A review of genetic modification for ex vivo cellular therapies Authors: Anh Dinh, Nan Zhang, David Stroncek Published: 2026 Abbreviations chimeric antigen receptor CRISPR‐associated protein clustered regularly interspaced short palindromic repeats double‐stranded break U.S. Food and Drug Administration gene of interest graft‐versus‐host disease fetal hemoglobin homology‐directed repair hematopoietic stem and progenitor cell lentiviral nicotinamide adenine dinucleotide phosphate nonhomologous end joining protospacer adjacent motif reverse transcriptase sickle cell disease transfusion‐dependent β‐thalassemia vaso‐occlusive episode γ‐retroviral INTRODUCTION Over the past decade, the field of biotherapies has made remarkable progress in treating diseases using genetic modification. Genetic modification components (e.g., viral vectors) can be administered directly to the patient (in vivo) or cells can be collected, modified externally, and reinfused (ex vivo). Gene delivery and editing techniques have advanced over time to meet the practical and functional needs of manufacturing cellular therapies. This review provides an introductory overview of various gene delivery and editing systems, specifically viral vectors, transposons, and clustered regularly interspaced short palindromic repeats (CRISPR)‐based platforms used in current and emerging ex vivo cellular therapies. These include chimeric antigen receptor T‐cells (CAR T‐cells) and hematopoietic stem and progenitor cell (HSPC)‐based gene therapies. For each, examples from U.S. Food and Drug Administration (FDA)‐approved products and from clinical research products, along with safety considerations, are provided. Ex vivo hematopoietic stem and progenitor cell‐based gene therapy HSPC‐based therapy originated as a method to reconstitute a patient's hematopoietic system with normal HSPCs that possess the ability for self‐renewal and differentiation into all hematopoietic lineages through transplantation. 1 Depending on the indication, HSPCs for transplant can be derived from the patient (autologous) or from a healthy donor (allogeneic). Conventionally, autologous transplants are performed to replenish the marrow in the setting of diseases requiring treatment with high‐dose chemotherapy, such as lymphoma or multiple myeloma. Allogeneic transplants are commonly performed to treat acute leukemia or to address hematologic (e.g., sickle cell disease) or metabolic disorders. Compared to autologous transplants, allogeneic transplants require the identification of a suitable donor and for recipients to undergo more intensive preparative chemotherapy regimens that include immunosuppression to prevent graft rejection but are associated with more severe toxicities. 2 The advent of gene modification, however, has made it possible to address certain diseases with autologous cells rather than allogeneic cells, eliminating the donor search and allowing for less intensive conditioning. 3 This genetic modification of autologous cells circumvents the need for allogeneic cells and is distinct from conventional autologous therapies, which presently do not involve genetic modification. With ex vivo gene therapies (Table 1), patients commonly receive mobilization agents such as granulocyte colony‐stimulating factor and/or a C‐X‐C chemokine receptor type 4 antagonist (e.g., Plerixafor) to increase the number of circulating CD34+ cells prior to apheresis collection. CD34+ cells are then isolated from the product, such as through immunomagnetic bead selection. The CD34+ enriched population is then genetically modified using viral or nonviral methods and expanded in culture. 4 The cells are then reinfused, reconstituting the patient's hematopoietic system and correcting their disease. TABLE 1Examples of Ex Vivo HSPC‐based Gene Therapy and Gene Modification Approaches.Product nameGene modification approachSponsorDisease indicationMechanismStatusBetibeglogene autotemcel (ZYNTEGLO)LV vectorBluebird bioTDTβA‐T87Q‐globin gene additionFDA‐approvedAtidarsagene autotemcel (LENMELDY)LV vectorOrchard TherapeuticsMLD ARSA gene additionFDA‐approvedElivaldogene autotemcel (SKYSONA)LV vectorBluebird bioCALD ABCD1 gene additionFDA‐approvedLovotibeglogene autotemcel (LYFGENIA)LV vectorBluebird bioSCDβA‐T87Q‐globin gene additionFDA‐approvedExagamglogene autotemcel (CASGEVY)CRISPR‐Cas9Vertex Pharmaceuticals/CRISPR TherapeuticsSCD/TDT BCL11A gene‐knockoutFDA‐approvedBEAM‐101BEBeam TherapeuticsSCD HBG1/2 gene promoter mutationPhase I/II – NCT05456880N/ABENIAIDX‐CGD CYBB missense mutation repairPhase I/II – NCT06325709PM359PEPrime Medicinep47 AR‐CGDdelGT mutation in NCF1 correctionPhase I/II – NCT06559176Abbreviations: AR‐CGD, Autosomal Recessive; Chronic Granulomatous Disease; BCL11A, B‐cell lymphoma/leukemia 11A; BE, Base