Emerging trends in gene and cell therapy: CRISPR in DNA editing and beyond

Summary

CRISPR-based gene and cell therapies are rapidly transitioning from experimental platforms to clinical reality, exemplified by the recent approval of CRISPR-derived treatments for β-hemoglobinopathies. This review highlights how advances in genome editing technologies, ranging from CRISPR-Cas nucleases to base and prime editors, are expanding the therapeutic landscape beyond traditional gene knockout approaches. We focus on the clinical translation of these tools, drawing on examples from ongoing and completed human trials to illustrate their potential across diverse disease areas. Furthermore, we discuss critical considerations such as delivery challenges, long-term safety, immune responses, and editing specificity, all of which are critical to the safe and effective integration of CRISPR technologies into modern medicine.

Graphical abstract

Ruijin Ji et al. review recent advances in CRISPR-based genome editing and their application in gene and cell therapies, highlighting clinical progress, delivery innovations, and safety considerations shaping the integration of CRISPR technologies into modern medicine.

Introduction

Gene and cell therapies have emerged as transformative strategies for treating devastating diseases. Gene therapy involves the delivery of nucleic acids to modify gene expression, correct genetic defects, or introduce novel functions, while cell therapy leverages the transplantation of engineered cells to restore or enhance biological activity. Advances in genome editing, particularly CRISPR-Cas (clustered regularly interspaced short palindromic repeats [CRISPR]-associated proteins) systems, have revolutionized the field by enabling precise and programmable genetic modifications.

Initially derived from bacterial defense mechanisms, CRISPR-Cas9 was the first widely adopted genome editing tool capable of efficiently and programmably cleaving the target DNA.^1,2,3 Since then, a broader family of CRISPR-associated enzymes, such as Cas12 and Cas13, has been characterized, and engineered Cas variants have expanded targeting scope and improved editing specificity.^4,5,6,7 Beyond nucleases that induce double-stranded breaks (DSBs), platforms such as base and prime editors enable precise nucleotide substitutions, insertions, or deletions without cutting both DNA strands.^8,9,10 Apart from DNA editing, RNA editing tools based on CRISPR-Cas and adenosine deaminases acting on RNA (ADAR) enzymes enable transient RNA base conversions, offering therapeutic potential for conditions where permanent DNA changes are undesirable.¹¹ This review summarizes the development and clinical translation of these CRISPR-based technologies, discusses current challenges in ongoing trials, and highlights future directions to enhance the therapeutic potential of gene and cell editing.

CRISPR toolbox

Since the discovery of the Cas9 system, CRISPR toolbox has blossomed substantially, advancing our understanding of basic biology and enabling breakthroughs in medicine, agriculture, and biotechnology. Here, we outline the core components of the CRISPR toolbox, including DNA and RNA editing systems, highlighting their specific features, benefits, and challenges in therapeutic applications.

DNA editing systems

Cas9 nucleases

Streptococcus pyogenes Cas9 (SpCas9) was the first Cas nuclease adapted for genome editing, establishing as a cornerstone technology due to its high activity and specificity.^1,12,13 Guided by a single-guide RNA (sgRNA) and requiring a protospacer adjacent motif (PAM) for recognition, SpCas9 induces DSBs through its HNH and RuvC nuclease domains (Figure 1). Selective inactivation of one nuclease domain generates a nickase, while mutating both yields a catalytically inactive Cas9 (dCas9), capable of precise DNA navigation.¹⁴ With the first CRISPR-Cas9-based therapy approved in 2023, genome editing has entered a new clinical era. Beyond SpCas9, a growing number of natural and engineered Cas9 variants, including those developed through generative artificial intelligence (AI), have been identified.¹² These variants differ in size, PAM recognition sequences, guide RNA structure, and editing efficiency, broadening the versatility of the CRISPR-Cas9 toolbox for therapeutic applications. Smaller nucleases, such as SaCas9, Nme2Cas9, and CjCas9,^15,16,17 offer important advantages for in vivo delivery using adeno-associated virus (AAV) vectors, which have limited cargo capacity. Variants with relaxed PAM requirements, such as SpCas9-NG, SpRY, SpG, and AtCas9,^18,19,20 offer greater targeting flexibility, which is especially valuable for correcting pathogenic mutations located in PAM-restrictive regions. While these variants offer important advantages, they may also show reduced editing efficiencies and an elevated risk of off-target activities, underscoring the need for careful optimization and validation in therapeutic development.

Figure 1.

CRISPR toolbox family

CRISPR-based editing encompasses a range of tools, including Cas nucleases, base editors, prime editors, large-scale genomic engineering platforms, RNA-targeting editors, and epigenetic editing systems. Representative substrates and editing outcomes are illustrated alongside the corresponding CRISPR-Cas systems used to achieve these modifications. PAM, protospacer adjacent motif; NHEJ, non-homologous end joining; HDR, homology-directed repair; nCas9, Cas9 nickase; PBS, primer binding site; RTT, reverse transcription template; dCas9, dead Cas9.

Cas12 nucleases

Classified as type V CRISPR systems, Cas12 employs a single RuvC-like nuclease domain to cleave both DNA strands, generating a staggered cut with 5′-overhangs at target sites distal to the PAM⁴ (Figure 1). Unlike the blunt cuts generated by Cas9, the staggered overhangs created by Cas12 may be useful to promote directional DNA insertion when donor templates with complementary overhangs are provided. Moreover, Cas12 can process crRNA arrays into individual functional crRNAs, a property that enables efficient multiplex editing using a single customized crRNA array.²¹ Furthermore, Cas12a exhibited strong trans-cleavage activity against RNA, a feature leveraged for nucleic acid detection.^{22,23,24,25,26,27} For a more detailed discussion of Cas12 applications in nucleic acid detection, readers are referred to other reviews that comprehensively cover this topic.^28,29

Acidaminococcus sp. Cas12a (AsCas12a), the first Cas12 orthologue demonstrated to be active in mammalian cells, recognizes a thymine-rich PAM sequence (TTTV, where V = A/C/G).^4,30 Although early versions of AsCas12a exhibited modest editing efficiency, engineered variants like AsCas12a ULTRA support efficient multiplex editing in primary T cells and was evaluated in clinical trials for sickle cell disease (SCD) and β-thalassemia.^31,32 To date, clinical trials of Cas12-mediated ex vivo therapies have shown increased fetal hemoglobin levels with a favorable safety profile. While these results are promising, the trials were suspended as commercial sponsors shifted priorities toward in vivo therapies. Beyond Cas12a, a growing family of Cas12 enzymes expands the editing toolkit: Cas12b has been optimized to cleave DNA at physiological condition, Cas12e offers a compact alternative, and Cas12f targets single-stranded DNA.³³ Importantly, a recent clinical study showed that Cas12b can be used to engineer allogeneic islet cells to evade immune rejection in a type 1 diabetic patient, enabling engraftment and glucose-responsive insulin secretion without systemic immunosuppression for 12 weeks.³⁴ Although based on a single case, this finding highlights the potential of Cas12b for immune-evasive cell therapies and regenerative medicine.

Cas9- or Cas12-nuclease-mediated genome editing relied on the generation of DSBs, which are repaired through endogenous DNA repair pathways: non-homologous end joining (NHEJ), microhomology-mediated end joining (MMEJ), or homology-directed repair (HDR), if given the presence of donor DNA template.^35,36 NHEJ and MMEJ, which operate without a repair template, often introduce small insertions or deletions (indels), making them valuable for gene disruption. In contrast, HDR uses a homologous DNA template to enable precise editing but is largely limited to dividing cells, as it requires factors active during the S and G2 phases of the cell cycle.³⁵ These repair pathways enable CRISPR-Cas systems to mediate both gene disruption and precise genome editing, expanding their research and therapeutic applications (Figure 1).

