be more efficient. Prime editing becomes particularly useful when missing sequence must be restored, a substitution is unsuitable for conventional base editors, or the desired sequence change needs to be specified more flexibly.[2,9]
PAH R408W is a useful example: base editing has already been investigated for this variant, so it should not be described as correctable only by prime editing. Tool selection must consider correction, byproducts, delivery feasibility, and functional recovery together.[9]
From PE2 and PE3 to PE6: Improving Proteins, RNA, and Repair
PE2 combines an engineered Cas9 nickase–RT with a pegRNA. PE3 adds an ngRNA to nick the opposite DNA strand and favor retention of the edit. PE3b designs the second nick to preferentially target the edited sequence, reducing unwanted outcomes. These names describe differences in reaction design.[2]
Another development path improves the editor and its interaction with cellular repair. Chen and colleagues introduced PEmax and showed that mismatch repair (MMR) influences editing outcomes. PE4 and PE5 add a dominant-negative MLH1 variant, MLH1dn, to transiently inhibit MMR and improve efficiency in some settings. These designs are options to evaluate, rather than a standard recipe for every application.[4]
In 2023, Doman and colleagues used protein engineering and phage-assisted evolution to develop the PE6 series. Different RT variants performed differently across editing tasks. Whether PE6c or PE6d is appropriate therefore depends on target-specific data.[5]
RNA engineering matters as well. Nelson and colleagues appended structured motifs to the pegRNA 3′ end to reduce degradation, creating engineered pegRNAs, or epegRNAs. In 2026, Sakai and colleagues used PE-PRISM to screen and optimize these motifs, producing variants including eSBRMV1-A. These studies underpin the RNA improvements in the new LNP paper.[3,6]
PE7 takes another approach by fusing part of the La RNA-binding protein to the prime editor, improving performance under certain conditions.[7] These names represent different engineering approaches; the numbers should not be treated as a universal performance ranking.
Why Prime Editing LNP Delivery Is More Complex Than Delivering One mRNA
The system in this study delivers three RNAs: a long mRNA encoding the prime editor, an epegRNA supplying targeting and editing instructions, and an ngRNA guiding the second-strand nick. Long mRNA and shorter guide RNAs differ in size, physical properties, and the timing of their roles inside cells.[1]
LNPs must be taken up by cells and release enough RNA from endosomes into the cytoplasm. The mRNA must then be translated into editor protein. The editor and guide RNAs also need to function together in the nucleus to access genomic DNA. If much of the epegRNA has degraded by the time sufficient editor protein is available, the reaction may lack its template. This timing mismatch is one of the central problems the authors address.
The researchers used OF-02 LNPs to encapsulate each RNA separately, then mixed the three preparations before administration. Each preparation carried one RNA component; mixing did not turn the particles into a formulation co-encapsulating all three RNAs. This makes it easier to adjust individual components and ratios, while leaving an important systems question: do all necessary components reach the same cell and remain available at the right time?
The work demonstrates an all-RNA delivery system improved through editor selection, guide RNA structure, RNA preparation, and component ratios. It should not be described as a study of three RNAs co-encapsulated in a single LNP.[1]
What Did the Nature Nanotechnology Study Improve?
The paper labels its optimization stages s0.2, s1, s2, and s3: s0.2 is the epegRNA-based starting system; s1 evaluates and selects a PE6 editor; s2 optimizes the epegRNA 3′ structural motif; and s3 adjusts the mRNA-to-guide RNA ratio. These are experiment-specific labels, not universal prime editing version names. Each target requires its own screening, with RNA preparation and purification assessed alongside these stages.[1]
Keeping the guide RNA functional
In the initial Pcsk9 experiments, a highly chemically modified pegRNA produced approximately 0.8% mean bulk-liver editing, compared with approximately 3.8% using an epegRNA. The authors subsequently tested new 3′ motifs; in one comparison, eSBRMV1-A increased editing from approximately 17% to 26%. These results come from different optimization stages and experimental groups, rather than one continuous comparison under otherwise identical conditions.[1, Figs. 1–2]
The stability advantage of eSBRMV1-A was more apparent when the editor was present. The authors proposed explanations involving complex formation, RNA protection, and reaction conformation. These remain mechanistic interpretations requiring further investigation, rather than a proven single explanation.
Selecting the editor for the target
The Pcsk9 insertion task favored PE6c, whereas correction of PAH R408W favored PE6d. What can be carried forward is the screening and optimization workflow; a single editor–RNA combination is not a universal optimum for every disease.[1, Figs. 2 and 5]
Interpreting RNA ratios and preparation quality together
After Pcsk9 optimization, the s3 system used a total RNA mass ratio of approximately 1:2 for mRNA to combined guide RNAs in the administered formulation. The epegRNA:ngRNA ratio was approximately 9:1, giving an overall mRNA:epegRNA:ngRNA ratio of 1:1.8:0.2.