Editing; CALD, Cerebral Adrenoleukodystrophy; CRISPR‐Cas9, Clustered Regularly Interspaced Short Palindromic Repeats‐CRISPR‐associated protein 9; CYBB, Cytochrome B‐245 Beta Chain; HBG1/2, Hemoglobin Subunit Gamma 1/2; LV, Lentiviral; MLD, Metachromatic Leukodystrophy; NCF1, Neutrophil Cytosolic Factor 1; NIAID, National Institute of Allergy and Infectious Diseases; PE, Prime Editing; SCD, Sickle Cell Disease; TDT, Transfusion‐Dependent Thalassemia; X‐CGD, X‐linked Chronic Granulomatous Disease. Examples of Ex Vivo HSPC‐based Gene Therapy and Gene Modification Approaches. Abbreviations: AR‐CGD, Autosomal Recessive; Chronic Granulomatous Disease; BCL11A, B‐cell lymphoma/leukemia 11A; BE, Base Editing; CALD, Cerebral Adrenoleukodystrophy; CRISPR‐Cas9, Clustered Regularly Interspaced Short Palindromic Repeats‐CRISPR‐associated protein 9; CYBB, Cytochrome B‐245 Beta Chain; HBG1/2, Hemoglobin Subunit Gamma 1/2; LV, Lentiviral; MLD, Metachromatic Leukodystrophy; NCF1, Neutrophil Cytosolic Factor 1; NIAID, National Institute of Allergy and Infectious Diseases; PE, Prime Editing; SCD, Sickle Cell Disease; TDT, Transfusion‐Dependent Thalassemia; X‐CGD, X‐linked Chronic Granulomatous Disease. Chimeric antigen receptor T‐cells CAR T‐cells are a type of immune effector cellular therapy in which T‐cells are genetically engineered to express a synthetic construct that combines the antigen recognition of a B‐cell antibody with the intracellular signaling of a T‐cell receptor, which allows for specific human leukocyte antigen (HLA)‐independent cell surface antigen targeting. 5 The starting material is typically collected from a patient via apheresis and undergoes T‐cell selection. T‐cells are then genetically modified to express the CAR construct and expanded in culture. CAR‐T therapy has been transformative for treating relapsed and refractory B‐cell malignancies (Table 2), with seven FDA‐approved products 6 and response rates ranging from 44 to 97% for large B‐cell lymphoma and chronic lymphocytic leukemia, 73–98% for multiple myeloma, and 77–82% for acute leukemia. 7 , 8 , 9 , 10 , 11 , 12 , 13 , 14 , 15 CAR‐T therapy is also being explored for other indications, such as solid tumors and autoimmune diseases. 16 , 17 TABLE 2Examples of Ex Vivo CAR T‐cells and Gene Modification Approaches.Product nameGene modification approachSponsorDisease indicationMechanismStatusTisagenlecleucel (Kymriah)LV vectorNovartis Pharmaceuticals CorporationR/R: Pediatric/young adult B‐cell ALL; LBCL, FLCD19 CAR transgene insertionFDA‐approvedAxicabtagene ciloleucel (Yescarta)γ‐RV vectorKite Pharma Inc.R/R: LBCL; FLCD19 CAR transgene insertionFDA‐approvedBrexucabtagene autoleucel (Tecartus)γ‐RV vectorKite Pharma, Inc.R/R: MCL; BCP‐ALLCD19 CAR transgene insertionFDA‐approvedLisocabtagene maraleucel (Breyanzi)LV vectorJuno Therapeutics, Inc./ Bristol Myers SquibbR/R: LBCL; FL; CLL/SLL; MCLCD19 CAR transgene insertionFDA‐approvedIdecabtagene vicleucel (Abecma)LV vectorCelgene Corporation/ Bristol Myers SquibbR/R MMBCMA CAR transgene insertionFDA‐approvedCiltacabtagene autoleucel (Carvykti)LV vectorJanssen Biotech, Inc.R/R MMBCMA CAR transgene insertionFDA‐approvedObecabtagene autoleucel (Aucaztyl)LV vectorAutolus TherapeuticsR/R CD19‐positive BCP‐ALLCD19 CAR transgene insertionFDA‐approvedCARCIK‐CD19 cellsTransposon (SB)Fondazione Matilde Tettamanti Menotti De Marchi OnlusR/R ALL Post HSCTCD19 CAR transgene insertionPhase I/II – NCT03389035; Phase II – 05252403CAR7 T‐CellsBEGreat Ormond Street Hospital for Children NHS Foundation TrustR/R T‐cell ALLDisrupt TRBC1, TRBC2, CD7, and CD52 genesPhase I – ISRCTN15323014Abbreviations: ALL, Acute Lymphoblastic Leukemia; BCMA, B‐Cell Maturation Antigen; BCP‐ALL, B‐Cell Precursor Acute Lymphoblastic Leukemia; BE, Base Editing; CAR, Chimeric Antigen Receptor; CARCIK‐CD19, anti‐CD19 CAR cytokine‐induced killer cells; CLL, Chronic Lymphocytic Leukemia; FL, Follicular Lymphoma; HSCT, Hematopoietic Stem Cell Transplantation; LBCL, Large B‐Cell Lymphoma; LV, Lentiviral; MCL, Mantle Cell Lymphoma; MM, Multiple Myeloma; R/R, Relapsed/Refractory; SB, Sleeping Beauty; SLL, Small Lymphocytic Lymphoma; TRBC1/2, T‐cell receptor beta constant 1/2; γR, Gamma Retroviral. Examples of Ex Vivo CAR T‐cells and Gene Modification Approaches. Abbreviations: ALL, Acute Lymphoblastic Leukemia; BCMA, B‐Cell Maturation Antigen; BCP‐ALL, B‐Cell Precursor Acute Lymphoblastic Leukemia; BE, Base Editing; CAR, Chimeric Antigen Receptor; CARCIK‐CD19, anti‐CD19 CAR cytokine‐induced killer cells; CLL, Chronic Lymphocytic Leukemia; FL, Follicular Lymphoma; HSCT, Hematopoietic Stem Cell Transplantation; LBCL, Large B‐Cell Lymphoma; LV, Lentiviral; MCL, Mantle Cell Lymphoma; MM, Multiple Myeloma; R/R, Relapsed/Refractory; SB, Sleeping Beauty; SLL, Small Lymphocytic Lymphoma; TRBC1/2, T‐cell receptor beta constant 1/2; γR, Gamma Retroviral. Despite these successes, challenges remain. Examples include post‐CAR‐T treatment relapse due to tumor cells that lack the target antigen (antigen escape) or targeting of non‐tumor cells that express the target antigen (on‐target/off‐tumor activity). 