Prime Editing Driven Functional Genomics: Bridging Genotype to Phenotype in the Post-Genomic Era

Abstract

The post-genomic era, defined by large-scale sequencing initiatives, has generated an unprecedented catalogue of human genetic variation. Yet, the vast majority of genetic variants remain classified as variants of uncertain significance or are located within poorly characterized non-coding regions, thereby hindering the effective translation of genomic data into meaningful biological understanding and clinical application. Bridging this genotype-to-phenotype gap requires precise, high-throughput functional genomics. Early CRISPR–Cas9 knockout and CRISPR interference/activation (CRISPRi/a) screens mapped gene-level functions but could not assess single nucleotide variants (SNVs). Bridging this genotype-to-phenotype gap demands precise, high-throughput functional genomics. Multiplexed assays of variant effect (MAVEs), like saturation genome editing, systematically test all possible mutations using CRISPR–Cas9 and donor libraries. Base editors allow targeted single-base changes without double-strand breaks but are limited in scope, while prime editing can introduce any small substitution, insertion, or deletion without double-strand breaks (DSBs) or donor templates. This review traces the evolution of functional screens from gene-level knockouts to saturation genomic editing (SGE), and highlights how prime editing is driving a new paradigm for the systematic functional characterization of thousands of variants across disease-relevant genes. We also detail the architecture, mechanism, and progressive optimization of PE systems and their delivery methods. Collectively, prime editing stands as a transformative platform poised to accelerate precision functional genomics and advance the diagnosis and treatment of genetic diseases.

1. Introduction

The completion of the Human Genome Project ushered in a new era of genomics, driving a global effort to rapidly sequence human genomes and generate vast amounts of data crucial for understanding how genetic variations influence health and disease. Resources such as the UK Biobank, which has completed whole-genome sequencing of approximately 490,000 participants, and the Genome Aggregation Database (gnomAD), which aggregates genomic data from 800,000 individuals, have uncovered millions of DNA variants [1]. For example, the UK Biobank has identified 1.5 billion genetic variants, including >600 million SNVs across coding and non-coding regions, and gnomAD has catalogued 786 million variants (from 76,000 whole genomes and hundreds of thousands of exomes), of which over 16 million are missense mutations [1]. Cancer genome consortia have sequenced over 2600 tumor genomes, revealing the landscape of somatic mutations across dozens of cancer types [2]. These vast datasets enable genome-wide association studies (GWASs) that link variants to traits, the discovery of rare disease mutations, and insights into gene function and disease mechanisms. Despite this trove of data, the majority of identified variants lack clear functional annotation. Most are classified as VUS because their impact on gene function or disease risk is uncertain [3]. This disconnect between variant discovery and biological understanding poses a critical challenge for genomic medicine. Functional characterization of variants in physiologically relevant systems is essential to determine which mutations are benign and which are pathogenic. Importantly, drug development experience shows that targets with robust human genetic evidence have far higher success rates in clinical trials on the order of a 2- to 3-fold improvement in approval likelihood. Therefore, a thorough examination of how genetic variations affect things can enhance genetic diagnosis. It can also help identify medication targets that are physiologically supported, which increases the chances of successful treatments.

To understand how a function works, it is important to investigate variations in the context of their native genetic environment. In contrast, traditional methods that overexpress genes or use cDNA might bypass natural regulatory processes, such as splicing, which can lead to potentially misleading results [4]. For example, recent research on TP53 demonstrated that various mutations linked to cancer have conflicting impacts on an organism’s traits, depending on whether the gene is exogenously overexpressed or endogenously present. This highlights the relevance of the gene’s normal amount and how it is controlled [4]. Similar discrepancies have been reported for BRCA1, where SGE revealed functional consequences of variants that were not captured by over-expression assays, and for PTEN, where endogenous editing uncovered dosage-sensitive effects that were masked in cDNA-based systems [5,6,7]. Collectively, these findings highlight the limitations of exogenous approaches and have driven a shift in functional genomics toward endogenous MAVEs. These are high-throughput experiments (often pooled screens) that introduce a library of mutations into the genome itself and measure their functional consequences in situ. A powerful approach within this paradigm is SGE, which uses genome editing tools to introduce all possible single nucleotide substitutions (or a comprehensive set of variants) across a locus of interest. By coupling precise editing with selection or phenotypic readouts, SGE can map genotype–phenotype relationships at unprecedented resolution, for example, classifying every possible missense mutation in a tumor suppressor gene as functional or deleterious.

CRISPRCas9 technology provided the initial engine for such functional screens. The standard Cas9 nuclease generates targeted DSBs, which, when repaired by error-prone non-homologous end joining (NHEJ), create gene knockouts, a strategy used in the first genome-wide CRISPR knockout screens to identify essential genes and drug resistance factors [5]. However, DSB-based editing has limitations, it often produces mixed insertions/deletions (indels) rather than specific nucleotide changes, and cleavage can trigger p53-mediated DNA damage responses that kill edited cells or select for p53-inactivated clones. Base editors (cytosine and adenine deaminases tethered to a Cas9 nickase (nCas)) addressed some issues by directly converting C•G to T•A or A•T to G•C without DSBs [8]. Base editing enabled many precise single-base changes and was successfully applied to functional screens (e.g., scanning BRCA1 regulatory regions). Yet, base editors are constrained to a subset of nucleotide changes (primarily C→T or A→G transitions, with newer variants for C→G/C→A). They can also produce bystander edits when multiple target bases lie in the editing window, and they require a nearby protospacer-adjacent motif (PAM), limiting the sites accessible for editing.

The advent of PE in 2019 opened a far broader solution. PE is a precise “search-and-replace” genome editing approach capable of introducing all 12 possible types of base substitutions, small insertions, or deletions at specified locations without requiring DSBs or donor DNA. This versatile technology uses a nCas9-reverse transcriptase (RT) fusion guided by a specialized pegRNA [9]. Its unparalleled precision and flexibility make it an ideal engine for functional genomics. PE combined with functional screening via high-throughput techniques offers a powerful insight for understanding genetic variation. PE enables precise genomic editing within the genome, this allows the researcher to detect the effects of variants on gene functions, the phenotype of cells, and disease pathways. This strategy helps to understand VUS, improve genetic testing, and develop targeted therapy. The remainder of this review outlines the evolution of functional variant screening approaches and the impact of PE on this field. We first survey the landscape of CRISPR-based functional screens, from initial conventional CRISPR screens to SGE, using CRISPR-Cas9, base editors, or prime editors for high-resolution variant-level analysis. We then describe the PE systemin detail, its mechanism, progressive improvements (PE1 through PE5), specialized variants, and delivery strategies.

2. Evolution of Functional Screens of Genetic Variants

Introducing genetic changes and detecting their phenotypic effects through functional characterization have become an essential technique to uncover gene functions, mechanisms, and biological pathways [10]. The advancement of CRISPR–Cas9 technology has transformed functional genomics, making it possible to study gene function on a large scale. Initially, the Cas9 protein was used to perform genome-wide knockout screening. These pooled and arrayed knockout screens systematically identified important genes and key regulators involved in several biological processes [11]. However, DSB-induced knockout screens had limitations, including off-target effects, chromosomal rearrangements, and the inability to determine the specific effects of individual variants [12]. To overcome the hurdles of nuclease-induced disruptions, methods like CRISPR interference (CRISPRi) and CRISPR activation (CRISPRa) were developed. These methods use an enzymatically inactive version of Cas9, called dCas9, which is coupled to either transcriptional repressor or activator domains. This allows for the targeted control of gene expression without changing the DNA sequence itself [13]. CRISPRi and CRISPRa have extended the scope of functional genomics by enabling the study of essential genes, regulatory elements, and dose effects, modifying the technique beyond traditional knockout-based genomic editing approaches. Despite these advances, CRISPRi/a approaches share an important limitation, they modulate gene expression at the transcript level but do not modify the underlying genomic sequence. As such, they cannot resolve the functional consequences of SNVs or small indels in a particular gene, which represent the majority of disease-associated mutations identified in human populations. To systematically understand how these specific variants influence particular gene function, new precision-editing strategies were required. SGE allowed the study of the functional effects of variant levels, moving beyond the gene level [14].

