The History, Use, and Challenges of Therapeutic Somatic Cell and Germline Gene Editing

Abstract

The advent of directed gene editing technologies now over ten years ago ushered in a new era of precision medicine wherein specific disease-causing mutations can be corrected. In parallel with developing new gene editing platforms, optimizing their efficiency and delivery has been remarkable. With their development, there has been interest in using gene editing systems for correcting disease mutations in differentiated somatic cells ex vivo or in vivo, or for germline gene editing in gametes or one-cell embryos to potentially limit genetic diseases in the offspring and in future generations. This review details the development and history of the current gene editing systems and the advantages and challenges in their use for somatic cell and germline gene editing.

Capsule

This review provides an overview of the development, use, and challenges of somatic cell and germline gene editing for treating genetic-based diseases.

Introduction

The term clustered regularly interspaced short palindromic repeats, better known by its acronym CRISPR, is synonymous with gene editing. CRISPR first appeared in the NCBI PubMed database in 2002, where it was coined in a paper describing a novel family of repetitive DNA sequences interspaced by similarly sized non-repetitive sequences that only exist in prokaryotes (1). From 2002 through 2006, there were a total of 15 papers published with the acronym CRISPR. In 2007, it was discovered that the array of repeats and non-repetitive sequences represented a prokaryotic adaptive immune system designed to prevent productive host infection and lysis by bacteriophages(2). It was five years later when the power of a CRISPR-based system for use in gene editing was revealed.

Charpentier and Doudna published a landmark paper in 2012 demonstrating that a CRISPR-associated nuclease (Cas9) bound to the RNA generated from repetitive sequences integrated into the bacterial genome became active and would cleave double-stranded DNA sequences targeted by complementary RNA transcribed from the non-repetitive CRISPR DNA sequences (3). More importantly, from a gene-editing perspective, this study demonstrated that the Cas9 endonuclease could be easily programmed with single RNA molecules known as guide RNAs (gRNA) to cleave any specific DNA target site. In the following year, these findings were extended whereby it was observed that including a single-stranded DNA template possessing homology to the Cas9-gRNA target site could be used to introduce a DNA sequence of interest into cell lines and the genome of mice through a process known as homology-directed repair (4) (HDR). Gene editing is now feasible in any cell of any organism for which the genome is known and gene editing CRISPR/Cas gene editing reagents can be delivered. Indeed, there was an explosion in the number of papers published in the subsequent years demonstrating the feasibility of using genome editing techniques in somatic cells and embryos to alter the genome in any species.

The pace of the transition from the discovery and development of CRISPR/Cas9 as a gene editing tool to clinical application has also been striking as over 110 gene editing-based therapies are undergoing clinical trials (5), with treatments for sickle cell disease and β-thalassemia currently undergoing Phase 3 trials. All clinical trials to date represent somatic cell gene editing approaches wherein the editing takes place in differentiated somatic cells. Genome editing that results in the transmission of a modified gene to subsequent generations, i.e., germline gene editing, is of interest as a treatment for inherited diseases but faces significant ethical and technical challenges that need to be addressed before it can be clinically applied. Although the technologies are based on the same molecular principles, the current landscape regarding somatic versus germline genome editing applications is substantially different, with the latter being stalled from further therapeutic implementation due to its use to generate “CRISPR babies” before being fully vetted for ethical and safety concerns. This review provides a history of the development of gene editing techniques, their utilization as somatic cell or germline-based therapies for treating diseases, and the challenges that both somatic and germline editing face in realizing their clinical potential.

The History and Development of Gene Editing Technologies

The first gene targeting in mammalian cells was successfully conducted in 1985 by Smithies (6). In this study, Smithies et al. demonstrated that exogenous DNA sequences could be introduced into a specific genomic locus by utilizing homologous recombination in cultured mammalian cells. The implications of this result proved substantial for the gene-editing field because it demonstrated that individual target genes could be disrupted either by removing exons or inserting an exogenous sequence that disrupts the translation of the native protein. By incorporating antibiotic selection into homologous recombination vectors, the isolation of mouse embryonic stem (ES) cells possessing a disrupted gene of interest became possible, allowing for the subsequent transfer of edited ES cells into mouse blastocysts and the generation and breeding of gene-edited mice. Being able to generate specific gene “knockout” mice allowed researchers to investigate gene function in vivo and develop rodent models of disease (7-9). However, this strategy was limited to mice and, relative to CRISPR/Cas9, has a low gene targeting efficiency.

