Novel methods for the generation of genetically engineered animal models

Abstract

Genetically modified mouse models have shaped our understanding of biological systems in both physiological and pathological conditions. For decades, mouse genome engineering has relied on transgenesis and spontaneous gene replacement in embryonic stem (ES) cells. While these technologies provided a wealth of knowledge, they remain imprecise and expensive to use. Recent advances in genome editing technologies such as the development of targetable nucleases, the improvement of delivery systems, and the simplification of targeting strategies now allow for the rapid, precise manipulation of the mouse genome. In this review article, we discuss novel methods and targeting strategies for the generation of mouse models for the study of bone and skeletal muscle biology.

1. Introduction

Our ability to precisely manipulate the mouse genome has transformed biomedical research. Over the past few decades, genome modifications in mice have relied on two main technologies: transgenesis and spontaneous gene replacement by homologous recombination (HR) in mouse embryonic stem (ES) cells [1]. While these approaches have provided a wealth of knowledge about the structure and function of biological systems, they remain imprecise, tedious, and expensive.

Transgenesis relies on the use of large and complex expression vectors directing the expression of complementary DNA (cDNA) under the control of viral or tissue-specific mammalian promoters delivered to pronuclear stage zygotes via microinjection (Fig. 1). While this approach provides a crude but effective way to engineer animal models expressing reporter genes, mutant forms of a gene, and conditionally regulated genes, it cannot be used to precisely modify endogenous genes. Moreover, the integration of transgenes within the mouse genome occurs at random and the location of the integration site as well as the number of integrations may affect expression of the transgene. In addition, if the transgene integration disrupts a gene or a transcriptional regulatory element, the integration site itself may induce a phenotype of its own. As the transgene integration site and the number of transgene integrations may vary from one mouse to another, multiple founders need to be expanded and examined both for transgene expression levels and the resultant phenotypes [1].

Figure 1. Current strategies to produce genetically modified mouse models.

Transgenesis consists of introducing foreign DNA into embryos via microinjection. These embryos are then transferred to pseudopregnant females and their progeny are analyzed for genetic modifications. Gene targeting in embryonic stem (ES) cells requires the culturing of mouse ES cells, which are transduced to introduce mutations. These cells are then subjected to selection and genotyping to isolate cells with desired mutations. Select clones are then injected into blastocysts, which are transferred to pseudopregnant females to produce chimeric mice. Chimeras are then outbred to wild-type animals and their progeny analyzed for desired mutations. Genome editing using RNA-guided nucleases is similar to transgenesis in that embryos are microinjected, transferred to pseudopregnant females, and their progeny analyzed for genetic modifications. In this case, however, instead of injecting a transgene that randomly integrates into the genome, the components of an RNA-guided nuclease system are microinjected into embryos, which introduce the desired genetic modification at a specific locus. In lieu of microinjection, RNA-guided nucleases can be delivered into pronuclear stage zygotes via electroporation.

Spontaneous gene replacement by HR, commonly referred to as gene targeting in ES cells, allows for the modification of endogenous gene loci but remains somewhat imprecise (Fig. 1). This technology relies on HR, a naturally occurring but infrequent event that requires the generation of complex DNA constructs containing large homology arms as well as selection markers whose expressions are driven by strong promoters. These constructs are typically difficult to engineer and leave exogenous genetic material at the target site. The presence of strong promoters may, in some cases, affect expression of surrounding genes [2]. Following transduction, thousands of clones must be manually picked and analyzed at the genomic level for the proper rearrangement. From these clones, only a handful will possess the correct modification and thus will be suitable for blastocyst injection. This technology is amenable to only a few mouse strains for which ES cells are available. The process of engineering mice using this technology is not always reliable and often requires several rounds of blastocyst injections and transfers to pseudopregnant females to produce mice capable of transmitting the mutant allele. Finally, lengthy backcrossing may be required if a specific genetic background is required for research purposes [1].

To get around some of the limitations associated with transgenesis and gene targeting in ES cells, most notably the lack of precision and efficiency, targetable nucleases were developed [3]. These enzymes function by introducing highly recombinogenic DNA double strand breaks (DSBs) at a specific location within a genome. These breaks in turn facilitate the engineering of mutations by activating endogenous DNA DSB repair pathways. These enzymes are so effective at introducing DNA breaks that they can be used to engineer mutations without the need for positive selection and thus can be delivered directly into pronuclear stage zygotes for the generation of mouse models (Fig. 1). Several targetable nucleases have been developed, including zinc finger nucleases (ZFNs), transcription activator-like effector (TALE) nucleases (TALENs) and Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR)-CRISPR-associated protein 9 (Cas9) systems [4–9]. By providing the precision, simplicity, and versatility needed to engineer virtually any genetic modification, targetable nucleases have transformed the way genome editing is performed.

Alongside advances in genome engineering technologies, the generation of mouse models has benefited from improvements in allele design strategies, delivery methods, and technologies allowing for the modulation of gene expression. These include the development of new recombinase systems for conditional activation or inactivation of genes; new systems for the modulation of gene expression at the transcriptional or posttranscriptional level; new genome editing technologies that do not rely on the generation of DNA DSBs; and new strategies to engineer conditional allele. In this review article, we discuss established and novel CRISPR-based technologies and targeting strategies for engineering genetically modified mouse models and for modulating gene expression including but not limited to the study of bone and skeletal muscle biology.

2. Genome editing and modulating technologies

Genome editing using targetable nucleases relies, for the most part, on the ability of these enzymes to introduce DNA DSBs at a precise location within a genome [3]. These highly recombinogenic breaks are then fixed by DNA DSB repair pathways, which can be hijacked to introduce desired genetic alterations. Recent technological advances now also allow for the engineering of simple modifications without the need for introducing DNA DSBs. These technologies, although less efficient than conventional targetable nuclease systems, may solve issues associated with therapeutic applications of targetable nucleases where introduction of random mutations at the target site may have deleterious effects.

2.1. Genome editing with DNA DSBs

In eukaryotic cells, several DNA repair pathways are involved in the resolution of DNA DSBs. The nonhomologous end joining (NHEJ) pathway typically results in the introduction of random insertions or deletions of genetic material at the break site. On the other hand, homology directed repair (HDR) pathways, which include HR, Fanconi anemia repair pathway, and many others, use DNA repair templates to resolve the break [10]. By providing an exogenous template, known as an HDR template, in the form of a double or single stranded DNA molecule, these pathways can be commandeered to introduce specific alterations at or near the break site (Fig. 2A) [3]. Introduction of DNA DSBs can be achieved by administration of targetable nucleases. For mouse genome engineering, two main platforms, CRISPR-SpCas9 [11–13] and CRISPR-FnCpf1[14, 15], have been developed.

Figure 2. Nuclease-based genome editing relies on the introduction of DNA DSBs.

A) Schematic representation of DNA DSB repair pathways and how they can be manipulated to introduce specific mutations or DNA elements. The most common DNA DSB repair pathways are nonhomologous end joining (NHEJ) and homology directed repair (HDR), which includes homologous recombination (HR) and others. Resolution of DNA DSBs by the error-prone NHEJ pathway typically results in the insertion or deletion of genetic material caused by misalignment of resected strands during repair. Resolution of the DSB by HR is thought to be error free as it uses the sister chromatid as a repair template. HDR pathways can be overtaken to introduce specific modifications by coadministration of an HDR template.

B) Schematic representation of the CRISPR-SpCas9 system for genome editing and modifications made for its use in mammalian cells. CRISPR-SpCas9 is one of the most commonly used genome editing systems and is made up of two main components: the endonuclease SpCas9 and an sgRNA. SpCas9 is guided to the target site by the sgRNA and induces a DNA DSB 3 nucleotides upstream of the PAM. For genome editing in mammalian cells, the open reading frame has been human codon optimized and a NLS has been added. The target genomic DNA is shown in black, with the PAM sequence (5’-NGG-3’) highlighted in green. The sgRNA is shown in orange. The RuvC-like and HNH-like cut sites are indicated by a black or a blue triangle, respectively.