Base editors

Base editors enable single-nucleotide changes without generating DSBs or requiring donor DNA templates. These systems typically consist of a Ruvc-inactivated Cas9 nickase (nCas9) fused to a nucleotide deaminase, allowing targeted base conversion while avoiding error-prone repair pathways and minimizing unintended genomic alterations (Figure 1).³⁷ Two major classes of base editors have been developed. Cytosine base editors (CBEs) mediate C⋅G-to-T⋅A conversions by deaminating cytosine to uracil, creating a U⋅G mismatch that is resolved as a T⋅A pair by DNA polymerase.⁸ An uracil glycosylase inhibitor (UGI) is often included to prevent uracil excision and enhance editing efficiency.³⁸ Adenine base editors (ABEs) convert A⋅T to G⋅C by using an evolved adenosine deaminase (derived from Escherichia coli TadA) to deaminate adenine into inosine, which is read as guanine DNA polymerase.⁹ Base editing occurs within a defined editing window of the single-stranded DNA (ssDNA), where the sgRNA hybridizes with the target strand, displacing the non-target strand and exposing PAM-distal nucleotides as ssDNA.³⁹ These nucleotides are accessible to the deaminase, and their compatibility with the catalytic domain determines the efficiency, position, and specificity.

Since their initial development, both CBEs and ABEs have undergone extensive optimization to enhance on-target efficiency, reduce off-target effects, and expand their targeting scope.^{7,12,40,41,42} Engineered variants now support broader base editing outcomes, including non-canonical conversions such as C-to-G,⁴³ A-to-C,⁴⁴ and A-to-T,⁴⁵ further enhancing their versatility. Base editing has advanced toward clinical translation, with several programs targeting hematologic disorders (Table 1) and inherited liver diseases (Table 2) currently in early-stage clinical trials. Despite these advances, base editors face important challenges in therapeutic applications. These include bystander editing of adjacent nucleotides and unpredictable, gRNA-independent RNA and DNA off-target activity compared with Cas9 nuclease.⁴⁶ While difficulties in achieving efficient and tissue-specific delivery are shared with CRISPR/Cas platforms, these challenges can be more pronounced for base editors due to their larger size. Ongoing engineering efforts aim to improve precision, mitigate off-target risks, and develop clinically viable delivery strategies to enable safe and effective therapeutic use.

Table 1.

Clinical trials for ex vivo CRISPR-based editing

Indication	Cell type	Delivery	Editing system	Targeted locus	Trial ID
β-Thalassaemia and/or sickle cell disease	HSC	electroporation	Cas9 knockout	BCL11A	NCT03655678
					NCT05577312
					NCT05356195
					NCT04211480
					NCT03745287
					NCT05329649
					NCT05951205
					NCT05477563
					NCT06300723
					NCT06287086
					NCT06287099
					NCT06647979
			Cas9 knockout	HBG1/2	NCT06506461
			Cas9 HDR	HBB	NCT04819841
					NCT04774536
			Cas12a knockout	HBG1/2	NCT04853576
					NCT05444894
			ABE	HBG1/2	NCT05456880
			tBE	HBG1/2	NCT06024876
					NCT06291961
					NCT06565026
					NCT06328764
					NCT06065189
Acute myeloid leukemia	HSC	electroporation	Cas9 knockout	CD33	NCT04849910
					NCT05662904
					NCT05309733
Chronic granulomatous disease	HSC	electroporation	PE	NCF1	NCT06559176
HIV	T cell	electroporation	Cas9 knockout	CCR5	NCT03164135
CD19+ leukemia and/or lymphoma	T cell	electroporation	Cas9 knockout	TRAC	NCT03232619
				HPK1	NCT04037566
				CBLB	NCT05169489
				TRAC, B2M	NCT03166878
				TRAC, CD52	NCT04557436
					NCT04154709
					NCT04227015
				TRAC, Power3	NCT06014073
		electroporation and AAV	Cas9 knockout, Cas9 HDR	PD1	NCT04213469
				TRAC	NCT05631912
				TRAC, PD1	NCT04637763
				TRAC, HLA-A/B, CIITA, PD-1	NCT06321289
				TRAC, B2M, TGFBR2, Regnase-1	NCT05643742
Multiple myeloma	T cell	electroporation and AAV	Cas9 knockout, Cas9 HDR	PD1	NCT05308875
		electroporation	Cas12a	TRAC, B2M	NCT05722418
CD5+ hematopoietic malignancies	T cell	electroporation	Cas9 knockout	CD5	NCT04767308
CD7+ hematologic malignancies	T cell	electroporation	Cas9 knockout	CD7, TRAC	NCT04264078
					NCT04984356
			CBE	CD52, CD7, TRBC	NCT05397184
				TRAC, CD52, CD7, PDCD1	NCT05885464
CD70+ hematologic malignancies	T cell	electroporation	Cas9 knockout	TRAC, B2M, CD70, TGFBR2, Regnase-1	NCT06492304
Acute myeloid leukemia	T cell	electroporation	CBE	CD52, CD7	NCT05942599
Solid tumors	T cell	electroporation	Cas9 knockout	CISH	NCT04426669
					NCT05566223
				PD1	NCT04417764
					NCT04842812
					NCT05812326
				TGFβR	NCT04976218
				PD1, ACE2	NCT04990557
				PD1, TRAC	NCT03545815
		electroporation and AAV	Cas9 knockout, Cas9 HDR	TRAC, B2M, CD70, TGFBR2, Regnase-1	NCT05795595
Autoimmune diseases	T cell	electroporation	Cas9 knockout	HLA-A, HLA-B, CIITA, TRAC, PDCD1	NCT05859997
Hemophilia B	B cell	electroporation and AAV	Cas9 knockout	CCR5	NCT06611436
Type 1 diabetes mellitus	PSC-derived pancreatic endoderm cells	electroporation and AAV	Cas9 knockout, Cas9 HDR	CD247, B2M	NCT05565248
B cell malignancies	iPSC	electroporation and AAV	Cas9 knockout, Cas9 HDR	TRAC	NCT04629729

Table 2.

Clinical trials for in vivo CRISPR-based editing

Targeted organ	Delivery	Indication	Editing system	Targeted locus	Trial ID
Liver	LNPs	HAE	Cas9 knockout	KLKB1	NCT05120830
		ATTRv-PN and/or ATTR-CM	Cas9 knockout	TTR	NCT04601051
					NCT06128629
			Base editor	TTR	NCT06539208
		Chronic hepatitis B	Cas9 knockout	Alpha-1 AAT	NCT06680232
		FH and/or CVD	Base editor	PCSK9	NCT05398029
					NCT06164730
					NCT06458010
		AATD	Base editor	SERPINA1	NCT06389877
		GSDIa	Base editor	G6PC	NCT06735755
		PH1	Base editor	HAO1	NCT06511349
Muscle	AAV	DMD	hfCas12max	DMD	NCT06594094
			Base Editor	DMD	NCT06392724
		FSHD	dCasONYX	D4Z4	NCT06907875
Eyes	AAV	LCA10	Cas9 knockout	CEP20	NCT03872479
		Retinitis pigmentosa	Cas9 knockout	RHO	NCT05805007
		nAMD	hfCas13Y	VEGF	NCT06623279
	VLPs	HSV-1-SK	Cas9 knockout	HSV1	NCT06474442
Brain	AAV	MDS	hfCas13Y	MECP2	NCT06615206
Cervix	Plasmid, Gel	CIN 1	TALEN, Cas9 knockout	E7, E6	NCT03057912
Urinary Tract	Bacteriophage	aUTI	Cas3	Drug Resistant E. Coli	NCT05488340
Bone marrow, Lymph nodes, Spleen	AAV	HIV-1-infection	Cas9 knockout	HIV-1	NCT05144386

Abbreviations for indications: HAE, hereditary angioedema; ATTRv-PN, hereditary transthyretin amyloidosis with polyneuropathy; ATTR-CM, transthyretin amyloidosis-related cardiomyopathy; FH, familial hypercholesterolemia; CVD, cardiovascular disease; AATD, alpha-1 antitrypsin deficiency; GSDIa, glycogen storage disease Ia; PH1, primary hyperoxaluria type 1; DMD, Duchenne muscular dystrophy; FSHD, facioscapulohumeral muscular dystrophy; LCA10, Leber congenital amaurosis 10; nAMD, neovascular age-related macular degeneration; HSV-1-SK, herpes simplex virus type I stromal keratitis; MDS, MECP2 duplication syndrome; CIN 1, HPV-related cervical intraepithelial neoplasia I; aUTI, acute uncomplicated urinary tract infection.