7 , 18 , 19 , 20 Solid tumor indications, in particular, have had limited success compared to hematologic malignancies. 21 CAR T‐cell resistance in solid tumors may be multifactorial, including an immunosuppressive tumor microenvironment that leads to CAR‐T dysfunction, T‐cell exhaustion due to tonic signaling or a high tumor burden, restricted CAR T‐cell trafficking and infiltration of the tumor site, and tumor antigen heterogeneity resulting in antigen escape. 22 Attempts to navigate these barriers include designing more advanced CAR T‐cells, such as those that require multiple antigen signals for activation (logic gating) to address on‐target/off‐tumor toxicities, 23 decoy receptors to bind immunosuppressive elements (e.g., transforming growth factor‐β), 24 , 25 , 26 and secretion of cytokines (e.g., IL‐7, −12, −15, −18) 27 , 28 , 29 , 30 or checkpoint inhibitors (e.g., anti‐PD‐L1, anti‐PD‐1) 31 to increase CAR‐T potency. The continued advancement of gene modification approaches will be critical for the optimal development of multifaceted CAR‐T therapies. Furthermore, given that all commercialized and most investigational CAR‐T therapies are autologous, the quality of the starting material may vary. Depending on the disease indication, patients may have received multiple lines of chemotherapy and immunotherapy and/or have a high tumor burden prior to starting material collection, which can result in fewer T‐cells, reduced T‐cell fitness, and a suboptimal product. 32 “Off‐the‐shelf” allogeneic CAR‐T therapies using starting material from healthy donors have been explored to enhance product quality, increase scalability, and reduce the time to treatment. 33 , 34 Allogenic CAR T‐cells require several simultaneous genetic modifications to prevent graft‐versus‐host disease (GVHD) and host rejection of the cells. 35 VIRAL GENETIC MODIFICATION METHODS γ‐Retroviral and lentiviral vectors Viral vectors have been the standard for genetically engineered cell therapies for decades and are derived from retroviruses, including γ‐retroviruses (e.g., Murine Leukemia Virus) and lentiviruses (e.g., human immunodeficiency virus [HIV]‐1). 36 Vectors leverage viral infectivity to deliver genes of interest (GOIs) into host cells for permanent integration into the host genome, allowing any progenitors of modified cells to possess the genetic change. Genes encoding core viral proteins for infectivity include gag, for capsid and matrix proteins, env, for viral envelope glycoproteins involved in cellular tropism, and pol, for reverse transcriptase and integrase. Lentiviruses have regulatory genes that facilitate viral replication, including tat, which encodes a transcriptional activator that supports transcriptional activation and RNA polymerase II elongation, and rev, which encodes the Rev protein that binds to viral RNA transcripts to facilitate their nuclear export. Entry of viral particles into host cells is mediated by the envelope glycoprotein binding to specific cell receptors, which prompts lipid bilayer fusion and unloading of viral cargo into the cytoplasm (Figure 1A). Reverse transcriptase transcribes viral RNA into DNA flanked by complementary long terminal repeats (LTRs) and, via integrase, is permanently integrated into the host genome. Viral Vector (A) and Transposon‐Transposase (B) Mechanisms. Created with BioRender.com. In addition to the viral elements required for transduction, viral vectors used for cellular therapy contain the GOI, such as a CAR construct or a desired modified gene. While transduction efficiency depends on several factors, including cell type, vector load, and enhancers, viral vectors used in clinical applications generally exhibit high efficiencies. 37 Viral vector‐based transduction of cellular therapies with γ‐retroviral (γ‐RV) or lentiviral (LV) vectors has been widely applied in clinical research, with subsequent FDA approvals for various clinical indications. Two LV‐based FDA‐approved gene therapies for hemoglobinopathies include betibeglogene autotemcel (Zynteglo) and lovotibeglogene autotemcel (Lyfgenia). Zynteglo was FDA‐approved in 2022 for treating transfusion‐dependent β‐thalassemia (TDT) based on clinical trial data showing that 89% of patients achieved transfusion independence for at least 12 months. 38 , 39 Lyfgenia became FDA‐approved in 2023 for patients with sickle cell disease (SCD) ages 12 and older with a history of vaso‐occlusive events (VOEs), based on data showing that 88% of patients achieved complete resolution of VOEs within 6–18 months of treatment. 