C) Schematic representation of the CRISPR-FnCpf1 system and the modifications for its usage in mammalian cells. CRISPR-FnCpf1 is another common system used for genome editing and is composed of two main elements: the endonuclease FnCpf1 (Cas12a) and a crRNA. Distinct from the CRISPR-SpCas9 system, FnCpf1 is guided to the target site by the crRNA and induces scattered DNA breaks at the 3’ end of the target sequence, more specifically at position 18 of the target strand and position 23 of the non-target strand, leaving a 5’ overhang at the cut site. For genome editing in mammalian cells, the open reading frame has been human codon optimized and a NLS has also been added. The target genomic DNA is shown in black, with the PAM sequence (5’-TTN-3’) highlighted in green. The crRNA is shown in red, and the two RuvC-like cut sites are indicated by black triangles.

D) PAM requirements for Cas9 and Cpf1 (Cas12a) from various species.

2.1.1. CRISPR-Cas9

The most commonly used and perhaps the most effective targetable nuclease system for genome editing in mice has been adapted from the Class 2 Type II CRISPR-Cas9 system from Streptococcus pyogenes [11–13]. Known as CRISPR-SpCas9 (Fig. 2B), this system comprises two main components: a large non-specific endonuclease called SpCas9, and a short 96 nucleotide-long RNA transcript, referred to as the single guide RNA (sgRNA). The sgRNA serves two main purposes: it interacts with SpCas9 and provides target specificity via Watson-Crick base pairing with its target DNA sequence. In this system, the simple modification of the first 20 nucleotides located at the 5’ end of the sgRNA is sufficient to program SpCas9 to target any genomic locus. The endonuclease contains two enzymatically active domains, a RuvC-like and an HNH-like domain, which generate blunt-ended DNA breaks 3 nucleotides upstream of a short sequence known as the protospacer adjacent motif (PAM). These PAM sequences are located at the 3’ end of the target sites and are required for Cas9 nucleases to be active. Although the requirement for PAM sequences has been viewed as a major limitation to the application of CRISPR-SpCas9 technology for genome editing, SpCas9 PAM sequences (5’-NGG-3’) are found, on average, every 8 nucleotides within the mouse genome, which allows for the targeting of virtually any locus. Moreover, several CRISPR-Cas9 systems with distinct PAM requirements have been identified in various bacterial species, expanding the scope of CRISPR-Cas9 systems for genome engineering (Fig. 2D) [3, 16]. For genome editing in mammalian cells, the SpCas9 open reading frame has been optimized for codon usage and a nuclear localization site (NLS) was added to promote nuclear translocation [8, 17, 18].

2.1.2. CRISPR-Cpf1

CRISPR-Cpf1 was adapted from the Class 2 type V CRISPR-Cas system from Francisella novicida (CRISPR-FnCpf1) (Fig. 2C) and comprises two main components: a large non-specific endonuclease called Cpf1 (also known as Cas12a) and a short CRISPR RNA (crRNA) transcript that provides target specificity [9]. Cpf1 is distinct from Cas9 in that it contains 2 RuvC-like endonuclease activities which introduce scattered DNA breaks at the 3’ end of the target sequence, more specifically at position 18 of the target strand and position 23 of the non-target strand, leaving a 5’ overhang at the cut site. The T-rich PAM sequence of Cpf1, 5’-TTN-3’, is located at the 5’ end of the target sequence and is found at a similar frequency as the SpCas9 PAM sequence within the mouse genome. Like CRISPR-Cas9 systems, multiple CRISPR-Cpf1 systems have been identified in various bacterial species, each with distinct PAM requirements, again expanding the number of possible target sites for CRISPR-Cas systems (Fig. 2D).

2.2. Genome editing without DNA DSBs

To avoid the potentially deleterious effects of introducing DNA DSBs, genome editing can also be performed using base editors or prime editors. These systems introduce specific mutations by directly changing the nucleotide composition at the target site, bypassing the need to activate DNA DSB repair pathways. However, the window in which they operate is narrow, undesired collateral nucleotide substitutions may occur, and their efficiency is not comparable to that of conventional CRISPR-Cas systems [19].

2.2.1. Base editing

Base editing is an RNA-guided-based genome editing technology that allows for the generation of nucleotide substitutions without the need for introducing DNA DSBs [19]. Three classes of base editors have been developed: cytidine base editors (CBEs) which allow the conversion of cytidine to thymine (C -> T) [20], adenine base editors (ABEs) which catalyze the conversion of adenine to guanine (A ->G) [21], and cytidine to guanine base editors (CGBEs) which catalyze cytidine to guanine (C ->G) base transversion [22]. Several iterations of these systems have been developed and the selection of a base editor depends on several factors including the nature of the substitution, the position of the target nucleotide relative to the PAM sequence, the possibility of introducing undesired bystander mutations, and the need to minimize off-target editing [19]. For simplicity, one example for each class of base editor is described.

BE4max is a CBE in which a catalytically impaired SpCas9 retaining only the HNH-like activity is fused to a cytidine deaminase (CD) at its N-terminus and a pair of uracil glycosylase inhibitors (UGIs) at its C-terminus (Fig. 3A). In this system, the CD catalyzes the conversion of a C into a uracil (U). The conversion is stabilized by the pair of UGIs which prevent the removal of the uracil by the mismatch repair (MMR) pathway and by the nicking of the non-target strand by the HNH-like domain of SpCas9, which promotes the replacement of the G for an A [20].

Figure 3. Base editors and prime editors allow for genome editing without DNA DSBs.

A) Schematic representation of BE4max, a cytidine base editor (CBE) that catalyzes the conversion of a cytidine into a uracil. This CBE is composed of a RuvC-deficient SpCas9 fused to a cytidine deaminase (CD) and a pair of uracil glycosylase inhibitors (UGIs). The target genomic DNA is shown in black, with the PAM sequence highlighted in green. The sgRNA is shown in orange, and the single HNH-like cut site is indicated by a blue triangle. The light gray box highlights the editing window for this CBE and the target nucleotide, C, is shown in green.

B) Schematic representation of BE7.10, an adenine base editor (ABE) that catalyzes the conversion of adenine to a guanine. This ABE is formed from the fusion of a catalytically impaired SpCas9 with a modified dimeric tRNA adenine deaminase (TadA*). The target genomic DNA is shown in black, with the PAM sequence highlighted in green. The sgRNA is shown in orange, and the single HNH-like cut site is indicated by a blue triangle. The light gray box highlights the editing window for this ABE and the target nucleotide, A, is shown in green.

C) Schematic representation of CGBE1, a cytidine to guanine base editor (CGBE) that catalyzes cytidine to guanine transversion. This CGBE is composed of a SpCas9 nickase fused to an uracil DNA N-glycosylase (eUNG) and the rat APOBEC1-R33A. The target genomic DNA is shown in black, with the PAM sequence highlighted in green. The sgRNA is shown in orange, and the single HNH-like cut site is indicated by a blue triangle. The light gray box highlights the editing window for this CGBE and the target nucleotide, C, is shown in green.

D) Schematic representation of a prime editor, which is composed of two main parts: an HNH-deficient SpCas9 fused to a reverse transcriptase (RT) and a prime editing guide RNA (pegRNA). In this system, the pegRNA both guides the catalytically impaired SpCas9 to the target site and acts as a template for the RT to incorporate the desired nucleotide substitutions upon nicking of the target site by the RuvC domain of SpCas9. The target genomic DNA is shown in black, with the PAM sequence highlighted in green. The pegRNA is shown in blue and the nucleotide substitutions encoded within the pegRNA are highlighted in red. The single RuvC-like cut site is indicated by a black triangle.

BE7.10 is an ABE formed from the fusion of a catalytically impaired SpCas9 with a modified dimeric tRNA adenine deaminase (TadA*) which efficiently catalyzes the conversion of A to G (Fig. 3B). Like the BE4max system, the nicking of the non-target strand by the HNH-like activity of Cas9 stabilizes the conversion [21].