Prime editors

Prime editor (PE) is a versatile CRISPR-based genome editing tool capable of introducing all 12 types of point mutations, as well as small insertions and deletions, without inducing DSBs or relying on HDR¹⁰ (Figure 1). The system consists of an HNH-inactivated nCas9 fused to a reverse transcriptase (RT) and is guided by a prime editing guide RNA (pegRNA). The pegRNA combines a single guide RNA (sgRNA) with a primer binding site (PBS) and a reverse transcription template (RTT). After nCas9 nicks DNA, the PBS anneals to the exposed 3′ end, allowing RT to synthesize DNA from the RTT template and install the desired sequence into the genome (Figure 1). PE efficiency has been enhanced through engineering of compact or high-activity RTs,⁴⁷ improved nuclear localization,⁴⁸ introduction of a second nick on the opposite DNA strand (PE3),¹⁰ and transient inhibition of the mismatch repair (MMR) pathway.^48,49 In parallel, pegRNA optimization through 3′-end stabilization,^50,51 AI-guided design,⁵² or chemical modifications^53,54 further improved efficiency and broadened the applicability of PE across different cell types.

Preclinical studies have shown that PE can effectively and safely correct disease-causing mutations in therapeutic relevant cells, such as those in SCD or p47^phox for chronic granulomatous disease (CGD). Despite this progress, editing efficiency remains variable across genomic loci and cell types, influenced by factors such as local sequence context, pegRNA stability, and cellular environment.⁵⁵ While PE largely circumvents the widespread genomic damage associated with DSBs, unintended editing events may still occur.⁵⁶ Current research focuses on developing next-generation PE systems to improve consistency, precision, and delivery, collectively advancing prime editing toward clinical application for the precise treatment of diseases.

Large-scale genome editing

While small-scale edits are essential for correcting point mutations, many therapeutic applications, particularly for monogenic diseases, require the insertion or replacement of larger DNA sequences.⁵⁷ These conditions often involve diverse mutations scattered throughout a single gene, making gene replacement a potentially universal approach that can benefit all patients regardless of the specific mutation site. Delivering functional gene cassettes at the kilobase scale could enable durable, one-time treatment for a broad spectrum of patients.

PE systems were limited to insertions of fewer than approximately 50 base pairs.¹⁰ By employing a pair of PE system, the dual pegRNAs extended this range to 100–250 bp, with isolated successes reaching up to 1 kb^58,59,60 (Figure 1). More recently, PE has been adapted to install landing sites for serine integrases such as Bxb1, facilitating targeted integration of large DNA payloads (Figure 1). Strategies include PASTE⁶¹ and PASSIGE^58,62 exemplify this concept. PASTE fuses Bxb1 to nCas9-RT, while PASSIGE employs engineered eeBxb1 with enhanced activity to enable large DNA insertion. These newly engineered systems significantly expand the potential for large-scale genomic modifications. Beyond insertions, PE also supports large-scale deletions with greater precision than conventional CRISPR-Cas9 approaches, which typically rely on dual sgRNAs and can result in unpredictable editing outcomes.^63,64 Emerging tools such as HOPE,⁶⁵ PRIME-Del,⁶⁶ Bi-PE,⁵⁹ PETI,⁶⁷ GRAND,⁶⁰ PEDAR,⁶⁸ and AE⁶⁹ expand the CRISPR toolbox by leveraging endogenous DNA repair pathway to mediate deletions, inversions, translocations, and amplifications with enhanced fidelity and efficiency. These innovations are expanding the boundaries of genome editing, enabling more complex and clinically relevant modifications, and paving the way for treating a wide array of disorders.

RNA editing systems

CRISPR-based RNA-targeting systems have emerged as important tools for transcriptome engineering. In RNA knockdown applications, the Cas13 family (type VI CRISPR systems) enables sequence-specific RNA cleavage guided by gRNAs^11,70 (Figure 1). Despite its utility, the collateral cleavage activity of Cas13 can lead to unintended degradation of non-target RNAs.^5,6 To address this limitation, engineered RfxCas13d variants, compact Cas13X/Y derived from extremophilic microbes, and the type III-E Cas7-11 effector have been developed.^70,71,72,73 These optimized systems effectively eliminated collateral activity while retaining compatibility with AAV delivery platforms, thereby enhancing their potential for in vivo therapeutic applications. In addition, multi-subunit type III-A Csm complexes demonstrate high knockdown efficiency in zebrafish and mammalian cells, underscoring the potential of multi-component systems in eukaryotic models.⁷⁴

In RNA base editing, reversible transcriptome modifications circumvent permanent genomic alterations. ADAR-deaminase-mediated tools (e.g., REPAIR¹¹ and RESCUE⁷⁵) integrate catalytically inactive dCas13 with engineered ADAR2 to enable site-specific adenosine-to-inosine (A-to-I) editing (Figure 1). REPAIRv2 incorporates the T375G mutation to reduce non-specific binding of ADAR2 to double-stranded RNA, significantly reducing off-target activity.¹¹ For cytidine-to-uridine (C-to-U) editing, RESCUE employs an evolved ADAR2 with acquired cytidine deaminase activity, while the CURE system utilizes apolipoprotein B mRNA editing enzyme catalytic subunit 3A (APOBEC3A) fused with dCas13 to enhance editing flexibility.⁷⁵ Compact RNA base editors, such as ceRBEs, have replaced dCas13 with the smaller EcCas6e protein, achieving high editing efficiency with a reduced transcriptome off-target, an advantage for in vivo delivery.⁷⁶

Epigenetic editing systems: CRISPRa and CRISPRi

Beyond permanent genome editing, CRISPR systems can also be repurposed for epigenetic editing, allowing transiently regulation of gene expression without altering DNA sequence. In this approach, catalytically inactive Cas9 (dCas9) is directed to regulatory elements of a gene and tethered to effector domains that modulate chromatin states or transcriptional machinery. In CRISPR interference (CRISPRi), dCas9 is fused to transcriptional repressors (e.g., KRAB domain,^77,78 MeCP2,⁷⁹ and SID4x⁸⁰) to silence target genes, providing a durable alternative to RNA interference. Conversely, CRISPR activation (CRISPRa) fuses dCas9 to transcriptional activators (e.g., VP64, p65, and Rta) or modular recruitment platforms such as SunTag or SAM to achieve robust and tunable upregulation of endogenous genes.^81,82,83,84

CRISPRa and CRISPRi have been widely adopted in functional genomics, enabling genome-wide gain- and loss-of-function screens to systematically interrogate gene networks.^85,86 Importantly, CRISPRa/i addresses therapeutic gaps not filled by conventional editing, particularly for disease caused by abnormal gene expression levels. For example, lipid-nanoparticle-delivered mRNA encoding an epigenetic editor targeting PCSK9 achieved durable silencing in mice and cynomolgus monkeys, with reductions of up to ∼90% in circulating PCSK9 protein and ∼70% in low-density lipoprotein (LDL) cholesterol levels after a single administration.⁸⁷ As CRISPRa/i regulates expression in a reversible and non-permanent manner, it offers a therapeutic option for conditions requiring transient correction.