SGE combines CRISPR-Cas9-mediated DSB induction with homology-directed repair (HDR). This approach uses a library of donor DNA templates containing specific nucleotide variants. This method makes it possible to introduce all possible SNVs or close mutations across a specific region of the gene. SGE enables comprehensive mapping of how genetic variants relate to a phenotype by combining the precise genomic editing technique with a high-throughput functional readout. This helps us understand the functional impact of genetic variants. Using CRISPR-Cas9-based saturation genome editing, researchers can quantitatively detect how different variants affect the coding or regulatory region of genes. This provides a powerful method for studying gene functional impacts of genetic variants and interpreting VUS in clinical genomics [15,16]. SGE is applied to both protein-coding and non-coding regions, such as promoters, enhancers, and splicing sites. This enables the detection of variants that influence transcription, translation, and post-translational regulations. The first application of SGE occurred in a human near-haploid HAP1 cell line, which was derived from KBM-7 chronic myelogenous leukemia cells [6,17]. Owing to their haploid nature, editing a single allele in HAP1 cells is sufficient to elicit a phenotypic effect, making them ideal for studying loss-of-function (LoF) mutations in tumor suppressor genes such as BRCA1. In contrast, in the normal diploid cell, unedited or wild-type alleles can mask the effect of mutations. Additionally, in HAP1 cells, suppression of the non-homologous end-joining (NHEJ) DNA repair pathway enhanced HDR-mediated editing, increasing the editing efficiency four times. Findlay and his coworker combined CRISPR-Cas9-mediated targeted cleavage with multiplex HDR, using a complex library of HDR donor templates to attain SGE in specific genomic regions in HAP1 cells. They introduced all possible SNVs into exons 13 and 18 of the BRCA1 and VHL (von Hippel–Lindau) genes. This allowed for a complete evaluation of the functional effects of these variants and also showed several variants significant impacts on transcript levels. This was caused by either nonsense mutation-mediated degradation or disruption of exonic splicing enhancer regions, thereby altering gene expression. Similarly, the application of SGE to the DBR1 gene revealed variant-dependent impacts on cellular proliferation, thereby highlighting the platform’s capacity to detect functionally significant variants with physiological implications. Collectively, these investigations show that SGE allows for comprehensive functional analysis of both protein-coding and cis-regulatory regions. Therefore, this provides a powerful approach for understanding the impact of genetic variants and enhancing the clinical interpretation of VUS [6,15,17]. Although HAP1 cells provide a reliable and efficient platform for genomic-scale functional assessment, their applicability is somewhat limited. Continuous culture of HAP1 cells results in spontaneous diploidization, therefore introducing heterogeneity and reducing the accuracy of experimental results. Moreover, some genes are either non-essential, dormant, or lowly expressed in HAP1 cells, with their functions often displaying cell-type specialization. Therefore, for these studies, diploid cell types, such as HEK293T or primary B lymphocytes, are frequently utilized. However, diploid-nature cells show analytical challenges due to the presence of two alleles, which may hinder the impact of variants. To attain a functional resolution similar to that of haploid systems, biallelic editing, which changes both alleles, is often required. The CARD11 gene is a critical regulator of B-cell receptor (BCR) signaling and the NF-κB pathway. Through biallelic editing, both gain-of-function (GoF) and LoF variants of CARD11 were functionally characterized, thereby demonstrating that diploid models can achieve interpretive resolution comparable to haploid systems [18]. Another way to get around the restriction of diploid editing is local haploidization. With this approach, the single allele of the target gene is removed by the CRISPR-Cas9 technique, thus, the gene is haploid at that locus, while the rest of the genome remains diploid. The method was successful in direct variant analysis in a biologically relevant environment when it was implemented in HEK293T cells for genes such as BRCA2 and NPC1 [19]. Although CRISPR-Cas9-mediated SGE has demonstrated success, it still faces technical challenges. This is mostly due to the limited efficiency of HDR, which occurs exclusively during the S and G2 phases of the cell cycle. Moreover, the HDR pathway, which competes with the more common NHEJ pathway, often introduces unintentional indels at the target site. This decreases both the precision and efficiency of the genomic editing process. To overcome this limitation, researchers developed alternative methods for precision genomic editing. These approaches allow for precise base substitution, addition, and deletion without inducing DSBs or using HDR donor templates. This led to the creation of base editing (BE), a technology that combines a nCas9 protein with a reduced ability to cut DNA (nCas9) with a nucleotide deaminase enzyme. There are two main types of base editors: cytosine base editors (CBEs), which substitute C•G pairs to T•A pairs, and adenine base editors (ABEs), which change A•T pairs into G•C pairs. Base editing represented a significant improvement in precise genome engineering. This method enabled efficient editing at the level of a single nucleotide, with few insertions or deletions. As a result, it facilitated large-scale functional screening of variants linked to diseases [20]. However, base editing is inherently limited to a subset of possible nucleotide changes, specifically four out of the twelve possible base substitutions. Moreover, base editors (BEs) can cause unintended edits in bystander bases when there are multiple editable bases in the editing window. This may affect the functional outcomes. The function of these systems depends on the presence of specific PAM sequences close to the target site, which restrict access to particular genomic regions. Due to these limitations, base editing screens represented a vital advancement between conventional SGE and PE [

Table 1. Comparison of CRISPR-based genome editing tools in functional genomics.

Genomic Editing Technique	Components	Mechanism of Action	Key Features andLimitations
CRISPR–Cas9	Cas9 nuclease + sgRNA	Generates a site-specific DSB that is repaired by NHEJ or HDR	Efficient and scalable; prone to indels; chromosomal rearrangements; off-target effects; and limited variant precision.
Base Editors (CBE/ABE)	nCas9 (D10A) + cytidine deaminase + sgRNA/ nCas9 (D10A) + adenosine deaminase + sgRNA	Deaminates cytosine (C) to uracil (U), which is repaired as thymine (T)/ Converts adenine (A) to inosine (I), read as guanine (G)	Precise single-base editing; without DSBs and donor DNA; limited to A→G edits and C→T edits; bystander editing within the editing window; PAM and window constraints.
PE	nCas9 (H840A) + RT+ pegRNA	Uses reverse transcription to directly incorporate new DNA sequences at the target site	Most versatile and precise; without double-strand breaks and donor DNA; enables all 12 base substitutions and small insertions or deletions.

].

This new method combines the precision and versatility needed for true “search-and-replace” genome editing. Liu and his team developed PE, which uses a nCas9 (H840A) connected to an RT. This system is guided by a PE guide RNA (pegRNA). The pegRNA has a primer binding site (PBS) and a reverse transcription template (RTT). This system allows for all twelve possible base substitutions, as well as small insertions and deletions [21] [Figure 1]. PE works independently of the cell cycle, which reduces chromosomal rearrangements, thus providing a new level of precision and flexibility. By changing the RTT and PBS of the pegRNA, edits can be made farther from the PAM site. This allows for a wider range of editing compared to base editing. Recently, PE has been applied in functional screening to investigate both disease-associated and random genetic variants. Building upon this, saturation PE (SPE) was advanced to expand the investigation of functional genomics at the variant level.