40 Both products use a LV vector, LentiGlobin BB305, which carries the human βA‐T87Q‐globin gene that encodes a β‐globin variant that exhibits the functionality of wild‐type β‐globin. 41 All currently FDA‐approved CAR T‐cells utilize viral vectors, with two employing γ‐RV vectors (Yescarta and Tecartus) and five employing LV vectors (Kymriah, Breyanzi, Abecma, Carvykti, Aucatzyl). 42 , 43 One concern with using viral vectors is the unintentional generation of replication‐competent retrovirus (RCR) during vector manufacturing, which can contaminate a culture of transduced cells and become amplified. 36 To minimize this risk, the design of viral vectors has undergone several generations. First‐generation HIV‐based LV vectors retained most viral components that were split across multiple plasmids to decrease the risk for recombination events. To facilitate transfection, pseudotyping the envelope protein with, for example, vesicular stomatitis virus glycoprotein allows viral entry independent of the host receptor, broadening tropism. 44 Second‐generation vectors enhanced safety by removing viral auxiliary genes involved in viral proliferation and infection. Third‐generation vectors lack the rev and tat genes involved in replication (rev may be kept for vector production, but not for transduction in cellular therapy manufacturing). Furthermore, self‐inactivating (SIN) vectors include a deletion in the 3′‐LTR to inhibit transcription of the full‐length virus. 45 Given these developments, the FDA guidance for retroviral vector‐based human gene therapy recommends that each lot of ex vivo transduced cells be tested for RCR, but a safe vector design and accumulated manufacturing or clinical experience with RCR‐negative products could justify the reduction or removal of testing. 46 Viral vectors also demonstrate semi‐random integration into the host genome with a preference for active gene expression regions, which poses a risk for insertional mutagenesis and oncogenesis. 47 Recent applications have shifted toward LV vectors, which appear to have a safer profile than γ‐RV vectors. While γ‐RV vectors tend to integrate into untranslated regions, promoters, and exonic regions, LV vectors demonstrate a preference for intronic and intergenic regions. 47 , 48 In an early clinical trial for X‐linked severe combined immunodeficiency using a γ‐RV vector, four of nine treated participants developed T‐cell leukemias due to insertional mutagenesis. 49 Additional cases have been reported in other studies using γ‐RV vectors, with a meta‐analysis indicating an increased risk associated with their use compared to LV vectors. 50 Despite a safer profile of LV vectors for ex vivo gene therapy, posttreatment malignancies have been reported. A subset of patients treated with the FDA‐approved LV vector‐transduced gene therapy elivaldogene autotemcel (Skysona) for cerebral adrenoleukodystrophy developed acute myeloid leukemia (AML) or myelodysplastic syndrome associated with clonal vector insertions within oncogenes and clonal evolution, including the acquisition of somatic genetic defects. 51 Since approval, 10 of 67 (15%) trial participants have developed hematologic malignancies, with an updated black box warning for the therapy. 52 The occurrence of these cases is thought to be at least in part due to an intact MNDU3 promoter‐enhancer element in the LV vector (Lenti‐D), which drives strong, ubiquitous constitutive transgene expression across hematopoietic cell lineages and has the potential to drive expression of nearby endogenous genes through transactivation. 51 MNDU3 was chosen for its broad expression, as at the time of the initial research, it was unknown which specific cell types were required to express functional and therapeutic levels of the ATP‐Binding Cassette Subfamily D Member 1 (ABCD1) protein. In contrast, the LentiGlobin BB305 LV vector used for Zynteglo and Lyfgenia employs gene regulatory elements derived from the endogenous β‐globin locus, lacking the transactivating capability and exhibiting erythroid lineage‐restricted expression. So far, no cases of insertional mutagenesis linked to LentiGlobin BB305 have been reported. Based on molecular analyses, one reported case of AML likely occurred independently of vector insertion and was due to other contributing factors. 53 For commercialized CAR T‐cells, the FDA issued a black box warning after receiving reports of 22 cases of posttreatment T‐cell malignancies after more than 27,000 treatments, with three cases in which the CAR transgene was detected in the malignant clone. 54 Some groups have since provided additional data on the rarity of these events. 