CRISPR-Cas9 base editors and their future role in human therapeutics

Abstract

Consistent progress has been made to create more efficient and useful CRISPR-Cas9-based molecular tools for genomic modification. CRISPR-Cas9 base editors are a useful system due to their high efficiency and broad applicability to gene correction and disruption. In addition, base editing has been suggested as a safer approach than other CRISPR-Cas9-based systems as they limit double stranded breaks during multiplex gene knockout and do not require a toxic DNA donor molecule for genetic correction. As such, numerous industry and academic groups are currently developing base editing strategies with clinical applications in cancer immunotherapy and gene therapy which this review will discuss, with a focus on current and future applications of in vivo base editor delivery.

Introduction

Genetic mutations that are causal of genetic disorders or predispose individuals to an enhanced cancer risks are abundant. Nearly half of these pathogenic mutations are single nucleotide variants (SNVs) [1], leading to a focus on understanding the occurrence and phenotypic consequences of particular SNVs. Parallel to this area of investigation, it is of high interest to correct or revert these SNVs to eliminate the problematic genotype completely [2-5]. Historically, all therapeutically relevant targeted nuclease strategies induce double stranded breaks (DSBs) in genomic DNA to achieve gene editing. Transcription activator-like effector nucleases (TALENs) and zinc finger nucleases (ZFNs) use a targeted DNA binding domain fused to an endonuclease domain of fok1 [6,7]. Meganucleases bind and create DSBs in palindromic, semi-palindromic (I-CreI) or non-palindromic (I-SceI) DNA sequences [8]. In contrast, the CRISPR-Cas9 platform uses a unique targeted single guide RNA (sgRNAs) which complexes with the constant endonuclease Cas9 to induce targeted DSBs [9]. These targeted nucleases can all produce off-target effects, unwanted DSBs at nucleotide sequences in the genome that are highly similar to the desired DNA target site. Further, simultaneous DSBs within one genome, such as one at the on-target site and one or more at off-target sites, can result in chromosomal translocation events between these sites, which can severely alter gene regulation and function [10]. DSBs have also been reported to induce cell cycle arrest [11], p53 activation [12], other complex rearrangements of chromosomes including large deletions of DNA [12-16], alteration of expression in genes neighboring the DSB [17,18], chromothripsis [19], and cancer induction [20]. Next generation Cas9 derived ‘digital’ genome editing tools, i.e. Cas9 base editors (BEs) [21,22], have the ability to perform single base conversion to correct a subset of SNVs, including C>T (G>A opposite strand) and A>G (T>C opposite strand). These tools provide a unique ability to correct pathogenic mutations and modify expression of genes (reactivate/suppress) without the need for creating a toxic double strand break (DSB) or a DNA donor molecule for genetic correction, which greatly decreases the occurrence of unintended effects in the genome and enhances efficiency and efficacy of repair [23,24].

This review focuses on advances in BE technology, current applications to human therapeutics, and the current and future landscape of in vivo delivery in clinical trials and beyond. A description of the mechanistic action of BEs will be followed by a summary of broadening the editing toolbox by reducing the stringency of the Cas9 protospacer adjacent motif (PAM). Recent base editing strategies in the field of cancer immunotherapy, as well as in genetic disorders that can be combated by editing single nucleotide polymorphisms (SNPs) will then be described. Therapeutic preclinical delivery of BE, both ex vivo and in vivo, will be highlighted with an emphasis on specific delivery methods; direct nucleic acid/protein delivery, lipid/polymer based delivery, and viral delivery. Finally, current and upcoming clinical trials involving in vivo delivery of BE will be outlined with emphasis and insight into the future applications and landscape of BE technology.

