In vivo genome editing in animals using AAV-CRISPR system: applications to translational research of human disease

Small Cas9 orthologues

Even though the AAV vector is an extremely attractive delivery vehicle for CRISPR, its relatively small viral genome-packaging capacity has limited its use for delivering large transgenes. Using dual-vector AAV system or smaller Cas9 orthologues along with truncated regulatory elements (promoter and polyadenylation signal) has greatly circumvented the transgene packaging issue (Figure 1A). Owing to a small AAV viral genome-packaging capacity (~4.8 kb), it has been technically challenging to co-package Streptococcus pyogenes-derived Cas9 (SpCas9) (4.1 kb) and multiple sgRNAs into all-in-one AAV vectors for multiplex genome editing. Recently, it has been demonstrated that when very small promoters were used (for example, miCMV promoter and H1 promoter to drive the expressions of SpCas9 and its gRNA, respectively), it was possible to package both SpCas9 and its gRNA into a single vector³². However, the number of promoters that can be selected for tissue-specific transduction is limited. To circumvent this issue, the previously reported dual-vector system was adopted to express SpCas9 nuclease from one vector and to express its gRNAs together with a fluorescent reporter gene from another vector²⁵. Recently, St1Cas9 (~3.3 kb) from Streptococcus thermophilus⁶⁶ and a rationally designed truncated form of SpCas9⁶⁷ were developed to fit SpCas9 and its gRNA into a single AAV construct. Unfortunately, St1Cas9 requires a more complex PAM sequence that limits the number of targetable loci⁶⁶, and truncated SpCas9 has a much lower efficiency than its wild-type counterpart⁶⁷.

Figure 1. CRISPR/Cas9-based in vivo genome editing by using small Cas9 orthologues, different routes of AAV administration, tissue-specific minimal promoters, and AAV serotypes.

(A) AAV-CRISPR vectors. Campylobacter jejuni-derived Cas9 (CjCas9) is the smallest Cas9 orthologue (984 amino acids, 2.95 kb). Staphylococcus aureus-derived Cas9 (SaCas9) (1,053 amino acids, 3.16 kb) is smaller than the most widely used Cas9 derived from Streptococcus pyogenes (SpCas9) (1,368 amino acids, 4.10 kb). Owing to the large size of SpCas9, the SpCas9 gene and two gRNAs are packaged into two separate AAV vectors. The SpCas9-based dual-vector system enables one vector to express SpCas9 and another to express multiple gRNAs and a fluorescent reporter gene. Two gRNAs can be packaged with the SaCas9 into a single AAV vector if minimal promoter and polyadenylation signal are used. In the all-in-one vector system, CjCas9 enables packaging with two gRNAs and a fluorescent reporter gene, enhanced green fluorescent protein (eGFP), into a single AAV vector. (B) AAV serotypes for tissue-targeted delivery. (C) Tissue-specific minimal promoters for driving CRISPR expression. (D) Different routes of AAV administration into various tissues of a mouse. AAV, adeno-associated virus; CRISPR, clustered regulatory interspaced short palindromic repeat; gRNA, guide RNA.

To overcome these drawbacks, other recently discovered small Cas9 orthologues, including Staphylococcus aureus-derived Cas9 (SaCas9, 3.16 kb)⁴ and Campylobacter jejuni-derived Cas9 (CjCas9, 2.95 kb)⁵, have been used to package the Cas9 and its gRNA into a single AAV delivery vehicle for in vivo genome editing. To date, at least 11 independent in vivo studies have used the AAV-SaCas9 system to edit disease-associated genes in a variety of tissues, including brain⁶⁸, muscle^54,55,57,69, retina⁵⁸, heart⁷⁰, and liver^4,49,71. More recently, the quadruplex gRNAs/SaCas9 vector consisting of SaCas9 and multiplex sgRNAs was successfully delivered using AAV-DJ/8 for in vivo excision of HIV-1 proviral DNA in various solid tissues/organs via a single intravenous injection in humanized bone marrow/liver/thymus (BLT) mice with chronic HIV-1 infection⁷². In the all-in-one vector system, CjCas9 can be packaged with multiple gRNAs and even a fluorescent reporter gene into a single AAV vector⁵. Even though CjCas9 represents the smallest Cas9 orthologue characterized to date, only one in vivo study has demonstrated the use of AAV vectors to deliver CjCas9 into the mouse’s retina for the treatment of age-related macular degeneration because of the recent development of CjCas9⁵. Excitingly, AAV-CjCas9 offers the possibility of editing multiplex genes and tracking the expression of fluorescent reporter genes at high efficiency by simply introducing a single type of AAV vector into the mouse body. The CRISPR-mediated genome-editing specificities can be further improved by adopting truncated gRNAs⁷³ and small chemical molecules to regulate Cas9 activity⁷⁴ or to enhance double-stranded break-induced homologous recombination efficiency⁷⁵ without sacrificing on-target genome-editing efficiencies.