55 , 56 Furthermore, similar to the LentiGlobin BB305 LV vector, the viral vectors employed for CAR T‐cells employ weak‐to‐moderate strength promoters (e.g., elongation factor 1‐α variant) that lack the transactivation capabilities of MNDU3. 57 While a higher number of viral vector integrations is associated with increased transduction efficiency, it may also increase the risk for mutagenesis. Therefore, release criteria testing of CAR T‐cells includes vector copy number (VCN), the average number of integrations per CAR‐positive cell. 58 The VCN should be below a specified threshold (generally five or fewer copies per transduced cell), which balances efficient transduction with minimizing risk. 59 Furthermore, patients treated with gene therapies require long‐term follow‐up for delayed adverse events related to viral vector integration. 60 Pending further data and post‐market surveillance, current CAR‐T clinical trials and commercial therapies use both γ‐RV and LV vectors. Although γ‐RV vectors carry a higher risk of genotoxicity than LV vectors, as was most prominently seen for HSPC‐based gene therapy (for which LV vectors are strongly preferred and have largely replaced γ‐RV vectors), there is currently no evidence that CAR T‐cells transduced with a given vector type portend a higher posttreatment malignancy rate. Given the very low overall rate of T‐cell malignancies and the potential for these therapies to greatly outweigh the risks, clinicians may continue to focus on the indication and efficacy regardless of vector type while maintaining appropriate posttreatment monitoring. In the product development setting, researchers may wish to explore new gene modification methodologies with the potential for a lower risk of genotoxicity. When selecting a vector type, rather than categorically avoiding γ‐RV vectors, future cell therapy developers may consider the specific application, cell type (e.g., HSPC vs. mature T‐cells), and clinical indication. Besides challenges related to safety, vector procurement and use are associated with the high cost of production under current good manufacturing practice (cGMP) conditions. For instance, the cost for purchasing sufficient LV vector for a clinical trial involving 50 participants has been reported to be as high as $800,000 in the United States. 61 The procurement and manufacturing times of clinical‐grade viral vectors can be lengthy, exceeding 18 months, and are often subject to delays. 62 The limited cargo capacity (6–7 kilobases + 2 kilobases of viral elements) and the inefficiency of using more than one viral vector for a given product may also prohibit the development of products requiring multiple GOIs or larger insert sizes. NONVIRAL GENETIC MODIFICATION METHODS In addition to viral vectors, nonviral methods have been used to achieve gene delivery and integration. These approaches include zinc finger nucleases, transcription activator‐like effector nucleases‐mediated gene editing, transposon‐transposase systems, and CRISPR‐based systems. This review will focus on the latter two systems, which are increasingly used. CRISPR‐based approaches include CRISPR/CRISPR‐associated protein (Cas), base editing, and prime editing. Transposons Transposons (“jumping” genes) are capable of moving locations within the genome either as RNA transposons, which use reverse transcription for a “copy and paste” approach, or DNA transposons, which use a “cut and paste” mechanism. 63 , 64 DNA transposon systems have been used for CAR T‐cell engineering and involve the use of donor DNA (e.g., plasmid) that contains the GOI (e.g., CAR construct) as the transposon, which is flanked by inverted terminal repeats (ITRs) (Figure 1B). A transposase enzyme binds to the ITRs to excise the transposon for random insertion into the host genome. Transposon‐based approaches, such as Sleeping Beauty (SB), piggyBac, Tol2, and TcBuster, have been evaluated in CAR‐T clinical trials. 65 , 66 , 67 A recent phase I/II clinical trial utilized SB to engineer donor‐derived anti‐CD19 CAR cytokine‐induced killer cells (CARCIK‐CD19) for the treatment of patients with B‐cell precursor acute lymphoblastic leukemia (ALL) who relapsed after allogeneic hematopoietic cell transplantation. 68 Cellular starting material was collected from the previous transplant donor. Thirty of 36 treated patients (83.3%) achieved complete remission, with 89% of responders achieving minimal residual disease negativity. As with viral vectors, one potential safety limitation is random integration, which poses a risk for insertional mutagenesis. Integration analyses indicate that piggyBac prefers transcriptional start sites, while SB and TcBuster integration patterns have been reported to be closer to theoretical randomness. 