In parallel with the discovery of homologous recombination, a different approach for introducing mutations in target genes involved creating double-stranded DNA (dsDNA) breaks at or near a DNA region of interest. The dsDNA repair mechanism non-homologous end joining (NHEJ) frequently leads to random insertion or deletion (indel) mutations that create frameshift mutations and premature stop codons. The ability to target a gene of interest to introduce indels via NHEJ was based on the discovery that certain proteins or domains of proteins could bind to specific nucleotides, including zinc-finger (ZF) proteins in Xenopus oocytes that can recognize and bind to three specific nucleotides (10) or transcription activator-like effector domains (TALEs) comprised of 33-35 amino acid sequences from the plant pathogen Xanthomonas that can recognize and bind to a single nucleotide(11, 12). ZF or TALE domains arranged in a continuous array that recognizes a specific DNA target were fused to the nuclease domain of the restriction enzyme FokI, which requires dimerization to cut DNA. Thus, a ZF or TALE array and FokI nuclease assembly (zinc finger nuclease, or ZFN; TALE nuclease, or TALEN) must be constructed that recognizes both DNA strands and places the FokI nuclease from each ZFN nearby such that it will dimerize and introduce double strand DNA breaks (DSB) (13). ZFNs and TALENs were used to effectively target specific loci within multiple species (14-18), making them a significant advancement in the gene-editing research field. However, several issues limit the usefulness of ZFNs and TALENs for gene editing, including their generation being time- and labor-consuming.

The practical application of CRISPR and Cas9 was not immediately apparent; taking over a decade to translate the discovery of the CRISPR/Cas system in bacteria to its adaptation for use in gene editing. The presence of palindromic repeats, from which the acronym CRISPR is derived, was first reported in 1987 when Ishino and colleagues found that 29 nucleotides were arranged as direct repeats with 32 nucleotide intervening spacers in Escherichia coli (19). It was suggested that the spacer sequences between the repeats in bacteria were remnants of a previous infection with bacteriophages or plasmids that serves as an anti-sense RNA defense mechanism to prevent subsequent infection (20, 21). However, the exact mechanism through which CRISPR functions as a bacterial defense system was not determined until 2007 when it was observed that a new spacer sequence was added to a host bacteria’s genome following bacteriophage infection. These studies revealed that bacteriophage resistance specificity is determined by the phage sequence similarity of the spacers in the CRISPR array and that the effector arm of the system was due to the associated cas nuclease genes (2).

Defining how the components of the CRISPR system work together to cleave a specific DNA target was central to its adaptation as a flexible and highly efficient gene editing platform. First, it was discovered that the CRISPR RNA (crRNA) transcribed from the spacer sequences binds to the Cas endonuclease produced by the Cas proteins, with the latter being translated as part of a larger CRISPR/Cas operon (22). An additional RNA molecule, the trans-activating CRISPR RNA (tracrRNA), was also found to be critical for Cas9-mediated DNA interference in vitro and in vivo (23) whereby Cas9 interacts with precursor crRNA (pre-crRNA) and tracrRNA. An RNase then processes the pre-crRNA-tracrRNA duplex, forming a ternary Cas9-crRNA-tracrRNA complex capable of cleaving DNA. The tracrRNA serves to bind Cas9 and activate its nuclease domain. The DNA target specificity is determined by 20 nucleotides of the crRNA that hybridizes to the complementary strand of the target DNA, which is termed the protospacer region (2, 24). A short nucleotide sequence termed the protospacer adjacent motif (PAM) located on the target DNA is required for Cas/crRNA DNA binding and is unique for each Cas protein (25).

The studies that defined the requisite molecular components of the CRISPR/Cas system were the basis for its development as a highly adaptable and powerful gene editing tool. In 2012, Jinek et al. successfully demonstrated that the Cas9 endonuclease can generate DSB in vitro by utilizing crRNA that recognizes the target DNA and a tracrRNA (3). It was also demonstrated that the system could be simplified in that a single guide RNA (sgRNA) possessing the essential conserved elements of the tracrRNA coupled to the protospacer sequence could direct the Cas9 protein to the target sequence to introduce DSB. Thus, the CRISPR/Cas9 system requires only the Cas9 protein and a relatively small and easy-to synthesize sgRNA that includes a 20 bp sequence complementary to the target sequence and is adjacent to a PAM sequence (Fig. 1A). With these advances, it became straightforward to disrupt single or multiple genes in different cell types through the simultaneous introduction of Cas9 and one or more sgRNAs (26, 27). It was also demonstrated that a donor DNA template, typically single-strand DNA (ssDNA), with a defined sequence or mutation and complementary to the target gene, can be inserted into the genome through HDR (28). Although HDR was adopted for use in creating cell lines and genetically modified model organisms, its therapeutic use is limited by low efficiencies in non-dividing or slowly dividing cells (29). After the success of CRISPR/Cas9 system in gene editing research, several Cas9 variants were discovered or engineered to possess unique features that would be beneficial for gene editing applications, including different editing protein sizes and more permissive PAM sequences. The discovery of Cas12, for example, has added to the versatility of the CRISPR/Cas system because it features a longer guide sequence, a ‘TTTV’ PAM sequence, and results in sticky-end DSBs. Collectively, these features may reduce off-target effects, increasing the flexibility of target loci selection and increasing the chances of HDR (30, 31). The list of engineered or Cas variants has rapidly expanded over the last few years (32), but they remain to be tested for efficacy and precision in somatic cell or germline editing contexts.