Base Editors

BEs are a fusion between a Cas9 nickase (nCas9) and a deaminase enzyme capable of performing base modifications through deamination, resulting in transition and transversion mutations. Similar to the CRISPR-Cas9 system, after nCas9 binds to its target DNA sequence, defined by a 20 base pair sgRNA, an ‘R-loop’ is generated (Figure 1A) [25]. This R-loop results from localized denaturation upon Cas9 binding, exposing a short stretch of single stranded DNA (ssDNA) which provides adequate space for the deaminase to act on their target base(s). Two main classes of BEs have emerged with high efficiencies and product purities, cytosine base editors (CBEs) and adenosine base editors (ABEs) which convert a C·G base pairs to a T·A base pairs (G·C to A·T opposite strand) and A·T base pairs to a G·C base pairs (T·A to C·G opposite strand), respectively [21,22].

Figure 1. Cytosine Base Editing.

A nCas9, which nicks the opposite target stand, induces a DNA repair pathway response. The target C in the editing window is deaminated by a fused APOBEC1 protein to a U. A fused UGI prevents the cell from resolving this mismatch through the BER pathway. During DNA replication or repair, the U matches with an A resulting in a C/G to T/A substitution.

Cytosine Base Editors (CBEs)

CBEs were the first class of BEs to be generated which initially convert target cytosine bases to uracil, followed by the edited strand being repaired to a thymine through DNA replication or repair (Figure 1B) [21]. The first CBE, BE1, was a catalytically dead version of Streptococcus pyogenes Cas9 (dCas9) linked to a ssDNA targeting cytidine deaminase enzyme from Rattus norvegicus (rAPOBEC1). BE1 induced low-level efficiency of cytosine to uracil conversion in HEK293T cells, an outcome that was hypothesized to be due to high rates of intracellular uracil excision by the endogenous base excision repair (BER) pathway [21]. Three improvements were then made to BE1 to generate the higher efficiency BE4. First, two phage-derived uracil glycosylase inhibitors (UGIs) were linked to the C terminus of BE1 by a 9 amino acid linker that functioned to inhibit uracil excision by uracil DNA glycosylase (UDG). The rAPOBEC1-dCas9 fusion was also optimized to 32 amino acids, and the dCas9 was converted to a D10A nCas9 to direct host repair to assume the nicked strand is the newly synthesized DNA strand, cloaking the strand deaminated by rAPOBEC1 (Figure 1A, B) [21,26]. Additional optimizations such as limiting the use of rare codons and enhancing/optimizing online design tools and changes to the NLS [27] led to BE4max, which has highly efficient (>95%) cytosine deaminase editing at positions 4-8 in the 5’ region of the protospacer (editing window) in human primary cells [28,29].

Adenosine Base Editors (ABEs)

ABEs were developed as a second class of BEs when it was recognized that adenosine deamination could generate an inosine intermediate [22]. The Liu Lab utilized directed evolution of the known bacterial dsRNA adenosine deaminase enzyme ecTadA to perform deamination of ssDNA. This evolution was achieved through incorporation of 14 amino acid changes in TadA, which was then fused to an additional wtTadA and nCas9 D10A to generate ABE7.10. Due to the complete absence of inosine in the genome, no known endogenous mechanisms of inosine removal exists, and thus a glycosylase inhibitor was not necessary (Figure 2) [22]. During DNA replication and cell division this inosine will be read as a guanine and lead to permanent incorporation of a guanine. Some limitations of ABE7.10 are its efficiency of deamination and its generation of off-target edits within the editing window. Subsequently, phage-assisted evolution of the wtTaDA domain was performed to overcome the shortcomings of ABE7.10. This identified 8 additional amino acid changes, resulting in a more processive and 590-fold more active form of ABE, named ABE8e [30]. ABE8e allows for efficient target adenosine base conversion to guanine at positions 4-8 of the protospacer target site. Resultant of this, re-screening of the public archive of SNPs on ClinVar showed that this version of BE could correct many of the known pathogenic transition mutations, which account for ~60% of all known pathogenic SNVs [31].

Figure 2. Adenosine Base Editing.