CGBE1, like CBEs and ABEs, is composed of a SpCas9 nickase fused to a uracil DNA N-glycosylase (eUNG) and the rat APOBEC1-R33A, a CD with reduced off-target editing activity, which catalyzes C to G transversion (Fig. 3C). Stabilization of this transversion is promoted by the nicking of the non-target strand by the HNH-like domain of SpCas9. Although this system has not been used in animals, this editor was shown to efficiently induce base transversion in cultured mammalian cells and may provide an alternative strategy to engineering single nucleotide changes [22].

For use in mammalian cells, the cDNAs encoding these base editors have been optimized for codon usage and NLSs have been added. Various base editors with distinct PAM requirements have also been engineered, including Cas12a base editors, increasing the scope of target possibilities [23].

2.2.2. Prime editing

Similar to base editing, prime editing is a CRISPR-based genome editing technology that allows for the generation of nucleotide substitutions without generating DNA DSBs [24]. Prime editors are composed of an HNH-deficient SpCas9 fused to a reverse transcriptase (RT) (Fig. 3D). In this system, a modified sgRNA called the prime editing guide RNA (pegRNA) is used to target SpCas9 to any location within a genome and provides a template for introducing nucleotide substitutions. Like conventional sgRNAs, the first 20 nucleotides located at the 5’ end of the pegRNA provide target specificity, while the following nucleotides act as a scaffolding and complex with SpCas9. Distinctly, however, the 3’ end of the pegRNA contains sequences complementary to the target site as well as the desired alterations. Prime editors function by first nicking the target strand with the RuvC-like domain of SpCas9, exposing a 3’ and a 5’ flap. The pegRNA then complexes with the exposed 3’ flap via Watson-Crick base pairing and serves as a primer site for the RT, which then extends the 3’ flap and incorporates the desired nucleotide substitutions. FEN1, an endogenous endonuclease, then excises the 5’ flap, allowing for hybridization of the edited 3’ flap, thus resulting in base editing via the DNA MMR pathway. If unedited strands remain, these may serve as substrates for a subsequent round of editing.

2.3. Modulation of gene expression

Although primarily used for genome editing, CRISPR–Cas9 systems have also been repurposed to allow for the regulation of gene expression by modulating epigenetic markings or bringing transcriptional activators or repressors to promoter or enhancer elements. Modulation of gene expression can also be achieved through the use of newly discovered CRISPR-based RNA interference systems as well as traditional tetracycline inducible systems. Several systems known as CRISPR activation (CRISPRa), CRISPR interference (CRISPRi), CRISPR epigenome editing, CRISPR-Cas13, and CRISPR-Cas7–11 have been developed, some of which have been used in vivo.

2.3.1. CRISPRa

CRISPRa system utilizes a catalytically inactive but sgRNA competent SpCas9 (dSpCas9) fused with a transcriptional activator (TA). Single point mutations in the RuvC-like and HNH-like domains inactivate the endonuclease function of SpCas9 rendering it catalytically inactive. This mutated SpCas9 is termed dead SpCas9 (dSpCas9). Like the wild-type SpCas9 used for genome editing, dSpCas9 is delivered to target sequences in the genome via sgRNAs complementary to that target region. As the catalytically inactive dSpCas9 is incapable of cutting DNA, the dSpCas9-sgRNA complex functions as an RNA-guided DNA recognition platform that delivers TAs to promoters of target genes. Various TAs have been successfully used in cultured cells, ex vivo and in vivo. These include VP64, a TA composed of four tandem copies of VP16 (Herpes Simplex Viral Protein 16) connected to each other by glycine-serine (GS) linkers; and VPR, a TA composed of the three activator domains VP64, p65, and Rta (Fig. 4A) [25, 26]. Transcriptional activation is achieved by tiling one or more dSpCas9::TA fusion proteins to the promoter of target genes with use of sgRNAs. The level of transcriptional activation varies depending on where the dSpCas9::TA fusion protein is placed in the gene’s promoter. Therefore, different levels of gene-induction can be obtained by using sgRNAs targeting different regions in the promoter or by using multiple sgRNAs, allowing tunable gain-of-function of endogenous genes.

Figure 4. Modulation of gene expression.

A) Schematic representation of CRISPRa systems, which are formed from the fusion of dSpCas9 to one or more transcriptional activators (TAs). Transcriptional activation is achieved by targeting one or more dSpCas9 fusion proteins to transcriptional start sites or distal enhancer elements. Differential regulation can be obtained by increasing or reducing the number of transcriptional regulators to a specific promoter. The target genomic DNA is shown in black, with the PAM sequence highlighted in green. The sgRNA is shown in orange. The transcription start site of the gene is indicated by the black arrow.

B) Schematic representation of CRISPRi systems, which are composed of a dSpCas9 fused to a transcriptional repressor (TR). Transcriptional repression is achieved by targeting one or more proteins to the promoter of a gene to actively repress gene expression by either recruiting corepressors, modifying the chromatin structure, or methylating DNA. Differential regulation can be obtained by tiling one or more repressors at a specific promoter. The target genomic DNA is shown in black, with the PAM sequence highlighted in green. The sgRNA is shown in orange. The transcription start site of the gene is indicated by the black arrow.

C) Schematic representation of the CRISPR-Cas13 system, CRISPR-LshC2c2, which recognizes ssRNA molecules and cleave them in a non-specific manner, at exposed uracil residues. This system comprises two components: a large ribonuclease called C2c2 and a crRNA. Rather than targeting double stranded DNA, CRISPR-LshC2c2 recognizes single stranded RNA molecules to promote its degradation. Unlike RNA-guided DNA-targeting systems, CRISPR-LshC2c2 does not require a PAM, but cleavage of the ssRNA may be affected by a protospacer-flanking site (PFS). The target RNA is shown in black with cleavage sites indicated by red triangles at exposed uracil residues. The crRNA is shown in red.

D) Schematic representation of the CRISPR-Cas7–11 system, Sb-gRAMP, which is comprised of a large modular protein that contains 4 Cas7-like and a single Cas11-like domains with intrinsic endoribonuclease activity that pairs with a crRNA to cleave ssRNA at positions 3 and 9 of the spacer. Unlike DNA-targeting systems, RNA-targeting systems do not require a PAM, but their activity may be affected by a PFS. The target RNA is shown in black with cleavage sites indicated by green triangles. The crRNA is shown in green.

E) Schematic representation of the Tet-Off configuration of tetracycline inducible systems, in which tTA binds to tetracycline responsive elements (TRE) located upstream of a target gene (or transgene) and promotes transcription. In the presence of tetracycline, tTA can no longer interact with TRE and thus prevents expression. Introns are shown as light gray boxes, exons as dark gray boxes, start sites as black arrows. The TRE, located upstream of the start site, is indicated in dark blue.

F) Schematic representation of the Tet-On configuration of tetracycline inducible systems, in which rtTA cannot bind the TRE element due to alterations in its amino acid composition. In the presence of tetracycline, rtTA can now bind TRE and promotes transcription. Introns are shown as light gray boxes, exons as dark gray boxes, start sites as black arrows. The TRE, located upstream of the start site, is indicated in dark blue.

To produce in vivo tools for cell type-specific activation of endogenous genes CRISPRa systems have been engineered in mice [27–30]. In these murine models, which we will refer to as CRISPRa mice, the expression of dSpCas9::TA fusion protein by a constitutive and ubiquitous promoter is guarded by a STOP cassette flanked by loxP sites (loxP-STOP-loxP). Therefore, expression of the dSpCas9::TA fusion protein requires Cre-mediated recombination of the loxP sites resulting in excision of the STOP cassette. To increase target gene expression in vivo using CRISPRa systems, CRISPRa mice are used in combination with a CRE source, which provides the system cell type specificity, and sgRNAs which provide the system target specificity.