Figure 1.

Overview of the current Cas9-based gene editing systems. (A) When bound to a sequence-specific guide RNA (gRNA), the wild-type Cas9 recognizes a target sequence and induces a double-stranded break (DSB). The cell then employs one of its major DNA repair mechanisms, either Non-Homologous End Joining (NHEJ) or Homology-Directed Repair (HDR), to repair the DSB. NHEJ can result in random insertion/deletion mutations that disrupt the target gene, while HDR in the presence of a donor DNA molecule results in the integration of the donor DNA. (B) Two different Cas9 nickases (Cas9n), the Cas9 D10A and Cas9 H840A, have mutations in either the RuvC or HNH domains, respectively, and induce a single-stranded break in the target or non-target DNA strand, respectively. (C) The two types of single-base editors utilize a catalytically dead or nickase Cas9 fused with a deaminase enzyme. The cytidine deaminase-based editor can convert the C:G base pair into a T:A base pair, while the adenine deaminase-based editor can convert a T:A base pair into a C:G base pair. The cytidine deaminase-based editor has a uracil glycosylase inhibitor (UGI) directly connected to the C terminus of Cas9, which prevents the repair of U:G mismatches back into C:G base pairs. (D) The prime editor is the latest engineered Cas9 variant that is fused with reverse transcriptase (RT). The prime editing guide RNA (pegRNA) has two extended sequences, one to bind the nicked DNA strand (primer binding site) and another to serve as a template for reverse transcription. The RT converts a portion of the pegRNA into the DNA template that will then be integrated into the target DNA.

Regardless of whether CRISPR/Cas editing is used in somatic cells or in germ cells and embryos, a major concern is the potential for off-target editing events that result from additional sequences in the genome that have a high degree of homology to the 20 bp protospacer sequence (3, 26). Fu et al. demonstrated that sequence mismatches at the 3’ side of the protospacer sequence and regions of homology with less than 5 mismatches to that of the sgRNA target sequence generate a significantly higher frequency of off-target editing (33). DSBs caused by Cas9 also result in increased chromosomal instability and abnormalities that include large deletions, translocations, and rearrangements (34, 35). In an attempt to limit such unwanted secondary outcomes, Cas9 was modified in a manner that affects its nuclease activity. The first modification of Cas9 resulted in its conversion to a nickase (Cas9n). Cas9 has two nuclease domains, with each being inactivated by introducing D10A or H840A amino acid substitutions, respectively (36) (Fig. 1B). Using Cas9n and two sgRNA that bind to each DNA strand, two single-strand breaks are introduced at adjacent target sites, which in turn leads to the formation of indels in the target gene through NHEJ and decreased off-target editing. Another notable change involved rendering Cas9 inactive or “dead” (dCAS9). The generation of dCas9 resulted in the development of several valuable downstream uses including coupling dCas9 to specific accessory proteins that regulate transcription and epigenetic marks, for example, in a targeted manner (37).

An additional advance in Cas9-based editing technology, termed DNA base editing (DBE), was reported in 2016. DBE utilizes a Cas9n protein that is linked to apolipoprotein B mRNA editing enzyme catalytic subunit 1 (APOBEC1) and uracyl glycosylase inhibitor (UGI). APOBEC1 activity leads to the deamination of cytidine to form uridine and UGI prevents the enzymatic removal of uracil from DNA. This approach effectively changes a C into a T within a 5 base pair window of the target sequence (38). Another base editing system is capable of converting A-T to G-C pairs through an engineered tRNA adenosine deaminase (39) (Fig. 1C). Both DBE approaches were reported to possess reduced off-target editing relative to Cas9 (40). In addition to reduced off-target events, DBE can be more precise and is less likely to cause hypomorphic mutations that arise from a partial loss of gene function through triplet indel mutations (41-43).

A newer gene-editing method also involving Cas9n nickase, termed prime editing (PE), is able to introduce specific insertions of at least 44 nucleotides or deletions of at least 80 nucleotides (44). Prime editors include the Cas9n protein fused to an engineered reverse transcriptase. A unique gRNA (pegRNA) includes a 30 nucleotide extension possessing the edit of interest that anneals to the nicked target DNA strand and serves as a primer-template for the reverse transcriptase. The newly formed DNA possessing the engineered edit then serves as a repair template and is incorporated into the genome (Fig. 1D). PE showed similar efficiency of Cas9 mediated HDR and over a 4-fold lower off-editing target frequency, but poses some limitations as well. For instance, the length of the template sequence, target GC content, and secondary structure of pegRNA greatly impact PE efficiency, thereby requiring careful consideration of target sites and optimization of pegRNA.