Utilizing SPE, researchers examined 426 BRCA2 and 978 NPC1 variants within HEK293T cells, creating targeted haploidization by employing the CRISPR GUARD technique to preserve the unedited allele. The study demonstrated that SPE is a dependable and broadly applicable method for large-scale functional variation analysis, which can transform ambiguous sequencing data into valuable biological insights [19]. Ren et al. (2023) [22] followed up to a great extent by inventing a pool of PE libraries that facilitate high-throughput, single-base resolution functional screening of genetic variants in natural genomic contexts. With this approach, they identified essential nucleotides in the 716 bp MYC enhancer, functionally characterized 1304 non-coding breast cancer-related GWAS variants, and looked into 3699 ClinVar variations, thereby disclosing 103 non-coding and 156 VUS variants that influenced cell fitness [22]. Recent innovations have extended the use of PE to explore GoF mutations and how cancer becomes resistant to drugs. Chardon et al. (2023) [23] engineered a multiplexed PE approach that enabled the investigation of the impact of SNVs in eight oncogenes at the same time. Consequently, the method revealed resistance mechanisms that covered a broad spectrum of EGFR tyrosine kinase inhibitors (TKIs) [23]. Similarly, Kim et al. applied PE to create and functionally evaluate 2476 SNVs, representing 99% of the EGFR tyrosine kinase domain, thus elucidating variant-specific resistance mechanisms for the most relevant TKIs, such as afatinib and osimertinib [24]. Moreover, several large-scale studies have chosen PE for the elucidation of mechanisms in tumor suppressor genes, as in the case of a 28,000-pegRNA library targeting TP53 created to determine the functional impact of a wide array of mutations [4]. In addition, a negative selection screen using 7500 pegRNAs targeting SMARCB1 resulted in a reduction in cells with LoF variants, thus demonstrating that PE-based pooled screens can accurately identify deleterious mutations [25]. With these technological advancements, functional genomics has undergone a significant transition from traditional gene-level knockouts to precise, variant-level editing [Figure 2]. CRISPR-Cas9-mediated SGE, base editing-based SGE, and saturation PE combined have made a robust and flexible framework for large-scale, precise functional analysis of human genetic variations. All of these advancements are accelerating progress in genome annotation, disease risk assessment, and the discovery of new therapeutic targets, which will eventually help in the realization of precision medicine.

3. Components and Mechanism of PE

The PE system has two key components: a PE protein consisting of a Cas9 nickase (H840A) from Streptococcus pyogenes that functions as a programmable DNA-targeting module to introduce a single-strand nick, fused to the RT, usually sourced from Moloney murine leukemia virus (M-MLV) and pegRNA that contains a spacer sequence that pairs with the complementary target DNA strand, and the Cas9-binding scaffold region plus a 3′ extension encoding a PBS and RTT specifying the desired edit [Figure 3A] [21].

The PE complex starts the PE process after it is delivered into the cell.

The PE complex binds to the DNA sequence, with the help of the spacer sequence of the pegRNA. After that, the nCas9 (H840A) protein makes a nick in the non-target DNA strand, resulting in an exposed 3′-single-stranded DNA (ss-DNA) end, which serves as a primer for reverse transcription.The PBS of the pegRNA anneals with exposed 3′-single-stranded DNA, thereby allowing the associated RT to elongate the 3′ end, utilizing the RTT encoded in the pegRNA.

This method produces two flaps of single-stranded DNA: a 5′ flap consisting of the original sequence and a 3′ flap with the desired edits. The 5′ flap is subsequently removed by endogenous structure-specific endonucleases, thereby ligating the 3′ edited flap and creating a heteroduplex. Then, cellular DNA repair processes, particularly the mismatch repair (MMR), resolve the heteroduplex pairs by using the edited strand as a template, thereby ensuring the permanent integration of the intended mutation into the genome [Figure 4] [21,26]. Because PE does not rely on donor templates or DSB-induced HDR, it typically results in fewer unintended mutations (indels or rearrangements) than CRISPR nuclease editing, and it is active in non-dividing cells, which are often poor at HDR.

4. Development of PE

Since the first PE1 came out in 2019, researchers have made several advancements in order to enhance editing efficiency, versatility, and precision. These developments encompass multiple domains, such as engineering the prime editor protein, optimizing the design of pegRNA, and modifying biological processes (such as MMR). This section summarizes significant advancements in prime editor generations and pegRNA engineering to improve stability and efficacy. In addition to these largely rational design-based improvements, directed evolution has emerged as a complementary approach to further optimize prime editor performance.

4.1. Variants of Prime Editors

The first-generation PE is composed of a Cas9 (H840A) nickase fused to the M-MLV RT, and this protein is guided by a pegRNA, a modified form of sgRNA, which specifies the genomic target site and encodes the desired edit. Its editing efficiency in human cells was relatively modest, typically less than 6% for point mutations and up to 17% for small insertions/deletions [21,27].

Anzalone and his team engineered five amino acid substitutions within the wild-type (PE1) RT (M-MLV RT) domain to increase its editing efficiency. The mutations D200N, L603W, and T330P were introduced to enhance thermostability, while T306K and W313F were added to strengthen the interaction between RT and the pegRNA. The second-generation PE (PE2) shows much more efficacy, demonstrating a 1.6- to 5.1-fold increase in editing efficiency relative to PE1, while retaining a low off-target rate (<0.5%) in human cells [Figure 3A;

Table 2. Comparison of PE1, PE2, PE3, PE3b, PE4, PE5, and PE max systems.

PE Type	Cas Protein	Components	Editing Efficiency	Key Features
PE1 (WT)	SpCas9 (H840A) nickase + wild-type M-MLV RT	pegRNA (spacer + PBS + RTT)	≤6% (point mutations); ≤17% (small insertions/deletions)	Without DSBs and donor DNA; enables all 12 base substitutions and small insertions or deletions; limited efficiency.
PE2	SpCas9 (H840A) nickase + engineered M-MLV RT (mutations D200N, T306K, W313F, T330P, and L603W)	pegRNA (spacer + PBS + RTT)	Moderate (up to ~20–50% for point mutation and deletions)	Improved efficiency.
PE3	Same as PE2	pegRNA + additional sgRNA to nick the nonedited strand	Higher (up to ~60%)	Increases editing efficiency but can introduce indels due to dual nicks.
PE3b	Same as PE2/PE3	pegRNA + nicking sgRNA (but nicking sgRNA targets only after successful editing)	Moderate to high	Reduces indel formation compared to PE3; enhances precision.
PE4	Same as PE2/PE3	pegRNA + MLH-1 dd	Higher in human pluripotent stem cells	Improves editing efficiency by suppressing MLH-1.
PE5	Same as PE2/PE3	pegRNA ± nicking sgRNA + MLH-1 dd	Higher in human pluripotent stem cells	Improves editing efficiency by suppressing MLH-1.
PEmax	Codon-optimized Cas9(H840A)-RT (PEmax)	Optimized pegRNA + (optional) nicking sgRNA	High (up to 70% in some systems)	Improves editing efficiency by adding bipartite SV40 NLS.

] [21]. After the incorporation of the edits into the 3′ flap and the 5′ flap is cut away, the major challenge of PE starts. This process creates a heteroduplex DNA structure, which contains a mismatch between the edited and unedited strands. The cellular MMR pathway can find a heteroduplex DNA structure and repair it. If MMR chooses the nonedited strand as the template, the edit is reverted, which limits overall PE efficiency [28].