Delivery strategies for gene editing

A major challenge in the clinical application of genome editing tools lies in the efficient and targeted delivery of CRISPR tools to specific cells or tissues. Effective delivery systems must ensure high intracellular uptake while minimizing cytotoxicity, immune activation, and off-target delivery. Current approaches can be broadly categorized into three classes: physicochemical methods, viral vectors, and non-viral platforms, each with distinct advantages and limitations (Figure 2).

Figure 2.

Overview of delivery strategies used for ex vivo and in vivo CRISPR-based editing

(Above) Delivery strategies for CRISPR-based editing can be broadly categorized into physicochemical methods, viral vectors, non-viral vectors, and combined strategies. (Below) CRISPR delivery is typically conducted via ex vivo or in vivo methods. Ex vivo approach requires the cell isolation from patients or donors. These cells are cultured, edited, expanded, and infused into the patient. In vivo delivery typically involves packaging CRISPR components into vectors that are administered directly into the patient, enabling gene editing to occur within the body. AAV, adeno-associated virus; LNP, lipid nanoparticles; VLP, virus-like particles.

Physicochemical delivery methods

Electroporation is widely used for ex vivo genome editing, as it transiently permeabilizes cell membrane to facilitate the entry of nucleic acids, proteins, or ribonucleoprotein (RNP) complexes.^88,89 It is particularly effective for ex vivo cell therapies involving hematopoietic stem cells (HSCs),⁹⁰ T cells,⁹¹ or induced pluripotent stem cells (iPSCs),⁹² which are edited outside the body and then re-infused into patients.

Electroporation itself imposes stress on cells. Several studies have reported that electroporation and the subsequent gene editing procedure can elicit heightened inflammatory responses and increase DNA damage response.⁹³ While electroporation of mRNA and proteins has proven effective, electroporation of DNA often activates the cGAS-STING innate immune pathway, triggering extensive cell death.^94,95 This presents a major challenge for applications requiring DNA donor templates for gene knockin, where electroporation is frequently combined with complementary delivery methods such as AAV vectors for donor DNA transfer.

Viral-vector-based delivery

Among viral vectors, AAVs are commonly used due to their favorable safety profile and ability to achieve tissue-specific targeting, including the liver, retina, muscle, brain, and heart.⁹⁶ However, their limited cargo size (∼4.7 kb) presents a challenge for delivering large genome editors, often necessitating dual-vector systems that reconstitute the full payload in vivo.⁹⁷ Nevertheless, the prolonged expression of genome editors via AAVs raises concerns about increased off-target activity and genotoxicity, highlighting the need for delivery systems that allow for controlled, transient expression.

Lentiviral (LV) and retroviral (RV) vectors have been extensively employed in cell therapies. Unlike AAV vectors, LV and RV vectors integrate their genetic payload into the host genome, enabling long-term transgene expression and accommodating larger cargos (∼8.5 kb) such as base editors or prime editors. While the integration feature confers long-term expression, it also entails significant genotoxic risks, including insertional mutagenesis, oncogene activation, and tumor suppressor gene inactivation. In addition, LV- or RV-mediated delivery of genome editors results in persistent nuclease expression, raising concerns of off-target activity and immunogenicity against Cas nucleases. Furthermore, due to their substantial immunogenicity, most LV- or RV-mediated gene therapies are performed in an ex vivo setting rather than in vivo.

In comparison, adenoviral (AdV) vectors offer distinct advantages, primarily due to their non-integrating nature, which eliminates the risk of insertional mutagenesis and associated off-target effects. Another notable strength is their remarkably high packaging capacity, reaching up to ∼35 kb, which allows the delivery of virtually any genome editing tool. Nevertheless, the clinical application of AdVs remains limited by two major challenges. First, capsid proteins of AdVs elicit robust innate and adaptive immune responses, often resulting in acute inflammation or severe immune-related toxicities. Second, the widespread prevalence of pre-existing neutralizing antibodies against AdVs in the human population significantly reduces in vivo transduction efficiency and persistence. Collectively, these constraints have restricted the broad clinical translation of AdV-based gene therapy approaches.

Non-viral delivery approaches

Lipid nanoparticles (LNPs) have become a leading non-viral delivery platform, especially for transient delivery of RNA or RNPs.^98,99 Their safety and clinical utility have been validated through widespread use in mRNA-based COVID-19 vaccines, which rely on systemic delivery.¹⁰⁰ Owing to the liver’s unique vascular structure and its natural role in nanoparticle clearance, LNPs efficiently target hepatocytes, enabling co-delivery of CRISPR mRNA and sgRNA for effective genome editing and treatment of liver-associated genetic disorders, as demonstrated in clinical studies. Supported by their favorable safety profile, including biodegradability, low toxicity, minimal immunogenicity, transient expression, and compatibility with repeat dosing, LNPs have become one of the most widely adopted delivery vehicles for in vivo gene therapy. In comparison, achieving effective delivery to non-hepatic tissues remains an active area of research, with recent engineering efforts focusing on improving tissue specificity and reducing off-target delivery.^101,102,103 Engineered virus-like particles (VLPs), which mimic the structure of viruses but lack viral genetic material, offer another promising transient delivery modality.^104,105 The first VLP-based gene editing therapy has progressed to clinical translation, with a phase I trial administering the investigational product to three patients with recurrent herpes simplex virus type 1 (HSV-1) stromal keratitis (NCT06474442), marking the first-in-human evaluation of a VLP-based delivery platform. Additional innovations include harnessing endogenous retroviral elements derived from the human genome, which may reduce immunogenicity,¹⁰⁶ as well as cell-penetrating peptides that facilitate delivery into difficult-to-transfect cells such as lymphocytes, neurons, and airway epithelial cells.^107,108,109 Emerging platforms such as bacterial contractile injection systems and computationally designed de novo protein cages are also under active investigation.^110,111,112 These platforms aim to transiently deliver CRISPR editors to otherwise inaccessible biological targets with high specificity and modularity.

Combined strategies and safety considerations

When precise editing requires a repair donor DNA template, its delivery remains a major challenge for non-viral modalities, because exogenous DNA can trigger strong innate immune responses and cytotoxicity.¹¹³ To overcome this, hybrid approaches are being developed that integrate multiple delivery modalities. For example, electroporation paired with viral vectors has enabled the co-delivery of genome editors and repair DNA templates, a strategy that has shown success in the therapeutic editing of T cells⁹¹ and shown promising results in HSPCs.¹¹⁴ Preclinical studies using LNP-encapsulated Cas9/sgRNA complexes co-delivered with AAV-cargo have shown target correction in hepatic tissue in vivo.¹¹⁵ Comprehensive safety evaluation remains essential, including assessments of cytotoxicity, immune responses, and off-target delivery. These factors can vary by cell type, tissue, and delivery context. Continued innovation and refinement are essential to support the safe and effective clinical deployment of gene editing technologies.

CRISPR technology in human clinical trials

The versatility and high efficiency of CRISPR-based tools have accelerated their translation into therapeutic applications across a broad spectrum of diseases.¹¹⁶ These include ex vivo editing, in which patient- or donor-derived cells are genetically modified outside the body and then reinfused, as well as in vivo editing, where genome editors are delivered directly to affected tissues within the body.

Ex vivo cell therapy

Gene editing for hemoglobinopathies

Transfusion-dependent β-thalassemia and SCD are monogenic disorders caused by mutations in the HBB gene, resulting in defective β-globin production.^117,118 Ex vivo gene editing of autologous HSCs has become one of the most clinically advanced applications of CRISPR technology for these conditions.