Since the discovery that the CRISPR/Cas system could be employed as a flexible and relatively simple tool to modify genes tied to a wide range of diseases, numerous studies designed to test its therapeutic possibilities have been completed. It was quickly realized that there were opportunities to target individual genes in somatic cells comprising different organ systems or compartments as a means to treat disease. Editing or modifying the genome of gametes or embryos as a means to correct disease-causing mutations in all subsequent generations also became possible. Both somatic cell and germline editing have been established as feasible, but each brings its own advantages and challenges. The application of germline gene editing to humans, resulting in the birth of gene-edited babies in China, shocked the world and led to debates regarding the ethics and acceptability of germline editing(45). Performing somatic cell gene editing avoids such ethical issues and has become a major research objective of the U.S. National Institutes of Health, leading to the creation of the Somatic Cell Genome Editing (SCGE) consortium. The consortium’s goal is to accelerate the development and application of somatic cell editing for clinical use, and NIH has allocated approximately 190 million US dollars over the first 6 years to support its efforts. The SCGE Consortium is focused solely on somatic editing, and studies on germline editing are not only excluded as a goal but are also considered to be an unacceptable outcome (46).

The Advantages and Challenges of Somatic Cell Gene Editing

The above-described gene editing systems are being deployed for therapeutic use in somatic cells to treat and, in some instances, prevent diseases. Because available gene editing approaches resulting in gene knockout have a long history and are better developed, their use commonly involves inactivating particular genes. An advantage of somatic cell gene editing is realized when cells can be removed, edited ex vivo, and then returned to the patient. Ex vivo editing focusing on the inactivation of target genes has had considerable success in treating hematologic disorders, including β-thalassemia and sickle cell disease (47), and in the generation of more effective chimeric antigen receptors (CAR) T cell products from autologous T cells to fight various forms of cancer (48). Strategies used to treat β-thalassemia and sickle cell disease involve targeting the transcription factor BCL11A that regulates the synthesis of fetal γ-globin in adult erythroid cells. CRISPR/Cas9 is used in CD34+ hematopoietic stem and progenitor cells to inactivate the BCL11A erythroid-specific enhancer, resulting in the generation of stem cells that will produce functional hemoglobin through the replacement of mutant β-hemoglobin with functional γ-globin (49). After myeloablative therapy, patients that received autologous CD34+ cells with a CRISPR/Cas9 edited BCL11A enhancer had high levels of editing in the bone marrow and blood, production of fetal hemoglobin did not require post-therapy transfusions and, in the patient with sickle cell disease, did not exhibit any subsequent vascular pathologies that are typically observed in patients. Approximately 80% of the BCL11A alleles were modified, with no evidence of off-target editing (49).

Clinical trials are also currently underway to utilize CRISPR/Cas editing in CAR T cells, which are patient-derived T cells that expresses a modified T cell receptor fusion protein comprised of an antigen-detecting domain that recognizes a specific tumor antigen and one or more intracellular signaling domain (50). Although cells expressing only the chimeric antigen receptor are effective treatments for specific types of cancers, primarily lymphomas, CRISPR/Cas mediated ex vivo knockout of additional specific genes are being tested for enhanced tumoricidal activities in these cells (48). CRISPR/Cas targets under investigation in CAR T cells include checkpoint inhibitors (PD-1) that suppress T cell activation (51-53), receptors (TGFBR) that inhibit T cell activation (54), or factors (GM-CSF) that lead to serious side effects such as graft-versus-host disease (55). Moreover, using CRISPR/Cas to insert the chimeric antigen receptor into the endogenous T cell receptor locus has been studied as a means to generate CAR T cells with reduced T cell receptor-induced autoimmunity and alloreactivity (56).

Somatic cell gene editing is also proving to be an effective therapy for treating diseases in vivo where gene editing components can be delivered to a site that is easily accessible and compartmentalized. For example, clinical trials are testing the direct delivery of editing reagents into the eye as a means to restore vision. In March 2020, the first patient received an in vivo CRISPR/Cas nuclease therapy to treat an inherited retinal disease caused by a point mutation in the centrosomal protein of 290 kDa (CEP290) gene. Mutations in CEP290 result in variable neurologic, renal, and retinal manifestations (57), and the specific target of this therapy is an intronic point mutation in CEP290 that leads to abnormal RNA splicing. Two sgRNAs were designed to remove the region containing the mutation and incorporated into a single adeno-associated virus (AAV) vector that also expresses Staphylococcus aureus Cas9 (58). Subretinal delivery of the AAV vector in humanized CEP290 mice and cynomolgus monkeys resulted in substantial CEP290 gene editing. The promising results led to a subsequent phase 1/2 clinical trial (NCT#03872479) that reported no major adverse events and visual improvements. Studies using a mouse model of Leber congenital amaurosis 2 (LCA2), which harbors a single nonsense mutation in exon 3 of the retinal pigment epithelium-specific 65 kDa protein (Rpe65) gene, demonstrated restoration of visual function using a DBE approach (59). The same group subsequently demonstrated that lentivirus delivery of an adenine base editor subretinally corrected 40% of Rpe65 transcripts, restored cone-mediated visual function, and prolonged the survival of cones (60). Using the mouse model of LCA2 that was used in previous base editing studies, a study by Jang et al. was the first to demonstrate that a dual-AAV delivery of a PE and pegRNA led to the correction of ~28% of mutant alleles of transduced retinal pigment epithelium cells (61). Importantly, no unintended edits were observed near the target site. A more recent in vivo somatic cell editing clinical trial reporting success includes infusion editing reagents, which then localize the liver where they act in concert to reduce the expression of kallikrein as a means to treat hereditary angioedema (62).