To overcome this limitation, Liu and coworkers adopted a strategy previously applied in base editing, in which a secondary sgRNA is used to introduce a nick in the nonedited DNA strand. This modification gave rise to the third-generation PE system (PE3). Similar to base editing, PE3 enhances editing efficiency by promoting the repair of the nicked strand using the edited strand as a template; however, unlike base editing, which employs a single, controlled nick, PE already introduces an initial nick during reverse transcription. In this strategy, nicking the nonedited strand at a distance of 70–100 base pairs from the pegRNA-induced nick was found to be optimal, boosting editing efficiencies by 1.5- to 4.2-fold over PE2 in HEK293T cells [Figure 5; Table 2] [21,26,29]. Nevertheless, the concurrent occurrence of two nicks on opposite DNA strands can occasionally lead to DSBs. This can lead to undesirable byproducts, including indels. To reduce these undesired byproducts, Liu and his coworkers created a modified PE system called PE3b. In the PE3b system, nicking sgRNA is designed to precisely recognize the edited sequence and induce a nick on the nonedited strand [Figure 5; Table 2] [21,29]. This approach ensures that the second nick is introduced only after the desired edit, hence reducing the frequency of indels. MMR is a significant challenge for PE because it often reverts to the original sequence by fixing mismatches. The PE3 and PE3b systems partially overcome this challenge by nicking in the nonedited strand, favoring the use of the edited strand as a template for repair. Another way to improve editing efficiency is to directly suppress MMR activity. Therefore, Chen et al. studied the effect of the MMR pathway by silencing MMR genes using siRNA, which enhanced the editing efficiency and decreased unwanted editing byproducts [28,30].

Building on this observation, Liu and colleagues co-transfected the PE2 system and different dominant-negative MMR protein (MLH1dn) variants into HEK293T cells to evaluate their impact on PE efficiency and precision. In MMR-proficient HEK293T cells, MLH1dn expression potently and dose-dependently increased PE efficiency; in contrast, this effect was absent in MMR-deficient cell lines, thereby confirming that the observed improvement was specifically attributable to MMR pathway inhibition [29]. This principle led to the development of the PE4 and PE5 systems. The PE4 system (PE2 + MLH1dn), which combines the PE2 editor with MLH1dn, demonstrated a substantial increase in editing efficiency, reportedly 6.5-fold over the PE2 system. Similarly, the PE5 system (PE3 + MLH1dn) showed a more modest but consistent 1.9-fold enhancement over PE3 across multiple loci in human cells [Figure 6; Table 2] [21,26,31].

4.2. Improving the PE Protein

Effective cellular gene manipulation is dependent upon the efficient nuclear import of PEs or genome editing tools. The nuclear pore complex governs macromolecular transport into the nucleus, a process mediated by nuclear localization signal (NLS). Research has demonstrated that the optimization of the (NLS) sequence can significantly influence the efficiency of gene editing. The original PE2 editor is structured with bipartite SV40 nuclear localization signals at both termini (N and C), which facilitate its transport into the nucleus. Its two core functional domains, the SpCas9 H840A nickase and the engineered M-MLV RT, are connected by a flexible 32-amino-acid peptide linker, allowing coordinated activity during genome editing [26,29]. Liu et al. introduced the PE2* variant by incorporating a c-Myc NLS at the N-terminal end and a hybrid bipartite SV40 NLS at the C-terminal region of the prime editor protein. This structure significantly increased the efficiency of nuclear import, which enhanced the overall performance of the system in mediating base substitutions and small insertions and deletions relative to the original PE2 construct [29,31]. Building upon these insights, a highly optimized variation called PEmax (prime editor max) was created. Here, the PE2 framework was further improved by the addition of a bipartite SV40 NLS in the linker region between the RT and Cas9 domains, the C-terminus was tagged with a c-Myc NLS, and both the RT and nCas9 sequences were codon-optimized. Compared to the original PE2 design, these combined changes resulted in a dramatic increase in editing efficiency [28,32]. In HeLa cells, the PE2max, PE3max, PE4max, and PE5max systems, all derived from the PEmax architecture, were able to demonstrate substantially higher editing efficiencies on a consistent basis, thus pointing out the importance of NLS configuration and codon optimization for PE performance enhancement [Figure 7; Table 2] [28].

4.3. Improving PegRNAs

4.3.1. Engineering of 3′ Extension

RNA molecules are naturally unstable and easily degraded by endogenous or exogenous RNases. In the case of the PE system, a pegRNA has exposed regions of the PBS and RTT that make it especially vulnerable to RNase degradation. When these regions are degraded, the pegRNA becomes shorter (truncated) and reduces editing efficiency. To address this issue, Nelson et al. developed engineered pegRNAs (epegRNAs). They incorporated RNA structural motifs (tevopreQ1 or frameshifting pseudoknot) into the 3ʹ end of a pegRNA to shield them from exonuclease-mediated degradation. This modification led to a 3- to 4-fold increase in editing efficiency across multiple mammalian cell types while preserving low levels of off-target activity and unintended byproducts [Figure 3B] [33]. Zhang et al. improved pegRNA stability by fusing viral exoribonuclease-resistant RNA (xrRNA) motifs at the 3′ end thus, the RNA was effectively protected from exonuclease degradation. The editing effectiveness of both PE2 and PE3 systems was raised across diverse genomic sites and cell types by this modification. Incorporating the xrRNA into the PE3 system (xrPE) in N2A cells was able to achieve a 9.1-fold increase in efficiency at six insertion and deletion loci, while in HEK293T cells, xrPE was 5.5 times more effective than the standard PE3 system at the same targets. Significantly, this modification did not cause any compromise in editing precision, as xrPE had similar insertion/deletion ratios and low off-target activities [34]. Feng et al. enhanced pegRNA stability by appending a 3′ stem–loop aptamer (MS2) to the 3′ end of pegRNAs, generating stem–loop prime editors (sPEs). This modification strengthened the interaction between PE2 and the pegRNA, resulting in an average 1.8-fold increase in editing efficiency in HEK293T and other cell types [Figure 3C] [35].

Adding a G-quadruplex motif in the 3′ extension of pegRNAs boosted editing efficiency to over 80% at the target sites [36]. Overall, integrating various RNA motifs into pegRNAs can help maintain the stability of their 3′ ends and serve as a general strategy to enhance editing efficiency however, the enlarged RNA size may make pegRNA synthesis and production more complex. To address the difficulty of producing long, modified pegRNAs, Feng and colleagues developed the “split pegRNA prime editor” (SnPE). This system separates the standard pegRNA into two parts, namely, a sgRNA and a prime RNA (pRNA), with the pRNA designed to bind to engineered RNA-binding proteins fused to the Cas9 nickase [Figure 3D] [35]. Beyond degradation, pegRNA circularization, resulting from self-annealing between partially complementary sequences at the 3′ and 5′ ends, can also compromise editing efficiency. To overcome this, Liu et al. developed the enhanced PE (ePE) system, in which both the pegRNA and nick-sgRNA are expressed together in a single RNA molecule. The Csy4 enzyme then precisely cleaves this transcript to release the functional RNAs. Furthermore, a U•A to C•G base pair flip in the crRNA scaffold enhances RNA stability and overall editing [37].

4.3.2. Optimization of PegRNA Sequences

As previously discussed, the efficiency and precision of PE are frequently obstructed by the cellular MMR system. However, the fact that the MMR pathway is less sensitive to single-base substitutions, in particular, G-to-C changes, has allowed scientists to figure out that the MMR system can be escaped if several mutations are introduced at once. To put it simply, this route is the least reactive to single-base substitutions and, in particular, to G-to-C changes. Based on these details, scientists came up with a hypothesis that the edited sequence would be less recognizable by the MMR system if it were corrected with the help of silent or harmless mutations introduced close to the target site. Eventually, they would obtain higher PE efficiency.