A nCas9, which nicks the opposite target stand, induces a DNA repair pathway response. The target A in the editing window is deaminated by a fused TadA protein to an I. During DNA replication or repair, the I is read as a G resulting in an A/T to G/C substitution.

Expanding BE targeting range through removal of PAM restrictions

While BEs are exciting for the future treatment of disease, both CRISPR-Cas9 and BEs require a specific PAM sequence adjacent to the 20 bp protospacer target sequence. The PAM sequence required for traditional Streptococcus pyogenes Cas9 (SpCas9) is NGG (N = standard IUPAC nucleotide code), which limits the number of SNVs that are targetable using BE or traditional CRISPR-Cas9 methods. Due to this restriction, sgRNAs must place the target base within the editing window located 12-16 bp away from the PAM. In order to expand the utility of these enzymes and increase the amount of targetable SNVs, BEs can have the nCas9 domain altered to permit different PAM requirements. Many Cas9 PAM variants exist, including VQR-Cas9 (NGAN), EQR-Cas9 (NGAG), VRQR-Cas9 (NGCG), Cas9-NG, [32,33] and the“near-pamless” (SpRY) Cas9 variant which recognizes NRN (R = adenine or guanine), and to a lesser extent NYN (Y = cytosine or thymine) PAMs [32]. This increased flexibility combined with the development of software that predicts base editing outcomes, such as BE-HIVE and Honey Comb, have and will continue to be critical to this expansion of base-targeting with BE [34,35].

Clinical Applications

Cancer Immunotherapy

One of the exciting applications of BEs for clinical therapy is in combination with ex vivo manufactured allogeneic chimeric antigen receptor (CAR) or T cell receptor (TCR) engineered lymphocytes. Autologous CAR-T cells (CAR-Ts) have been used with great success clinically to mount an immune response against leukemia and lymphoma [36,37]. Allogeneic or ‘off-the-shelf CAR-T’ cell therapies are of significant interest due to the cost-effective mass manufacturing and viability of lymphocytes from healthy donors. In theory, allogeneic T cells could be isolated from healthy donors, genetically engineered to express a CAR, and then infused into unrelated cancer patients as a CAR-T therapy. Unfortunately, it has been well documented that without manipulation, a patient's immune cells reject allogeneic T cell products, or the allogeneic T cells can attack the patient’s healthy cells, a phenomenon known as graft-versus-host-disease (GvHD) [38].

Several methods have been employed in an attempt to circumvent the most harmful effects of allogeneic CAR-T cell therapy, i.e. GvHD. For instance, Cas9 nucleases have been deployed for gene knockout (KO) to generate allo compatible CAR-T by inactivating the endogenous TCR and/or removing beta-2-microglobulin (B2M), a structural component of the major histocompatibility complex (MHC), which has shown some effectiveness in evading GvHD in preclinical models [39]. These multiplex gene KO approaches using Cas9 nuclease can be highly efficient (>90% KO) but have the potential to induce translocations and other unwanted genomic insults. In light of this concern, researchers have begun to use BE to achieve robust multiplex gene KO using premature stop codon induction [40,41] and splice site disruption [34]. Beyond engineering for allogenic compatibility, multiplex KO of checkpoint regulators, such as CTLA4, PDCD1, SOCS3 and CISH, have been used in T cells to enhance in vivo tumor clearance [28,42]. This multiplexed gene KO approach has demonstrated enhanced tumor clearance in comparison to CAR integration alone in multiple murine models of cancer [28,34]. The utilization of BEs will most likely represent the future for multiplexed gene KO as they have demonstrated greatly reduced to undetectable levels of DSBs and translocations, and thus may provide a way to tailor specific gene edit combinations to target individual cancers as the next generation of precision medicine [12,16,28,43,44].