This is a total RNA mass ratio, not a molecular ratio or a description of individual particles or cells. Measured encapsulation efficiencies were approximately 63%, 87%, and 92%, respectively. Based on these values, the authors estimated an encapsulated RNA mass ratio of approximately 1:2.5:0.3. Neither ratio establishes the intracellular ratio of active components.[1, Fig. 3]
In the mouse Pcsk9 ratio comparison, differences between the tested groups were not statistically significant. Ratio optimization remains worth considering, but these results do not establish it as the main breakthrough or prove that 1:2 is uniquely optimal.
Switching to commercially prepared mRNA increased mean editing from approximately 26% to 49% in one comparison. The coding sequences were identical, but the UTRs and purification methods differed. The results support the importance of RNA design and manufacturing quality without isolating the contribution of either factor.[1]
The PAH development work also used epegRNA purified by high-performance liquid chromatography (HPLC). The authors hypothesized that stricter purification could reduce nonfunctional truncated products and increase the fraction of effective guides. Whether truncation was the main cause of the performance difference still needs dedicated testing.
Reading the Results: 49% Liver Editing and PKU Correction Are Different Experiments
Pcsk9 served as a platform performance test
At Pcsk9, the authors designed a TTAC insertion to produce a frameshift and premature termination. After one dose of 2 mg/kg total RNA, the optimized system achieved approximately 49% mean target editing in bulk-liver DNA. This is the proportion of sequencing reads carrying the specified edit, not evidence that 49% of liver cells were edited.[1, Figs. 2 and 4]
Editing was predominantly hepatic. The tested extrahepatic tissues and sorted liver nonparenchymal cell populations did not show comparable target editing, supporting functional delivery biased toward hepatocytes. This does not establish an absence of extrahepatic exposure.
In a repeat-dose experiment, one 1 mg/kg dose produced approximately 28% editing; a second dose one week later increased editing to approximately 43%. This supports cumulative editing after redosing in mice, but does not establish unlimited repeat dosing in patients.[1, Fig. 4]
PAH R408W was the disease-correction experiment
The PKU model in this study has deficient phenylalanine hydroxylase (PAH) function. PAH helps convert phenylalanine to tyrosine; when its activity is insufficient, phenylalanine can accumulate to harmful concentrations. The authors targeted PAH c.1222C>T, or p.R408W, using an optimized version of a previously developed PE3b correction strategy.[1,8]
Humanized mice carrying this variant received a single retro-orbital intravenous injection of 4 mg/kg total RNA. One week later, mean bulk-liver genomic DNA correction was approximately 15% with s2 and 12% with s3. Correction in cDNA derived from Pah transcripts was approximately 23% and 19%, respectively. DNA and cDNA measurements have different denominators, and transcript measurements are influenced by the cells expressing the gene and their expression levels.[1, Fig. 5]
The s2 group showed an approximately 90% reduction in circulating phenylalanine by day 3. By day 7, mean phenylalanine concentrations in all PE-LNP groups except the initial s0.2 formulation were below the study’s 360 µM comparison threshold. This threshold was drawn from recommendations for clinical intervention; it is not a universal definition of normal concentrations or an individual treatment target. Each PKU condition included four mice, making this a small preclinical experiment.[1]
To me, this functional result is more interesting than the 49% headline. The correction rate required for benefit depends on disease biology and functional thresholds. In this model, correcting a fraction of liver DNA was sufficient for substantial metabolic improvement. Other diseases need their own evidence.
The PKU experiment mainly followed outcomes through day 7. It did not establish long-term metabolic control, neurological benefit, or permanent cure after one dose. The eight-week Pcsk9 follow-up cannot substitute for long-term PKU efficacy data.
Also, mean DNA correction with s3 was not higher than with s2 in this disease model. Additional optimization steps do not guarantee improvement in every endpoint.
Safety and the AAV Comparison: What Does Transient Expression Offer?