Li et al. applied this concept by designing two modified pegRNAs for the PE3 and PE5 systems. In the experiment, they integrated synonymous (silent) mutation within the RTT region, resulting in a new PE system named same-sense (silent) mutation-enhanced pegRNA (spegRNA). The optimized spegRNA significantly boosted editing efficiency by 353 times compared to the normal pegRNA. Additionally, they discovered that by simply changing less than four bases in the RTT region, the editing results could be greatly improved. In another design, they developed an altered structure pegRNA (apegRNA), in which C/G base pairs were either inserted or substituted for non-C/G pairs within the small hairpin structure of the pegRNA. This modification stabilized the pegRNA’s secondary structure. As these strategies were aimed at different parts of the pegRNA, the combination of them resulted in a very significant increase in editing efficiency at multiple genomic sites, without accuracy being affected [38].

4.4. PE for the Large Gene

Prime editing has changed the way genomes can be engineered by making it possible to perform very precise small-scale changes, such as point mutations, short insertions, and deletions, without cutting the DNA DSBs or without the need for donor templates. Nevertheless, the initial version of the tool was capable of only relatively small changes; thus, it could not be used for large-scale genomic manipulations like long deletions, large insertions, or chromosomal rearrangements. To overcome these limitations, paired pegRNA strategies, an approach in which two pegRNAs are simultaneously used to direct coordinated editing at two distinct and often distal genomic sites, have been developed. Paired pegRNA strategies expand the editing window, allowing large deletions, sequence replacements, and complex rearrangements with higher accuracy, without relying on conventional DSB-dependent repair pathways.

PRIME-Del (PE-Mediated Deletion), developed by Jay Shendure and colleagues, is an advanced PE approach designed to mediate large genomic deletions of up to 10 kilobases, and short insertions (typically <30 bp) can be added at the deletion junction if desired [39]. In this strategy, two pegRNAs are designed to target two different sites on the same DNA sequences that can be even a few kilobases apart. Each pegRNA carries a homology arm that goes beyond the opposing nick site, thus allowing coordinated editing over a large genomic interval. In a situation where both target sites are nicked and reverse transcription is started, the DNA sequence that lies between the two pegRNA-directed nicks is removed. Optionally, both pegRNAs can encode a short insertion that becomes incorporated at the junction formed after deletion [Figure 8]. A key mechanistic insight of this system is that each 3′ flap representing the newly synthesized single-stranded DNA produced by reverse transcription can independently initiate DNA repair. This means that both flaps do not need to resolve simultaneously, thereby increasing the efficiency, robustness, and flexibility of the editing process compared with other multi-pegRNA systems. PRIME-Del has achieved editing efficiencies of up to 25% at endogenous loci, which is quite exceptional for large-scale genomic changes [39]. Twin PE, a PE tool developed by David Liu and his team, is a robust PE strategy aimed at replacing or inserting relatively long DNA sequences up to 113 base pairs between two nick sites directed by two pegRNAs. In the case of TwinPE, two pegRNAs are used, each of which produces complementary 3′ flaps carrying the new DNA sequence. These 3′ flaps are not homologous to the genome rather, they are intended to anneal with each other, thereby creating a double-stranded DNA segment that carries the code for the insertion or replacement. Mechanistically, the two pegRNAs direct the editing of two opposite DNA strands at different locations, thus generating complementary 3′ flaps that hybridize to form a duplex with the desired edit. The DNA originally between the two nick sites is removed, and the newly made DNA duplex is joined to the genome. Since the 3′ flaps in twinPE do not have to re-anneal with the genomic DNA, the method does not face competition from the endogenous sequence and thereby gets rid of several inefficient DNA repair steps, such as 3′ flap invasion, 5′ flap displacement, and heteroduplex resolution that can greatly slow down traditional PE [Figure 9]. Therefore, TwinPE reaches very high editing efficiency (up to 90%), practically no indel byproducts are generated, and it is very powerful for large sequence replacements and accurate insertions [40]. Building on TwinPE, GRAND allows large insertions and deletions of DNA fragments up to 250 bp in non-dividing cells, thereby expanding PE’s therapeutic potential in quiescent tissues [41]. Xue and team introduced a PE variant called PEDAR (PE-Cas9-based deletion and repair), which is an inventive method to accomplish large genomic deletions (up to 10 kb) along with precise short insertions (<60 bp) in a single step [42]. PEDAR works differently from the PRIME-Del and TwinPE system, which depend on a nickase Cas9–RT fusion. This method utilizes a Cas9 nuclease–RT fusion, resulting in two DSBs being introduced at target sites by paired pegRNAs. The RT domain of each pegRNA, during repair, produces short 3′ overhangs that contain the sequence of the desired insertion. The overhangs thus formed are then linked by microhomology-mediated end joining (MMEJ) or single-strand annealing (SSA) repair pathways, thereby deleting the DNA fragment between the overhangs and inserting the new sequence at the same time. While PEDAR can reach efficiencies of up to 27%, it is still frequently associated with a higher proportion of indel byproducts due to the error-prone nature of DSB-based repair [42]. In the same way, PETI (prime editor nuclease-mediated translocation and inversion), an advanced PE approach invented by Kim and coworkers, leverages the capabilities of PEDAR by not only enabling targeted deletions and insertions but also precise inversions and inter-chromosomal translocations [Figure 9] [43]. Similar to PEDAR, PETI employs a Cas9 nuclease–RT fusion, which, upon two pegRNA-defined sites, introduces DSBs. The reverse transcription carried out by each pegRNA results in a 3′ single-stranded flap, and this can be constructed to be homologous to the sequence for inversion or translocation. These tailored flaps serve as ligands for the recombination of DNA ends in a predetermined way, thus effectively accomplishing the targeted rearrangement [Figure 9]. Consequently, PETI can link DNA fragments from different chromosomes or simply flip a genomic region without changing the rest of the genome, therefore delivering a potent instrument for the study and manipulation of large-scale genome architectures. Nevertheless, like all other DSB-based systems, PETI is sometimes characterized by low editing purity and the likelihood of off-target indels or rearrangements, due to the fact that the occurrence of multiple DSBs can lead to the activation of error-prone repair pathways [43]. Most paired-pegRNA methods are still limited by the availability of the PAM site, the variability in the editing efficiency in different cell types, and the dependence on endogenous DNA repair pathways, despite these advances. These inventions, in combination, substantially widen the limits of PE, thus making it possible to carry out large and accurate genome changes of a different kind while also pointing out that it is necessary to make further improvements in order to increase the safety, the adaptability, and the efficiency of different cellular environments.

5. Delivery Methods for PE Systems

Precise genome modification and the therapeutic potential of PE in genetic disease correction depend on the efficient delivery of PE components into target cells. Since PE depends on the coordinated delivery of the nCas9–RT fusion protein along with pegRNAs and, when necessary, secondary sgRNAs, introducing this large molecular complex into cells both in vitro and in vivo remains a significant challenge [29]. Current approaches can be broadly classified into three different trade-offs in terms of efficiency, specificity, and clinical applicability.

5.1. Physical Methods

5.1.1. Electroporation

Electroporation is a widely used technique for introducing gene-editing components into cells. It uses short electrical pulses to temporarily open the cell membrane, allowing these molecules to enter. Owing to its simplicity and reproducibility, it is extensively utilized to assess PE efficiencies in vitro [Figure 10A]. For example, Kim and colleagues employed the PE3 and PE3b systems to rectify a pathogenic G→A mutation within Fah exon 8, which is implicated in hereditary tyrosinemia type 1 (HT1). In chemically derived hepatocytes (CdHs), they observed an average correction efficiency of 2.3%, with no discernible bystander effects furthermore, transplantation of the corrected CdHs substantially extended survival in HT1 mice [44]. Likewise, Hong et al. used the electroporation technique to deliver the PE3 editing tool for the correction of COL7A1 mutations in primary fibroblasts obtained from patients with recessive dystrophic epidermolysis bullosa (RDEB). The efficiency of editing reached up to 10.5% with minimal off-target activity (~1%) and the restoration of functional type VII collagen [45]. Sousa et al. [46] recently optimized PE to correct the CFTR F508del mutation in cystic fibrosis cells. They accomplished this by using epegRNAs, PEmax, and MLH1dn co-expression, along with silent edits. This technique yielded correction efficiencies of 58% in cystic fibrosis cells, with minimal off-target effect [46]. Although electroporation is efficient and easy to use, it can cause transient mechanical stress and alter the phenotype of cells, thereby limiting its direct clinical applicability.