Natural and engineered AAV serotypes

Multiple naturally occurring AAV serotypes have been harnessed to deliver the CRISPR/Cas9 complex into different tissues and organs in mice (Figure 1B). AAV1 is particularly efficient to drive CRISPR transgene expression in the brain^25,65,76, whereas AAV8 appears well suited for the transduction of liver^{4,49,61,71,77} and muscle⁵⁵ in mice. AAV9 has the general ability to transduce all major tissues, including muscle^54,56,69, retina⁵, heart^62–64,78, and lung⁶⁵, in mice. Recently, the AAV9-CRISPR system has been successfully used to identify functional tumor suppressors in glioblastoma by performing high-throughput mutagenesis in the brain of conditional-Cas9 mice to recapitulate human glioblastoma⁷⁹. AAV2⁵⁹ and AAV5⁵⁸ have also been used to transduce retina, whereas AAV6⁵⁷ enables CRISPR-mediated gene editing in muscle tissue.

In order to alter tropism, reduce blockage by neutralizing antibodies, or enhance transduction efficiency, AAV capsids have been chemically or genetically modified to produce hybrid capsids by combining the properties of multiple serotypes, capsid shuffling, directed evolution, rational mutagenesis, or carrying peptide insertions that introduce the novel receptor-binding activity. AAV variants with engineered capsids (AAV2-retro, AAV2g9, AAV-DJ, and AAV-DJ/8) have been used to package CRISPR for transduction in the mouse brain. An in vivo-directed evolution-engineered AAV variant, rAAV2-retro, permits robust retrograde access to projection neurons in functionally connected and highly distributed networks with a high efficiency for neural circuit dissection and in vivo genome editing⁶⁸. An AAV chimera derived from AAV2 and AAV9, AAV2g9, enables preferential, robust, and widespread neuronal transduction within the mouse brain for CRISPR/Cas9-mediated gene deletion for the treatment of neurological disorders⁸⁰. To provide substantially higher infectivity rates across a broad range of tissues than any known native serotypes, the AAV-DJ was engineered by DNA family shuffling to create a hybrid capsid from eight different native serotypes⁸⁷. AAV-DJ/8 was created by introducing two point mutations in the heparin-binding domain of AAV-DJ to increase uptake in brain tissue in vivo, leading to a similar ability to infect heart and brain tissues with AAV8 and AAV9^87,88. In fact, AAV-DJ/8 has been successfully used alongside the CRISPR/Cas9 system with doxycycline-dependent gRNA expression for inducible genome editing of neurons in vivo within the mouse brain⁸¹. AAV-DJ/8 has also been successfully used to package and deliver the CRISPR/SaCas9 vectors for in vivo excision of HIV-1 proviral DNA in animal models⁷².