69 , 70 In the SB‐modified CARCIK‐CD19 trial mentioned above, based on integration site and clonal expansion analyses, the investigators concluded that the SB transposon was not biased toward regulatory and exonic regions, maintained a close‐to‐random insertion profile post‐infusion, and had an absence of clonal dominance. 68 Thus far, there have been two cases of posttreatment lymphoma after piggyBac‐transposed CAR‐T treatment for non‐Hodgkin lymphoma, with no reported cases for SB. 71 Compared to viral vectors, transposon‐based approaches have higher cargo capacities (up to ~200 kilobases) and are therefore more amenable to multiplexing at lower costs. 72 In the above CARCIK‐CD19 trial, the authors reported that the estimated cost of goods for this product was five to ten times lower than that of viral vectors, despite comparable average transduction efficiency and final cell viability. 73 CRISPR‐Based Tools CRISPR/CRISPR‐associated protein (Cas)9 The CRISPR/Cas gene‐editing platform was developed from a bacterial adaptive immune defense mechanism against viruses, in which viral DNA elements are captured and integrated into the bacterial genome as “spacers” within a CRISPR array (a “memory bank” of prior viral infections) that is transcribed into RNA and processed into shorter fragments known as guide RNAs (gRNA). 74 These gRNAs bind to endonucleases and guide them to complementary invading viral DNA for cleavage. The CRISPR/Cas platform for gene editing, therefore, consists of an endonuclease, most commonly Cas9, and a gRNA that binds to the endonuclease, forming a ribonucleoprotein complex (Figure 2A). 75 The gRNA guides the complex to a 20‐base pair protospacer adjacent motif (PAM) sequence located next to the target DNA sequence. The endonuclease cleaves the DNA, introducing a double‐stranded break (DSB). Endogenous DNA repair mechanisms are activated through two primary pathways: nonhomologous end joining (NHEJ) and homology‐directed repair (HDR). NHEJ is a default cellular repair mechanism that is error‐prone and joins broken DNA ends, often causing insertions and deletions (indels) that can lead to frameshift mutations and premature stop codons. HDR occurs only in actively dividing cells and, for gene editing, requires a donor DNA repair template with the desired gene edit flanked by sequences homologous to the target site. HDR is preferred for stable integration of gene edits, such as insertions, deletions, and sequence changes. However, because HDR is usually outcompeted by NHEJ in DSB repair, precise edits by HDR are frequently outnumbered by indels, reducing overall editing efficiency. 76 As a result, the CRISPR/Cas platform may be better suited for gene disruption or deletion applications rather than precise edits. Furthermore, there is a risk of off‐target edits, which are discussed below. CRISPR/Cas9 (A), Base editing (B), and Prime editing (C) Mechanisms. Created with BioRender.com. CRISPR/Cas9 editing is currently being used for both CAR‐T and ex vivo gene therapy applications. Exagamglogene autotemcel (Casgevy) is the first FDA‐approved cellular therapy to use CRISPR/Cas9 to knock out BCL11A, a gene involved with repressing fetal hemoglobin (HbF) production. BCL11A disruption reactivates HbF production in adult erythrocytes. 40 , 77 , 78 Casgevy received approval in late 2023 for treating patients aged 12 and older with SCD and recurrent VOEs, as well as for TDT. The approval was based on clinical trial data showing that 93.5% of patients with SCD were free of severe VOEs for at least 12 months after a 24‐month follow‐up, and 91.4% of patients with TDT achieved transfusion independence at 12 months. Targeted introduction of GOI poses less risk for insertional mutagenesis and is being investigated for CAR T‐cell indications such as relapsed/refractory B‐cell non‐Hodgkin lymphoma. 79 Base editing Base editing is a method based on the CRISPR/Cas system that combines sequence recognition with the ability for single‐nucleotide conversion and can be used to correct pathogenic point mutations or to disrupt genes of interest. 80 Base editors are composed of a modified Cas9 variant, commonly Cas9 nickase (nCas9), linked to a deaminase enzyme (Figure 2B). Compared to the Cas9 used in canonical CRISPR/Cas9 editing, nCas9 introduces a single‐stranded DNA nick rather than DSBs and retains its ability to bind to gRNA, which directs the base editor complex to the target DNA site for the deaminase to chemically alter a base. DNA repair and replication mechanisms then incorporate the change. Cytosine base editors facilitate cytosine‐to‐uracil conversions (and with subsequent replication, thymidine), while adenine base editors mediate adenine‐to‐guanine conversions. Besides transition edits (purine‐to‐purine, pyrimidine‐to‐pyrimidine), newer base editors can perform some transversion edits (e.g., adenine to cytosine). 