A Cre-dependent CRISPRa model produced by Zhou et al. makes use of 3 main components: a dSpCas9-SunTag fusion, a scFv- p65-HSF1 fusion, and a sgRNA. SunTag is a polypeptide that comprises multiple GCN4 epitopes which can be recognized by a single chain antibody (scFv). The fusion of scFv to a TA such as p65-HSF1 provides a means to recruit several TAs to dSpCas9-SunTag. Co-expression of sgRNAs allows for the targeting of dSpCas9-SunTag to a gene’s promoter where it can recruit several scFv-p65-HSF1 TAs to induce that gene’s transcription [27]. This murine model when used with appropriate sgRNAs and cell-type specific Cre expression, can individually or simultaneously induce the expression of multiple genes in vivo.

2.3.2. CRISPRi

CRISPRi makes use of a catalytically inactive Cas9 (dSpCas9) fused to a transcriptional repressor (TR) (Fig. 4B). Various transcriptional repressors have been successfully used in cultured cells and in vivo, including Krüppel associated box domain (KRAB) or methyl CpG binding protein 2 (MECP2) [25, 31]. In this system, one or more dSpCas9::TR fusion proteins are directed to the promoter of target genes via sgRNAs to suppress target gene transcription. Transcriptional repression is thought to be achieved by various means such as blockade of the transcriptional machinery, recruitment of corepressors, or chromatin modifiers. The level of target gene suppression by CRISPRi varies depending on the location to which dSpCas9::TR fusion protein is directed by the sgRNAs. Therefore, screening of multiple sgRNAs is required to determine the use of which sgRNA or sgRNAs would facilitate the highest level of target gene suppression. IN addition, utilization of multiple sgRNAs to tile multiple dSpCas9::TR fusion proteins at the target gene promoter may also increase the level of transcriptional repression by CRISPRi system.

Several CRISPRi mouse models have been developed [30, 32]. Murine models that express dCas9::KRAB under constitutively expressed or temporally controlled (tetracycline-inducible) promoters are available in multiple genetic backgrounds from Jackson Laboratories as tools for global loss of function studies [32]. In the temporally controlled models, expression of the dSpCas9::KRAB fusion protein is mediated by a tetracycline-inducible promoter (TRE or tetO). These mice also carry a M2rtTA transactivator located in the Rosa locus. In this system, doxycycline administration allows global expression of the dSpCas9::KRAB fusion protein, which is directed to target genes via sgRNAs for target gene suppression.

To produce a new tool for cell type-specific suppression of endogenous genes Gersbach and colleagues have created Rosa26-LSL-dSpCas9-KRAB knock-in mouse model [30]. This model expresses dSpCas9::KRAB fusion protein under the control of the CAG promoter and a floxed STOP cassette. Co-expression of a sgRNA and Cre allows for the targeting of the dCas9-KRAB fusion protein to promoters or distal enhancers of target genes and silences expression of the target locus. When used with a cell type-specific Cre source and appropriate sgRNAs, this murine model can be used for conditional loss-of-function studies of one or more genes in vivo.

2.3.3. CRISPR epigenome editing

Catalytically inactive dSpCas9 can also be fused with epigenetic modifiers, such as p300, TET1, DNMT3A, to allow programmable epigenome engineering. In these systems, target gene expression is altered by modification of DNA via methylation/demethylation of gene promoters, or by modification of chromatin structure via acetylation or demethylation of histones. While this approach allows transcriptional activation or suppression of target genes like CRISPRa and CRISPRi systems, they differ from these in their persistence and inheritance. Gene regulation by CRISPRi and CRISPRa systems is reversible, meaning transcriptional modulation by these two systems is alleviated upon the removal of the system components [33]. Gene regulation by epigenetic modifications may not be as easily reversible and may even be inherited by daughter cells [34].

R26-LSL-dCas9-p300 mice created by Gemberling et al. is a murine model that can be used to activate the transcription of target genes via CRISPR epigenome editing. These mice express dSpCas9 fused to the histone acetyltransferase domain of CBP/p300 (p300core) under the control of the CAG promoter in a Cre-dependent manner. Upon Cre-mediated excision of the floxed STOP cassette, dSpCas9-p300core is expressed and co-expression of a sgRNA drives the fusion protein complex to proximal promoters or distal enhancers of target genes where the acetyltransferase catalyzes the acetylation of lysine 27 of histone 3 (H3), thereby promoting robust transcriptional activation of the target genes [30].

2.3.4. RNA interference systems

Regulation of gene expression can also be achieved by promoting RNA downregulation using RNA interference [35] or newly developed CRISPR systems including CRISPR-Cas13 [36, 37] and Cas7–11 systems [38, 39].

CRISPR-Cas13 systems comprise two main components, a large RNA endonuclease and short crRNA. These systems typically recognize ssRNA molecules and cleave them non-specifically at exposed uracil residues. For example, Type VI CRISPR-Cas13a from Leptotrichia shahii (also known as CRISPR-LshC2c2) comprises a large ribonuclease containing two Higher Eukaryotes and Prokaryotes Nucleotide (HEPN) binding domains called C2c2, and a short 54 nucleotide-long crRNA (Fig. 4C) [36]. Rather than targeting double stranded DNA, CRISPR-LshC2c2 recognizes ssRNA molecules and promotes its degradation. Of the 54 nucleotides, 28 residues provide target specificity (antisense to the protospacer, also referred to as target site). The other 26 residues pair with C2c2. Cleavage of the ssRNA is sensitive to the nucleotide composition at the 3’ end of the protospacer (also known as protospacer-flanking site or PFS). Spacers with a A, C or T immediately flanking the 3’ end of the protospacer were cleaved more efficiently that those with a G.

Cas7–11 systems are also single-protein effector systems comprised of a large modular protein containing several Cas7-like domains and a single Cas11-like domain with intrinsic endoribonuclease activity that pairs with a short crRNA to cleave ssRNA at specific locations along the spacer (Fig. 4D) [39]. For example, Type III-E CRISPR-Cas7–11 from Candidatus Scalindua brodae (Sb-gRAMP) contains 4 Cas7-like domains and a single Cas11-like domain that pairs with a 47 nucleotide-long crRNA to cleave ssRNA at position 3 and 9 of the spacer. The 5’ most 27 nucleotides of the crRNA encode the direct repeat segment of the crRNA, whereas the subsequent 20 nucleotides provide target specificity.

2.3.5. Tetracycline inducible systems

Gene expression can also be modulated at the transcriptional level by using the well-established tetracycline inducible systems [40]. Tetracycline inducible systems are dual transgenic systems in which expression of a transgene is dependent on the expression of tetracycline binding TAs, tTA or reverse tTA (rtTA). These TAs are formed from the fusion of the repressor DNA binding protein from the Escherichia coli tetracycline resistance operon transposon Tn10 (TetR) with the transactivating domain of VP16 from Herpes simplex virus. Two tetracycline modulating configurations exist. The Tet-Off configuration makes use of tTA, which binds to tetracycline responsive elements (TRE) located upstream of a target gene (or transgene) and promotes transcription (Fig. 4E). In the presence of tetracycline, tTA can no longer interact with TRE and thus no longer supports expression. Conversely, the Tet-On configuration makes use of the rtTA, which cannot bind the TRE element due to alterations in its amino acid composition (Fig. 4F). In the presence of tetracycline, rtTA can now bind TRE and promote transcription.

3. Allele design

Virtually any type of mutation can be engineered in mice using CRISPR-Cas technologies [11, 41]. These include targeted random insertion or deletion of genetic materials also known as indels; nucleotide substitutions; insertion of large DNA elements; deletion of large DNA fragments; inversion of DNA elements; and translocations (Fig. 5).

Figure 5. Genome modifications.

Schematic representation of the different types of alleles that can be engineered with a single or a pair of cut sites and the engineering systems used to create them. With one cut site, three types of alleles can be produced: targeted random insertion or deletion of genetic material (indels), nucleotide substitutions, and insertion of DNA elements. With two cut sites, three types of alleles can be produced: deletion of DNA elements, inversions, and chromosomal translocations. Introns are shown as light gray boxes, exons as dark gray boxes, start sites as black arrows. Indels are shown as a red box, nucleotide substitutions as a blue box, and insertion of DNA elements as a green box. Black arrowheads indicate cut sites within an intron or exon.