Recently, the capsids of AAV9, which is the most efficient native serotype characterized to date for in vivo transduction of the brain, have been further engineered for better performance in transductions and tissue targeting specificity. These include AAV-PHP.B (a chimeric AAV9 variant generated by inserting a 7-mer sequence, TLAVPFK, on AAV9 capsid)¹³, AAV-AS (a chimeric AAV9 variant generated by inserting a poly-alanine peptide on AAV9 capsid)¹⁴, and AAV9/3 (double tyrosine-mutant form of AAV9)⁸⁹, all of which were shown to be more efficient in transductions of the adult mouse central nervous system after intravenous injection compared with the wild-type AAV9. Furthermore, some of these engineered AAV capsids could de-target the native AAV tropism without affecting vector production, capsid assembly, infectivity, and gene transfer. Therefore, these novel engineered vectors offer promising options to package the CRISPR/Cas9 complex to efficiently edit disease-associated genes in the mouse brain or any other tissues. Additionally, self-complementary AAV9 (scAAV9) can be used to significantly decrease the time required for the onset of CRISPR expression in the central nervous system of mice⁸⁴. Second-strand synthesis is a rate-limiting step for efficient transduction by recombinant AAV vectors³¹. The use of double-stranded self-complementary AAV (scAAV) has improved the transduction efficiencies by bypassing the second-strand synthesis step, but cloning capacity was further reduced to accommodate the complementary strand⁹⁰.

Tissue-specific minimal promoters

A tissue-specific promoter is important to ensure the expression of the CRISPR transgene in a tissue- or organ-specific manner. Various truncated or minimal tissue-specific promoters have been used to drive the expression of CRISPR in specific tissues upon injection of the AAV vectors into the mouse body (Figure 1C). For example, the 235 base pair (bp) mouse pMecp2 (methyl CpG binding protein 2) and 485 bp human hSyn1 (human synapsin I) promoters have been used to efficiently and specifically drive the expression of CRISPR and other transgenes in the mouse brain²⁵ and retina⁵⁹. A striated muscle-restricted promoter, CK8, enabled specific expression of SpCas9 and SaCas9 for gene editing in the muscle of the mdx mouse model of Duchenne muscular dystrophy upon intramuscular (tibialis anterior) or systemic (retro-orbital) injection of AAV6 vectors⁵⁷. The liver-specific TBG (thyroxine binding globulin)^4,49 and cardiac-specific Myh6 (myosin heavy polypeptide 6)⁶³ promoters have been successfully used to drive the expression of CRISPR for in vivo genome editing in mouse liver and heart, respectively. More general promoters such as CMV (cytomegalovirus)^54,58,60,62 and EFS (elongation factor-1 alpha short)^5,65 can be used to express the CRISPR in multiple tissues, including muscle, retina, heart, and lung, for simultaneous gene editing of various tissues in mice. Compared with the CMV promoter, the elongation factor-1 alpha (EF1a) promoter tends to give more consistent and prolonged expression regardless of cell type or physiology^91,92. EF1a is more stable in long-term culture because it maintains high transgene expression and stability and the copy number of episomal vectors⁹¹. Thus, despite a stronger gene expression driven by the CMV promoter in many mammalian cell types, EF1a is preferred for efficient transgene expression in mouse embryonic stem cell lines⁹³ and for engineering stable tumor cell lines⁹². The library of minimal tissue-specific promoters alongside AAV vectors used for controlling the expression of CRISPR or other transgenes in vivo is expanding with the development of more tightly regulated synthetic hybrid promoters, such as widely used CAG, CAGGS, CASI, CBh, GFAP, TRE3G, SMVP, Spc512, H1/T0, CK7-miniCMV, pICAM2, HCRhAATp, and Tight promoters (Table 2).