Although the aforementioned examples of somatic cell gene editing in both animal models and in initial clinical trials show tremendous promise, there are challenges that remain. Regarding ex vivo editing, it carries substantial costs and logistical challenges based on its need for significant cell culture space and expertise. In vivo somatic cell gene editing must meet a certain level of efficiency, which is likely to be unique to each disease and cell type, as well as limited toxicity. Additionally, because the effector nucleases needed for NHEJ, HDR, DBE, or PE all possess some form of the bacterial Cas protein, there is concern that an immune response will be generated when these components are administered outside of an immune-privileged site such as the eye. Local inflammatory or innate immune activation must be minimized to avoid acute side effects, whereas adaptive immune response to Cas proteins must be curtailed in cases where multiple treatments of the editor are required to avoid reducing gene editing efficacy. A study by Charlesworth and colleagues, reported 78% and 58% of human subjects possess antibodies to Staphylococcus aureus and Streptococcus pyogenes Cas9 (SauCas9 and SpCas9), respectively (63), which indicates a rapid immune response would be initiated following Cas9 delivery to a large proportion of the population. In mice previously immunized against Cas9, editing still occurred in the liver but was accompanied by increased numbers of cytotoxic T cells and the complete loss of genome-edited cells (64). The potential problem of an anti-Cas immune response is being addressed by eliminating immunodominant epitopes through targeted mutation while preserving its function and specificity (65, 66).

An additional challenge facing the implementation of somatic cell gene editing is developing systems that will effectively deliver gene editing components to cells and tissues of interest. Fortunately, the years of advances and innovation in delivery platforms for gene therapies have provided a solid foundation for developing somatic cell gene editing delivery systems. Nucleic acid and protein delivery systems suitable for use in gene therapy and gene editing have been extensively detailed elsewhere (46, 67) and are beyond the scope of this review. Gene editing delivery platforms are primarily comprised of viral vectors and lipid-based nanoparticles (LNPs).

Numerous viral vectors have been developed and tested (68), with the most common being lentivirus, adenovirus, and AAV. AAV has become arguably the most commonly used viral gene editing delivery system because it is the best studied in terms of safety and efficacy, does not generally integrate into the host genome, which avoids potential insertional mutagenesis, and it has lower immunogenicity relative to adenovirus (67). Multiple AAV serotypes exist, each with different tissue tropism that allows for the transduction of a wide range of tissue types. A major challenge in using AAVs for the delivery of gene editing components includes the limited size of the cargo it can carry (~4.7 kb). For example, SpCas9 and an associated gRNA are too large to fit within a single AAV vector. Current strategies to overcome this limitation include the use of dual AAV vectors that effectively carry the nuclease into transduced cells in two parts, which subsequently results in its reconstitution (67). The recent discovery of smaller Cas proteins provides another potential means to circumvent the cargo limitations imposed by AAV vectors. SauCas9 and one or two gRNAs will fit within a single AAV, as was noted above in the use of an AAV-SauCas9 editing system to correct the CEP290 mutation in the eye. Other examples of compact Cas9 genes that theoretically can be packaged along with gRNA into a single AAV vector include those from Neisseria meningitidis (Nme2Cas9, 3.24 kb), Campylobacter jejuni (CjCas9, 2.95 kb), and Staphylococcus auricularis (SauriCas9, 3.18 kb) (69-72). An adenine base editor was created using Nme2Cas9 that, along with a gRNA, can be packaged into a single AAV. It was demonstrated that in vivo delivery of this single AAV vector resulted in the efficient editing of a mutant mouse fumarylacetoacetate hydrolase (Fah) gene that gives rise to a mouse model of tyrosinemia (73). Although the use of dual AAV vectors and compact Cas9 editors have overcome challenges presented by the limiting cargo capacity of AAV, their sustained expression leads to antiviral immune responses that also limit the effectiveness of a viral delivery system. According to a recent study, a dual AAV system (one AAV encoding Cas9 and the other AAV encoding two guide RNAs) typically integrated within the viral inverted terminal repeats (ITRs); however, insertions within the viral genome were also detected (74). The study also found a higher frequency of AAV integration into the CRISPR-induced DSB than the desired target site deletions. Moreover, integrations were detected at sgRNA off-target loci. Although AAV has a good safety record in over 100 clinical trials, the potential impact of genome editing constructs on AAV integration and genotoxicity must be considered when developing vectors for therapeutic purposes. To overcome this issue, engineered viral-like particles (eVLPs) were developed to allow for in vivo delivery of genome editing components. Because the eVLP payload is comprised of ribonucleoprotein complexes, it has a short half-life that the potential for DNA integration and persistent expression is needed for initiating an immune response. Subretinal injection of eVLPs possessing a BE into LCA2-model mice resulted in editing efficiencies comparable to what was observed with lentiviral delivery, but with lower off-target editing (75).