3.1. Indels

Indels result from the repair of CRISPR-mediated DNA lesions via the error-prone NHEJ repair pathway. Insertion or deletion of a few nucleotides may introduce nonsense, missense, silent or frameshift mutations. The insertion of frameshift or nonsense mutations will likely result in inactivation of the allele and degradation of the transcript via the nonsense-mediated mRNA decay pathway [42]. Indels have been used to engineer several knockout mouse models and their characterization at the genomic level is quite simple and involves PCR amplification of the locus and Sanger sequencing [11–13]. Specific primers, based on the genetic modification, can also be designed for routine genotyping [11].

3.2. Nucleotide substitutions

Nucleotide substitutions, such as silent, missense, and nonsense mutations, can be engineered using various technologies including CRISPR-Cas9 systems, CRISPR-Cpf1 systems, base editors and prime editors. Using CRISPR-Cas9 or CRISPR-Cpf1 systems, specific nucleotide substitutions can be engineered via the co-administration of short ssDNA molecules encoding the various mutations flanked by short homology arms of approximately 30 to 50 nucleotides [11–13, 41, 43]. Using prime editing, substitutions are incorporated within the pegRNA [24]. For base editing, substitutions must fall within the editing window and do not require HDR templates. Countless mouse models have been engineered using these technologies. Characterization of these models at the genomic level is straight forward and involves the amplification of the locus by PCR and analysis via Sanger sequencing. Whenever possible, silent mutations encoding a restriction enzyme site can also be engineered to facilitate identification of founder animals and for routine genotyping [41, 43].

3.3. Insertion of DNA elements

Insertion of genetic elements, such as epitope tags, selection markers, reporter genes, site-specific recombinases (SSRs), and SSR sequences (SSRSs) can be engineered by inserting one or two DNA DSBs using CRISPR-Cas systems and the coadministration of HDR templates encoding the desired inserts in the form of short ssDNA molecules, linear dsDNA molecules or plasmids. A single break site is usually used to insert epitope tags either at the 5’ or 3’ end of a coding sequence [11, 41, 43]. Similarly, a single break can be used to insert larger DNA elements, such as SSRs, reporter genes or selection markers[11–13]. Alternatively, entire genes can be removed and replaced by large DNA segments. Such an approach is commonly used to engineer cell type or tissue specific reporter alleles or SSRs where the promoter of a specific gene is used to drive expression [11–13].

Spatiotemporal regulation of gene expression can be achieved using various strategies including the use of Cre-loxP [44] and related systems such as FLP-Frt [45], Dre-rox [46], Vika-vox [47], Nigri-nox [48] and Panto-pox [48], as well as tetracycline inducible systems (Fig. 4 and Fig. 6). The most common strategy to engineer conditional alleles involves the flanking of one or more critical exons of a gene with two SSRSs such that upon recombination between the two SSRSs, the intervening DNA segment is removed, thereby inactivating the gene (Fig. 6B) [11–13]. Other strategies include the use of large artificial introns such as conditionals by inversion (COIN) or CRISPR-FLIP (Fig. 6C and 6D) [49, 50]. In these systems, large artificial introns containing splice donor sites and selection markers flanked with SSRSs are inserted within an exon. In the absence of their cognate recombinase, the large introns are recognized by the splicing machinery and removed, allowing for proper expression (Fig. 6C and 6D). In the presence of the recombinase, the intronic sequence is flipped head to tail, revealing a strong splice acceptor followed by a stop cassette. A simpler version of this strategy, which makes use of a short artificial intron was recently developed and implemented in mice (Fig. 6E) [51, 52]. In this case, the intron, which is 201 nucleotides in length, is inserted within an exon of a gene. The intron contains both a splice donor and a splice acceptor site, as well as a branch point and a polypyrimidine tract flanked by 2 loxP sites. In the absence of Cre, the intron is recognized by the splicing machinery and removed, allowing for proper expression. In the presence of Cre, a fragment of the intron is removed, leaving a string of stop codons in all three frames, thereby inactivating the gene.

Figure 6. Spatiotemporal regulation of gene expression.

A) Site-specific recombination systems and their cognate sequences.

B) Schematic representation of flanking an exon with two loxP sites (flox) such that upon Cre-mediated recombination between the two loxP sites, the intervening DNA segment is removed, and the gene is inactivated. Introns and sequences flanking genes are shown as light gray boxes, exons as dark gray boxes, start sites as black arrows. loxP sites are indicated as black triangles and the creation of a premature stop codon upon Cre-mediated recombination is highlighted as a red box within exon 3.

C-D) Schematic representation of the CRISPR-FLIP and COIN systems, in which a large artificial intron containing splice donor sites and selection markers, or reporter genes flanked with SSRSs are inserted within an exon of a gene. In the absence of Cre, the large introns are recognized by the splicing machinery and removed, allowing for proper expression. In the presence of Cre, the intronic sequence is flipped head to tail, revealing a strong splice acceptor followed by a stop cassette. The splicing of the endogenous splice donor site into the newly revealed splice acceptor site results in expression of the reporter cassette and inactivates the target gene. Introns and sequences flanking genes are shown as light gray boxes, exons as dark gray boxes, start sites as black arrows. loxP sites are shown as black triangles, lox66 sites as red triangles, lox71 sites as orange triangles, lox72 sites as pink triangles, lox5171 sites as purple triangles, and Frt sites as light blue triangles. Splice donors are shown as green boxes, splice acceptors sense of the gene are shown as red boxes, and splice acceptors antisense of the gene are shown as maroon boxes within an intron. The puromycin resistance gene is shown as a blue arrow, the GFP open reading frame (ORF) is shown as a green arrow, the PGK promoter is shown as an orange arrow, and polyadenylation signals are shown as yellow arrows. Small blue and green arrows indicate branch points sense or antisense of the gene, respectively.

E) Schematic representation of the DECAI system, which makes use of a short artificial intron which contains both a splice donor and splice acceptor site, as well as a branch point and a polypyrimidine tract flanked by 2 loxP sites. In the absence of Cre, the intron is recognized by the splicing machinery and removed, allowing for the proper expression of the gene. In the presence of Cre, much of the intron is removed, leaving a string of stop codons in all three frames and inactivating the gene. Introns and sequences flanking genes are shown as light gray boxes, exons as dark gray boxes, start sites as black arrows. The splice donor is shown as a green box, the loxP sites are shown as black triangles, the sense branch point is shown as a small blue arrow, the polypyrimidine tract is shown as a yellow box, and the splice donor is shown as a red box within the intron.

3.4. Deletion of DNA elements

Deletion of genetic material can be used to inactivate genes, remove several contiguous genes or for precisely excising protein domains in vivo (Fig. 5) [41, 53, 54]. Removal of DNA elements can be performed by the coadministration of 2 or more gRNAs together with their cognate endonuclease. Up to several hundred kilobases have been successfully removed from various loci via microinjection of pronuclear stage zygotes with CRISPR-SpCas9 reagents (unpublished). The addition of an HDR template to promote in frame recombination between the two breaks can be used to remove protein domains [41].

3.5. Chromosomal translocations

Chromosomal translocations can also be engineered by the insertion of 2 DNA DSBs (Fig. 5). However, because chromosomal translocations are, in most instances, not compatible with embryonic development [55], translocations are typically engineered via insertion of 2 SSRSs onto distinct chromosomes. Chromosomal translocations can then be driven by expression of SSRs at a later stage of development or in adult animals.

4. Delivery

Various delivery methods have been developed for the engineering of animal models. For transgenic animals, transgenes, in the form of linearized DNA encoding the promoter and a transgene of interest followed by a polyadenylation signal, are typically delivered into pronuclear stage zygotes via microinjection. Typically, approximately 150 zygotes are injected with the transgene. 70% of the zygotes will survive microinjection and transferred to pseudopregnant females. 30% of surviving zygotes will develop and 10 to 20% of those will carry the transgene [1]. Similar efficiency can be achieved when using large DNA constructs such as bacterial artificial chromosomes (BACs) and yeast artificial chromosomes (YACs). As an alternative to microinjection, transgenic mice can also be engineered via lentiviral delivery of transgenes. In this case, 60–70% of mice produced express the transgene [56]. However, special requirements for handling viral particles and mice produced via lentiviral infection must be taken.