5.1.2. Lipofection

Lipofection utilizes the cationic lipid-based transfection reaction to enable the delivery of gene-editing components. These techniques involve the formation of lipoplexes with negatively charged nucleic acids, which enhances cellular uptake and protects the cargo from enzymatic degradation. Commercially available reagents, such as Lipofectamine 2000 and Messenger MAX, are frequently utilized for PE delivery [21,29] [Figure 10B]. Anzalone et al. utilized Lipofectamine 2000 to deliver the PE3 system across various mammalian cell lines, resulting in a targeted modification efficiency exceeding 50%. In the HEXA 1278 + TATC mutant (4-nucleotide insertion) HEK293T cell (Tay–Sachs disease model), 43 pegRNAs were evaluated, resulting in a median editing efficiency exceeding 20% (maximum 33%) [21,27]. Building on this foundation, Nelson et al. later enhanced the structural design of pegRNAs by epegRNAs that improved PE efficiency in plasmid lipofection of Tay–Sachs patient-derived fibroblasts [33]. Despite its utility for in vitro optimization, lipofection presents major challenges for PE delivery. The large size of PE constructs (nCas9–RT fusion + pegRNA ± sgRNA; >6–7 kb) limits transfection efficiency and increases endosomal degradation. Lipid-based reagents also perform poorly in primary, stem, and non-dividing cells, such as neurons and hepatocyteskey targets for therapeutic genome correction. Moreover, high reagent concentrations can cause cytotoxicity and stress responses, and transient, heterogeneous expression restricts editing duration and consistency. Finally, standard lipid formulations exhibit poor stability, biodistribution, and tissue targeting in vivo, confining their use to experimental contexts rather than clinical application [21,33,47].

5.1.3. Hydrodynamic Injection

Hydrodynamic injection offers an effective approach for directly introducing gene editing tools and genetic materials into target organs or tissues via intravenous (i.v.) administration. This method has shown promise for in vivo gene delivery applications by successfully delivering plasmid DNA to the mouse liver [48,49,50] [Figure 11C]. Building on this basis, Liu et al. corrected the SERPINA1 PiZ mutation, which is linked to α₁-antitrypsin deficiency (AATD), within a murine model using hydrodynamic injection. After 45 days, delivery of PE2 and an optimized PE2* plasmid produced a 6.7% gene repair rate with minimal indel formation (0.4–2.7%), outperforming the 2.1% repair rate of the traditional PE2 system [31]. Similarly, Jang et al. delivered plasmids encoding the PE3 system to HT1 mice, resulting in 61% FAH⁺ hepatocytes and 11.5% precise edits by day 40, furthermore, off-target edits were significantly diminished (0.78%) when contrasted with Cas9-HDR (26%) and ABE (1.9%) [51]. Hydrodynamic delivery offers an efficient experimental platform for assessing in vivo PE in animal models, but its clinical translation is still limited, requiring the development of safer and more targeted delivery modalities. The technique can cause toxicity and hemodynamic stress, yields variable and liver-biased transfection efficiency, and requires large-volume, high-pressure injections incompatible with human use. These factors limit precise tissue targeting and reproducibility [50,52,53,54]. Therefore, despite its utility as an effective experimental tool for assessing in vivo PE within animal models, the clinical translation of hydrodynamic delivery is currently restricted, thereby underscoring the need for the development of safer and more targeted delivery strategies.

5.2. Viral Delivery Vectors

Viruses have an in-built mechanism to replicate very quickly when they enter a host, thereby producing a multitude of copies. Scientists have capitalized on these inherent viral traits to engineer synthetic viral vectors. These vectors are engineered to transport therapeutic agents or components for genomic modification into cells. Consequently, viral vectors continue to be a potent approach for gene delivery, and they are extensively employed in both in vivo gene therapy and genome editing investigations.

5.2.1. Lentiviral Vectors

Lentiviral vectors (LVs) are derived from the Human Immunodeficiency Virus (HIV), a single-stranded RNA virus capable of integrating its genetic material into the genomes of both dividing and non-dividing cells. They accomplish this by integrating foreign genes into the host genome in a stable manner through reverse transcription of exogenous RNA. By eliminating unnecessary viral sequences, LVs can transport comparatively large genetic cargos (~10 kb). Pseudotyping, in which the viral envelope is substituted with glycoproteins from other viruses (such as vesicular stomatitis virus G [VSV-G] protein), can increase the host cell range and targeting specificity of conventional LVs despite their limited natural tropism. Jang et al. demonstrated the viability of LV-mediated PE library delivery [51] [Figure 11A] by building and delivering a lentiviral pegRNA library with hundreds of guide variants for functional screening in HEK293T cells. Despite these benefits, integrating LVs into host genomes raises safety concerns for clinical translation due to the possibility of insertional mutagenesis, chimeric transcript formation, and long-term off-target effects. To address safety concerns associated with the genomic integration of LVs, integration-deficient lentiviral vectors (IDLVs) have been developed through the introduction of mutations in the viral integrase. IDLVs enable efficient transient delivery of PE components without stable integration into the host genome, thereby substantially reducing the risk of insertional mutagenesis. Although editing efficiencies achieved with IDLVs are generally lower than those obtained using integrating lentiviral systems, IDLVs represent a safer alternative for PE applications, particularly in therapeutic and in vivo contexts. Notably, a recent study evaluated the feasibility of IDLVs for the transient delivery of all-in-one PE constructs driven by a single EF-1α promoter in human induced pluripotent stem cells however, the observed editing efficiencies were low, underscoring the need for further optimization of IDLV-based PE delivery strategies [55].

5.2.2. Adenoviral Vectors

Adenoviral vectors (AdVs), which are double-stranded DNA viruses from the Adenoviridae family, deliver genetic material that remains episomal and does not integrate into the host genome. Because of their high transduction efficiency and large packaging capacity, AdVs have been studied as delivery systems for PE in human cells. Wang and his team developed a fully gene-deleted adenoviral vector carrying a PE2 construct. This vector showed strong editing efficiencies in HeLa and HEK293T cells, thus demonstrating the potential of AdV-mediated PE delivery [56] [Figure 11B]. Böck and colleagues, in a similar vein, employed an AdV-based PE3 system to rectify a pathogenic mutation linked to phenylketonuria (PKU), demonstrating up to 22.2% correction in vitro and 14.4% in vivo, with no observable off-target effects, and ultimately normalizing phenylalanine levels in a murine model [57]. Notwithstanding these encouraging outcomes, the systemic delivery of AdVs can elicit robust innate immune responses and the production of inflammatory cytokines, potentially diminishing therapeutic effectiveness. Consequently, current investigations are centered on the development of vectors derived from rare or non-human AdV serotypes, such as AdV-26, AdV-35 (species B), and AdV-48 (species D), which exhibit diminished immunogenicity, thus enhancing safety and performance in vivo [58].