Based on decades of research using LNPs to deliver DNA and RNA to target cells, they were quickly adapted for use in delivering gene editing reagents to somatic cells (46, 67). LNPs are generally composed of four kinds of lipids that include an ionizable lipid (usually a cationic lipid), polyethylene glycol lipid, helper phospholipid, and cholesterol (76). The pH-dependent cationic lipid is neutral prior to entering target cells but becomes cationic within the acidic pH of the endosome, which in turn leads to the dissociation of the particles and release of the cargo into the cytoplasm. Because the elements that comprise LNPs are synthetic, they can be modified to alter the pharmacokinetic profiles and ability to target different cell types. LNPs were critical for the delivery system of SARS-CoV-2 mRNA vaccines, with millions of doses being administered with a good safety profile (77). Thus, LNPs are safe, and large-scale production is feasible. Systemically administered nanoparticles accumulate primarily in the liver due to their association with the low-density lipoprotein receptor (LDLR) on hepatocytes (78). Several animal studies using LNPs demonstrated that injection of LNPs with gene editing components resulted in significant target gene editing in the liver (reviewed in (67)). LNP delivery of mRNA for gene editing was recently tested in patients with hereditary transthyretin amyloidosis, wherein the goal was to knockdown liver transthyretin (TTR) protein levels (79). After LNP formulation was administered intravenously, reductions in serum TTR levels of up to 87% at the highest dose tested were observed. Thus, LNP delivery holds considerable promise for somatic cell gene editing and avoids the issues associated with viral delivery systems. Although progress has been made in developing tissue or cell-selective LNPs, including for gene editing use (67), advances in this area are still needed to allow for the therapeutic use of somatic cell gene editing in a wider range of diseases.

The Advantages and Challenges of Germline Gene Editing

Shortly after the discovery that CRISPR/Cas9 could be used to target nearly any gene of interest, it was adapted to introduce mutations that would be transmitted through the germline in many different animal species, from rodents to large animals, primarily as a means to create models of human diseases (80–85). In 2014, Niu reported the first use of CRISPR/Cas in monkey zygotes to target three different genes, by co-injecting gRNAs for nuclear receptor subfamily 0 Group B Member 1 (Nr0b1), peroxisome proliferator-activated receptor gamma (PPARγ), and recombination activating gene 1 (86) (RAG1). Analysis of the resultant cynomolgus macaque embryos revealed editing in one or a combination of two of the three target genes. Subsequent transfer of Cas9-gRNA injected embryos into surrogate dams resulted in 10 pregnancies (34% pregnancy rate). This initial report of efficient CRISPR editing in a primate model was followed by several additional examples of germline editing in macaque and marmosets (87–91). Additional evidence that the power of gene editing technologies, including CRISPR and TALEN systems, could be used to potentially correct disease-causing mutations in humans came from studies in mice and nonhuman primates where HDR was used to incorporate specific DNA sequences into genomic targets (28, 92, 93). Collectively, these initial studies demonstrating germline modification of individual genes led support to undertaking studies to determine safety and efficacy of germline gene editing in human embryos.

In practice, germline gene editing to correct disease mutations would have advantages over somatic cell gene editing, with the most obvious being that it would prevent the disease from developing in the immediate offspring and eradicate it from all subsequent generations. Moreover, with standard in vitro fertilization (IVF)/Assisted reproductive technology (ART) procedures, the simplicity and flexibility of the CRISPR system would allow for it to be easily adapted to the specific mutations of parents that are carriers of a wide range of diseases. Gene editing in humans was first reported in 2015 in human zygotes with three pronuclei (3PN), which can occur following in vitro fertilization due to polyspermy (94). The use of 3PN zygotes allows for testing of gene editing outcomes and reduces the ethical concerns associated with testing in human embryos because they have limited developmental potential. Zygotes were injected with Cas9 and gRNA targeting the β-globin gene (HBB) gene along with a ssDNA template to assess HDR. Analysis of outcomes revealed 52% of the embryos possessed some level of editing, and of those, only 14% had undergone HDR. Moreover, 25% of the embryos possessed the sequence of the closely related delta subunit gene (HBD), indicating that this homologous gene served as the HDR template. Additional studies in 3PN embryos confirmed the ability of HDR to insert a specific sequence into a target gene (94) and that a base editing (BE) was able to effectively introduce single-nucleotide substitutions (95). In both of the studies, off-target editing in homologous regions in the genome was absent (94) or only observed in a single site (95). Subsequent studies in “normal” 2PN human embryos were conducted whereby injection of Cas9 protein complexed with the appropriate sgRNAs and ssDNA templates resulted in HDR- mediated correction of point mutations in HBB and glucose-6-phosphate dehydrogenase (G6PD) genes (96). Although the sample size was low, the HDR efficiency was 50%, but also yielded embryos that were mosaic, with half of the cells containing the HDR corrected gene and the other half possessing a 4 bp deletion.

