Genetically engineered hypoimmunogenic iPSCs: Strategies and considerations for clinical applications

Abstract

Induced pluripotent stem cell (iPSC)-based therapies are rapidly advancing toward clinical implementation, with increasing numbers of trials and approvals for allogeneic iPSC-derived cell products. However, immune rejection remains a major barrier to the broad application of “off-the-shelf” allogeneic cell therapies for various diseases and injuries. With current advances in genetic engineering technologies, including genome editing, it is possible to establish a low-immunogenic cell product by disrupting immune recognition molecules and by overexpressing immunosuppressive factors. This review provides a comprehensive overview of current genetic engineering strategies for generating hypoimmunogenic iPS cells. Because multiplex genome editing and transgene integration may raise safety concerns, we further outline quality-control considerations for clinical-grade hypoimmunogenic iPSC products, including assessment of on-target outcomes, off-target mutations, structural variants, chromosomal abnormalities, residual editing components, and vector-related risks. Finally, we review current clinical development of genetically engineered hypoimmunogenic iPSC-derived products and discuss future challenges, including the importance of safety profiling and international regulatory harmonization. Understanding the advantages and limitations of each engineering and validation strategy will be critical for translating hypoimmunogenic iPSCs into safe, effective, and globally accessible regenerative therapies.

Highlights

• •Hypoimmunogenic iPSCs can be generated by disrupting immunogenic genes and/or overexpressing immunomodulatory factors.
• •Diverse CRISPR-Cas systems and their derivatives provide versatile tools for engineering hypoimmunogenic iPSCs.
• •Quality-control assays are essential for evaluating the safety of genome-edited cells intended for clinical applications.
• •Hypoimmunogenic iPSC-derived products are emerging as platforms for universal regenerative medicine.

1. Introduction

Several clinical trials of induced pluripotent stem cell (iPSC)-based cell therapy are currently underway, aiming to treat a range of diseases and injuries, such as Parkinson's disease, heart failure, macular degeneration, spinal cord injury, thrombocytopenia, cancer, GVHD, and articular cartilage defect. Accumulating evidence supports the safety and efficacy of cell therapies utilizing iPSCs derived from either autologous sources or allogeneic donors [1]. However, several challenges continue to impede the broader clinical application of iPSCs. One significant obstacle is the time and cost associated with generating clinical-grade tailored iPSC lines, as each line requires rigorous safety validation, and the differentiation process further extends preparation time [2]. When compared to autologous cell sources, allogeneic cells offer several advantages in both manufacturing efficiency and cost-effectiveness, because a single or a few pre-established cell lines can be utilized for treating multiple patients. Therefore, the allogeneic approach is increasingly favored as a viable strategy for expanding the accessibility of iPSC therapies while maintaining economic feasibility [3].

Recent advancements in genetic engineering tools, such as Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR) systems, have opened the door to precise engineering of target cells, either by overexpressing immunosuppressive genes or by knocking out specific immunogenic cell surface proteins. However, for researchers just stepping into the cell therapy field, it may be unclear which genes to target, which CRISPR-Cas type to use, how to ensure the edited cell product is clinically safe, and many other concerns. This review aims to guide the selection of hypoimmunogenic strategies and the appropriate genetic tools for their implementation. Additionally, we outline recommended procedures for quality control of genome-edited cells to ensure their safety in clinical applications.

2. Gene editing strategies for hypoimmunogenic iPSCs

In the following sections, we will first provide an overview of cutting-edge genetic engineering technologies and then discuss target genes for immune rejection evasion, which we classify into knockout and overexpression categories.

2.1. Genetic knockout methodologies

CRISPR was originally discovered in diverse bacteria and archaea as an adaptive defense mechanism against bacteriophages [4]. The classification of CRISPR-Cas systems is complicated by their diversification, but they can be divided into two main classes based on their CRISPR-associated (Cas) effector genes, with Class I systems using multi-Cas protein complexes and Class II systems relying on a single Cas effector protein [5]. A list of CRISPR-Cas nucleases for genome editing in mammalian cells is summarized below and in Fig. 1.

Fig. 1.

Variety of CRISPR-Cas systems for genome editing in mammalian cells.

To select a suitable CRISPR-Cas, researchers should consider the PAM, as this determines the cutting site. Moreover, the size of Cas effectors and gRNA is important, as different delivery cargoes or methods have different size limitations. A) Class I CRISPR-Cas consists of multiple subunits, hence it is generally larger in size. Class I type I-B, I-C, I-D, and I-E have been reported for gene editing in mammalian cells. B) Class II CRISPR-Cas only consists of one effector. Reported use in mammalian cells includes type II, type V-A, V-B, V-F, and V-I. Some Class II Cas are also used in multiplex, which could be convenient for editing multiple targets at once.

2.1.1. CRISPR class I type I

CRISPR Class I is the most abundant system in bacteria and archaea in nature, with multiple effector subunits needed to form a Cascade complex in the Class I system [5,6]. Among the seven subtypes, E. coli-derived type I-E has been most widely investigated [6]. Type III and type IV also belong to the class I system; however, type III mainly targets RNA and transcription-dependent DNA, while the function of type IV is not yet fully understood, so this review does not cover these types (Fig. 1A).

The type I system contains 7 diverse subtypes, classified as I-A to I-F and I-G (also known as I-U) [7,8]. Most subtypes have Cas3 as the effector, which contains nuclease and helicase activities. The Cascade complex of each subtype differs in number and protein variants. Cas3 is characterized by its ability to induce broad genomic deletions spanning kilobase-scale regions. The gRNA of the type I system requires only CRISPR RNA (crRNA), not trans-activating CRISPR RNA (tracrRNA).