Instead of directly restoring β-globin expression, a common therapeutic strategy is to reactivate γ-globin as a functional substitute for defective β-globin. Among the regulators of γ-globin expression, BCL11A serves as a key transcriptional repressor of HBG1/2 (the gene encoding γ-globin).¹¹⁹ Because complete knockout of BCL11A results in severe developmental abnormalities in HSCs due to its essential roles in hematopoiesis, the discovery of an erythroid-specific BCL11A enhancer has enabled lineage-restricted disruption of BCL11A.⁹⁰ In 2023, exagamglogene autotemcel (Casgevy), which uses SpCas9 to disrupt this erythroid-specific BCL11A enhancer and thereby achieve sustained activation of γ-globin, became the first ex vivo CRISPR-Cas9-based therapy approved in the United Kingdom and the United States for the treatment of SCD and β-thalassemia. Other trials have explored the use of AsCas12a nuclease to disrupt the HBG1/2 promoter regulatory region to reactivate γ-globin (Table 1). Because of gene duplication within this region, DSB-based editing strategy generates a mixture of editing outcomes, each leading to variable levels of γ-globin reactivation. Meanwhile, HDR-based correction of the sickle mutation using viral or synthetic (ssODN) donor in HSCs has been explored. However, HDR is intrinsically constrained by its dependence on S/G2 phase activity, while long-term engrafting HSCs remain largely quiescent during the short ex vivo culture period. These challenges were exemplified in the CEDAR clinical trial, which combined Cas9, HDR, and AAV6 donor templates for SCD but was paused after the first patient experienced severe pancytopenia. These limitations highlight that while HDR remains a conceptually precise strategy, its low efficiency, cytotoxicity risks, and potential impairment of HSC repopulating capacity pose major barriers to clinical translation.^{120,121,122,123}

Base editing provides a more precise and efficient alternative. ABEs targeting the HBG promoter have demonstrated robust editing with increased HbF expression and minimal off-target effects in HSCs.¹²⁴ This strategy introduces A-to-G substitutions in the BCL11A binding motif within the HBG1/2 promoters, thereby disrupting BCL11A binding and reactivating γ-globin, a therapeutic approach currently being evaluated in clinical trials. CBEs have also shown potential to induce γ-globin, though their higher risk of off-target remains a concern.¹²⁵ Newer CBE tools like the transformer base editor (tBE) have demonstrated highly specific editing of transcription factor motifs in HSCs without detectable DNA or RNA off-target alterations.¹²⁶ This next-generation platform is currently being evaluated in phase I/II clinical trials to assess its safety and therapeutic potential.

While CRISPR-based therapies for hemoglobinopathies are progressing rapidly, emerging technologies such as prime editing offer additional precision.¹²⁷ Further preclinical and clinical data are required to assess the long-term safety, efficiency, and durability of these advanced gene editing strategies.

Protective epitope editing for AML immunotherapy

In contrast to the corrective paradigm, gene editing in acute myeloid leukemia (AML) aims to protect normal cells from immunotherapy-related toxicity. AML is a hematological malignancy characterized by the clonal expansion of immature myeloid progenitors. Immunotherapeutic agents, such as CAR-T cells and antibody-drug conjugates (ADCs) targeting surface antigens like CD33 and CD123, show therapeutic potential; their clinical translation is limited by severe on-target, off-tumor toxicity, as these antigens are also expressed on normal hematopoietic cells.^128,129

Early gene editing efforts focused on CD33 knockout in HSPCs to confer resistance to CD33-targeting therapies (e.g., gemtuzumab ozogamicin).^{130,131,132,133} Preclinical studies in mice and non-human primates showed the safety and efficacy of CD33-null HSPCs, which maintained multilineage reconstitution and enabled selective AML cell clearance.¹³⁰ Clinical trials evaluating this CD33 knockout approach are currently underway (Table 1), marking a key step toward integrating gene-edited HSPCs into AML immunotherapy.

To overcome the limitations of full antigen deletion, precision epitope editing has emerged as a refined alternative. This approach introduces minimal alterations at antibody-binding sites, rendering the antigen undetectable by therapeutic agents while preserving its native function. For instance, CD123, a critical interleukin-3 (IL-3) receptor subunit and AML marker, has been edited using Cas9-HDR,¹³⁴ ABEs,¹³⁵ and PEs¹³⁶ to generate epitope variants resistant to the CAR-T cells, without disrupting protein function. This strategy has since been extended to other functionally essential antigens, including CD45, CD117, and FMS-like tyrosine kinase 3 (FLT3), illustrating a broader paradigm shift toward antigen-sparing immunotherapy enabled by high-precision editing.^137,138

Resistance-conferring editing for HIV therapy

Antiretroviral therapy (ART) can effectively suppress HIV replication for decades, enabling long-term disease control. However, these regimens are not curative and require lifelong adherence. Unlike traditional gene therapies aimed at correcting pathogenic mutations, gene editing for HIV focuses on introducing genetic resistance to viral infection. The proof-of-concept study came from the “Berlin patient,” who achieved durable HIV remission after receiving an allogeneic HSC transplant from a donor homozygous for the C-C chemokine receptor type 5 (CCR5) Δ32 allele, a naturally occurring loss-of-function mutation that abolishes HIV entry.¹³⁹ Inspired by this, ex vivo CRISPR-Cas9 editing of CCR5 in allogeneic HSPCs has been attempted in HIV patient with coexisting hematological malignancies.¹⁴⁰ These trials showed stable engraftment of edited cells, but the low editing efficiency limited clinical efficacy in eradicating HIV reservoirs.

Recent studies have employed CRISPR-based technologies to disrupt CCR5 or C-X-C chemokine receptor type 4 (CXCR4), the two main co-receptors used by different HIV strains, in primary cells, achieving resistance to HIV infection in animal studies.^141,142 However, the dispensability of CCR5 in humans remains controversial, and CXCR4 plays an essential role in HSC homing and engraftment, posing challenges for clinical translation.^143,144 To circumvent these limitations, epitope editing is being explored as an alternative strategy, targeting viral interaction domains on host receptors to render host resistance while preserving receptor function.^145,146 These approaches, combined with advances in delivery systems and precision editing, hold promise for achieving a functional cure for HIV.

Improving T cell function for immunotherapy through ex vivo CRISPR-based editing

CAR-T cell therapy involves engineering T cells with antigen-specific receptors to eliminate malignant cells. While autologous CAR-T products targeting CD19 or BCMA have achieved clinical success, their individualized manufacturing and integration-based delivery pose challenges in scalability and safety.^147,148 To overcome these issues, CRISPR-edited allogeneic (“off-the-shelf”) CAR-T cells are being developed.¹⁴⁹ Using CRISPR-mediated HDR, typically facilitated by AAV-mediated donor delivery, CAR transgenes can be precisely inserted into specific genomic loci such as the T cell receptor alpha constant (TRAC) locus.¹⁵⁰ This strategy eliminates endogenous TCR expression and reduces the risk of graft-versus-host disease (GvHD). To further improve CAR-T cell function and persistence, additional knockouts of immune checkpoint and regulatory genes—such as programmed cell death protein 1 (PDCD1), Regnase-1, and transforming growth factor β (TGF-β) receptor type 2 (TGFBR2) are being investigated.^{91,151,152,153} For example, dual knockouts of TRAC and beta-2-microglobulin (β2M), or triple knockouts of TRAC, β2M, and PDCD1, have been shown to generate universal CAR-T cells with reduced immunogenicity but facing persistence issue, as quickly cleared by natural killer (NK) cells.