81 Base editing is being used in several cellular therapy clinical trials. As an example for CAR T‐cells, base editing was used in a phase I study to generate universal off‐the‐shelf allogeneic anti‐CD7 CAR T‐cells to treat a patient with relapsed T‐cell ALL, who had molecular remission within 28 days after infusion. 82 Base editing was used to inactivate several genes, including those encoding CD52 (for CAR‐T resistance against anti‐CD52 lymphodepletion, which is used to prevent host rejection of the allogeneic cells), CD7 receptors (to prevent fratricide, in which anti‐CD7 CAR T‐cells target other CAR T‐cells that express the CD7 T‐cell antigen and induce cytotoxicity), and the β chain of the αβ T‐cell receptor (to prevent GVHD mediated by donor T‐cell receptors). An LV vector was used to express the anti‐CD7 CAR construct. Base editing is also being used in ex vivo gene therapy trials for hematologic and immune deficiency indications. In a phase I gene therapy trial for patients with SCD and severe VOE, base editing is used to introduce targeted point mutations to the promoter regions of the HBG1 and HBG2 genes, which encode γ‐globin, to prevent binding of the BCL11A repressor protein and increase HbF production. 83 Seventeen patients achieved HbF induction of >60%, sustained sickle hemoglobin reduction to <40%, and had no VOEs. 84 Given that base editing occurs without DSBs, this method results in greater efficiency and fewer indels compared to CRISPR/Cas editing for the same purpose. 76 For gene‐knockout, a recent study demonstrated that base editing resulted in a 210‐fold lower rate of chromosomal translocations. 85 However, base editing has some limitations. Despite the development of newer base editors, base editing currently cannot perform all 12 possible base changes. Furthermore, because canonical base editors function in a small window of five nucleotides, nearby nontarget nucleotides may also be deaminated, leading to ‘bystander editing’. Although bystander editing often results in silent mutations, efforts to optimize the method and reduce undesired edits may help to avoid clinically consequential edits. Off‐target edits in DNA and RNA, which are discussed below, have also been reported with this method. 86 Prime editing Prime editing is the most recently developed gene‐editing platform to overcome the limitations of CRISPR/Cas and base editing, with the ability to introduce insertions, deletions, and all possible nucleotide changes with high precision and without introducing DSBs. 87 A prime editor complex consists of nCas9, reverse transcriptase (RT), and prime editing guide RNA (pegRNA) (Figure 2C). The pegRNA contains the programmable RNA template for the desired DNA edit and guides the prime editor complex to the target site. The nCas9 introduces a single‐stranded DNA nick, and the RT extends the new DNA sequence using the pegRNA as a template. The newly synthesized DNA strand containing the desired edit is incorporated into the genome through natural cellular processes. Updated versions of the prime editor include additional elements to increase the efficiency of incorporating desired edits. Prime editing was recently used in a phase I/II first‐in‐human trial to treat a patient with a variant of chronic granulomatous disease (CGD) caused by a defect in the p47 subunit of the nicotinamide adenine dinucleotide phosphate (NADPH) oxidase complex required for neutrophilic antibacterial and antifungal activity. 88 The product, PM359, was corrected for the most common mutation of the p47 variant and the autosomal recessive form of CGD, a two‐nucleotide deletion in the NCF1 gene, which encodes the p47 subunit. Treatment with PM359 restored NADPH oxidase activity in 58% and 66% of neutrophils by days 15 and 30, respectively, with rapid engraftment of neutrophils (day 14) and platelets (day 19). Based on preclinical studies, despite its high efficiency of correction, off‐target editing, unintended editing, and chromosomal arrangements were not detected. 89 While prime editing is not designed to introduce DSBs and the risk of indels is significantly lower, inadvertent DSBs can still occur. 86 Other limitations of the original prime editing system include: lower efficiency relative to other methods, especially for larger edits; the large size, complexity, and susceptibility to degradation of the pegRNA, making delivery to cells and optimization challenging; and a competing endogenous mismatch repair system, which can correct desired mismatches introduced by prime editing. 