Type I-E is composed of a Cascade complex including Cas5, Cas6, Cas7, Cas8, and Cas11, which then recruits Cas3 to cleave DNA at the non-target strand containing the PAM site, in the 3′-to-5′ direction [9]. Although the structure and mechanism are thoroughly examined, their application to human cells has received limited attention until recent years. Dolan et al. used the T. fusca type I-E system in the form of ribonucleoprotein (RNP) to generate long-range genomic deletions in human embryonic stem cells and myeloid leukemia cell line-derived HAP1 cells [10]. Mashimo group applied a plasmid-based E. coli K-12 type I-E system to induce broad-range deletion in HEK293T reporter cells [11]. Our group also successfully utilizes the type I-E system to remove a 4 kb gene fragment from the human dystrophin gene, or to knock out the β2-microglobulin (B2M) gene in iPSCs [11]. Implementation of the dual crRNA approach with the type I-E Cas3 system can induce a deletion of more than 300 kb in both HEK293T cells and iPSCs [12].

Recent research has begun investigating the potential of other type I subtypes as genome editing tools, including I-B, I-C, and I-D. Lu et al. demonstrated the type I-B CRISPR-Cas3 from Synechocystis sp. deleted up to 4.5 kb of the TRAC locus in human CD3⁺ T cells by RNA electroporation [13]. A type I-C CRISPR-Cas3 from Neisseria lactamica is reported to delete around 20 kb HPRT1 locus in HAP1 leukemia cell line and around 2 kb deletion in HeLa, HEK293T, and human embryonic stem cells [14]. Type I-C has a smaller Cascade as Cas6 is not needed, which could reduce cargo delivery burden. The type I-D system uses Cas10d instead of Cas3 for the nuclease function and induces bi-directional large deletion. Testing with HEK293T cells, the type I-D system induced over 19 kb deletion in the AAVS locus [15].

Type I CRISPR-Cas3 could be an attractive alternative to generate hypoimmunogenic iPSCs when aiming for a broad range of genomic deletions, such as the complete deletion of individual human leukocyte antigen (HLA) genes (HLA-A, B, C: 3.4 kb). Although Cas9 can also induce large genomic deletion on some rare occasions [16] or by using multiple single gRNA (sgRNA) [17], a comparison study between dual-Cas3 and dual-Cas9 indicated Cas3 has higher efficiency in large-gene deletion and does not induce unintended inversion at the edited site [11,12]. However, even with large-deletion capability, complete deletion from HLA-A to HLA-B will still require approximately 1.4 Mb, which remains challenging even for Cas3.

One possible drawback of Cas3 is the lack of precise control in the deletion pattern [11,14]. Firstly, the starting point of gene editing for type I is not as fixed as CRISPR-Cas9; instead, initiation sites disperse around the protospacer adjacent motif (PAM)-proximal region, and deletions extend from 3′ to 5’ on the non-target strand [11,14]. More importantly, the endpoints of deletion under the same gRNA are also variable, based on the deletion patterns observed [10,11,13]. This may increase the difficulty of selecting the correct cell colonies that have the desired deletions. Li et al. have reported a method to prevent long-range cleavage by Cas3 by introducing dCas9 into HEK293T cells, mouse embryonic stem cells, and mouse zygotes [18]. However, this approach requires introducing not only the full type I-E components, but also dCas9 and the corresponding gRNA. To achieve a highly precise large deletion, further advancement will be necessary.

2.1.2. CRISPR class II type II

The CRISPR Class II system can be further divided into three types, including type II (Cas9), type V (Cas12), and type VI (Cas13). Because type VI has RNase activity and targets only single-stranded RNA, we focus here on DNA-targeting types II and V systems (Fig. 1B).

The use of Streptococcus pyogenes-derived CRISPR-Cas9 as a gene-editing tool was first reported in 2012 [19]. Due to its high activity in mammalian cells and simple mode of use, it is widely used to knock out various gene targets for hypoimmunogenic induction, including B2M, HLAs, and CIITA genes. CRISPR-Cas9 uses crRNA and tracrRNA, or a fusion of both small RNAs into a sgRNA, to target specific DNA sites in the presence of “NGG” PAM. To simplify the gRNA design process, many online tools have been developed and described in detail in other reviews [20]. Multiplex editing using CRISPR-Cas9 is also reported. Ren et al. used electroporation to deliver Cas9-mRNA and multiple gRNAs into T cells to knock out TRBC, B2M, and PD1 genes. 65% of the cells showed deletion in both CD3 and HLA class I markers before selection, and reached 99% after purification with paramagnetic beads [21].

2.1.3. CRISPR class II type V

Cas12 belongs to the CRISPR-Cas class II system, like Cas9, but has a different protein structure and genome-editing activity. Cas12 contains the RuvC endonuclease domain but not the His-Asn-His endonuclease domain [22]. It recognizes a T-rich PAM sequence, such as 5′-TTN or 5′-TTTV, and produces a staggered end upon cleavage. Moreover, Cas12a requires only crRNA, not tracrRNA. The smaller size and unique crRNA processing of the CRISPR-Cas12 system make it easier for multiplex gene editing [23,24].

2.1.4. Various derivative technologies: Base editor and prime editor

Since the outcome of genome editing via DNA cleavage induction depends on the stochastic DNA repair mechanisms of the host cell, several more precise genome editing technologies that do not require DNA double-strand break (DSB) induction have been developed.

Base editing is a popular derivative technology developed by David Liu's group in 2016, utilizing Cas9 nickase and base deaminase [25]. Base editing can induce single-nucleotide alteration of interest [26]. The incorporation of cytidine deaminase with Cas9 nickase formed the cytosine base editors (CBEs), which can mediate the nucleotide change from C to T (or G to A at the complementary strand). In addition, adenine base editors (ABEs) are developed using adenine deaminase, which can convert A-to-G (or T-to-C). The Cas9 nickase generates a nick at the gRNA binding strand, which forces the DNA repair process to use the modified nucleotide on the opposite strand as the repair template. Recently, not only SpCas9 but also other Cas families, such as Cas12 [27,28] and Cas3 [29] variants, have been utilized to provide broader PAM selection and hence, more available sites for base editing.

