Regulating CRISPR/Cas9 Function through Conditional Guide RNA Control

Abstract

Conditional control of CRISPR/Cas9 has been developed using a variety of different approaches, many focusing on manipulation of the Cas9 protein itself. However, more recent strategies for governing CRISPR/Cas9 function are based on guide RNA (gRNA) modifications. They include control of gRNAs by light, small molecules, proteins, and oligonucleotides. These designs have unique advantages compared to other approaches and have allowed for precise regulation of gene editing and transcription. Here, we discuss strategies for conditional control of gRNA function and compare effectiveness of these methods.

Graphical Abstract

Different strategies have been developed for the conditional control of guide RNA function in the CRISPR/Cas9 system. Both external and cellular triggers, including light, small molecules, proteins, and oligonucleotides, were utilized for convenient and precise regulation of genome editing and transcriptional activation in cells. These methods provide increased precision for gene editing with spatial and temporal resolution and may find future therapeutic applications.

1. Introduction

The clustered, regularly interspaced, short palindromic repeat (CRISPR)/CRISPR-associated protein 9 (Cas9) endonuclease is part of a microbial adaptive immune system used by bacteria and archaea against invading genetic elements.^[1] The Cas9 effector protein functions by first recruiting CRISPR RNA (crRNA) and trans-activating crRNA (tracrRNA) to form an active DNA surveillance complex. These two RNAs have been combined into a convenient single guide RNA (gRNA).^[2] gRNA binding results in a major structural rearrangement of Cas9 for target DNA recognition.^[3] The recognition process is directed by the search for a protospacer adjacent motif (PAM, 5’-NGG-3’) in the DNA sequence. Watson-Crick base pairing between the DNA adjacent to the PAM and the 5’−20-nt spacer sequence of the gRNA confers specificity of genomic localization.^[4] Once bound to target DNA, Cas9 undergoes a further conformational change that directs the HNH nuclease domain to the target strand of DNA and the RuvC nuclease domain to the complementary DNA strand, producing a predominantly blunt-ended double-strand break at a specific site 3 bp from the NGG PAM sequence.^{[5, 2b]} Since its initial discovery^[6] and biochemical characterization,^{[5b, 2b]} CRISPR/Cas9 has been harnessed for precise genome editing in mammalian cells,^[7] engineered for higher editing efficiency^[8] and precision,^[9] and redesigned for broader PAM compatibility.^{[9c, 10]} It has also been repurposed for transcriptional modulation,^[11] epigenetic modification,^[12] genome labelling,^[13] and base editing.^[14]

Among the abovementioned advances, initiating and/or terminating CRISPR/Cas9 function with temporal and spatial precision have been of particular interest, and systems for the conditional control of its function by small molecules, light, and other physiological signals have been developed.^[15] Confining genome manipulation capability to a defined period of time not only helps reduce off-target effects and genotoxicity due to prolonged nuclease activity,^[16] but also provides a vital tool for the interrogation and dissection of spatially and temporally dynamic regulatory networks, as seen in embryonic development and tumorigenesis.^[17] Acutely activated CRISPR activation/interference (CRISPRa/i)^[11] would play a pivotal role in addressing similar biological questions.

Several methods for conditional control of the Cas9 protein have been developed, such as small molecule-induced activation of the Cas9 protein^[18] or the recruitment of transcriptional modulators to catalytically dead Cas9 (dCas9).^[19] Optical control includes triggering of photochemically caged Cas9,^[20] reconstitution of a split-Cas9,^[21] switchable Cas9 activation based on dimerization,^[22] and also the recruitment of transcriptional modulators to dCas9.^[23] Though these efforts have greatly enhanced the CRISPR/Cas9 gene editing toolbox, the recent developments of conditionally controlled gRNA have further broadened CRISPR/Cas9 technology. As a mediator between Cas9 and the target DNA, engineered gRNAs offer direct manipulation of the searching, recognition, and binding of dsDNA. We hereinafter review the recent progress in the conditional control of gRNAs for regulating CRISPR/Cas9 function.