LNP-delivered RNA generally produces transient editor expression, while completed DNA changes can persist. This creates an opportunity to shorten the editor’s activity window and reduce unnecessary editing.[1]
In the eight-week Pcsk9 comparison, mean bulk-liver target editing was approximately 44% with PE-LNPs and 46% with dual-AAV delivery. The difference was not statistically significant. This supports similar editing levels in that experiment, rather than formal proof of equivalence.[1, Fig. 6]
Among fourteen previously nominated candidate off-target sites, both delivery methods showed editing at OT13, with higher editing in the AAV group. The supported claim is reduced off-target editing at a tested site under those conditions, not zero off-target activity or comprehensive genome-wide safety.[1]
In the 2 mg/kg experiments in standard mice, ALT rose mildly and transiently at day 1 and returned to levels near the controls by day 3. In the PKU experiment, day-3 ALT and AST did not increase relative to baseline. That measurement alone cannot exclude earlier changes.[1]
Fiumara and colleagues’ work in human hematopoietic stem and progenitor cells also shows why avoiding a double-strand break as a required reaction step does not eliminate the possibility of double-strand breaks or genotoxic byproducts. Those findings are not a direct toxicity conclusion for this mouse liver system, but they reinforce the need to evaluate each cell type, editor, and delivery context.[11]
Prime Editing Applications and Early Human Evidence
Direct correction of disease-causing sequences is an obvious application. Prime editing can also create disease models, test the function of specific variants, and introduce precise research labels. The required efficiency, product purity, and safety depend on the intended use.[2]
Prime Editing for p47phox-Deficient Chronic Granulomatous Disease, published online in 2025 and appearing in the 2026 New England Journal of Medicine, reported two participants receiving PM359. Their autologous CD34+ cells underwent ex vivo prime editing to correct a two-nucleotide deletion in NCF1, followed by infusion after busulfan conditioning. The study observed restoration of neutrophil NADPH oxidase function. This is early human evidence for ex vivo prime editing, rather than direct administration of liver-targeted PE-LNPs.[10]
The widely discussed personalized treatment for CPS1 deficiency used LNP-delivered base editing. It supports the translational feasibility of personalized RNA-LNP genome editing, but should not be described as a clinical prime editing result.[12]
In 2025, Rothgangl and colleagues reported a transient prime editing approach for a metabolic liver disease in mice, evaluating both AAV-delivered guide RNA with LNP-delivered editor mRNA and an all-RNA LNP strategy.[13] The 2026 paper therefore builds on earlier delivery work by improving the efficiency and optimization workflow of all-RNA PE-LNPs; it is not the first demonstration that LNPs can deliver prime editing components.
This study advances the evidence for direct in vivo PE-LNP use. Other organs still require their own solutions for functional delivery. Success in mouse liver cannot be assumed to extend to the brain, lungs, or immune cells.
My Perspective: Turning an Editing Reaction into a Controllable Product
From an RNA and LNP development perspective, my next question is whether the performance can be reproduced across batches, suppliers, and target sequences. An effective research formulation needs measurable relationships between quality and function before it can support further development.
The paper provides a useful starting point. Different RNA cargos have different encapsulation efficiencies and particle characteristics; the total RNA ratio in the administered formulation also differs from the encapsulated RNA ratio. Beyond particle size, developers need to consider mRNA integrity, double-stranded RNA (dsRNA) impurities, intact versus truncated epegRNA, and the functional potency of the combined components.[1]
I see these as quality and functional assessments to establish during product development, rather than evidence that the paper has completed GMP or regulatory validation. Separating the effects of mRNA untranslated-region (UTR) design and purification would be another useful experiment. Otherwise, we may know which preparation performs better without knowing which attributes actually need to be controlled.
Prime editing expands the range of DNA changes we can specify. This study shows that making those changes inside the body requires coordinated engineering of proteins, RNA, and nanoparticles. It demonstrated clear functional improvement in a mouse model. Important next evidence includes human primary cells, larger animals, long-term disease-model follow-up, and broader assessment of unintended editing and immune safety.
About LuTra Studio
LuTra Studio helps biotechnology teams develop technical strategies for RNA therapeutics, LNP delivery, genome editing, and cell therapy, connecting experimental findings with translational development and CMC planning. If your team is evaluating editing tools, delivery platforms, or a product development pathway, you are welcome to contact LuTra Studio.