By introducing single-nucleotide editing, a premature stop codon can be inserted in the target gene [30]. For example, CGA, CAG, and CAA in the target gene can be converted to the stop codons TGA, TAG, and TAA, respectively, by CBE. The base editing window should be considered when selecting suitable base editors, as it defines the region where bases will be modified. Bystander effects occur when nucleotides other than the target base are modified. Earlier versions of ABE, such as ABE7.10, have a 4-nucleotide editing window [31], which can result in bystander editing. Another challenge for ABE is bystander cytosine editing, as ABE is known to catalyze C-to-T conversion at certain sequences [32], which could interfere with the intended A-to-G conversion. Better base editors are established to tackle the above problems, such as ABE9, which is reported to minimize the window to one nucleotide without cytosine editing activity, although the editing activity tends to be lower than ABE8e [33]. Another variant, minimally evolved ABE (ME-ABE), was developed using favorable mutations in ABE7.10 [34]. ME-ABE has better on-target editing and lower off-target activity compared to ABE7.10, ABE8e, and ABE8.20, though the editing window is similar to ABE7.10 [34].

Base editors enable efficient base substitutions but are not capable of introducing arbitrary sequence changes or inducing small insertions. To address this limitation, the group led by David Liu developed an innovative genome editing tool known as Prime Editing [35]. This system employs a prime editing guide RNA in which the 3′ extension encodes a primer binding site and a reverse transcription template containing the desired edit. A reverse transcriptase fused to Cas9 nickase extends the 3′ end of the nicked DNA strand using this template, thereby enabling the installation of a wide range of precise edits, including substitutions and small indels at the target site.

2.2. Gene overexpression methodologies

2.2.1. Nuclease-based targeted knock-in

Nuclease-based knock-in uses various site-directed nucleases and enables targeted insertion, although the insertion structure and copy number require verification. Early nucleases for gene integration include zinc finger nucleases [36] and TALEN, followed by the emergence of the CRISPR system, such as Cas9, Cas12a, and Cas12b. These early systems mainly rely on creating a DSB at the target site, allowing the donor gene template to be inserted via homology-directed repair (HDR) (Fig. 2A). However, HDR efficiency is much lower compared with non-homologous end joining (NHEJ) and is only feasible in cycling cells [37].

Fig. 2.

Genetic strategies for overexpressing immune-protective genes.

Gene integration can be mediated by the CRISPR-Cas system. The insertion method depends on the repair pathway after inducing a double-strand break, which can be A) HDR-based or B) NHEJ-based. Gene delivery by C) viral vectors. AAV delivers a transgene as ssDNA form, which generally becomes double-stranded and persists in episomal form. Since the episome does not integrate into the genome, the copy number of the transgene decreases with each cell division. On rare occasions, ssDNA can also integrate into the cell genome. Lentiviral vectors deliver a gene in ssRNA form, which is reverse transcribed into dsDNA. As a lentiviral vector contains integrase, the dsDNA can be inserted into the genome. D) Transposon vectors. Transposase binds to the ITR region of the plasmid, forming a synaptic complex that then cuts and releases the GOI. The complex then recognizes a preferred motif and ligates the GOI into the genome.

To tackle the problem associated with DSB, prime editing can be utilized to insert a small sequence, typically less than 50 bp [35]. Improvements aimed at inserting larger fragments successfully increase the insertable size to around 800 bp in a template jumping PE design [38], while PE combined with a serine integrase can enable insertions of up to 36 kb [39].

As NHEJ is a more common DNA repair pathway and less dependent on a specific cell cycle stage, another approach is to use the NHEJ event for site-directed knock-in (Fig. 2B). Strategies that employ the NHEJ pathway include homology-independent targeted insertion (HITI) [40] and knock-in blunt ligation (KiBL) [41]. HITI showed promising editing efficiency across different human cell types, including the insertion of a transgene into hematopoietic stem and progenitor cells [42], and the insertion of a chimeric antigen receptor (CAR) into T cells [43] or iPSCs [44].

Finally, there are attempts to merge transposases with nucleases, which confer genomic integration capacity at or near the targeting site. Transposase-CRISPR-mediated targeted integration uses Cas9 and HITI, which achieves even higher editing efficiency than HITI alone in HEK293T cells [45].

2.2.2. Viral vectors

Conventional DNA integration methods involve the use of viral vectors (Fig. 2C). Lentiviral/retroviral vectors contain integrase that can insert target DNA into the genome of host cells [46]. However, they have limitations, including a size limit of around 8-10 kb [47], and a risk of insertional mutagenesis or oncogenesis [48], due to uncontrolled random integration [49].

The adeno-associated virus (AAV) vector can deliver a transgene as single-stranded DNA (ssDNA), which can enter the nucleus, become double-stranded, and persist as episomes in non-dividing cells. Though there are still small chances that the ssDNA will integrate into the genome by random integration [50], the chance is much lower than retroviral/lentiviral vectors. If needed, an AAV vector can be engineered to serve as a donor DNA for HDR [51]. When Cas9 or similar enzymes are combined with an AAV vector, the unintended integration of AAV vector fragments at the Cas9-targeting site can occur at a relatively high frequency[[52], [53], [54]]. The limitations of AAVs include inconsistent transduction efficiency in mammalian stem cells [55,56]. Another issue is the cargo size limitation of around 5 kb. Strategies such as splitting the transgene into multiple AAV vectors [57] or minimizing the cargo size are being developed.