2. Light-control of gRNA function through optical decaging

Light has been harnessed as a powerful tool to study and manipulate biological processes, and provides acute, non-invasive, and precise control with high spatiotemporal resolution.^[24] Even though different types of photo-inducible CRISPR/Cas systems have been reported, they inevitably involve intricate engineering efforts, including directed evolution,^[18e] the screening of mutations and split sites,^{[21, 18f]} or expansion of the genetic code.^[20] As an alternative approach, optical control of the gRNA has been independently reported by the Bhatia group,^[25] the Deiters group,^[26] and the Stevens group.^[27] All of these designs followed a similar general strategy: masking the spacer region until light-triggered photolysis releases the native gRNA and restores target hybridization (Figure 1a). The Bhatia group introduced a method that relies on a photocleavable “protector” ssDNA complementary to the spacer region of the gRNA, blocking its hybridization with the target dsDNA until 365 nm irradiation induces cleavage of the photolabile linkers in the protector. This lowers the melting temperature (T_m) between the protector and gRNA, thus unveiling the native gRNA for target surveillance (Figure 1b). Varying lengths and positions of the linkers in the protector were screened by a DNA cleavage assay. A 24-nt ssDNA with photocleavable linkers spaced 6-nt apart was found to completely block DNA cleavage in the dark and fully restore CRISPR/Cas9 nuclease activity after rapid light activation. Silencing of EGFP in Cas9-expressing HeLa cells was demonstrated by inducing random insertions or deletions (indels) in the EGFP gene, resulting in a 2-fold increase in the fraction of genetically modified Cas9/EGFP-expressing HeLa cells after light activation. However, undesired background editing of EGFP in the dark compared to a no gRNA control (7.4% versus 1.7% indel rate) was observed. With 15 min exposure to 365 nm light, 78% editing capability was restored as compared to a native gRNA control. The authors attributed the background activity in the dark to possible spontaneous disassembly of the protector-gRNA hybrid in cells. While perfect off-to-on switching was not achieved, this approach constituted the first advance in optical triggering of gRNA function. The nitropiperonyloxymethylene (NPOM) caging group has found extensive application in the control of nucleic acid function.^[28] In 2020, the Deiters group and the Stevens group independently published strategies utilizing NPOM-installation at specific positions across the 20-nt spacer sequence for blockade of gRNA:dsDNA hybridization until photolysis of the caging group restored nucleic acid base-pairing.^[29] The Deiters group introduced NPOM-caged uridines and guanosines spread evenly across the spacer region (Figure 1c) of a single gRNA molecule, based on the finding that incorporation of NPOM groups every 5–6 bases of a DNA oligonucleotide completely abolished base-pairing with the complementary strand.^[30] The introduction of two caged oligonucleotides (NPOM-U and NPOM-G) enables broad applicability to a wide range of target sequences. The Deiters group demonstrated that the Cas9:caged gRNA complex was preserved, and that binding of the ribonucleoprotein (RNP) complex to the target DNA was completely blocked until light exposure rapidly reinstated this interaction. The RNP can be conveniently preassembled and delivered into cells and animals.

Figure 1.

a) General method for optical control of gRNA function. b) Chemical structure of the photocleavable linker (shown in red) and the decaging process of the CRISPR-plus strategy, with photocleavable linkers represented by red rectangles. Adapted from [15c] with permission from Wiley-VCH. c) Structure of NPOM-caged uridine and guanosine and the decaging process of caged gRNA, with NPOM-caging groups shown in red. d) Structure of NPOM-caged thymidine and the decaging of the caged crRNA:tracrRNA complex, with NPOM-caging groups shown in red. While their overall designs were similar, the Stevens group distributed the two caged thymidines evenly throughout the spacer, while the Ha group localized them in the PAM-distal end of the spacer. e) Structure of DMNEC-caged ribose and the decaging process of gRNA, with the DMNEC caging group shown in red. f) Representative images of zebrafish embryos microinjected with Cas9 and caged gRNA targeting the slc24a5 gene, with one kept in the dark and one irradiated with 365 nm light. Adapted from [26] with permission from Wiley-VCH. g) Representative images of zebrafish embryos microinjected with Cas9 and caged gRNA targeting the slc45a2 gene, with one kept in dark and one irradiated by a 405 nm laser in a single eye. Adapted from [27] with permission from ACS Publications.