References
Jiang AY, Cristian A, Brooks DL, et al. Efficient prime editing in vivo and in vitro using lipid nanoparticles. Nature Nanotechnology. 2026;21(7):1022–1035. doi: 10.1038/s41565-026-02200-6
Anzalone AV, Randolph PB, Davis JR, et al. Search-and-replace genome editing without double-strand breaks or donor DNA. Nature. 2019;576(7785):149–157. doi: 10.1038/s41586-019-1711-4
Nelson JW, Randolph PB, Shen SP, et al. Engineered pegRNAs improve prime editing efficiency. Nature Biotechnology. 2022;40(3):402–410. doi: 10.1038/s41587-021-01039-7
Chen PJ, Hussmann JA, Yan J, et al. Enhanced prime editing systems by manipulating cellular determinants of editing outcomes. Cell. 2021;184(22):5635–5652.e29. doi: 10.1016/j.cell.2021.09.018
Doman JL, Pandey S, Neugebauer ME, et al. Phage-assisted evolution and protein engineering yield compact, efficient prime editors. Cell. 2023;186(18):3983–4002.e26. doi: 10.1016/j.cell.2023.07.039
Sakai HA, Pierce SE, Jiang AY, et al. Directed evolution of small RNA-stabilizing motifs that improve prime-editing efficiency. Nature Biotechnology. 2026. Advance online publication. doi: 10.1038/s41587-026-03123-2
Yan J, Oyler-Castrillo P, Ravisankar P, et al. Improving prime editing with an endogenous small RNA-binding protein. Nature. 2024;628(8008):639–647. doi: 10.1038/s41586-024-07259-6
Brooks DL, Whittaker MN, Qu P, et al. Efficient in vivo prime editing corrects the most frequent phenylketonuria variant, associated with high unmet medical need. The American Journal of Human Genetics. 2023;110(12):2003–2014. doi: 10.1016/j.ajhg.2023.10.005
Brooks DL, Whittaker MN, Said H, et al. A base editing strategy using mRNA-LNPs for in vivo correction of the most frequent phenylketonuria variant. Human Genetics and Genomics Advances. 2024;5(1):100253. doi: 10.1016/j.xhgg.2023.100253
Gori JL, Haddad E, Frangoul H, et al. Prime Editing for p47phox-Deficient Chronic Granulomatous Disease. New England Journal of Medicine. 2026;394(12):1195–1203. doi: 10.1056/NEJMoa2509807
Fiumara M, Ferrari S, Omer-Javed A, et al. Genotoxic effects of base and prime editing in human hematopoietic stem cells. Nature Biotechnology. 2024;42(6):877–891. doi: 10.1038/s41587-023-01915-4
Musunuru K, Grandinette SA, Wang X, et al. Patient-Specific In Vivo Gene Editing to Treat a Rare Genetic Disease. New England Journal of Medicine. 2025;392(22):2235–2243. doi: 10.1056/NEJMoa2504747
Rothgangl T, Tálas A, Ioannidi EI, et al. Treatment of a metabolic liver disease in mice with a transient prime editing approach. Nature Biomedical Engineering. 2025;9(10):1705–1718. doi: 10.1038/s41551-025-01399-4
Jinek M, Chylinski K, Fonfara I, et al. A Programmable Dual-RNA–Guided DNA Endonuclease in Adaptive Bacterial Immunity. Science. 2012;337(6096):816–821. doi: 10.1126/science.1225829
Makarova KS, Shmakov SA, Wolf YI, et al. An updated evolutionary classification of CRISPR–Cas systems including rare variants. Nature Microbiology. 2025;10(12):3346–3361. doi: 10.1038/s41564-025-02180-8
U.S. Food and Drug Administration. CASGEVY (exagamglogene autotemcel): Prescribing information. Accessed October 10, 2026. FDA prescribing information
Zetsche B, Gootenberg JS, Abudayyeh OO, et al. Cpf1 Is a Single RNA-Guided Endonuclease of a Class 2 CRISPR-Cas System. Cell. 2015;163(3):759–771. doi: 10.1016/j.cell.2015.09.038
Zetsche B, Heidenreich M, Mohanraju P, et al. Multiplex gene editing by CRISPR–Cpf1 using a single crRNA array. Nature Biotechnology. 2017;35(1):31–34. doi: 10.1038/nbt.3737
Abudayyeh OO, Gootenberg JS, Essletzbichler P, et al. RNA targeting with CRISPR–Cas13. Nature. 2017;550(7675):280–284. doi: 10.1038/nature24049
Cox DBT, Gootenberg JS, Abudayyeh OO, et al. RNA editing with CRISPR-Cas13. Science. 2017;358(6366):1019–1027. doi: 10.1126/science.aaq0180
Qi LS, Larson MH, Gilbert LA, et al. Repurposing CRISPR as an RNA-Guided Platform for Sequence-Specific Control of Gene Expression. Cell. 2013;152(5):1173–1183. doi: 10.1016/j.cell.2013.02.022
Gilbert LA, Larson MH, Morsut L, et al. CRISPR-Mediated Modular RNA-Guided Regulation of Transcription in Eukaryotes. Cell. 2013;154(2):442–451. doi: 10.1016/j.cell.2013.06.044
Shi P, Murphy MR, Aparicio AO, et al. Collateral activity of the CRISPR/RfxCas13d system in human cells. Communications Biology. 2023;6(1):334. doi: 10.1038/s42003-023-04708-2
Komor AC, Kim YB, Packer MS, et al. Programmable editing of a target base in genomic DNA without double-stranded DNA cleavage. Nature. 2016;533(7603):420–424. doi: 10.1038/nature17946
Gaudelli NM, Komor AC, Rees HA, et al. Programmable base editing of A•T to G•C in genomic DNA without DNA cleavage. Nature. 2017;551(7681):464–471. doi: 10.1038/nature24644
Gao L, Cox DBT, Yan WX, et al. Engineered Cpf1 variants with altered PAM specificities. Nature Biotechnology. 2017;35(8):789–792. doi: 10.1038/nbt.3900
Comments