Engineering mouse models via conventional gene targeting in ES cells can be achieved via electroporation of mouse ES cells with large DNA constructs containing the desired genetic changes together with positive and negative selection markers flanked by large homology arms. Following electroporation, several hundreds or thousands of clones must be manually isolated to identify a select few with the proper alterations. The advent of CRISPR technology has significantly improved this process and simplified the design of targeting vectors as shorter homology arms suffice to properly target a locus and selection markers are, in most cases, no longer required due to the efficiency at which Cas systems cleave their target sites. This strategy, however, still requires the generation of chimeric animals. Using conventional targeting in ES cells, approximately 1% of clones will be properly targeted and half of these will produce chimeric animals capable of transmitting the allele to their progeny [1]. Using CRISPR technology, approximately 10% of clones will possess at least one allele with the proper genetic alteration. This number varies depending on the complexity of the allele to be engineered.

Given the high efficiency at which CRISPR-Cas systems can promote insertion of mutations, CRISPR reagents can be delivered directly into pronuclear stage zygotes together with HDR templates. Delivery of these reagents can be performed via intracytoplasmic injections of pronuclear-stage zygotes [11–13, 41, 51], or electroporation [57–59]. On average, indels occur in approximately 70%, large in-frame deletions occur in approximately 40%, insertion of small DNA elements occur in approximately 30%, large deletions and point mutations occur in approximately 20%, floxed alleles and double point mutations occur in approximately 10%, large deletions replaced with the insertion of large DNA elements occur in approximately 5%, and insertions of large DNA elements occur in less than 5% of pups produced from microinjection (this data comes from the generation of ~200 mouse models in collaboration between the Pelletier Lab and the St. Jude Children’s Research Hospital Transgenic Core Facility). Electroporation of zygotes can be performed ex vivo, using a method known as CRISPR-EZ, or in vivo, using a method known as iGonad. CRISPR-EZ consists of isolating pronuclear stage zygotes for ex vivo electroporation using conventional electroporation systems. Using the CRISPR-EZ system, indels were observed in up to 100% of pups produced, whereas point mutations occurred in approximately 25% and large deletions occurred in approximately 40% of pups [57]. iGonad, on the other hand, does not require the isolation of embryos as electroporation is performed directly within the oviduct using specialized electrodes known as electrokinetic tweezers [59]. In this system, CRISPR reagents are delivered directly in the oviduct via microinjection. Using the iGONAD system, indels were found to occur in nearly 100% of pups produced, whereas point mutations and large deletions occurred in approximately 50% of pups produced [59]. While electroporation strategies have been shown to efficiently introduce simple modifications, insertion of larger DNA fragments at a precise location within a genome has proved to be more difficult. A recently developed strategy, known as CRISPR-READI, uses a combination of electroporation and delivery of HDR templates via viral delivery [58]. Using this method, zygotes are first electroporated and subsequently infected with AAVs containing an HDR template. From pups produced using the CRISPR-READI system, approximately 90% had indels, 70% harbored point mutations, and 25% were found to bear large insertions.

Numerous plasmids encoding CRISPR-Cas systems have been developed and are available for purchase from Addgene (www.addgene.com). These include conventional expression vectors expressing various Cas effectors with or without their associated crRNAs or sgRNAs. For the generation of mouse models using CRISPR-assisted gene targeting in ES cells or direct delivery into pronuclear stage zygotes, CRISPR-Cas systems can be delivered in the form of RNA transcripts encoding the effector proteins together with RNA transcripts encoding the crRNAs or sgRNAs. Alternatively, Cas effector proteins can be produced in vitro (or purchased from various vendors) and combined with their cognate crRNA or sgRNA to produce ribonucleoprotein (RNP) complexes.

5. CRISPR-Cas technologies in orthopedic research

The impact of novel methods for engineering animal models, in particular CRISPR-based strategies, has made its mark in the field of orthopedic research. The ability to inactivate genes, modify enhancer sequences, promote gene expression, or engineer disease-associated mutations, have all deepened our understanding of the physiological and pathophysiological processes involved in bone and muscle development, homeostasis, and disease (Table 1).

Table 1.

CRISPR-engineered models in orthopaedic research.

Gene	Mutation	Type	Method	Summary
Atp6v1h	Deletion of 5 nucleotides within exon 2	Indel	Conventional CRISPR-Cas9 using 1 sgRNA	Mice deficient for Atp6v1h gene exhibited low bone mass due to decreased bone formation [60]
Lamp2ac	Deletion of Exon 9	Large deletion	Conventional CRISPR-Cas9 using 2 sgRNAs	Deletion of both Lamp2 A and C isoforms resulted in lower vertebral cancellous bone volume in adult mice [61]
Bglap/Bglap2	Deletion of Bglap exon 2 up to Bglap2 exon 4.	Large deletion	Conventional CRISPR-Cas9 using 3 sgRNAs	Unlike previously described Bglap/2deficient mice, mice bearing the Bglap/2 deletion do not show bone mass reduction nor several other phenotypic changes originally reported. [63]
Slc7a7	Deletion of exons 3 and 4	Large deletion	Conventional CRISPR-Cas9 using 2 sgRNAs	Mice deficient for Slc7a7 exhibit delayed skeletal development and delayed development in the kidneys, the lungs and liver [64]
Ghrh	Partial deletion of exons 2 and 3	Large deletion	Conventional CRISPR-Cas9 using 2 sgRNAs	Ghrh null mice show significant decrease in lean mass, bone mineral content and density, and increase fat mass [62]
Npy	Conditional allele by flanking exons 2 and 3 with loxP sites	Floxed allele	Conventional CRISPR-Cas9 using 2 sgRNAs and 1 single donor template	The tissue-specific removal of neuropeptide Y (NPY) in various bone compartments was used to illustrate its role in the development and homeostasis of the mammalian skeleton [65]
Tnfrsf11b	Conditional allele by flanking exon 2 with loxP sites	Floxed allele	insertion of two DSBs flanking exon 2 of the Tnfrsf11b gene and insertion of loxP sites using two HDR templates	Mice bearing the conditional allele were used to identify functionally relevant cellular sources of OPG in vivo [66]
Fgf23	Deletion of regulatory elements	Indels	Conventional CRISPR-Cas9 using multiple sgRNAs	Removal of specific enhancers within the promoter of Fgf23 modulates Fgf23 expression in response to a variety of factors including inflammatory inducers like LPS, IL-1 beta and TNF alpha, Phosphate and Vitamin D [67, 68]
Cyp27b1	Indels in regulatory elements	Indels	Conventional CRISPR-Cas9 using multiple sgRNAs	Deletion of enhancer modules within introns of genes upstream of Cyp27b1 has severe effects on Cyp27b1 regulation and skeletal phenotypes similar to those of Cyp27b1-null mice [69]
Vdr	Several hundred base pair deletions	Large deletions	Conventional CRISPR-Cas9 using 2 sgRNAs (for each deletion)	In vitro and in vivo studies illustrated the diverse nature of VDR binding activity on Vdr gene expression [70]
Notch3	Insertion of stop codon	Insertion of a nonsense mutation	Conventional CRISPR-Cas9 using 1 sgRNA and 1 HDR template	truncation in mice by engineering a premature stop codon, mimicking the human NOTCH3 variant [71]
Mab21l2	R51C missense mutation	Insertion of a missense mutation	Conventional CRISPR-Cas9 using 1 sgRNA and 1 HDR template	Similar to humans with R51C missense mutation, mice heterozygotes for the mutation develop skeletal abnormalities including joint fusions and eye abnormalities [72]
Dio2	SNP	Insertion of single nucleotide polymorphism	Conventional CRISPR-Cas9 using 1 sgRNA and 1 HDR template	Dio2 SNP provides protection against early onset osteoarthritis [73]
Wnt10bm	Increased expression	N/A	CRISPRa with expression cassettes encoding 3 sgRNAs	CRISPR engineered stem cells were shown to have increased osteogenic potential and improve calvarial bone healing upon implantation to calvarial defects [74, 75]
Foxc2	Increased expression	N/A	CRISPRa with expression cassettes encoding 3 sgRNAs	CRISPR engineered stem cells were shown to have increased osteogenic potential and improve calvarial bone healing upon implantation to calvarial defects [74, 75]
DANCR	Increased expression	N/A	CRISPRa together with multiple sgRNAs	CRISPR engineered stem cells were shown to have increased osteogenic potential and improve calvarial bone healing upon implantation to calvarial defects [74, 75]
Noggin	Increased expression	N/A	CRISPRi with expression cassettes encoding 2 sgRNAs	CRISPR engineered stem cells were shown to have increased osteogenic potential and improve calvarial bone healing upon implantation to calvarial defects [76]
Sox9	Increased expression	N/A	CRISPRi with expression vector encoding 1 sgRNA	CRISPR engineered stem cells were shown to have increased osteogenic potential and improve calvarial bone healing upon implantation to calvarial defects [77]
Pparg	Increased expression	N/A	CRISPRi with expression vector encoding 1 sgRNA	CRISPR engineered stem cells were shown to have increased osteogenic potential and improve calvarial bone healing upon implantation to calvarial defects [77]
Tnfsf11		N/A	Transgene expressing CRISPRi and 1 sgNRA (selected from 3 sgRNAs tested in vitro)	Mice with high transgene expression exhibited phenotypic changes similar to those associated with the deletion of Tnfsf11 gene [78]
Lmna	Correction of c.1824 C>T mutation	Point mutation	Base editors	Missense correction restored LmnA expression [79]
Dmd	Correction of a nonsense mutation	Point mutation	Base editors	Missense mutation correction restored Dmd expression [80]