2.2.3. Transposon vectors

DNA transposon vectors, including piggyBac, Sleeping Beauty, and Tol2, are widely used to insert a transgene into target cells, where efficient DNA transfection is feasible (Fig. 2D). In general, transposon vectors have greater cargo capacity than AAV. For example, the piggyBac vector has been shown to deliver a 100 kb fragment into mouse embryonic stem cells [58], but efficiency tends to decline as cargo size increases [[59], [60], [61]]. By controlling the amount of transfected DNA and transposase activity, the copy number of the transposon vector may be roughly estimated depending on the vector type and size. The integration sites of the transposon vectors are essentially random, and site bias is weaker than that of retroviral/lentiviral vectors. However, each transposon vector shows a preference for certain genomic regions. piggyBac vectors integrate into the 5′-TTAA-3′ sequence and tend to prefer within transcription units. Sleeping Beauty vectors integrate into the 5′-TA-3′ motif and have a weak preference to insert at an open chromatin region. Tol2 has no particular target motif, but prefers to integrate near the transcriptionally active regions [61,62].

Interestingly, some CRISPR-Cas subtypes found in certain bacterial strains also encode a transposase, known as the CRISPR-associated transposase (CAST). The CRISPR-Cas system being utilized included type I-B [63,64], type I-F [65,66], and Cas12k [67,68]. CAST components mainly include the CRISPR effector, either cascades in the type I system or a single effector in Cas12k, and some conserved transposition proteins like TnsA, TnsB, TnsC, and TniQ. Different CASTs have their own composition and integration mechanisms. Type I-B CASTs activities are reported only in bacteria [63,64], while type I-F [65] and Cas12k [69] CASTs have successfully inserted the DNA fragments into HEK293T cells, although the insertion efficiency is still less than 1-5%.

2.3. Delivery of genetic engineering tools into iPSCs for clinical-grade manufacturing

Delivery of genome editing tools and/or hypoimmunogenic transgene(s) into iPSCs is a critical step, as it affects genetic editing efficiency and safety. The major delivery methods for iPSCs include viral vectors, electroporation, and lipid-based transfection, with the choice of cargo types (DNA, RNA, or ribonucleoprotein) [70,71].

For transgene overexpression strategies, viral vectors are efficient, established means of delivery into various cell types for ex vivo genetic engineering. Lentiviral and retroviral vectors are approved for generating CAR-T cell products. However, there are still concerns about the risk of creating replication-competent viruses. The use of a transposon vector or nuclease-based knock-in would avoid this risk; however, an additional delivery method, such as electroporation or lipofection, is needed to deliver transposon DNA and transposase into the target cells.

As for gene knockout strategies, genome editing tools should be delivered transiently to minimize off-target mutagenesis risks. Considering this, the editing tools are typically delivered as mRNA or ribonucleoproteins, but not DNA, which could undergo random integration of a fragment.

2.4. Gene targets for hypoimmunogenic engineering

Strategies for avoiding immune rejection through genetic engineering can be broadly classified into two categories: suppression of immune recognition through gene disruption and activation of immunosuppressive pathways through transgene overexpression. In the former approach, mismatched HLA-related molecules expressed on the cell surface, which constitute major triggers of immune rejection, are targeted for genetic deletion. In the latter approach, genes encoding molecules that activate immunosuppressive pathways are introduced or overexpressed to attenuate host immune responses. Fig. 3 summarizes the genes that have been targeted by either gene disruption or transgene overexpression in previous studies, together with the corresponding immune cell populations affected by each strategy.

Fig. 3.

Candidate genes for developing hypoimmunogenic cells.

One approach is by the overexpression of target genes. SERPINB9 is an inhibitor for granzyme B that is released from CD8⁺ T cells or NK cells. PD-L1 is a ligand that binds to the receptor PD-1, which inhibits T cell activation. FASL is another ligand that induces apoptosis in T cells when it binds to the FAS receptor. Expression of IgG receptor CD64 can protect engrafted cells from antibody-dependent cellular cytotoxicity. CD47 binds to SIRPα and acts as a “don't eat me” signal. CD200 binds to CD200R1 to inactivate NK cells or macrophages. HLA-C, HLA-E, and HLA-G can suppress NK cells through their respective inhibitory receptors. Cytokines like TGF-β and IL-10 can promote the differentiation of macrophages into M2 macrophages rather than M1. CCL21 can promote a tolerogenic microenvironment. CTLA4-Ig binds to CD80/86 and acts as a competitor for T cells' CD28. The other approach is to knock out genes. To inhibit CD8⁺ T cells, HLA class I molecules (HLA-A, HLA-B, HLA-C) or B2M can be knocked out to avoid antigen presentation. Similarly, CD4⁺ T cells can be inhibited by HLA class II molecules (DP, DQ, DR) or CIITA knockout. Some populations of NK cells express the receptor CD226, which binds to the ligands CD112 and CD155, so knockout of these two genes can inhibit CD226⁺ NK cells.

2.4.1. Gene knockout strategies

iPSCs and their progeny express multiple types of HLA proteins, also known as major histocompatibility complex (MHC) molecules, on their surface. More specifically, Classical HLA class I molecules (HLA-A, -B, and -C) are broadly expressed in most nucleated cell types, forming a heterodimeric complex with B2M. In contrast, HLA class II molecules (e.g., HLA-DP, -DQ, -DR), which are composed of alpha and beta chains, are primarily expressed on antigen-presenting cells, including dendritic cells, macrophages, B cells, and inflammatory endothelial cells. The primary function of HLA class I is to present antigenic peptides to CD8⁺ cytotoxic T cells, whereas HLA class II molecules are responsible for antigen presentation to CD4⁺ helper T cells. Although cell surface HLA expression levels vary by cell type, they are significantly upregulated upon stimulation with inflammatory cytokines, such as interferon-gamma.