In a preclinical study, a dual-function CRISPR-Cas9 system enabled non-viral generation of anti-CD19 CAR-T cells by integrating the CAR cassette into the AAVS1 or PDCD1 loci, the latter simultaneously disrupting PD-1 and enhancing antitumor activity in xenograft models.¹⁵⁴ In a phase I clinical trial (NCT04213469), these non-viral CAR-T cells induced high complete remission rates in relapsed or refractory B cell lymphoma without severe adverse events.¹⁵⁴ These findings highlight the promise of precise, non-viral CAR-T engineering, though larger studies are needed to confirm long-term safety and efficacy.

Beyond oncology, CRISPR-based CAR-T therapies are being explored in autoimmune indications. In a landmark study, an allogeneic CAR-T product (TyU19) was developed by transducing healthy donor T cells with a lentiviral anti-CD19 CAR construct, followed by CRISPR-Cas9-mediated knockout of human leukocyte antigen (HLA)-A, HLA-B, class II major histocompatibility complex transactivator (CIITA), TRAC, and PDCD1.¹⁵⁵ These multiplex-edited “off-the-shelf” CAR-T cells were administered to patients with severe myositis and systemic sclerosis. The infused cells persisted for over 90 days and led to complete B cell depletion within 2 weeks. Remarkably, treated patients demonstrated reversal of tissue inflammation and fibrosis, along with clinical improvement across multiple functional evaluations. No cytokine release syndrome (CRS) or other severe adverse events were observed. These results highlight the expanding therapeutic scope of CRISPR-engineered CAR-T cell products beyond hematologic malignancies.¹⁵⁵ Nevertheless, it is important to note that multiplex CRISPR-Cas9 editing carries a risk of megabase-scale deletions and chromosomal translocations.¹⁵⁶ These safety concerns have motivated exploration of base editors or prime editors as an alternative strategy to enable safer multiplex engineering.

CRISPR is also employed to eliminate self-antigens in CAR-T cells, enabling therapies for T cell malignancies where targets like CD5 and CD7 are shared by normal and malignant T cells. For example, CD7 knockout via CRISPR prevents fratricide and supports robust expansion of CD7 CAR-T cells, which have shown clinical efficacy in relapsed or refractory CD7⁺ malignancies.^157,158 In a separate trial (NCT05397184), a cytosine base editor was used to disrupt T cell receptor beta constant (TRBC), CD7, and CD52, generating BE-CAR7 cells, which were administered to pediatric patients with relapsed T cell leukemia, marking the first clinical use of base editing in humans.¹⁵⁹ In contrast, CD5-targeted CAR-T cells exhibit minimal fratricide, likely due to rapid internalization of CD5.¹⁶⁰ Nonetheless, researchers are exploring CRISPR-mediated CD5 disruption to further increase CAR-T activity for the treatment of relapsed or refractory CD5⁺ hematologic malignancies.^161,162

Beyond cancer and autoimmunity, CRISPR gene editing is also being explored in the context of xenotransplantation. Preclinical studies have shown that targeted modification of immunogenic loci in pig lung grafts can improve compatibility and reduce immune rejection in non-human primate recipients,¹⁶³ underscoring the broad potential of genome editing to modulate immune responses across diverse therapeutic contexts.

CRISPR in induced-pluripotent-stem-cell-based therapies

Type 1 diabetes mellitus is characterized by the destruction or dysfunction of insulin-producing pancreatic β cells. While exogenous insulin therapy can alleviate hyperglycemia, it does not prevent long-term complications or restore endogenous glucose regulation.¹⁶⁴ A promising approach in regenerative medicine integrates ex vivo CRISPR genome editing with iPSC technology. Unlike HSCs, iPSCs are highly proliferative, making them well-suited for HDR-based gene correction. Current clinical trials are investigating the transplantation of pancreatic endoderm cells derived from allogeneic iPSC as a potential therapy for type 1 diabetes.¹⁶⁵ A representative strategy includes ex vivo genome editing of iPSCs, followed by directed differentiation into pancreatic endoderm cells, encapsulation within biocompatible devices, and implantation into patients to restore insulin production and glucose homeostasis. Within this workflow, CRISPR can be used to remove immunogenic genes to promote graft tolerance, enhance cellular resilience to stress, and optimize insulin secretion.¹⁶⁶

Beyond metabolic diseases, CRISPR-edited iPSC-derived immune cells are gaining traction in cancer immunotherapy. iPSC-derived NK or T cells serve as renewable, off-the-shelf platforms for generating universal CAR-T or CAR-NK therapies. For example, CD38-knockout iPSC-derived NK cells have progressed to clinical trials for acute myeloid leukemia, multiple myeloma, relapsed/refractory B cell lymphoma, and solid tumors (Table 1). Additional strategies include editing immune evasion pathways, such as β2M knockout paired with transgenic expression of PD-L1, to reduce immune rejection and prolong graft persistence. Recently, the first-in-human use of iPSC-derived CD19/BCMA dual-targeting CAR-NK cells in a patient with severe systemic sclerosis resulted in B cell depletion and reversal of fibrosis over a 6-month follow-up, highlighting their potential beyond oncology and into autoimmune disease therapy.¹⁶⁷ Also noted, these iPSCs were heavily engineered using CBE to knockout β2M, CIITA, and CD16 and to knockin HLA-E and HLA-G to minimize the risk of GvHD and host-versus-graft response (HvGR). While these clinical results are very promising, additional studies involving more patients and longer follow-up are needed to fully assess safety and durability. As with other emerging cell therapies, iPSC-based therapies face several challenges, including variable differentiation efficiency, cellular heterogeneity, tumorigenic potential, and suboptimal engraftment in vivo. Addressing these limitations will be critical to ensure a safe and effective clinical application of CRISPR-engineered iPSC therapies.

Large-fragment CRISPR insertion strategies for treating genetically diverse primary immunodeficiencies

Primary immunodeficiencies (PIDs) comprise a diverse group of rare genetic disorders resulting from mutations in genes essential for the development and function of immune cells. Among these, severe combined immunodeficiency (SCID) represents the most life-threatening form, often leading to fatal outcomes within the first year of life without treatment.¹⁶⁸ Recent advances have demonstrated that CRISPR-Cas technologies, combined with delivery platforms such as AAV6 and electroporation, can facilitate the targeted insertion of full-length therapeutic gene cassettes into relevant genomic loci.¹⁶⁹ This HDR-based approach has restored gene function in patient-derived HSPCs across multiple PIDs, including SCID-X1,¹⁶⁹ Wiskott-Aldrich syndrome,¹⁷⁰ and X-linked chronic granulomatous disease.¹⁷¹ These preclinical studies support large-fragment gene insertion as a versatile and broadly applicable strategy to overcome the genetic heterogeneity of monogenic diseases. By circumventing the need for mutation-specific corrections, this method allows for consistent therapeutic outcomes through the integration of functional genes into native or safe harbor sites. While promising, challenges remain, including limited insertional efficiency in quiescent HSCs and the low prevalence of individual PID subtypes, which complicate clinical challenges. Nonetheless, large-fragment CRISPR-based insertion holds strong potential as a scalable, curative platform for treating genetically heterogeneous and previously untreatable immunodeficiencies.