CRISPR-based technologies have significantly reduced the labor, cost, and time requirement for production of in vivo loss of function models. CRISPR-Cas9 genome editing is becoming the main genetic approach to produce global loss-of-function models. The role of Atp6v1h in bone homeostasis, for example, was illustrated by the generation of a small five base pair deletion introduced by CRISPR-Cas9. Mice deficient for Atp6v1h gene exhibited low bone mass due to decreased bone formation [60]. The role of chaperone-mediated autophagy (CMA) in skeletal homeostasis under physiological conditions was studied using Lamp2ac knockout mice. Mice deficient for both Lamp2 transcript variant A and C was accomplished by introducing DNA DSBs flanking exon 9 of the Lamp2 gene. Deletion of both isoforms resulted in lower vertebral cancellous bone volume in adult mice which was likely due to an inhibitory role of CMA on osteoclast genesis or a positive role of CMA in osteoblast formation and/or function [61]. Other examples of global loss-of-function models used for skeletal research were created by CRISPR-Cas9 mediated deletion of essential exons of Bglap, Slc7a7 and GHRH [62–64].

The engineering of mice bearing conditional alleles using CRISPR-Cas9 technology has also been reported. Conditional alleles in mice, as discussed above, can be engineered using various strategy including CRIPSR-assisted gene targeting in ES cells. The tissue-specific removal of neuropeptide Y (NPY) in various bone compartments was used to illustrate its role in the development and homeostasis of the mammalian skeleton and highlighted the complex nature of NPY signaling. In this case, the mouse model was engineered using CRISPR-assisted gene targeting in ES cells. More specifically, a large DNA construct containing loxP sites flanking exons 2 and 3 was built using recombineering and co-administered with a sgRNA targeting intron 2 of the NPY gene [65]. More recently, a mouse model bearing a conditional allele of Tnfrsf11b, which encodes osteoprotegerin (OPG), was engineered via direct administration of CRISPR reagents into pronuclear stage zygotes. The strategy involved the insertion of two DSBs flanking exon 2 of the Tnfrsf11b gene and insertion of loxP sites using two HDR templates in the form of ssDNA molecules. Mice bearing the conditional allele were used to identify functionally relevant cellular sources of OPG in vivo [66].

CRISPR-Cas9 technology has also been utilized as a tool to study transcriptional regulation in vivo. Recent advances in genome-wide epigenetic studies have allowed us to appreciate complex regulation of gene transcription via multiple distal regulatory elements. Transcriptional regulatory elements can be identified by epigenetic marks in genome-wide epigenetic studies. However, deletion of these regulatory elements from the mouse genome is necessary to address their roles in transcriptional regulation and examine how transcriptional regulation of genes play a role in physiology and pathophysiology of skeleton. While such studies were previously performed by traditional methods, the use of CRISPR-Cas9 genome-editing has made these studies cost-and time-efficient, thereby allowing expansion of such studies. For example, CRISPR-Cas9 technology was used to produce 4 murine models each lacking a potential regulatory region of Fgf23 gene [67, 68]. These studies demonstrated the complex regulation of Fgf23 expression in response to phosphate, 1,25-Dihydroxyvitamin D3, and lipopolysaccharides (LPS). Additionally, similar approaches that use CRISPR-Cas9 genome-editing have been undertaken to examine transcriptional regulation of Cyp27b1 [69] and Vitamin D Receptor [70] and have expanded our understanding of calcium-phosphate metabolism.

To better understand disease pathogenesis, disease causing variants identified in humans can be engineered in animal models using CRISPR-Cas technologies, which provide increased editing precision compared to conventional editing. An example of this was the engineering of Notch3 truncation in mice by engineering a premature stop codon, mimicking the human NOTCH3 variant. Similar to patients with NOTCH3 truncating mutations, mice bearing this premature stop codon developed osteopenia [71]. Another interesting example is the engineering of mice bearing the R51C missense mutation in the Mab21l2 gene. Human patients with MAB21L2 R51C mutation develop eye abnormalities, intellectual disabilities and skeletal abnormalities including dysplasia. Similar to humans affected by this rare disease, mice heterozygotes for the mutation develop skeletal abnormalities such as joint fusion between the humerus and radius as well as the femur and tibia. These mice also presented with eye abnormalities, as seen in human patients, and failed to develop properly [72]. CRISPR/Cas9 genome editing can also be utilized to study SNP function in vivo. For example, in a recent publication, Butterfield and colleagues have utilized a murine model in which an osteoarthritisassociated polymorphism in the Dio gene, which is orthologous to the human variant, was recapitulated via CRISPR/Cas9. These studies showed that this Dio SNP provides protection against early onset osteoarthritis [73].

CRISPRa and CRISPRi have also been applied to basic orthopedic research, mostly for in vitro and ex vivo studies. Several calvarial defect studies have used these methods to increase the osteogenic potential of mesenchymal stem cells (MSCs). For example, CRISPRa was to induce expression of endogenous Wnt10bm, Foxc2 and DANCR of mesenchymal stem cells (MSCs) [74, 75]. Another study used CRISPRi to suppress Noggin expression of MSCs which also overexpress BMP2 [76]. In a fourth study, CRIPSRa and CRISPRi were simultaneously used in MSCs to activate expression of Sox9 and inhibit expression of PPAR-γ, respectively [77]. In all four studies the CRISPR engineered stem cells were shown to have increased osteogenic potential and improve calvarial bone healing upon implantation to calvarial defects.

CRISPRi was recently used for in vivo loss-of-function (LOF) studies in bone. In a proof-of-concept experiment, a single transgene broadly expressing a sgRNA targeting Tnfsf11 and dCas9::KRAB fusion protein was shown to effectively suppress Tnfsf11 expression in vivo. In this global CRISPRi model, mice with high transgene expression exhibited phenotypic changes similar to those associated with the deletion of Tnfsf11 from the mouse genome [78]. This approach was further refined by producing two knock-in mouse models to control Tnfsf11 gene expression. The first model expresses the dCas9::KRAB fusion under the control of Dmp1 regulatory elements and the other expresses a sgRNA targeting the promoter of the Tnfsf11 gene under the control of the U6 promoter. These CRISPRi mice exhibited phenotypic changes similar to those observed in mice lacking Tnfsf11 in Dmp1-Cre expressing (unpublished studies by Onal and colleagues).