This not only increases gRNA stability,^[31] but also enables the use of cationic lipids^[32] or injection for easy cellular delivery across different species. The Deiters group then achieved optical triggering of gene editing in HEK293T cells with spatial resolution. They also temporally controlled editing of the slc24a5 gene important for pigment development at different stages of zebrafish embryonic development with high off-to-on ratios, producing zebrafish embryos with significant loss of pigment (Figure 1f, 3.1% indel rate in the dark versus 75.4% with a 5 min irradiation, which is comparable to the 87.3% indel rate achieved with non-caged gRNA). Varying the duration of light exposure elicits tunable gene editing in zebrafish embryos (14.6% indel rate for 30 sec, 27.9% for 1 min, and 75.4% for 5 min of irradiation).

Simultaneously, the Stevens group reported a similar approach by installing two NPOM-caged thymidines in lieu of regular uridines into the spacer region of a crRNA (Figure 1d), based on previous reports that crRNAs tolerate deoxynucleotide substitutions in the spacer, provided that the A-form-like helix structure of RNA is not altered.^[33] Complexed with its cognate tracrRNA, the caged crRNA functions similarly as the caged gRNA developed by the Deiters group and can be readily decaged by 365 or 405 nm irradiation for 5 min without any cytotoxicity observed. The Stevens group chose to utilize the two shorter RNAs for synthetic reasons. However, the Cas9:crRNA:tracrRNA ternary complex may not offer the same ease-of-use (pre-annealing of crRNA and tracrRNA) and editing efficiency as compared to a single gRNA.^[34] Stevens first applied this strategy in HEK293FT cells in the form of an RNP, achieving restoration of over 70% of the gene-editing efficacy, as compared to non-modified gRNAs at multiple genomic loci. These authors then applied the caged crRNA in zebrafish embryos as well, demonstrating temporally controlled editing of the slc45a2 gene at different timepoints after fertilization. They also demonstrated spatially precise induction of an albino phenotype in one of the eyes (Figure 1g). However, even though mismatch-based mutation assay of the targeted locus was assessed by gel, indel rates were not reported by Stevens. The versatility of optically controlling nucleic acid hybridization also allowed for spatially defined triggering of a dCas9-transcriptional activator, inducing patterned expression of a fluorescent protein.

Later in the same year, the Zhou group revealed a different strategy for masking gRNA activity through carbonate formation at the 2’-OH of in vitro transcribed gRNA with nitrobenzyl groups (Figure 1e).^[35] However, the exact number of DMNEC-protection moieties per gRNA was not determined. The 2’-OH caged gRNA was unable to induce Cas9 target DNA cleavage until 365 nm light activation. Targeting a genomic locus in Cas9-expressing HeLa cells showed only 13.2% indel formation in the HBEGF gene after 6 minutes of UV irradiation. Indel frequencies before light exposure and with non-modified gRNA were 0% and 24.2%, respectively. The Zhou group also extended the application of their DMNEC-caging strategy to crRNA-guided RNA cleavage by Cas13a, thereby demonstrating the general applicability of photocaged guide RNAs in other CRISPR/Cas systems. Gel shift assays carried out with catalytically inactive Cas13a and DMNEC-caged crRNA suggest that acylation of multiple 2’-hydroxy groups abolishes formation of the RNP complex until light-triggered decaging.