In vivo organ-oriented therapy

Liver diseases

CRISPR-Cas9-based gene therapy is making significant strides in the treatment of genetic disorders, with several in vivo gene-editing programs now entering pivotal clinical trial stages (Table 2). The liver has emerged as a privileged target organ for therapeutic genome editing, due in large part to major advances in LNP delivery innovations.¹⁷² These advances include the development of hepatocyte-tropic LNP systems with improved biodistribution, sustained endosomal escape, and compatibility with repeat dosing, enabling transient and precise gene modulation in vivo.¹⁷² Current clinical efforts are primarily focused on treating monogenic liver-related disorders, including hereditary transthyretin amyloidosis (hATTR), hereditary angioedema (HAE), familial hypercholesterolemia (HeFH), α-1 antitrypsin deficiency, glycogen storage disease type Ia, chronic hepatitis B, and hyperlipidemia.^173,174,175

In hATTR, pathogenic mutations in transthyretin (TTR) gene destabilize the transthyretin TTR tetramer, leading to misfolded monomers that aggregate into amyloid fibrils, causing progressive multisystem dysfunction. Accordingly, TTR silencing has become a key therapeutic strategy. LNP-based delivery of Cas9 mRNA and sgRNA targeting TTR achieves sustained serum TTR reduction after a single dose,^173,174 with phase III trials underway. A similar strategy targeting the KLKB1 gene has also advanced to late-stage clinical evaluation for the treatment of HAE.¹⁷⁵ This approach leverages CRISPR/Cas9-mediated disruption of the kallikrein enzymatic pathway, inhibiting kininogen release and thereby preventing bradykinin-mediated angioedema attacks.¹⁷⁵

Recent developments in base editing further exemplifies the precision and durability of genome engineering. Hepatocyte-targeted GalNAc-LNP delivery of ABE mRNA and sgRNA against PCSK9 achieved durable suppression (>6 months) and significant LDL-cholesterol reduction in early clinical studies (NCT06164730).¹⁷⁶ Early-phase clinical data indicate a favorable safety profile, with no signs of hepatotoxicity, supporting progression to dose-escalation studies. Notably, in an earlier clinical trial employing a different LNP formulation, a serious cardiovascular adverse event was observed (NCT05398029).¹⁷⁷ Specifically, one patient receiving a 0.45 mg/kg dose experienced a myocardial infarction on the second day post-infusion, which was considered potentially treatment-related. Although concerns remain regarding the long-term efficacy and safety, the overall outcomes of this trial are nevertheless encouraging. Furthermore, a personalized base editing therapy delivered via mRNA-LNP platform was developed and clinically implemented within 6 months of molecular diagnosis for an infant with carbamoyl phosphate synthetase 1 (CPS1) deficiency. Following two escalating doses, blood ammonia levels normalized, demonstrating the feasibility of rapid, individualized base editing interventions for rare metabolic disorders.¹⁷⁸ Together, these studies underscore the accelerating translation of in vivo genome editing, enabled by optimized delivery systems and precise molecular tools, paving the way for durable, one-time treatments of liver-associated genetic diseases.

Non-liver diseases

Achieving efficient delivery of gene-editing tools to extrahepatic tissue remains a major hurdle in therapeutic development. While LNP platforms have enabled mRNA delivery to organs such as lungs, spleen, and pancreas, their efficacy in these tissues remains limited compared to that in the liver and is generally insufficient for clinical translation.¹⁷⁹ As a result, alternative strategies, such as local administration or use of non-LNP-based vectors, are actively being pursued to enhance tissue-specific editing outside liver.

Inherited retinal diseases represent one of the advanced applications. Leber congenital amaurosis type 10 (LCA10), caused by biallelic loss-of-function mutations in the CEP290 gene, is one such example.¹⁸⁰ A targeted therapy using AAV5 delivers SaCas9/sgRNA complex via subretinal injection has been tested in clinical trials.¹⁸¹ This strategy employs dual sgRNAs to flank the pathogenic intronic region, enabling direct excision or inversion of the mutated sequence and restoring physiological CEP290 expression.¹⁸¹ In a phase I/II clinical study, 11 of 14 patients with hereditary blindness showed measurable improvements in visual function following treatment.¹⁸² Complementary approaches are also being explored for infectious ocular disease. For example, VLPs have been developed to deliver Cas9/sgRNA targeting HSV-1, achieving viral genome disruption and therapeutic clearance of HSV-1-associated keratitis in clinical studies.¹⁸³ This VLP platform, derived from lentiviral HSV-1, leverages the interaction between the MS2 RNA-binding protein and its cognate stem loop to enable selective RNA packaging. By incorporating MS2 into the GagPol scaffold and fusing the MS2 stem loop to target RNAs, both Cas9 mRNA and dual sgRNAs against HSV-1 can be co-packaged and delivered in a single vector, overcoming AAV limitations of cargo capacity and prolonged Cas9 expression.

Neuromuscular disorders have also seen promising preclinical and early clinical advances. Duchenne muscular dystrophy (DMD), an X-linked recessive neuromuscular disorder, arises from mutations in the DMD gene encoding dystrophin.¹⁸⁴ A common therapeutic strategy aims to restore the reading frame via exon skipping. A single AAV vector was deployed to deliver the CRISPR/hfCas12Max system targeting the splice donor site of exon 51, allowing functional dystrophin restoration in patients with exon 52–63 deletions. Clinical observations report motor function improvements within 3 months of treatment. A concurrent study utilizes dual AAV9 vectors to deliver an optimized trans-splicing adenine methyltransferase (oTAM) editor under muscle-specific promoters targeting exon 50 for exon skipping. This system also includes co-expression of γ-actin (ACTG1) to enhance muscle cell cytoskeletal stability. In an ongoing investigator-initiated trial (IIT), two pediatric recipients exhibited progressive motor improvement 6 months post-administration.

Different than exon-skipping strategies, a preclinical study employed a compact dCasMINI-VPR epigenetic activator, efficiently packaged into a single AAV vector, together with sgRNAs targeting the UTRN promoter.¹⁸⁵ This approach upregulated utrophin expression as a functional surrogate for defective dystrophin, and a single AAV injection increased utrophin expression and improved muscle function in both DMD mouse models and non-human primates, highlighting the therapeutic potential of epigenetic activation. In a recent clinical study, a dSaCas9-VP64-based transcriptional activator was used to upregulate DMD expression as a CRISPR-based gene activation therapy.¹⁸⁶ Unfortunately, the patient experienced fatal toxicity following high-dose AAV9 administration, underscoring the risks associated with systemic delivery and the need for improved safety in epigenetic activation approaches. Another example of epigenome editing is found in facioscapulohumeral muscular dystrophy (FSHD), a disorder caused by hypomethylation of the D4Z4 repeat array that permits aberrant expression of DUX4. Abnormal expression of DUX4 in muscle is toxic and drives progressive degeneration. An epigenetic strategy re-methylates the D4Z4 region to silence DUX4 and prevent muscle cell death and has recently advanced into clinical evaluation (NCT06907875).

MECP2 duplication syndrome is a rare, X-linked dominant neurodevelopmental disorder caused by genomic copy-number gains of the MECP2 locus, leading to dosage-dependent overexpression of methyl-CpG-binding protein 2 (MeCP2).⁷² Leveraging upon this gain-of-function pathology, a CRISPR-Cas13Y-mediated RNA-targeting approach has been developed to selectively reduce MECP2 transcript levels.⁷² In preclinical studies, intracranial administration of AAV9-mediated hfCas13Y delivery achieved CNS-restricted degradation of MECP2 mRNA.⁷² By fine-tuning MeCP2 expression, this strategy aims to restore neurodevelopmental balance.⁷² In interim clinical data from two pediatric patients, both individuals exhibited favorable tolerability profiles. Notably, the first patient demonstrated marked neurological improvement by 3 months post-treatment, including enhanced expressive communication, sustained eye-contact responsiveness, and measurable gains in gross motor coordination. Long-term monitoring will be essential to assess sustained Cas13Y expression and evaluate potential immunogenicity and safety risks.