Base editors have been used to correct pathogenic mutations in mouse models of monogenic diseases such as Hutchinson-Gilford progeria syndrome (HGPS) [79] and Duchenne muscular dystrophy (DMD) [80]. In a mouse model of HPGS which harbors the human LMNA c.1824 C>T mutation, base editors were delivered via a single retro-orbital injection of AAV9 encoding ABE7.10max-VRQR and an sgRNA targeting the pathogenic locus. This resulted in efficient correction of the pathogenic mutation, restoration of normal RNA splicing, and reduction of progerin, the toxic protein produced from mis-splicing of LMNA. Similarly, delivery of a split ABE gene, encapsulated in trans-splicing AAVs, into muscle cells in a mouse model of DMD resulted in the correction of a nonsense mutation in the Dmd gene and restored dystrophin expression, demonstrating the therapeutic potential of base editing. It is unclear, however, whether correcting the mutation resulted in improved muscle function [80].

6. Conclusion & perspectives

While CRISPR-Cas based technologies for genome manipulation have simplified the process of engineering animal models, a number of issues remain that must be addressed. These include the need to improve the efficiency at which intended genetic alterations are engineered and to limit off-target editing.

To improve desired on-target editing efficiency, several approaches may be envisioned. One of these approaches would be to limit or even prevent NHEJ activity in target cells or embryos. To this end, small molecule inhibitors or RNA interference targeting Ku70/Ku80, DNA-PK, and/or LigaseIV could be co-administered during the engineering process. Such an approach has been previously shown to be successful in cultured cells [81, 82] but, to our knowledge, has not been applied to engineering animal models. Complementary to inhibiting components of the NHEJ pathway, strategies to promote HDR can also be employed. One such strategy would be to co-administer RNA transcripts or cDNAs encoding various components involved in HR or HDR, such as Rad51, Brca2, and RPA1–4. Such a strategy has been used in cultured cells where co-administration or fusion to Cas9 of homologous recombination factors improved on-target editing [83–86]. Another approach to improve on-target editing efficiency is the use of Cas9 shuttles to facilitate the delivery of HDR templates into the nucleus. Cas9 shuttles are 23 nucleotide-long DNA sequences added to both ends of a dsDNA donor template composed of the PAM sequence preceded by 16 nucleotides corresponding to the target sequence. The remaining 4 nucleotides are altered such that they do not pair with the sgRNA, allowing for the Cas9 RNP complex to bind the shuttle but not enable cleavage by Cas9 itself; this promotes translocation of the HDR template into the nucleus. In cultured cells, the use of Cas9 shuttles was found to increase the efficiency of HDR by up to four-fold [87].

While these strategies may be used to improve the overall efficiency of on-target editing, new CRISPR-Cas systems have been developed to promote the integration or deletion of large DNA elements, which remains somewhat challenging using conventional genome editing systems. These include Cascade-Cas3 systems, which make use of the RNA-guided DNA helicase-nuclease Cas3 that is capable of removing large DNA segments [88], and CRISPR-associated transposases (CAST), which allow RNA-guided insertion of large DNA cargos to any location within a genome [89, 90]. While these technologies may aid with the generation of large insertions or deletions of DNA elements, both have limitations. For Cascade-Cas3, the main drawback of this technology is the apparent uncontrollable processivity of the helicase-nuclease Cas3, which can create deletions ranging from several hundred to several thousand base pairs. Not only does this make it so that this technology cannot be used to create precise deletions, but it also makes the characterization of models using this technology more problematic. The main drawback of using CRISPR-CAST systems, on the other hand, is that this technology does not allow for scarless or in-frame integration of genetic material. Consequently, this technology cannot be used for precise DNA insertion, but could be employed to facilitate the insertion of transgenes at safe harbor sites. While Cascade-Cas3 and CRISPR-CAST systems have not yet been used for the generation of animal models, the implementation of these systems for their use in mammalian genome engineering is imminent and provides an alternative to conventional CRISPR-Cas systems.

To limit off-target editing, several strategies have been established, including bioinformatic tools to identify highly specific target sequences; the modification of SpCas9 to improve its specificity or duration of action; and the development of delivery methods and formats to limit the duration of SpCas9 activity. Highly selective guide sequences can be identified through the use of guide selection applications, including Cas-Designer, quick guide-RNA designer for CRISPR/Cas derived RNA guided nucleases (http://www.rgenome.net/); CRISPR Design (http://crispr.mit.edu/); E-CRISP (http://www.e-crisp.org/E-CRISP/); and ZiFit (http://zifit.partners.org/ZiFiT/). While implementing CRISPR-SpCas9 technology for mouse genome editing, we developed a stringent guide selection procedure which makes use of Cas-Designer and Cas-Offinder from CRISPR RGEN Tools (http://www.rgenome.net/) [11, 41]. To control the duration of SpCas9 activity, genetically encoded inducible systems have been developed [91]. Self-cleavable CRIPSR systems promote SpCas9 degradation upon its expression through the use of a sgRNA targeting sequences encoding the endonuclease [92]. Alternatively, the duration of SpCas9 activity can be controlled by delivering components of the CRISPR-SpCas9 system in the form of RNA transcripts or RNP complexes, which have a much shorter half-life compared to the use of DNA plasmids. Various additional strategies have also been developed to improve target specificity. These include the use of paired SpCas9 nickases that introduce scattered DNA DSBs, guided by a pair of sgRNAs recognizing juxtaposed sequences, which doubles the length of the target sequence and thus increases target specificity [93]. Similarly, pairs of dSpCas9 fused to the non-specific endonuclease FokI have been shown to reduce off-target activity [94]. SpCas9 target specificity has also been increased through the use of directed evolution to engineer improved enzymes that have been shown to increase on-target over off-target activity by several folds [95, 96]. Finally, the modification of gRNAs themselves has also been used to reduce off-target cleavage. Although counterintuitive, previous studies have shown that shortened guide sequences increase specificity without diminishing on-target activity [97].

The field of genome editing is expanding and evolving rapidly. The application of both conventional and emerging CRISPR-based technologies has already begun to advance our understanding of fundamental biology and holds great promise for the treatment of monogenic musculoskeletal disorders.

Highlights.

This review article describes established and emerging technologies used to engineer animal models with an emphasis on bone and skeletal muscle biology.

Both conventional and emerging CRISPR-Cas-based technologies are discussed.

Novel targeting strategies and allele designs are also presented.

Abbreviations:

HR

homologous recombination

ES

mouse embryonic stem

DSBs

DNA double strand breaks

CRISPR

Clustered Regularly Interspaced Short Palindromic Repeats

Cas9

CRISPR-associated protein 9

HDR

Homology directed repair

ZFNs

zinc finger nucleases

TALE

transcription activator-like effector

TALENs

nucleases

sgRNA

single guide RNA

NLS

nuclear localization site

CBEs

cytidine base editors

ABEs

adenine base editors

CGBEs

cytidine to guanine base editors

RT

reverse transcriptase

CRISPRa

CRISPR activation

CRISPRi

CRISPR interference

TA

Transcriptional activator

TR

Transcriptional repressor

COIN

Conditionals by inversion

OPG

steoprotegerin

MSCs

Mesenchymal stem cells

NPY

Neuropeptide Y

CRISPR-FnCpf1

Class 2 type V CRISPR-Cas system from Francisella novicida

CRISPR-SpCas9

Class 2 Type II CRISPR-Cas9 system from Streptococcus pyogenes

CD

Cytidine deaminase

TadA*

tRNA adenine deaminase

eUNG

Uracil DNA N-glycosylase

pegRNA

Prime editing guide RNA