90 Ongoing research efforts continue to increase efficiency and address these limitations. Challenges and Safety Considerations for CRISPR‐Based Methods CRISPR‐based approaches involve precise targeted edits and, therefore, have a lower risk of insertional mutagenesis compared to untargeted integration approaches of viral vectors and transposons. These approaches, however, present unique challenges, such as off‐target effects, in which editing occurs at a nontarget site, often due to high‐sequence homology to the target site, which may result in adverse effects. 91 Cas9‐induced DSBs can cause indels, which can lead to gene disruption, or large structural variations such as translocations, large deletions, inversions, chromosomal loss, or insertions of donor DNA at nontarget sites. 92 Reducing the risk of off‐target edits depends on multiple factors, such as optimizing the gRNA to have minimal specificity for off‐target sites, selecting a delivery vehicle for genome editing components that minimizes the time of activity in cells, and choosing the optimal Cas nuclease for the specific application. 92 Base editing and prime editing do not introduce DSBs and, therefore, are less susceptible to these issues. In a recent comparison of multiplexed edits using CRISPR/Cas9 or base‐edited CD19 CAR T‐cells, both achieved high editing efficiencies; however, the CRISPR/Cas9 approach exhibited translocations and had higher activation of p53 and DNA damage response pathways. 93 The base‐edited cells demonstrated superior in vitro proliferation and in vivo antitumor activity. Although base editing and prime editing address the issue of DSBs in the CRISPR/Cas approach, they still have risks of off‐target effects and genotoxicity, albeit to a lesser extent. 86 Genome editing, therefore, requires monitoring throughout the development and clinical application stages, as well as long‐term patient follow‐up. The FDA guidance for products incorporating human genome editing recommends that assessments of off‐target editing frequency, intra‐ and interchromosomal rearrangements, as well as residual gene‐editing components, may be necessary for product release, depending on the results of nonclinical studies. 94 CONCLUSION AND FUTURE DIRECTIONS Genetic modification approaches for ex vivo cellular therapies initially used viral vector‐based methods and have since expanded to include nonviral techniques, such as transposon‐based and CRISPR‐based methods. Their advancement has enabled the treatment of a broader range of diseases, from oncologic indications to inherited immune disorders. While these methods offer tremendous therapeutic potential, they also carry the risk of genotoxic effects. Ongoing development will require a rigorous assessment of their safety. Ultimately, selecting a gene modification approach for a given cell therapy requires balancing safety with therapeutic goals, engineering capabilities, and practical considerations, including cost, manufacturing feasibility, and regulatory pathways. For example, if the therapeutic goal is a simple transgene addition (e.g., a CAR construct), a viral vector approach may suffice. However, for multiplex engineering that requires both transgene addition and gene‐knockout, a combinatorial approach using a viral vector and CRISPR or base editing could be employed. Manufacturing scalability and cost considerations of these approaches vary. Viral vector production is complex, time consuming, and requires specialized cell lines, extensive quality control for RCR, and costly GMP facilities. 95 Transposon and CRISPR reagents can be produced with high precision under GMP conditions in a scalable, modular fashion, with lower material costs and reagent interchangeability. That said, the cost of viral vector‐ and CRISPR‐based approved ex vivo gene therapies ranges approximately from $300,000 to $4 million USD, posing a significant affordability challenge for patients. 96 While multiple factors beyond manufacturing contribute to these high costs, continued research and development of sustainable gene modification methods may help reduce financial barriers to treatment. 97 As the efficacy and cost‐effectiveness of these methods improve, logistical and regulatory considerations will further determine the accessibility of these treatments. CONFLICT OF INTEREST STATEMENT This research was supported by the Intramural Research Program of the National Institutes of Health (NIH). The contributions of the NIH author(s) are considered Works of the United States Government. The findings and conclusions presented in this paper are those of the author(s) and do not necessarily reflect the views of the NIH or the U.S. Department of Health and Human Services. DATA AVAILABILITY STATEMENT Data sharing not applicable to this article as no datasets were generated or analysed during the current study. REFERENCES Associated Data Data Availability Statement Data sharing not applicable to this article as no datasets were generated or analysed during the current study.