Classical HLA genes exhibit extensive polymorphism, particularly in the peptide-binding domains encoded by exon 2 and 3 for class I and exon 2 for class II. The sequence diversity at each HLA locus is described in terms of alleles, and the combination of HLA alleles across multiple HLA loci is referred to as the HLA haplotype. For example, in the Japanese population, there are more than 60 major HLA-A alleles, over 100 HLA-B alleles, and more than 30 HLA-C alleles [72]. The vast number of possible haplotypes, as a combination of multiple alleles, enables T cells to distinguish between self and non-self. Mismatched HLA class I and II haplotypes trigger the activation of CD8⁺ and CD4⁺ T cells, respectively [73]. To mitigate T cell activation, HLA-A, HLA-B, HLA-C, or B2M knockout can inhibit CD8⁺ T cells, while HLA-DP, HLA-DQ, HLA-DR, or CIITA depletion can inhibit CD4⁺ T cells.

Natural killer (NK) cells are granular lymphocytes belonging to the innate immune system. Some NK cell subsets express the receptor CD226, which binds to the ligands CD112 or CD155 commonly found in transformed or infected cells [74,75]. Hence, reducing CD112 and CD155 expression can avoid NK cell killing [76].

2.4.2. Gene overexpression strategies

One way to neutralize immune cell response is by tackling their mechanism of action. CD8⁺ T cells and NK cells kill target cells by releasing granzyme B, so overexpressing SERPINB9, an inhibitor of granzyme B, can act as protection [77,78]. B cells release antibodies, leading to antibody-mediated killing. CD64 can capture IgG F_c and prevent interactions with Fc receptor-bearing effector cells or the complement system [79]. CCL21 can trap regulatory cells to suppress immune responses. Moreover, a synthetic chimeric receptor known as alloimmune defense receptor (ADR) is designed to selectively target and eliminate activated host T and NK cells that express 4-1BB (CD137), thereby protecting allogeneic cell grafts from rejection and enhancing their functional persistence [80,81].

Immune cells can also be suppressed by overexpressing inhibitory ligands. PD-L1 and FASL inhibit T cells and B cells. CD200 inhibits NK cells and macrophages. Anti-inflammatory cytokines like TGF-β and IL-10 promote macrophages to differentiate into M2 macrophages, which suppress immune responses. Killer-cell immunoglobulin-like receptors (KIRs) are primarily expressed on NK cells and are critical regulators of NK cell activity, which interact with HLA class I molecules to modulate immune responses. Inhibitory KIRs recognize cell-surface HLAs and suppress NK cell cytotoxicity [82], whereas activating KIRs associate with HLAs in a peptide-dependent manner to activate NK cells [83]. Major ligands for inhibitory KIRs include HLA-C, with KIR2DL1 recognizing the C2 epitope and KIR2DL2/3 binding to C1 [84,85]. These receptor-ligand interactions enable NK cells to detect “missing self” signals on HLA-deficient cells, leading to activation of NK cells to eliminate them. Consequently, complete depletion of HLA class I molecules, such as through B2M knockout, renders cells vulnerable to NK cell attack [86,87]. To circumvent this issue, strategies include overexpressing inhibitory ligands, such as HLA-E (to engage NKG2A/CD94) and/or selectively retaining specific HLA-C allotypes (to engage inhibitory KIRs KIR2DL1/2/3).

Importantly, immune rejection is not mediated by a single immune cell type, such as CD8⁺ cytotoxic T cells alone, but rather is orchestrated through the coordinated actions of multiple layers of the immune system. Therefore, modification of a single target gene is generally insufficient to achieve robust immune evasion, and combinatorial engineering of multiple target genes is considered essential for effective hypoimmunogenic cell design. For more details on each immunological pathway, please refer to our recent review paper [88].

3. Quality control of genome-modified iPSC lines

3.1. Overview of quality-control considerations

Because hypoimmunogenic engineering often requires multiplex genome modifications and transgene integration, rigorous quality control becomes essential for clinical translation. In this section, we will focus on key quality-control considerations during the manufacturing process for genetically modified hypoimmunogenic iPSC cell products intended for clinical use (Table 1). In addition to the safety assessment criteria for clinical cell manufacturing by the CiRA Foundation in Japan [89,90], the FDA issued draft guidance in April 2026 on the use of next-generation sequencing methods to assess the safety of human genome-editing therapy products [91]. This draft guidance provides recommendations for NGS-based methods used in nonclinical studies to support the development of investigational human genome-editing products.

Table 1.

Recommended assays for quality control of clinical-grade genome-edited iPSCs.

Category	Quality check	Experimental method	Original iPSCs	Genome-edited iPSCs	Differentiated cell product
Quality	Identity	STR analysis	STR identical	STR identical	STR identical
	Purity	Flow cytometry	Pluripotency marker + ve	Pluripotency marker + ve	Differentiation marker + ve
	Potency	TBD	Tri-lineage differentiation	Tri-lineage differentiation	TBD
Safety	Sterility	Bacteria test	−ve	−ve	−ve
		Mycoplasma test	−ve	−ve	−ve
		Virus test	−ve	−ve	−ve
	Endotoxin	Endotoxin test	Within specification	Within specification	Within specification
	Tumorigenicity	TBD	/	/	−ve
	Edited Gene	qRT-PCR, RNA-seq, western blot, ELISA	/	Target gene knockout or expression	Target gene knockout or expression
	Editor persistence	qRT-PCR, RNA-seq, western blot, ELISA	/	Residual editing components within specification	Residual editing components within specification
Genetic	Karyotyping	G-banding	Normal karyotype	Normal karyotype	Normal karyotype
	CNV	WGS, SNP array	No unacceptable copy-number alterations	No unacceptable copy-number alterations	No unacceptable copy-number alterations
	Oncogenic mutation	WGS, WES	No unacceptable cancer-associated variants	No unacceptable cancer-associated variants	No unacceptable cancer-associated variants
	On-target editing	WES, Amp-seq, Sanger-seq	/	+ve	+ve (for genome edited products)
	Off-target mutagenesis	WGS, other	/	TBD	TBD
	Vector integration	PCR, Amp-seq, WGS	/	TBD	TBD
Traceability	Certificate of Analysis/Origin	/	Culture and Reprogramming processes	Genome editing processes	Differentiation processes

STR: Short Tandem Repeat. CNV: Copy Number Variation. TBD: To Be Determined (based on the properties of the final cell product). WGS: Whole Genome Sequencing. SNP: Single nucleotide polymorphism. COSMIC CGC: Catalogue of Somatic Mutations in Cancer-Cancer Gene Census. WES: Whole Exome Sequencing. Amp-seq: Amplicon-sequencing. +ve: Positive. −ve: Negative. Acceptance criteria and assay detection limits should be predefined and justified for the product.