Chimeric gRNA expression is frequently driven by an RNA polymerase III promoter, most often by the human U6 small nuclear (U6) promoter. The human U6 promoter has been shown to be more efficient than its murine homolog for short hairpin RNA (shRNA) expression in both human and murine cells⁹⁴. The human U6 promoter was also proven more efficient than the human H1 promoter in driving shRNA expression for the long-term inhibition of gene expression in vitro and in vivo⁹⁵. However, the use of the U6 promoter to overexpress shRNAs in vivo can induce severe hepatotoxicity and inflammatory side effects due to abundant shRNA production⁹⁶. The weaker H1 promoter may offer a safer option in this case, but its therapeutic efficacy is dependent on the sequence of the shRNA⁹⁶. Some promoters require a particular nucleotide as the transcription start site (TSS) to initiate transcription. For instance, the human or mouse U6 promoters require a guanine nucleotide (G) at the 5′ end of the transcript for effective transcription⁹⁷. To use the U6 promoter for driving gRNA expression, the first nucleotide of the transcribed gRNA should be a G to maximize U6 promoter activity⁹⁷. If the G is absent at the TSS of the gRNA sequence, the human RNA polymerase III promoter H1 can be used to drive the expression of the gRNA⁷⁶. The use of H1 promoter-expressed gRNAs could enhance the versatility of the CRISPR technology by expanding the targetable genome⁹⁸. For multiplex CRISPR/Cas9-based genome editing such as large genomic DNA deletion or synergistically enhanced gene activation and repression, four independent polymerase III promoters (human U6 promoter, murine U6 promoter, 7SK, and H1) can be used to express four gRNAs in a single expression cassette²¹. The use of distinct polymerase III promoters can avoid the problem of genetic recombination in a viral vector and the loss of the proximal gRNA due to identical DNA sequences of the promoters⁹⁹. The doxycycline-dependent promoters such as H1/TO and U6/TO can also be used for inducible in vivo genome editing by supplying animals with Dox-containing food to regulate the expression of the gRNA in a Dox-dependent manner⁸¹. The introduction of Dox inhibits Tet repressor (TetR) binding and induces gRNA expression⁸¹.

Local and global delivery of AAV vectors

The route of administration of AAV-CRISPR vectors into the mouse body also determines the efficacy and specificity of gene editing for in vivo therapeutics (Figure 1D). Given the inability of the most native AAV serotypes (except for AAV9¹⁰⁰, AAVrh8¹⁰¹, and AAVrh10¹⁰²) to cross the blood-brain barrier, stereotactic^25,65,76 and intrathecal⁸⁴ injections are still the preferred routes to systemic injection in order to introduce the AAV-CRISPR vectors into specific regions of the mouse brain or central nervous system for therapeutic gene editing. For the stereotactic injection, AAV-CRISPR vectors are intracranially injected into the brain by using a stereotaxic apparatus and the injection micropipette. In the intrathecal injection approach, AAV-CRISPR vectors are injected into the spinal canal or subarachnoid space to reach cerebrospinal fluid. For efficient therapeutic gene editing in the mouse muscle tissue, intramuscular (via tibialis anterior), intraperitoneal (via intraperitoneal space), or intravenous (via the tail vein, temporal vein, and retro-orbital) injections are the preferred methods to introduce AAV-CRISPR vectors, as have been demonstrated for the mdx mouse model of Duchenne muscular dystrophy^55–57. Subretinal^58,60 or intravitreal^59,85 injections are used to deliver AAV-CRISPR vectors into the mouse retina for gene editing. As most of the AAV vectors will accumulate in the mouse liver tissue upon systemic injection, intravenous injection by tail vein or temporal vein can efficiently edit the genes associated with liver diseases^4,49,71. Systemic injection remains preferred to local injection in the liver tissue because the delivery is more uniform and the distribution of viral vectors is broader. It is also a good delivery mode for gene editing in multiple tissues, as it is almost non-invasive to the body. On the other hand, intracardiac⁶⁴, subcutaneous⁷⁸, intraperitoneal⁶³, or systemic^62,70 injection has been used to introduce AAV9 vectors carrying CRISPR transgene into the mouse heart for gene editing. Intranasal and intratracheal injections are used to locally deliver the AAV-CRISPR vehicles into the mouse lung tissue⁶⁵. Therefore, the selection of an effective route of injections is dependent on the target tissue, AAV serotypes, and tissue-specific promoters used.