5.2.3. Adeno-Associated Viral Vectors

Adeno-associated viral vectors (AAVs), which are single-stranded DNA viruses belonging to the Parvoviridae family, demonstrate favorable biocompatibility and minimal immunogenicity, thereby rendering them ideal for in vivo gene delivery. However, a significant limitation is their restricted packaging capacity, which is around 5 kb. This is a problem because the combined size of the prime editor (PE) and its corresponding pegRNA exceeds this limit [59] [Figure 11C]. To overcome this problem, Liu and his team developed a split prime editor (sPE) system that uses two AAVs. In this system, one vector carries the nCas9, while the other vector contains the RT, pegRNA, and nicking sgRNA components. Following co-transduction into the target cells, the two AAV vectors undergo recombination to reconstitute a functional prime editor, thus facilitating efficient genome editing [59,60]. High-capacity AdVs, which lack viral coding sequences and can package about 30 kb of DNA, have also been used to deliver prime editors, leading to strong editing results in cell cultures [61]. The dual-AAV strategy has also been successfully used in animal studies. In a mouse model of tyrosinemia type I, delivering split PEs together corrected the mutation causing the disease and improved the related symptoms [60]. Using this approach, Liu and colleagues reported a 3.1% correction of α1-antitrypsin deficiency through dual-AAV delivery. In contrast, Böck and his team achieved 14.4% G-to-C editing at the Dnmt1 locus [32]. Further improvements to AAV-PE systems, including codon optimization of PE, stabilization of pegRNAs, and changes to the AAV genome structure, allowed for efficient editing in the mouse brain (42%), liver (46%), and heart (11%), with few off-target effects, as shown by Davis et al. Sub-retinal administration of AAV-PE2 yielded a 6.4% restoration of visual function in RPE65-mutant (LCA) mice [62]. Notwithstanding these advancements, the wider implementation of AAV-mediated PE delivery faces several obstacles. These encompass the limited packaging capacity, the reduced recombination efficiency observed in dual-AAV systems, the immune reactions against viral capsids, the potential for vector genome integration at DSBs, the presence of pre-existing anti-AAV antibodies in humans, and the technical intricacies associated with the production of dual-AAV vectors.

5.3. Nonviral Delivery Vectors

Nonviral delivery systems present a potentially advantageous approach for administering PE components, offering several benefits including greater packaging capacity, diminished immunogenicity, adaptable design, targeted tissue delivery, and scalability for clinical use. A range of nonviral carriers, such as lipid nanoparticles (LNPs), virus-like particles (VLPs), polymeric, peptide-based, and inorganic nanoparticles, have been investigated for the delivery of CRISPR-Cas9 and base editors, although optimized systems specifically for prime editors are still in development. In accordance with other genome editing platforms, prime editors can be delivered as plasmid DNA (pDNA), messenger RNA (mRNA), or ribonucleoprotein (RNP) complexes [63]. Delivering the PE complex as an RNP is difficult because of its large size. Therefore, most studies use mRNA or DNA delivery methods. mRNA is preferred because it works quickly, produces temporary expression, reduces unintended effects, and avoids the risks associated with DNA integration [64,65,66]. Among nonviral methods, LNPs are the most advanced and widely used for mRNA delivery. LNPs are made up of ionizable lipids, cholesterol, phospholipids, and PEG lipids that self-assemble with mRNA to form nanoparticles [66].

Herrera-Barrera and his coworker developed an improved RNA delivery by enhancing LNPs (eLNPs) with β-sitosterol, which achieved 54% PE efficiency in cell models [67]. Another nonviral delivery approach involves the use of VLPs, derived from lentiviral or retroviral sources, that show promise for delivering genome-editing tools. These VLPs can carry proteins or mRNAs, and they can be modified (pseudotyped) with different envelope proteins to target specific cells or improve stability. In addition, VLP delivery reduces unwanted indels, off-target effects, and the activation of p53-dependent apoptosis by limiting the prolonged expression of nucleases. GAG-VLPs can carry prime editor ribonucleoproteins (PE-RNPs), since they can deliver RNA-bound proteins up to 286 kDa, even though the exact packaging limit is not known. Although VLPs are composed of viral structural proteins and exploit virus-like cellular entry pathways, they lack viral genetic material and replication capacity and are therefore functionally classified as nonviral delivery vectors. Meirui An et al. engineered VLPs to deliver the PE system both in vitro and in vivo by optimizing pegRNA packaging and adding nuclear export signals. These engineered PE-VLPs (eVLPs) achieved 15% editing efficiency in a mouse model of ocular disease [68]. Further, Halegua et al. developed a PE-VLP system, called Nanoscribes, which was optimized both in the prime editor itself and in the design of the particle. Nanoscribes achieve gene editing efficiencies comparable to DNA transfection, with higher target fidelity. They support multiplexed editing and can successfully deliver prime editors into primary cells, human iPSCs, and their derivatives, including hematopoietic stem cells. In HEK293T cells, Nanoscribes reached up to 68% editing at the HEK3 locus, and in myoblasts, iPSCs, and iPSC-derived hematopoietic stem cells, editing efficiencies reached up to 25% [69]. Haldrup and colleagues developed an advanced lentivirus-derived nanoparticle system, LVNP3.0, for the delivery of PE and epegRNA ribonucleoprotein (RNP) complexes. By removing RT and integrase, LVNP3.0 eliminates insertional mutagenesis risk while retaining pseudotyping for cell-specific targeting. Delivering RNPs directly bypasses viral packaging limits, and optional vector genome inclusion allows further flexibility, making LVNP3.0 a modular, non-integrating platform with strong potential for in vivo genome editing applications [70]. Despite these advances, nonviral delivery systems still encounter important barriers.

Their short stability of time in vivo can restrict the duration of gene-editing activity. Many times, their production is costly, and they are very sensitive to freezing and thawing. In order to solve such problems, scientists are investigating chemical modifications, pH-responsive materials, and lyophilized formulations, all of which are directed at stabilizing, facilitating, and increasing the therapeutic potential of nonviral prime-editing platforms. Significant advances have been made in PE, but the efficient and safe delivery of the editing tool is still a big challenge.

Physical methods can be useful experimental platforms, but they are limited by cytotoxicity and the small area of tissue that can be targeted. Viral vectors ensure efficient transduction but are limited by issues such as the size of the cargo, immunogenicity, and safety. Nonviral technologies, like lipid nanoparticles, virus-like particles, and engineered lentivirus-derived nanoparticles (for example, Nanoscribes, LVNP3.0), provide a transient, tissue-targeted delivery with less immunogenicity and better safety. Nevertheless, their lower effectiveness in comparison with viral vectors and the problem of stability require that they be further refined before they can be used for safe, accurate, and scalable clinical applications.

6. PE for Functional Interrogation of Non-Coding Regulatory Variants

The high precision and low off-target activity of PE make it well-suited for studying non-coding regulatory elements in their natural genomic context. Prime editing can introduce any single nucleotide change, as well as small insertions or deletions, enabling detailed analysis of enhancers, promoters, splice sites, and untranslated regions (UTRs). Prime editing allows researchers to examine enhancer sequences one DNA base at a time to determine which bases are important for gene regulation. Ren et al. (2023) developed a pooled prime-editing screen to introduce every single nucleotide substitution called PRIME across a 716 bp enhancer near the MYC gene [22]. This approach identified specific nucleotides that are essential for enhancer function and strongly influence cell fitness, revealing which bases are required for proper MYC regulation. The study demonstrated that editing near transcription factor sites induces stronger phenotypic effects. As a result, this approach can precisely map critical enhancer motifs one base at a time and is particularly well suited for studying c-Myc super-enhancers that contribute to drug resistance in leukemia-derived K562 cells [71]. These examples illustrate how prime editing can systematically evaluate enhancer variants, directly linking specific base changes to regulatory activity. Similarly, prime editing has been used to dissect core promoter regions and their transcription factor (TF) binding motifs [72]. Prime editing can also be used to study DNA changes that affect RNA splicing. In a 2025 study, Lopes da Costa et al. used prime editing to introduce a disease-causing splice-donor mutation into the PRPH2 gene in human iPSCs, creating cell models that showed the same abnormal splicing patterns observed in patients. A single base change activated an incorrect splice site and caused intron retention in the mRNA. When the mutation was corrected using prime editing in the same cells, normal splicing and mRNA production were restored. Importantly, the edited cells showed no detectable off-target effects or chromosomal abnormalities. Their work demonstrates that prime editing can precisely model and correct splice-site mutations, making it a powerful tool for studying splicing regulation and for developing potential therapies [73]. Prime editing has also been used to study DNA changes in the 5′ and 3′ untranslated regions (UTRs), which control mRNA stability and gene expression after transcription. In a recent study, researchers identified rare 3′UTR variants that were predicted to affect mRNA stability and then used prime editing to introduce these variants directly into the genome of human cells [74]. Overall, the study demonstrates that prime editing can precisely link non-coding UTR mutations to changes in gene expression and cellular function. Together, these features make prime editing a transformative technology for regulatory genomics, as it enables functional dissection of enhancers, promoters, splice junctions, and UTRs at single-nucleotide resolution.