Candidate off-target sites should be verified using appropriate targeted sequencing or enrichment-based approaches, ideally in the same or representative cell type used for product manufacturing, with editing rates comparable to those observed in the final product. The sequencing strategy should be selected based on the editing modality and expected genomic alterations. Long-read sequencing may be particularly useful when large insertions, deletions, or other structural changes are anticipated.

For ex vivo hypoimmunogenic iPSC-derived cell products generated by genome editing and/or vector-mediated transgene expression, quality testing should address both general cell product quality attributes and risks specific to genetic engineering. FDA guidance emphasizes assessment of on-target editing outcomes, off-target editing activity, chromosomal integrity, sample representativeness, delivery-related risks, editor persistence, and vector-related impurities or replication-competent virus when viral vectors are used.

3.2. Evaluation of genomic integrity

Genome mutations can range from single-nucleotide changes to large mutations that affect chromosome structure. Small-scale mutations can be regarded as changes that affect fewer than 50 base pairs of nucleotides, such as SNVs or indels. On the other hand, large-scale mutations, structural variants (SVs) that involve more than 50 base pairs, including copy number variations (CNVs), inversions, and translocations [92,93], sometimes span ranges greater than megabase scale. Depending on the location of the mutations, the biological significance may vary.

Tumorigenicity represents one of the most serious health risks for recipients because even a rare event can be amplified by a growth advantage. Activation of oncogenes or suppression of tumor suppressor genes should be carefully monitored. Such cancer-causing genes can be found in the Catalogue of Somatic Mutations in Cancer (COSMIC) Cancer Gene Census (CGC) database [94]. As of May 2026 (v104), 591 genes are classified as Tier 1, comprising genes with published evidence showing oncogenic function and mutation patterns, and 171 genes are classified as Tier 2, comprising genes that lack proof for either functional impact or mutation pattern [94]. Tier 1 genes are usually prioritized as the benchmark for tumorigenicity evaluation. Additionally, different countries may have their own regulations for evaluating genetically edited cell products. For example, the “Shibata list” is a curated panel of cancer-related genes widely used in Japan for genomic safety assessment of human pluripotent stem cells and their derivatives. The list consists of roughly 242 genes based on familial cancer-causing genes in the Online Mendelian Inheritance in Man (OMIM), genes presumably responsible for primary somatic cancer mutations found in the literature, and frequently found (top 20) genes in the COSMIC database [95]. The above lists serve only as references; currently, no definite regulations specify which oncogenes or mutation types would prohibit administering a cell product to a patient. It is also difficult to determine the exact effects of those mutations without experimental or clinical validation.

In 2022, Han et al. reported a clinical case in which a teratoma formed 2 months after transplantation of autologous iPSC-derived beta cells into a type II diabetic patient at Chennuo Medical Hospital in China, without prior genetic testing of the transplanted cells. A follow-up study performed cancer-related gene panel sequencing on the teratoma sample and identified several missense and frameshift mutations in 16 genes, including Tier 1 cancer driver genes such as DNMT3A, KMT2D (MLL2), MSH2, and MSH6 [96]. It is unclear when those massive mutations occurred, because no pre-transplant genomic data were available.

Another case also indicates the importance of a genetic test. For patients with bullous keratopathy, allogeneic iPSC-derived corneal endothelial cells (CLS001) were transplanted as a first-in-human clinical study. After transplantation into three patients, genetic testing of the transplanted corneal cells revealed an in-frame deletion in exon 22 of the EP300 gene (Lysine acetyltransferase) in one recipient. Since the mutation was not detected in the parental iPSCs, it most likely arose during differentiation. Importantly, no tumorigenic or adverse events were observed during the 1-year follow-up period, while visual acuity and corneal stromal edema improved [97]. If the final cell product cannot be cryopreserved after differentiation, it is challenging to wait for genomic test results before transplantation. However, prior genetic testing enables rigorous patient monitoring if any significant mutations are detected post-transplantation.

With advances in DNA sequencing technologies, various platforms offering different sequencing methods have evolved (Fig. 4A). It is crucial to select a suitable platform for accurately interpreting different mutation types. This depends on several factors, including mutation size, frequency, genomic position, sequence complexity, and more [98]. For detecting SNVs and small indels within the protein-coding sequences, short-read sequencers are most widely utilized. On the other hand, for CNV detection, whole-genome sequencing (WGS) with short-read sequencers or DNA microarray assays (CGH or SNP arrays) have been used. More recently, SNVs, small indels, and SVs can be detected simultaneously with advances in the accuracy and throughput of long-read sequencers (Nanopore, PacBio).

Fig. 4.

Methods for verifying the genetic integrity of genome-edited cells.

A) Genome sequencing methods include whole genome sequencing (WGS), whole exome sequencing (WES), and targeted sequencing. All three methods are suitable for detecting single-nucleotide variants (SNVs) and small insertions and deletions. WGS and, with more limited sensitivity and genomic coverage, WES can also detect copy number variants (CNVs) and structural variants. B) Chromosome abnormalities can be analyzed by karyotyping, which can identify translocation of chromosome parts and aneuploidy.