A subsequent study by Ma et al. reported increased editing efficiency in human embryos that were heterozygous for a 4 bp mutation in the cardiac myosin-binding protein C (MYBPC3) gene that is responsible for familial hypertrophic cardiomyopathy (97). A male carrier of the MYBPC3 mutation provided sperm, and healthy oocyte donors were used for IVF to generate wild-type and heterozygous mutant MYBPC3 zygotes. Two approaches were employed for correcting the MYBPC3 mutation. One approach included injecting gRNA, Cas9 protein, and a ssDNA repair template into zygotes 18 hours post-fertilization (i.e., S-phase injected embryos), with the resultant individual blastomeres of 4 to 8 cell stage embryos being isolated for genetic analysis. The outcomes from the injection of gene editing components into 2PN zygotes were as expected, based on preceding nonhuman primate (NHP) and human studies, and included unedited mutant, NHEJ-modified, and HDR-repaired blastomeres. Within the mosaic embryos, it was observed that a majority (35/67; 52.2%) of individual blastomeres were homozygous for the wild-type MYBPC3 allele. Because the embryos originated from heterozygous wild-type/mutant MYBPC3 zygotes, the authors speculated that the individual blastomeres repaired the 4 bp deletion using the maternal MYBPC3 wild-type allele as an HDR template instead of the ssDNA template (97). Moreover, the mosaic embryos contained a mixture of multiple indels and mutant MYBPC3 that were hypothesized to arise from the completion of DNA replication and the generation of two mutant alleles, or CRISPR-Cas9 remained active, continuing to target MYBPC3 after the first cell division. To permit MYBPC3 editing prior to DNA replication and cell division, when only one copy of the mutant allele is likely available, CRISPR-Cas9-ssDNA was co-injected with sperm into the M-phase oocyte during fertilization by intracytoplasmic sperm injection (ICSI). Individual blastomeres of embryos were isolated and analyzed for MYBPC3 editing, where it was observed that 72.4% were wild-type. This was a significant increase the incidence of wild-type MYBPC3 blastomeres in M-phase injected oocytes relative to S-phase injected zygotes (52.2%) and control IVF embryos (47.4 %). Collectively, the results of the study suggest that the injection of M-phase oocytes with gene editing systems has the potential to correct heritable mutations in human embryos (97)

The paper by Ma et al. provided valuable insight into the importance of the timing of the delivery of gene editing reagents for germline editing and that HDR correction of a disease-causing mutation through the presence of a wild-type allele was possible. However, unedited or uncorrected alleles may possibly maintain the disease, and random indel mutations on disease-causing alleles, as well as chromosomal rearrangements, may lead to other unexpected consequences (98). Finally, the results of the potential off-target events are also important issues to consider. Using a mouse model, the observations by Ma were extended, wherein it was demonstrated that Cas9-mediated conversion of heterozygous alleles into homozygous alleles could occur without exogenous templates (99). This process, termed interhomolog repair (IHR), was promoted by the action of the strand exchange protein RAD51 recombinase (RAD51). An alternative explanation was proposed; however, that would also account for the increase in the incidence of wild-type MYBPC3 embryos, an outcome that was demonstrated to be a significant safety issue that needs to be addressed before germline gene editing can be used therapeutically. Questions were raised with regard to the assessment of gene editing outcomes, whereby it was proposed that larger chromosomal rearrangements and deletions would not be captured by the assay that was used to determine the sequence of the target gene. Based on the PCR assay used for genotyping, it is possible that larger-scale chromosomal rearrangements would remove a PCR primer binding site, leading to amplification of only the maternal allele and an overestimation of the IHR rate (100).