Non-specific delivery and transgene expression

Tissue-specific promoters and AAV serotypes might still be able to transduce and induce transgene expression in healthy tissues or other non-target organs if a strong promoter or a high dose of viral vectors is introduced into the body by intravenous injection. Despite a significant improvement in the efficiency and specificity of newly discovered natural and engineered AAV variants for systemic delivery in mice, the toxicity and side effects associated with the non-specific delivery of the transgenes to the non-target tissues and organs remain a concern^13,89. It is easier to get delivery vehicles taken up by liver than other organs upon systemic injection. Therefore, it is more challenging to specifically target non-liver organs than liver via a systemic delivery approach. For instance, recently discovered synthetic vectors, AAV-PHP.B¹³ and tyrosine-mutant AAV⁸⁹, were found to transduce the adult mouse central nervous system more efficiently and widely than the natural AAV9 after intravenous injection. However, these synthetic vectors can also transduce the liver and other peripheral organs upon systemic delivery^13,89. In this case, local injections such as stereotaxic injection into the brain can minimize the possible complications or adverse side effects associated with non-specific expression of CRISPR in healthy tissues or other non-target organs.

Off-target effects

Even though CRISPR/Cas9 predominantly recognizes the intended target sites, a series of high-throughput genome-wide methods such as multiplex Digenome-seq^109,110, ChIP-seq^111,112, GUIDE-seq¹¹³, and whole-genome sequencing¹¹⁴ as well as targeted deep sequencing¹¹², Cas9 toxicity screens¹¹⁵, and SITE-seq biochemical methods¹¹⁶ have revealed evidence of off-target effects due to target mismatch tolerance of CRISPR/Cas9. As inter-individual natural genetic variation can affect CRISPR/Cas9 specificity¹¹⁷, a recently developed CIRCLE-seq approach could be used to identify genome-wide off-target mutations of CRISPR/Cas9 that are associated with cell type-specific single-nucleotide polymorphisms to provide personalized specificity profiles¹¹⁸. Notably, given the same target sequence of gRNA, off-target sites of Cas9 before and after being fused to a catalytic enzyme (for example, cytidine deaminase base editor and chromatin modifiers) could be different; therefore, independent assessment of their genome-wide specificities is recommended¹¹⁹. Imprecise repair of Cas9-induced DNA double-stranded breaks can give rise to deleterious structural chromosomal rearrangements such as deletions, inversions, and translocations, which in turn may activate oncogenes or cause genome instability¹²⁰. Hence, high-throughput screenings of CRISPR/Cas9 off-target activity and further improvement in the fidelity of CRISPR/Cas9 on-target activity are essential for safety in clinical gene transfer applications.

There are many ways to minimize CRISPR/Cas9 off-target effects in the human genome. For example, a recently developed RNA-targeting Cas9 (RCas9) system could avoid permanent off-target genetic lesions in DNA-mediated CRISPR-based therapeutics¹²¹. The RCas9 system consists of a fusion of rationally truncated dCas9 protein and PilT N-terminus (PIN) domain that can be packaged into the AAV vector for eliminating toxic microsatellite repeat expansion RNA or reducing repetitive RNA level without targeting the DNA¹²¹. With a similar strategy, systemic delivery of dCas9/gRNA by AAV9 significantly reduced pathological RNA foci, rescued chloride channel 1 protein expression, and decreased myotonia in myotonic dystrophy mice by impeding the transcription of expanded microsatellite repeats¹²². Another way is to deliver purified Cas9 RNPs rather than plasmid expression vectors^10,123. Despite some technical constraints in non-viral delivery methods for in vivo administration, the use of Cas9 RNPs can limit the duration of Cas9 expression and decrease the chance of Cas9 nuclease cleaving at non-specific sites in a genome because of the rapid clearance from the cell^10,123. Complementing the CRISPR-based editing capability with conditional genetic manipulation tools such as photoactivatable Cas9^124,125, chemical-inducible Cas9^126,127, or multiple inputs logic gate genetic circuits^128,129 enables the precise spatial and temporal control of Cas9 activity inside the cell, which in turn leads to the reduction in off-target activity. Alternatively, a pair of Cas9 nickases^130–132, dCas9-FokI¹³³, or high-fidelity Cas9 variants such as SpCas9-HF1¹³⁴, eSpCas9¹³⁵, and HypaCas9¹³⁶ can be used to minimize undesired off-target mutagenesis.