Regardless of the sequencing technology, appropriate sequencing depth must be considered to balance detection sensitivity and cost. Sequencing depth is defined as the number of times a specific base is sequenced, and so the higher the depth, the more confident the base call is. Given a fixed total sequencing output, narrowing the target genomic regions allows for a proportional increase in sequencing depth. WGS can cover the entire genome, but a significant amount of sequencing is required. Whole-exome sequencing (WES) sequences approximately 19,000 protein-coding genes, which account for only around 1-2% of the genome [99,100], hence enabling higher depth coverage. Some researchers may utilize cost-effective amplicon sequencing for investigating several hundred target genes (e.g., tumorigenic genes) or predicted off-target sites. As practical reference points, WGS is often performed at a depth of 30×, WES at 50-150×, and targeted amplicon sequencing at 500× or more, although the appropriate depth should be justified based on the intended sensitivity and risk assessment. The greater the sequence depth, the greater the sensitivity to detect rare mutations in the cell population. Hence, amplicon sequencing is effective for validating the detected off-target sites identified by WGS or WES [101].

3.2.1. Detection of small indels and SNVs

Because CRISPR class II enzymes primarily induce indels and SNVs, detecting these variants is important for assessing the risk of off-target mutagenesis. SNVs and indels can be detected by WGS and WES with high accuracy [102], but which method is better remains debatable. One big advantage of WGS is the non-biased evaluation of non-coding regions. Studies have shown that some cancer driver variants are located in non-coding regions, such as regulatory regions and non-coding RNAs, although at a lower frequency [103,104]. When applying WES, include sufficient regulatory and intronic regions to cover pathogenic variants [98]. It is reported that the inter-center variation of WES is greater than that of WGS [102], so WGS might be a better option if reproducibility is the priority, rather than cost and sequencing depth. Since WES involves capturing specific DNA target regions using probes, certain areas may not achieve sufficient sequencing depth due to variations in GC content and sequence complexity [105,106]. Therefore, it is crucial to perform WES using appropriately designed probes and/or long-read sequencing to ensure comprehensive coverage [106].

3.2.2. Detection of large indels

CRISPR-Cas editing may induce large mutations like SVs[[107], [108], [109], [110], [111]] and CNVs [112,113]. Microarray-based assays such as comparative genomic hybridization (CGH) arrays and single-nucleotide polymorphism (SNP) arrays are conventional methods for detecting CNVs but rely on probe design and exhibit bias in the detectable regions. NGS-based CNV detection has less bias and can detect other small mutations.

3.2.3. Evaluation of chromosomal abnormalities

Karyotyping is the standard assay for detecting abnormalities in chromosomal number and structure (Fig. 4B). The International Stem Cell Biobanking Initiative (ISCBI) recommended analyzing a minimum of 20 metaphase cells with G-banding [114]. The cell that displayed clonal abnormalities (identical structural abnormality in multiple metaphases) should be reanalyzed and excluded if the recurrent mutations known for PSCs (e.g., gain in chr12, 17, or 20q11.21) or growth advantage-associated abnormality are confirmed. Furthermore, G-banding alone is insufficient to detect low-frequency mosaicism or sub-chromosomal CNVs. Hence, additional assays, such as SNP array, CGH array, FISH, or NGS-based assays, should be combined.

3.3. Evaluation of residual delivery vectors and editing components

Delivery vehicles, including viral and non-viral vectors or genome-editing components, may pose a safety risk if retained in edited cells. Immune responses have been reported against viral vector components [115] and the Cas9 protein [116,117]. Moreover, continuous expression of genome editing tools may increase off-target editing and the risk of oncogenesis when tumor suppressor genes are disrupted or oncogenes are activated as a consequence of genomic instability. In line with FDA recommendations for human genome editing products, including CMC considerations described in the January 2024 guidance, residual genome-editing components should also be assessed, such as Cas9 mRNA/protein, gRNA, donor DNA, vector particle, or transfection reagents. The individual components can be examined by different methods, such as real-time qPCR or droplet digital PCR (ddPCR) for residual plasmid DNA, and the corresponding reverse-transcription assays for Cas9 mRNA and gRNA[[118], [119], [120]], and ELISA, LC-MS/MS, or Western blot for Cas protein detection [119].

For products generated using viral vectors or donor DNA, additional testing for vector-derived sequences, insertion-site risks, replication-competent virus, and transgene expression stability is also important.

4. Current clinical development of genetically engineered hypoimmunogenic iPSCs

Clinical translation of iPSC-derived products reached a pivotal milestone in Japan with the recent conditional and time-limited regulatory approval of both iPSC-derived cardiomyocyte sheets and dopaminergic progenitors in March 2026, marking a new era in regenerative medicine. In patients with ischemic cardiomyopathy, allogeneic iPSC-derived cardiomyocyte sheets were evaluated with a brief three-month immunosuppressive regimen, with no severe adverse events attributed to the cell product during one year of follow-up after transplantation (jRCT2053190081) [121]. Breakthrough progress was observed in the treatment of Parkinson's disease, where transplantation of allogeneic dopaminergic progenitors remained free of inflammatory sequelae, even after the discontinuation of immunosuppressants at 15 months (jRCT2090220384) [122].

Following these successes, a number of iPSC-derived cell products featuring hypoimmunogenic engineering are being evaluated in human clinical trials, predominantly within the United States. A summary of these investigational products is presented in Table 2.

Table 2.

Clinical and preclinical applications of hypoimmunogenic iPSCs.