Correction of SNVs in Genetic Disease

Many rare autosomal diseases are caused by SNVs which lead to nonfunctional or dysfunctional protein expression. Many SNVs that could be targeted with BE have been defined for diseases such as severe combined immunodeficiency (SCID), Crigler-Najjar syndrome, sickle cell disease (SCD), β-Thalassemia, cystic fibrosis, Fanconi anemia (FA), and many others [45,46]. In the case of hematological diseases like SCD (the most common inherited genetic disorder worldwide) or beta-thalassemia, hematopoietic stem and progenitor cells (HSPCs) can be isolated from patients, base edited ex vivo, and infused back into the individual post preconditioning [47]. Targeting HSPCs can also theoretically be done in vivo by first mobilizing the patient’s HSPCs into the bloodstream, and then delivering BE reagents intravenously after which time the HSPCs would re-engraft into the bone marrow (BM) compartment [47,48].

As an example of BE phenotypic alleviation in SCD, the most common SCD SNV (A>T) in the beta-hemoglobin gene (HBB) causes a glutamine (GAG) to valine (GTG) substitution. This single amino acid change causes polymerization of hemoglobin proteins which in turn cause red blood cells (RBCs), or erythrocytes, to form their pathogenic sickled shape. These mutant RBCs hemolyze frequently and block arteries leading to pain crisis, organ failure, and vaso-occlusion related issues [49]. Because BE technology cannot convert T>A directly, several strategies have been employed to alleviate symptoms without directly correcting the pathogenic SNV back to wild-type (WT). One strategy uses ABE to convert the pathogenic valine (GTG) to alanine (GCG), which has been shown to be a benign mutation in individuals from the Makassar region of Indonesia [49]. Given that this conversion could not be achieved by the current base editors, a PACE-evolved ABE (ABE-NRCH) was employed to deaminate the opposite strand A base to a G base generating a C base through DNA replication resulting in an alanine codon. This method produced high-levels of targeted base editing in human SCD HSPC CD34+ cells and resulted in decreased levels of sickled RBCs [48]. Corrected human HSPCs CD34+ cells were then transplanted into irradiated mice where edits persisted after 16 weeks and healthy HBB proteins were achieved, demonstrating pre-clinical therapeutic relevance [48].

A separate strategy utilizes BE to alleviate SCD symptoms by reactivation of fetal hemoglobin (HbF), which has a much higher affinity for oxygen than adult hemoglobin (HbA) [50,51]. In adults, circulating hemoglobin is composed of alpha-hemoglobin (HBA) and HBB subunits, whereas HbF is composed of HBA and gamma-hemoglobin (HBG) subunits. Three months postnatally, HbF levels are reduced due to the silencing of HBG transcription by BCL11A, a transcriptional repressor which normally binds to the promoter region of HBG [52]. It was previously shown that creating small insertions or deletions (indels) with CRISPR-Cas9 in the erythroid specific enhancer region of BCL11A halts the transcriptional repression by BCL11A in the erythroid compartment, leading to re-expression of HBG, and, in turn, the assembly of HbF [53].

Further examination of the genetic sequence of this enhancer revealed several “hot-spots” were the most relevant for disruption of subsequent BCL11A enhancement. Fortunately, some of these areas could be targeted with CBE by making single C>T edits. Ex vivo genetic modification of SCD HSPCs CD34+ cells with CBE targeted to these erythroid specific enhancers of BCL11A “hot-spot” regions led to the reactivation of HBG transcription and thus HbF expression [53]. Again, these corrected human HSPCs CD34+ cells were transplanted into irradiated mice where edits persisted after 16 weeks and HbF was highly expressed, out-competing the assembly of HbA with mutated HBB, leading to phenotypic improvement [29].

A second strategy for reactivation of HbF is to target the binding of BCL11A and other transcription silencers on the promoter of HBG directly. In this in vivo mouse study, CBE and ABE constructs were packaged into CD46-targeting helper-dependent adenovirus (HDAd5/35⁺⁺) and targeted single base conversion on the HBG promotor where lymphoma-related factor (LRF) normally binds for transcriptional silencing. These “hot-spot” sites are a form of naturally occurring mutations called hereditary persistence of fetal hemoglobin (HPFH) as these individuals have abnormally high levels of HbF [54]. After injecting mice with virus it was found that moderate levels of HSPC base editing (15-25% in both HBG1 and 2 promoter regions) equated to >40% HBG induction in the peripheral blood erythrocytes [55]. On the commercial side, Beam Therapeutics is pursuing both the Makassar BE strategy (BEAM-102) as well as strategies to increase the expression of HbF (BEAM-101) as first-in-human clinical applications of BE.