Recently, the Ha lab applied NPOM nucleobase-caging of gRNAs for fast introduction of double strand breaks at second timescales. The group replaced two or three uracils in the PAM-distal end of the crRNA with NPOM-caged thymidines. The PAM-proximal portion of the crRNA could still bind its target, but nuclease activity was inhibited until the PAM-distal nucleobases were decaged and full crRNA-target hybridization was restored. Using a similar design to Deiters and Stevens, the Ha group uniquely exploited the fast NPOM photolysis to achieve precise gene editing on a single allele with only 30 sec of UV irradiation.^[36] Their results shed light on the kinetics of endogenous DNA repair factors 53BP1 and H2AX with high resolution in space, time, and genomic coordinates. This approach broadened the application of photocaged gRNA for interrogation of DNA repair dynamics, characterizing multi-cycle 53BP1 repair foci formation and dissolution, with the first cycle being longer than the subsequent. Additionally, caging group placement in different localtions of the 20 nt spacer region could be contemplated for either disrupting binding to the DNA or simply blocking target cleavage, as inspired by Ha’s design principle of controlling DNA cleavage by localizing NPOM caging groups in the PAM-distal region. The PAM-proximal 9–10 nt region of the gRNA governs DNA binding while mismatches in the PAM-distal region (10–20 nt) abolish target cleavage through inteference with DNA unwinding and Cas9 conformational changes.^[37]

A recent study by Tang revealed that the installation of a vitamin E moiety to the 5’-terminus of crRNA with a nitrobenzyl photocleavable linker also conferred optical control over the interaction between RNP and target dsDNA.^[38] Targeting the vegfa gene with photocaged crRNA in HEK293T cells, the Tang group was able to achieve a 45.5% indel rate after 4 min of 365 nm UV irradaition, compared to 3.1% editing efficiency in the absence of light and 60% in case of non-modified crRNA. Based on these results, we envision that a commercially available, 5’-amine-modified crRNA could be blocked with other sterically demanding groups as well, such as a photocleavable PEG,^[39] circumventing the need for custom phosphoramidite synthesis.

The advances by the Bhatia, Deiters, Stevens, Ha, and Tang groups share a practical feature in that the optically controlled oligonucleotides are readily produced through either commercial or in-house solid-phase nucleic acid synthesis, providing broad accessibility to these approaches and universal applicability to a wide range of target DNA sequences. Depending on the caging approach, employment of Cas9:caged gRNA RNPs were possible, facilitating application in a wide range of cells and organisms. Importantly, high recovery rates of gene editing efficacy with light-activated caged gRNA were observed, as evidenced by indel frequencies, and spatially and temporally restricted Cas9 function was demonstrated.

3. Small molecule-control of gRNA function through chemical decaging

Similar to light-induced decaging of gRNA, small molecule-induced decaging allows for temporal activation of CRISPR/Cas9 function by treatment with an external stimulus. Though lacking the localized activation provided by light, and requiring consideration of toxicity, membrane permeability, and stability, small molecules enjoy other advantages: easy application without specialized laboratory equipment, penetration into deep and opaque tissues, and dose responsiveness. Staudinger reduction of aryl azides triggered by phosphines has been proven to function in living organisms with high specificity, efficiency, and bioorthogonality^[40] and has been recently utilized by the Kool group^[41] and the Zhou group^[42] for deprotection of acylated gRNA. In this approach, multiple ribose 2’-OH groups of a gRNA are first randomly acylated by an azide-substituted acyl imidazole reagent, resulting in blocked RNA folding^[43] and disrupted Cas9-gRNA interaction. The covalently attached azidomethylnicotinyl (AMN; the pyridine ring enhances water solubility) groups are then efficiently removed upon treatment with a phosphine (R₃P) through Staudinger reduction (Figure 2a). Here, the primary amine undergoes cyclization and promotes deacylation, releasing functional gRNA, as well as lactam and phosphine oxide byproducts.^[43] The caging reaction through acylation can be conveniently carried out with in vitro transcribed gRNA, with up to 8 AMN-modified 2’-OH groups within a 102-nt gRNA. Several phosphines were screened by both groups for optimal decaging efficiency (Figure 2b). By testing cleavage of target DNA in vitro, Kool revealed that THPP and DPBS could quantitatively restore caged gRNA activity to the level of native gRNA, while Zhou showed that DPBM completely reinstated caged gRNA activity, but DPBA and DPBS were not as efficient, matching earlier reports.^[44] In GFP-expressing HeLa cells co-transfected with the acylated gRNA and Cas9 mRNA, Kool achieved silencing of GFP in ~80% of cells with 1 mM of DPBS for 17 hours, compared to 95% by non-modified gRNA (Figure 2c). No background activity in the absence of phosphine trigger was detected, and recovery was largely complete after 6 hours. In the meantime, the Zhou group demonstrated that 200 μM of DPBM triggered ~70% recovery of gene editing (represented by indel rate in the targeted hbegf gene) as compared to non-modified gRNA (Figure 2d) in HeLa-OC cells stably expressing Cas9 protein. Furthermore, the Zhou lab also showed that DPBM treatment could be shortened to 2 hours without reduced effectiveness of chemical decaging.