7. In Vivo Applications of PE

Several studies have now shown that in vivo prime editing can correct diseasecausing mutations in many organ systems; however, efficiency and delivery challenges remain. The liver has been a major focus because of its central role in metabolic diseases and the availability of robust mouse models. In a phenylketonuria (PKU) mouse model, prime editing of the PAH gene using a dual-vector (split-intein) PE2 system delivered by AAV8 achieved 11% allele correction, which was sufficient to restore phenylalanine metabolism, demonstrating that low-level precise editing can be therapeutically effective [57]. In HT1, Jang et al. (2022) used a PE3 system to correct a FAH mutation in adult mice, resulting in complete liver recovery and long-term survival, with ~61% FAH-positive hepatocytes and minimal off-target effects [51]. Prime editing has also corrected mutations in α₁-antitrypsin deficiency (15% editing) and enabled lipid nanoparticle-mediated editing of PCSK9, highlighting its therapeutic versatility in liver disease [31,75]. Prime editing has shown promise in the eye and brain, two tissues traditionally considered difficult to edit. Early studies demonstrated the feasibility of retinal prime editing in a mouse model of Leber congenital amaurosis, although editing efficiencies were low [51]. More recently, dual-AAV delivery achieved ~26% correction of a pathogenic mutation in a mouse model of retinitis pigmentosa, restoring PDE6β expression, reducing photoreceptor degeneration, and preserving visual function [76]. In the central nervous system, Sousa et al. (2025) reported the first robust therapeutic application of in vivo prime editing by treating Alternating Hemiplegia of Childhood (AHC) caused by dominant ATP1A3 mutations. AAV9-mediated delivery to neonatal mouse brains partially corrected the mutant allele, restored protein expression, improved neurological function, and more than doubled survival, highlighting that partial neuronal correction can yield substantial clinical benefit [77,78]. Prime editing is also being explored for muscle, cardiovascular, and hematological disorders, although these applications remain in the early stages. In Duchenne muscular dystrophy, precise mutation correction has been achieved in patient-derived cells, but efficient in vivo delivery remains challenging [79]. Similarly, while prime editing holds potential for inherited heart and blood diseases, effective delivery to cardiomyocytes and hematopoietic stem cells remains a major hurdle [80,81]. Overall, despite delivery limitations, partial high-fidelity in vivo editing can yield meaningful therapeutic benefits, highlighting prime editing as a promising platform for precision medicine.

8. Conclusions, Limitations, and Future Directions

PE is quickly becoming a strong and adaptable genome editing technology. The combination of this technology with high-throughput functional genomics is expected to radically change the way we understand the relation between genotype and phenotype. By enabling precise, programmable introduction of virtually any variant into a genome, PE overcomes many limitations of earlier CRISPR methods and provides a clean readout of variant function. We now have the technique to conduct “SGE” on specific genes and regulatory elements in their endogenous genomic context, thereby realizing what was once a theoretical goal, the comprehensive functional evaluation of all possible mutations within a given sequence. This will greatly accelerate the interpretation of variants of uncertain significance, which currently predominate in clinical genetic testing. As functional data from PE screens accumulate, we can expect more VUS to be reclassified as either benign or pathogenic, improving diagnostic yield for conditions like hereditary cancers and cardiomyopathies.

Prime editing screens provide a powerful approach in research for the systematic identification of functionally significant regions within proteins, including enzyme active sites and interaction domains, through targeted mutagenesis, as well as the identification of regulatory nucleotides in enhancers and non-coding RNAs critical for gene expression. Such insight not only deepens our understanding of basic biological processes but also aids in the discovery of therapeutic targets. In addition, PE makes it possible to find GoF mutations as well as LoF changes. Cancer research will especially benefit from this the most, where the screening of oncogene variants can identify mutations that lead to drug resistance or sensitivity. However, the issues that challenge us still linger behind this massive breakthrough. Editing efficiencies, while much improved (often 20–50% or higher with PE5 in cell line), can still vary widely by locus and cell type. PegRNA design is more art than science it often requires optimization of the best PBS and RTT lengths for each target. Delivering prime editors into cells, especially primary or in vivo systems, is another challenge because the PEmax construct is quite large (~6.3 kb) and cannot easily fit into AAV vectors.

In order to achieve this, scientists are looking into dual-AAV systems and nonviral delivery methods like nanoparticles. The off-target effects of PE are usually very low, especially if we compare them to Cas9 nuclease, but more experiments in different cell types will be needed to confirm that the technique is safe enough for clinical use. Since PE depends on cellular DNA repair, the results can be different for dividing and non-dividing cells or cells with different p53 or MMR statuses. It may be necessary to carefully optimize (e.g., transient p53 inhibition or MMR modulation) to a certain extent to be able to achieve the maximum level of editing without any lasting side effects.

From a functional genomics perspective, another consideration is throughput and readout. Creating very large pegRNA libraries (tens of thousands) is feasible with current oligonucleotide synthesis technology, and coupling PE to sequencing-based readouts (such as counts of edited vs. nonedited cells after a selection) is straightforward. However, for more nuanced phenotypes (e.g., measuring enzymatic activity or morphology), one might need innovative reporter systems or single-cell readouts that can link genotype to phenotype. In the coming years, we anticipate PE will be increasingly applied to functionalize non-coding variants, which are abundant in GWASs but hard to interpret. By prime-editing candidate variants in cell models or organoids, researchers can directly test if a non-coding SNP affects gene expression, chromatin state, or cellular phenotype. This will bring much-needed mechanistic insight to the non-coding genome, which has so far lagged behind coding regions in functional annotation. Furthermore, PE could facilitate in vivo screens, for example, delivering a pooled pegRNA library to a mouse model to see which edits drive tumor formation or confer drug resistance in an in situ cancer model. Such in vivo functional screens would be incredibly powerful, essentially modeling evolution or disease progression via programmed mutations.

In conclusion, PE represents a new paradigm in functional genomics one where our ability to write genetic changes is finally catching up with our ability to read them. As the technology matures, it is closing the loop between genotypes and phenotypes: variants discovered in sequencing can be rapidly installed and tested for function, creating a virtuous cycle that will refine genetic databases and clinical variant interpretation. The convergence of PE with other advances (like induced pluripotent stem cells, 3D organoids, and single-cell omics) provides a rich toolkit to probe human biology. Challenges notwithstanding, the trajectory suggests that comprehensive, base-by-base functional maps of disease-related genes (and even whole genomes) are an attainable goal in the foreseeable future. The knowledge gained will pave the way for improved genetic diagnostics, where a patient’s variants can be functionally assessed and acted upon, and for precision medicine strategies that target the specific genetic vulnerabilities of diseases. PE and functional screens thus stand as key enablers in the quest to translate the human genome into actionable biology and tailored healthcare.