Targeting disease	Therapy cell type	Company	Clinical phase	Genetic engineering for hypoimmunity (and other purposes)	Reference
B-cell malignancies (NCT05336409) or B-cell-mediated autoimmune diseases (NCT06255028)	iPSC-derived CAR-NK cells (CNTY-101)	Century Therapeutics	Phase 1; sponsor-led activities discontinued	B2M and CIITA knockout HLA-E knock-in (Secreted IL-15 for NK cell growth) (Truncated EGFR safety switch) (CD19 CAR knock-in)	[123]
B cell lymphoma (NCT05950334)	iPSC-derived CAR-NK cells (FT522)	Fate Therapeutics	Phase 1	ADR targeting 4-1BB overexpression CD38 knockout (hnCD16 to enhance ADCC) (IL-15 and IL-15R fusion for NK cell growth) (CD19 CAR overexpression)	[81]
Type 1 diabetes	iPSC-derived islet (CNTY-813)	Century Therapeutics	Preclinical	B2M and CIITA knockout CD300a TASR and IdeS overexpression	[124]
Type 1 diabetes	iPSC-derived islet (SC451)	Sana Biotechnology	Preclinical (In preparation)	B2M and CIITA knockout CD47 overexpression (expected)	[125]

Century Therapeutics has advanced CNTY-101, an allogeneic iPSC-derived CD19 CAR-NK cell product, into Phase I clinical evaluation for B-cell malignancies and B-cell-mediated autoimmune diseases. CNTY-101 contains hypoimmunogenic edits, including B2M and CIITA knockouts and HLA-E expression, together with a CD19-targeting CAR, secreted IL-15, and a truncated EGFR safety switch. Clinical and translational findings from CNTY-101 studies have provided early evidence supporting repeat dosing and immune-evasion potential [123].

Fate Therapeutics is also evaluating FT522, an off-the-shelf iPSC-derived CD19 CAR-NK cell product, in patients with relapsed or refractory B-cell lymphoma (NCT05950334). FT522 is engineered to express a CD19-targeting CAR, a 4-1BB-targeting alloimmune defense receptor (ADR), high-affinity non-cleavable CD16 (hnCD16) to enhance antibody-dependent cellular cytotoxicity, and an IL-15/IL-15R fusion protein to support NK-cell persistence, with CD38 knockout to prevent CD38-mediated fratricide and improve compatibility with anti-CD38 antibody therapy. The ADR is designed to enhance persistence by eliminating alloreactive host immune cells, and early clinical findings suggest that FT522 can maintain anti-tumor activity with a favorable safety profile [81].

Although these are not iPSC-derived cells, a genetically modified primary pancreatic islet cell product UP421 has been evaluated in a single patient with type 1 diabetes. UP421 is genetically engineered by CRISPR-Cas12b-mediated B2M and CIITA knockout and lentiviral CD47 overexpression. The cells were transplanted into the forearm muscle without systemic immunosuppression, and detectable C-peptide responses suggested graft survival and insulin secretion [126].

Currently at preclinical stage, Century Therapeutics is developing the CNTY-813, an iPSC-derived beta islet replacement therapy for type 1 diabetes. CNTY-813 is genetically engineered to reduce MHC class I and II expression and to express an engineered ligand CD300a TASR (trans-antigen signaling receptor), which inhibits NK cells through CD300a binding, and IdeS, which is an IgG antibody-specific protease derived from S. pyogenes, aiming to evade T cell-, NK cell-, and antibody-mediated immune rejection [124].

Similarly, Sana Biotechnology is developing SC451, an iPSC-derived pancreatic islet cell therapy for type 1 diabetes. SC451 is expected to be modified similarly to UP421, to reduce HLA class I and II antigen presentation and to overexpress CD47, although exact strategy remains to be published [125].

While these initial clinical results are promising, larger-scale trials will be essential to confirm the long-term safety and broader efficacy of these hypoimmunogenic platforms.

5. Conclusion

This review summarizes a diversity of genome editing tools for creating hypoimmunogenic cells, and the crucial assays for quality control of the iPSC products. The generation of hypoimmunogenic iPSC stocks can make iPSC therapy more applicable to the wider population, but this ability can also be a double-edged sword. Extensive genetic editing for immune evasion may also raise safety concerns, including impaired immune surveillance against transformed [127] or infected cells [128]. A well-considered hypoimmunogenic strategy, careful selection of the genome engineering methodology, optimization of the manufacturing process, meticulous cell handling, thorough quality control, and reliable clinical operations will be necessary.

As a safety measure, a fail-safe system, such as inserting a suicide gene [129], genetic engineering to make cell proliferation dependent on an external uridine supply [130], or employing an antibody that specifically targets genome-edited cells, can be considered. Ensuring that such a complex system functions reliably in patients and completely eliminates rare abnormal cells remains technically challenging, and further research is needed.

Although guidance is available for generating clinical-grade iPSCs and for genome editing in cell therapies, the rapid evolution of genome editing techniques and validation assays underscores the need for ongoing updates. Since hypoimmunogenic cells have the potential to deliver the benefits of cell therapy to patients worldwide, we envision that cell preparations developed in one country will be utilized in other countries in the future. It will be challenging to maximize the potential of hypoimmunogenic cells within the current regulatory framework, where regulations vary by country. Therefore, it will be necessary to promote global standardization of regulations, starting with essential tests, such as sterility testing methods.

The year 2026 marks the 20th anniversary of the discovery of iPSCs. Significant progress has been made over the past two decades, with the establishment of various differentiation protocols and the initiation of clinical trials utilizing iPSC-derived cells. To further promote the use of iPSCs, the generation of hypoimmunogenic iPSCs may be a significant milestone for the next generation in broadening their clinical applications. Technologies are evolving rapidly, and more options will become available for hypoimmunogenic strategies, genome engineering tools, delivery methods, manufacturing processes, and quality control. It will be essential to understand the benefits and drawbacks of each choice to determine the most suitable protocol and ultimately create high-quality iPSC therapies for patients. Establishing internationally harmonized standards for genome-edited hypoimmunogenic iPSC products may become a critical foundation for the global distribution of universal regenerative medicines.

Author contributions

J.K.M.C. drafted the manuscript and prepared the figures and tables. A.H. conceived the review, supervised manuscript preparation, and revised the manuscript. Both authors approved the final version of the manuscript.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.