Figure 2.

a) General method for small molecule control of gRNA function through chemical decaging. The AMN caging group is shown in red. b) Structures of phosphines (R₃P) tested for gRNA activation. c) Gene editing efficiencies (from flow cytometry analysis of EGFP-expressing HeLa cells) of non-transfected, non-caged, or caged gRNA with or without phosphine triggers. Bars represent the mean and error bars represent s.d. p values: *** < 0.001, * < 0.05.) Adapted from [41] with permission from the Royal Society of Chemistry. d) Quantification of T7E1 assays of Cas9-expressing HeLa-OC cells transfected with non-caged gRNA or caged gRNA and treated by different concentrations of DPBM. Bars represent the mean and error bars represent s.d. Adapted from [42].

The Kool group investigated the mechanism of how gRNA inactivation occurs, finding that acylation of the ribose 2’-OH blocks interaction of the gRNA with the target DNA, due to modifications to the riboses in the spacer region. Both Kool and Zhou also provided evidence for interrupted Cas9:caged gRNA interaction, which would prevent cellular delivery in the form of RNPs. Though AMN-caged gRNA cannot benefit from increased stability by complexing with Cas9, the Zhou group showed that it is resistant to degradation by different RNases (RNase I and RNase T1) due to the chemically modified 2’-OH groups. Zhou also demonstrated in vitro conditional control of Cas13a, a crRNA-guided RNA nuclease, extending this strategy to other CRISPR/Cas systems. This suggests that even though acylated gRNAs cannot be deployed as RNPs, their stability may still be comparable due to resistance to endogenous RNases.

The efficacy of small molecule control of gRNA function has been demonstrated in mammalian cells, and holds promise for temporal gRNA control in other biological systems. While spatial control would require cell-specific expression of Cas9, the advantages of small molecules such as tissue penetrance and easy dose control makes them an attractive option for applications in animal models where tissues may not be accessible for light-activation.

4. Aptamer- and aptazyme-control of gRNA function

Aptamers are RNA or DNA oligonucleotides that have been selected to bind small molecules or proteins with high affinity and specificity, similar to how an antibody binds its antigen.^[45] One of the first examples of aptamers being appended to gRNAs was reported by the Church group, who used an MS2 bacteriophage coat protein-binding hairpin to recruit a VP64-MS2 fusion to the target locus for activation of transcription.^[46]

Besides recruitment of proteins, aptamer binding can function as a switch that induces changes in gRNA structure which may promote Cas9 binding or spacer-target recognition, based on how the gRNA was designed. The most straightforward example was developed by appending a spacer-complementary sequence and theophylline-triggered aptamer at the 3’ end of the gRNA, with the intention of blocking spacer hybridization with the target DNA until theophylline addition (Figure 3a).^[47] The Yang group showed activation of gene editing in mammalian cells transfected with plasmids encoding Cas9 and the aptamer-gRNA in the presence of theophylline (2 mM) 48 hours post-transfection, but significant background cleavage in the absence of theophylline was observed as well. In contrast, the Cai group introduced a system where a small molecule or protein binding aptamer was fused to the 3’ end of the gRNA, and the aptamer stem was designed to be complementary to a portion of the spacer thereby blocking its interaction with the DNA target (Figure 3b).^[48] Binding of the small molecule or protein ligand to the aptamer disrupts inhibitory stem base-pairing, opening up the spacer for target hybridization. The Cai group used dCas9 or dCas9-VP64 for transcriptional silencing or activation, respectively, in mammalian cells transfected with plasmids encoding the dCas9 construct and the aptamer-gRNA. The group used aptamers that responded to tetracycline or theophylline and controlled expression of endogenous genes such as vegf with no background activation 48 hours after transfection. They also elicited a dose-response of transcriptional regulation. They saw a similar result when the small molecule-responsive aptamer was switched out for the MS2 aptamer and MS2 expression was induced with doxycycline. The most intriguing part of this study however, was the construction of Boolean logic gates in live cells by swapping out aptamers in their gRNA design to recognize endogenous proteins such as NPM, and rewire these cellular signals to promote transcriptional changes (Figure 3b).