The use of a scarless genome-editing strategy for targeted point mutation knock-in can also minimize unwanted mutation formations to favor the desired clean base editing outcomes^137,138. For a point mutation knock-in, the efficiency of precise sequence replacement by CRISPR-mediated homology-directed repair (HDR) could be significantly increased by using asymmetric donor DNA¹³⁹, HDR enhancer¹⁴⁰, or short ssDNA donor oligonucleotides as a donor template instead of long plasmid donor¹³⁹ or by inhibiting non-homologous end joining (NHEJ) activity¹⁴¹. Also, the artificial chimeric RNAs¹⁴²—truncated^73,113 and chemically modified¹⁴³ gRNAs—were shown to have lower off-target activity than the original gRNAs. A number of bioinformatics analysis tools such as GuideScan¹⁴⁴, CRISPRdirect¹⁴⁵, and Cas-OFFinder¹⁴⁶ permit the specific design of gRNAs to avoid binding at the non-intended target sites in the human genome. The specificity of the gRNA designed can be improved by selecting a target sequence with two Gs at the 5′ terminus¹³¹ and avoiding potential mismatches at the seed sequence or base-pairing adjacent to the PAM¹⁴⁷. Another innovative way to improve the specificity and reduce the toxicity of the CRISPR/Cas9 is co-delivery of DNA decoys or competitive inhibitor oligonucleotides bearing all possible off-target sequences that can sequester and prevent the CRISPR/Cas9 from binding to the off-target sites within a host genome. A similar concept has been successfully demonstrated to alleviate the off-target effects of RNA interference by using RNA decoys to reduce the sense strand activity of shRNAs¹⁴⁸, while artificial microRNA (miRNA) sponges have been used to inhibit miRNA function and its ability to regulate natural mRNAs¹⁴⁹. In addition, a recently developed miRNA-responsive CRISPR/Cas9 switch could be useful for cell type-specific genome editing by sensing endogenous miRNA activities¹⁵⁰.

Immunogenicity

Owing to the exogenous nature of AAV and CRISPR components, host immune responses can attenuate therapeutic effects and cause side effects. Thus, the toxicity and immunogenicity associated with AAV and CRISPR components should be circumvented for safer and higher efficacy in clinical gene therapy applications. Even though the AAV generally elicits a very mild immune response and does not induce extensive cellular damage in vivo, certain AAV serotypes such as AAV9 may evoke humoral immune responses, as indicated by the presence of capsid-specific antibodies⁵⁰. While transient immunosuppression is one of the possible ways to mitigate the host immune response following the delivery of AAV vectors¹⁵¹, it is not feasible for long-term therapeutic treatments of chronic diseases and is prone to adverse complications such as infections and malignancies. Empty capsid mutants can be used as decoys to overcome pre-existing humoral immunity by adsorbing antibodies in the bloodstream upon systemic delivery of both empty and functional AAVs¹⁵². Alternatively, the AAV capsids can be genetically engineered or mutated to reduce the binding affinity and the neutralizing effects of AAV antibodies^15,153–155. In addition to AAVs, the presence of Cas9-specific antibodies in Cas9-exposed animals indicated that Cas9 could evoke deleterious cellular and humoral immune responses in vivo⁵⁰. Expression of Cas9 in vivo could affect the transcription of the genes associated with the immunological processes. This in turn may destabilize the host immune system, elicit significant cellular infiltration or expansion, and induce enlargement of the draining lymph nodes with increased immune cell counts⁵⁰. The Cas9-responsive T-cell clonotype described previously could serve as a distinctive biomarker to assess Cas9-specific immunity before clinical implementation of the CRISPR system⁵⁰. To minimize the immune response due to prolonged expression of Cas9, conditional genome editing with the self-limiting CRISPR/Cas9 system⁵⁸ or light-inducible^124,125 or chemical-inducible^126,127 Cas9 can be used to minimize the duration of Cas9 expression in the body.