Clinical Delivery

Direct Delivery of BE Reagents

Clinical delivery of CRISPR-Cas9 and BE reagents have commonly been through purified protein and chemically modified sgRNAs in the form of ribonucleoprotein (RNP) (NCT03745287), yet this approach has proven more difficult with BEs due to the size and folding properties of the molecule [56]. Thus, many groups are using mRNA encoding BEs in combination with chemically modified sgRNAs because they can be rapidly produced and do not require complex protein manufacturing [28,57,58]. Direct ex vivo delivery of BE reagents is a process that involves isolation of target cells from the patient, delivery of the BE reagents to the cells (typically via electroporation), and re-infusion of modified cells to the patient (Figure 3). Ex vivo delivery of BE to human primary cells has become a common method for groups working with immune cells, stem cells, epithelial cells, and retinal cells [28,48,59]. This approach is attractive for several reasons: 1) a smaller starting population requires less cost and complexity of manufacturing, and 2) after expansion and genetic modification ex vivo, the BE protein or mRNA will be dissipated by the time of infusion and gene editing, resulting in a modified cellular product that limits exposure of innocuous exogenous BE to the host immune system; thus, reducing the concern of immune activation against the foreign protein or mRNA [60].

Figure 3. Ex Vivo or In Vivo base editing.

Schematic representation of a clinical workflow for ex vivo manipulation of human leukocytes using electroporation and the base editor system. In vivo delivery of AAV to the target organ or systemically represented by a split base editor system due to cargo size limits of AAV.

Viral Delivery

Viruses, such as adenovirus, recombinant adeno-associated virus (rAAV), and lentivirus can be used to deliver BE reagents both in vivo and ex vivo. rAAV has emerged as a popular candidate due to its low immunogenicity and largely transient, non-integrating nature [61]. Obtaining viruses with strict tropism can be challenging and efforts to characterize rAAV serotypes to various tissues are ongoing [62]. Another limitation is that viruses have packaging limits (~4.7kb with rAAV), to deliver cargo larger than this limit, multiple viruses need to be produced, increasing the overall cost of the therapy. Successful injection of rAAV containing BEs that have been split into two viral preparations have been performed successfully. These separate rAAVs generate precursor proteins that are then spliced together after transduction using the intein system [59]. While these methods have proven to be effective in mice, these mice live in specific pathogen free colonies and have little to no exposure to rAAV or Cas9. Groups have demonstrated that humans can produce antibodies against spCas9 and various rAAV serotypes, which is a potential concern for translation [63,64]. While viral delivery continues to be an efficient method of gene transfer in mice or as a template for homology directed repair (HDR), much optimization is needed before the in vivo delivery method can become standard of care.

Lipid/Polymer Based Delivery

Lipid-like nanoparticles (LLNPs), virus-like particles (VLPs) polymer-based nanoparticles overcome the limitations of in vivo viral delivery in part by improving in vivo tissue specificity, and toxicity [65,66]. Traditionally, this method relies on a negatively charged RNP being packaged within an exosome, liposome or polymer for systemic infusion [66,67]. These LLNPs, VLPs and polymer-based nanoparticles have the ability to deliver large molecular payloads to tissues in the body, and are traditionally less immunogenic than viruses [68]. As this field has grown, charged liposomes that contain sgRNAs and protein have been used as a vehicle for intracellular delivery [1,56,66,69]. With RNP delivery methods, the BE is able to penetrate and act on the target cell faster and will be cleared from the host rapidly. As optimization of this delivery method continues, the tissue specificity and delivery efficiency will also improve and will likely emerge as the predominant form of delivery for in vivo studies and trials in years to come.