Innovations in cancer immunotherapy are leveraging in vivo genome editing to overcome limitations of traditional ex vivo CAR-T therapy. Conventional CAR-T production requires labor-intensive ex vivo manipulation, expansion, and quality control steps, contributing to high costs and manufacturing complexity.^187,188 In contrast, in vivo CAR-T therapies aim to reprogram endogenous T cells within the patient, offering simplified logistics and reduced cost.^187,188 Current strategies include the use of viral vectors, such as lentivirus, to deliver CAR constructs into circulating T cells. This approach is being evaluated in hematologic malignancies, including B cell acute lymphoblastic leukemia (B-ALL), autoimmune, multiple myeloma, and B cell non-Hodgkin lymphoma (B-NHL).^187,188 In parallel, LNP-based platforms are being used to deliver mRNA encoding CAR constructs, allowing transient but potent CAR expression while avoiding risks associated with permanent genomic integration.^187,188 These platforms have entered phase I clinical trials, and insights gained from these early studies will be critical in guiding the future development of in vivo CAR-T therapies.¹⁸⁷

Challenges and future perspectives

Over the past decade, CRISPR has evolved from a basic genome-editing tool into a versatile platform for both research and therapy. With over 300 CRISPR-Cas systems now characterized, increasingly precise and versatile editors are being developed to meet diverse clinical needs. CRISPR-based gene and cell therapies have entered into clinical trials across multiple diseases with promising early outcomes, yet key challenges in efficiency, specificity, delivery, and safety must still be overcome to fully realize their therapeutic potential.

Delivery barriers for in vivo applications

Safe, efficient, and tissue-specific transient delivery of CRISPR components remains one of the most formidable hurdles, particularly for in vivo therapies. Considerable efforts are being devoted to engineering non-hepatic LNPs by altering lipid composition or conjugating tissue-specific antibodies to enable delivery to organs such as the brain, lung, spleen, or bone marrow.^{189,190,191,192} While promising, these strategies raise additional questions, including whether antibody-mediated uptake could transiently perturb receptor function and thereby affect cell physiology. Moreover, efficient payload release from endosomes remains a persistent barrier that demands continued innovation.

Viral vectors such as AAV and lentivirus are widely used, but they suffer from limited cargo capacity, immunogenicity, persistent expression of genome editor associated with off-target effects, and potential insertional mutagenesis. Engineering AAV vectors to deliver mRNA, or employing non-integrating lentiviral systems, may enable more efficient and transient targeting of diverse tissues while minimizing risks associated with long-term expression and genomic integration.

Editing efficiency and precision

Achieving higher efficiency and precision remains a central pursuit in the field, as the therapeutic thresholds vary across diseases. Over the past decade, the development trajectory has moved from Cas nucleases to BEs and PEs, with precision improving but efficiency often declining. For instance, PEs enable highly precise edits but currently work effectively only with SpCas9, while smaller Cas variants linked to PEs show much lower activity. Ongoing engineering efforts, however, are expected to bring about editors that combine both high efficiency and precision in the near future.

While BEs and PEs theoretically provide solutions for a wide range of point mutations, the diversity of mutation types and rarity of individual patient cases make it unrealistic to pursue a “one-drug-per-patient” strategy. Thus, site-specific large-fragment gene integration represents a critical step toward fully enabling genome editing as a broadly applicable therapeutic modality. Emerging technologies such as PASSIGE,⁶² evoCAST,¹⁹³ or R2¹⁹⁴ provide proof of concept for this approach, but the complexity of their multi-component systems highlights the need for significant optimization before translation to clinical use.

Higher precision also translates into greater safety, with reduced off-target events and byproducts. Compared to traditional small-molecule drugs, genome editing already offers remarkable specificity, particularly given that most therapeutic strategies employ transient delivery, thereby limiting nuclease persistence and associated risks. As long as increasingly sensitive off-target detection methods are developed and sgRNAs with carcinogenic potential are systematically excluded, the safety assessment of genome editing therapies can be substantially strengthened.

Long-term safety and cellular persistence

The first gene editing drug, Casgevy, was approved in 2023, with few patients monitored for ∼3–5 years so far, a relatively short time frame to fully evaluate long-term safety. Encouragingly, because current therapies are ex vivo edited cells with transient nuclease exposure, no secondary malignancies akin to those observed in early gene therapy trials have been reported. Although emerging studies in HSPCs indicate that prolonged ex vivo culture exacerbates Cas9-nuclease-induced chromosomal deletions or rearrangements, clinical protocols that minimize ex vivo culture and manipulation may substantially reduce these risks.^156,195 Moreover, multiplex gene-edited universal CAR-T for treating autoimmune diseases has shown follow-up beyond 1 year without evidence of serious editing-related adverse events, suggesting that cell-based editing strategies may carry a reassuring safety profile when carefully controlled.

In vivo genome editing presents a more complex safety landscape. Achieving a balance between persistence, efficacy, and immune safety remains the central challenge for clinical translation. Delivery vectors, particularly viral systems such as AAV, can drive long-term expression of editors, which may enhance efficacy but also raises the possibility of neoantigen formation and immune-mediated clearance of edited cells. Similarly, epigenome editors delivered for durable effects must be evaluated for potential effects on neighboring gene regulation, and prolonged nuclease activity increases the chance of off-target mutations with unknown long-term consequences. Thus, while the early safety record of ex vivo cell editing is encouraging, sustained vigilance is required as the field advances toward more complex in vivo applications, where the interplay between durability, efficacy, and immune safety will be critical to therapeutic success.

Manufacturing complexity and scalability

Many current CRISPR therapies, particularly autologous cell therapies, rely on complex and individualized manufacturing processes that are costly and logistically challenging. Viral vectors commonly used to deliver therapies, such as AAV, VLPs, and lentiviruses, also require complex manufacturing processes and incur high production costs, which limit the broader accessibility of these therapies to broader patient populations. Recent advances have introduced rapid manufacturing protocols capable of engineering patient-derived T cells or HSPCs in under 24 h, significantly reducing ex vivo culture time.^196,197 Minimizing ex vivo manipulation not only decreases cost and production burden but may also mitigate nuclease-driven genotoxicity and improve cell quality. Moreover, integrated “all-in-one” prototype platforms have recently been developed, which further streamline the cell processing workflow. Allogeneic “off-the-shelf” approaches and iPSC-derived products offer more scalable alternatives but require advanced gene editing to prevent immune rejection and standardize differentiation. Streamlining production pipelines while preserving safety and efficacy is crucial for widespread clinical adoption.

Ethical and legal considerations

There is broad international consensus that germline genome editing should not be performed in clinical settings. For somatic editing, challenges persist around equitable access, informed consent, and the potential for misuse. Regulatory frameworks often lag behind technological progress, creating uncertainty for clinical developers. In addition, the possibility of off-target in germline cells is an important safety concern. iPSC-based editing introduces additional concerns—such as immune compatibility, tumorigenicity, and donor consent—that must be addressed through robust ethical oversight.

In summary, CRISPR has transformed molecular medicine, but its full therapeutic potential depends on overcoming challenges in delivery, specificity, persistence, scalability, and regulation. Future progress will hinge on interdisciplinary advances, particularly AI-driven innovation. AI accelerates nuclease engineering for greater efficiency and compactness, enables de novo design of functional protein binders to enhance editing, and guides the creation of optimized delivery platforms with improved tropism and safety. The convergence of CRISPR and AI is poised to shape the next decade of precision medicine.

Acknowledgments

This work was kindly supported by National Key R&D Program of China (2022YFF1002800), the Noncommunicable Chronic Diseases National Science and Technology Major Project (2023ZD0500600), National Natural Science Foundation of China (82450105, 82525033, 824B2050, and 32501185), Fundamental Research Funds for Central Universities (2042022dx0003), Natural Science Foundation of Wuhan (2024040801020223), the China Postdoctoral Science Foundation (2024M762492), Hubei Provincial Natural Science Foundation of China (2025AFB125), the Postdoctoral Project of Hubei Province under grant number 2004HBBHCXA056, and the Postdoctoral Fellowship Program of CPSF (CZC20251894).

Author contributions

R.J. and Q.C., and Y.Z. wrote the manuscript with input from all authors. R.J. drew the illustrations.

Declaration of interests

The authors declare no competing interests.