A study was conducted to directly determine CRISPR/Cas editing outcomes with regard to detrimental on-target chromosomal rearrangements in human embryos. A paternal 1 bp mutation in the eye shut homolog (EYS) gene, resulting in blindness, was targeted for Cas9 editing in human zygotes or 2 cell embryos (101), wherein the double-stranded DNA break would be repaired using the maternal wild-type EYS gene as a template if IHR occurred at a significant level. Flanking maternal and paternal single-nucleotide polymorphisms (SNPs) were assessed in the genotyping of the embryos, allowing a conclusive assessment of whether the maternal EYS gene sequence was incorporated into the mutant paternal allele. It was demonstrated that the most common repair outcome was due to NHEJ and microhomology-mediated end joining (MMEJ) at the Cas9 EYS target site, which resulted in the detection of the flanking SNPs of the paternal chromosomes. MMEJ repair occurs as a result of serendipitous microhomologies that are present near the Cas9 cleavage site. Of the embryos with a repaired EYS gene, only 7% were corrected due to gene conversion, with the wild-type maternal allele serving as the template for correction. Strikingly, in 50% of all embryos analyzed, there was no gene repair but instead possessed substantial losses of chromosome 16, the chromosome where the human EYS gene is located (101, 102). Loss of the individual arms of chromosome 16 as well as loss of the entire chromosome (monosomy) was observed. These findings were confirmed by an independent study utilizing Cas9-mediated editing to study the function of the pluripotency gene POU5F1 (i.e., OCT4) on embryo development (103). It was observed that the genotype of individual blastomeres for a number of the embryos could not be determined because the PCR method used to amplify the on-target genomic fragment had failed. It was also noted that 42% of the samples possessed a homozygous wild-type genotype. These findings could have been due to PCR amplification of short genomic fragments and the inability to discern larger chromosomal abnormalities induced by Cas9 editing. In a subsequent followup study designed to specifically to determine the incidence of chromosomal abnormalities due to CRISPR/Cas9 editing, additional embryos were generated or samples from the previous study were reanalyzed for more complex genomic rearrangements (103, 104). From the whole genome sequencing of single cell or trophoblast biopsies, it was determined that 32% of samples exhibited evidence of chromosomal abnormalities. Segmental chromosomal losses/gains of ~16% of the samples were adjacent to the POU5F1 locus, spanning 4 kb to at least 20 kb (104). Large chromosomal rearrangements, deletions, or loss of heterozygosity was also observed in Cas9 edited mouse and bovine embryos, induced pluripotent stem cells (iPSCs), primary hematopoietic progenitor/stem cells, and primary human T cells (100, 105–108). Chromosomal rearrangements, including large deletions, insertions, inversions, and translocations, pose significant challenges in detecting gene editing outcomes, as most results are identified through PCR products of limited size (<1kb). Consequently, the absence of detection of editing may occur if it involves an inversion or large deletion at the target locus.

Although proof-of-principle studies have demonstrated that germline gene editing is feasible, recent data indicates that the issue of large chromosomal rearrangements are a formidable hurdle that needs to be overcome before acceptable for clinical use. Additional studies, especially in clinically relevant NHP or human embryos, are needed to determine if newly developed editing approaches (Cas9 nickase, BE, PE) that do not rely on dsDNA cuts lead to similar large scale chromosomal rearrangements. PE in mouse embryos has been reported to lead to significant levels of large deletions but remains to be assessed in primate embryos (61). Moreover, editing kinetics and efficiency must also be optimized to avoid mosaicism. Currently, the level of mosaicism in germline editing outcomes is considerable (94, 96, 109), which would be difficult to detect or to determine the level of complexity with current preimplantation genetic testing methods. Although there are reports of minimal off-target editing events in offspring derived following gene editing, additional studies using newly developed methods to interrogate off-target editing (110, 111) or in-depth whole genome sequencing of edited embryos or offspring are needed to conclusively demonstrate that the available editing platforms will not also introduce potentially harmful mutations into the germline (112). Germline editing safety and efficacy have to meet high standards because current preimplantation genetic testing methods can be used to exclude the transfer of embryos that carry disease-causing mutations. Moreover, once the risk of off-target editing and mosaicism are addressed, the ethical aspects of what constitutes an acceptable target gene will need to be defined.

Conclusions

Studies conducted after the development of gene editing systems demonstrated the feasibility of performing both targeted somatic cell and germline gene editing. It has also become clear that each approach faces some shared (i.e., avoiding off-target editing), but mostly unique challenges that they must overcome prior to being used broadly for treating diseases. The pace at which somatic cell gene editing has advanced to clinical trials is remarkable, with apparent success. Further development of somatic cell gene editing therapies requires overcoming the practical challenge of developing adaptable delivery systems to target a wider range of cell and tissue types, editing systems that have increased on-target efficiencies with little to no off-target activity, and a means to avoid immune-mediated neutralization of the delivery or gene editing systems. Germline gene editing faces additional hurdles that include the ethics and acceptability of manipulating embryos and permanently altering the genome of future generations. From a safety perspective, understanding the cause and how to prevent chromosomal instability and off-target editing is paramount before gamete or germline editing could even be considered a viable therapeutic option.