The Future of Base Editor Applications

Current and Upcoming Clinical Trials

Several preclinical studies that utilize in vivo delivery of base editors to combat genetic disorders have been described in recent publications [70], press releases, and conference abstracts. The only in vivo trial thus far that has entered the clinic is the heart-1 trial that aims to treat heterozygous familial hypercholesterolemia (HeFH), uncontrolled hypercholesterolemia, and atherosclerotic cardiovascular disease (ASCVD). The novel gene therapy, VERVE-101, which consists of ABE mRNA and sgRNA targeting the PCSK9 gene encapsulated in an LLNP for complete silencing of protein expression in the liver [70]. As of July 5th 2022, one patient has been infused with an estimated total of 40 patients to be included over the next several years (NCT05398029). The dosing of this first patient represents an important milestone for the future of base editor rooted therapies which are trending in the direction of in vivo delivery.

The remaining upcoming preclinical trials that have yet to move into phase 1b approval for human subject testing range in delivery method from LLNP, rAAV, to a novel subcutaneous injection of BE mRNA [71]. Verve Therapeutics has an additional preclinical trial in the pipeline which also utilizes LLNP containing BE to combat familial hypercholesterolemia, but by the silencing of the ANGPTL3 gene which also regulates circulating cholesterol levels [72]. In addition to the aforementioned ex vivo base editor IND approved or IND enabling studies to treat SCD and beta-thalassemia, Beam Therapeutics is pursuing preclinical models of in vivo base editing. The first approach targets the R83C gene for glycogen storage disease 1a [73], the second targets the G1961E gene for Stargardt disease, and the third targets SERPINA1 mRNA for alpha-1 antitrypsin deficiency [71,74]. These preclinical studies that are being pushed into the clinic are exciting and provide evidence that the field of BE delivery is shifting into in vivo delivery methods.

Broadening the Base Editor Toolbox

As base editors have entered the clinic and their utility both in translation and basic science has been solidified, several groups have been working towards expanding their application both by optimizing and creating new iterations as well as by designing creative strategies for mutation targeting. Already, BEs have been continuously evolved to allow for varying PAM requirements based on the location of a specific disease mutation that were unreachable with the original enzymes. These include in vivo murine delivery of an ABE-VRQR requiring PAM to target the LMNA gene in the case of Hutchinson–Gilford progeria syndrome [75], and ex vivo murine editing with a ABE-NRCH requiring PAM in the aforementioned Makassar targeting strategy for SCD [48]. Besides evolving and creating new iterations of BEs, targeting strategies are being employed that do not correct disease mutations back to WT, but instead target a base within the mutated codon to restore protein translation by removal of a stop codon. As an example of this, we recently targeted the most common FA mutation worldwide, the Spanish founder mutation (FANCA c.295C>T). Here, we used ABE to convert the mutated stop codon into a different amino acid that was conserved among mammals, which restored the reading frame and led to phenotypic recovery in patient primary cells [76]. New enzyme creation for expanded mutational targeting as well as creative codon targeting strategies continue to further the possibilities of disease correction with BE.

Conclusions

Since the first BE was described in 2016, there has been an explosion of research aiming to bring this technology into the clinic for cancer immunotherapy and to address genetic disorders. Much of the attraction of BE technology in the clinical setting relies on circumventing the need to create a toxic DSB, removing the need for a DNA donor molecule, and other concerns such as high off-target editing and insertional mutagenesis. As improvements are concurrently being made in the space of in vivo delivery of nucleic acids and protein, in vivo BE delivery will dominate the field for the treatment of genetic disorders and cancers once a robust delivery system is defined. Parallel to this, newer generations of Cas9 will continue being created that have loosened PAM requirements, reduced off-target effects, and will ultimately expand the list of targetable diseases that may undergo direct corrections.