Figure 3.

a) Aptamer-based control of gRNA function through ligand binding. b) Control of gRNA function using a protein binding aptamer, in this case NPM. c) Aptazyme controlled gRNA (agRNA). Function of the tool was assessed by editing GFP in mammalian cells and measuring loss of GFP fluorescence with flow cytometry. Catalytically dead hammerhead ribozyme-blocked spacer control gRNA (dHHR-bsgRNA). All aptamers are shown in red. Bars represent the mean and error bars represent s.d. Adapted from [50].

The Crabtree group designed a rapamycin-inducible system for recruiting epigenetic modifying complexes to DNA loci targeted by dCas9.^[49] They added two MS2 binding aptamers to the gRNA and expressed an MS2-Fkbp fusion along with the epigentic modifier-Frb fusion protein. Upon rapamycin treatment, they demonstrated 90% reduction in target gene expression due to transcriptional repression by histone methylation in HEK293T cells after 5 days of treatment. Despite the long incubation times for these experiments, they observed rapid recruitment of epigenetic complexes by 5 minutes post-rapamycin treatment.

A ribozyme-based approach to gRNA activation has been recently developed by the Liu group which couples aptamer binding to ribozyme activation, termed an “aptazyme”.^[50] Their design consisted of an aptamer modified hammerhead ribozyme attached to the 5’ end of the gRNA, coupled with a spacer-complementary strand that blocked target recognition. Theophylline binding to the aptamer activates the self-cleaving ribozyme, thus releasing the inhibitor strand from the spacer and inducing genome editing in mammalian cells (Figure 3c). Plasmids containing the Cas9 construct and the gRNA were delivered by transfection and analysis of EGFP editing was conducted 7 days post-transfection. However, significant background activity of their aptazyme-gRNA (22% loss of mean cell GFP fluorescence) was observed in the absence of theophylline. Liu and coworkers reported similar results with dCas9-based transcriptional activators and base editors.^[51] They showed four-fold activation of gene editing with theophylline addition as measured by mean cell fluorescence, but did not report any indel rates. While the theophylline riboswitch typically has very low background in the absence of ligand, substituting it into the ribozyme may not fully suppress ribozyme self-cleavage, as demonstrated earlier.^[52] The theophylline aptazyme needs further optimization in order to reduce background CRISPR/Cas9 activity.

A couple aptamer-based approaches have been designed that do not rely on spacer blockade. The Kortemme group placed the aptamer in several locations within the upper stem, nexus, or hairpin regions of the gRNA to disrupt complexation with Cas9.^[53] After creating a library of 86 gRNAs, they found one that showed strong target DNA cleavage only in the presence of theophylline. The mechanism was predicted to be strand displacement and stabilization of the hairpin upon aptamer binding, promoting a Cas9-binding conformation. During their selections, they serendipitously came across a gRNA design that was deactivated upon theophylline addition. This gRNA had the aptamer placed in the nexus, and was predicted to prevent an important uracil from hydrogen bonding to Cas9 when bound by theophylline. They showcased the ability to orthogonally regulate GFP and RFP expression with two different aptamers in the same bacteria, but the method suffered from not being functional in eukaryotic systems. They hypothesized this was because of insertion of the aptamer in locations known to be important to Cas9 binding, thereby presumably reducing binding affinity. Thus, the required concentration of functional Cas9:gRNA complexes may have never reached a threshold for detection of transcriptional changes in eukaryotic cells. The Batey group reported a design strategy where the tetraloop used to fuse the crRNA and the tracrRNA into a single gRNA was replaced with the theophylline aptamer in order to disrupt the REC1 domain, crucial for Cas9 nuclease activity.^[54] The group randomized 14 nucleotides in the region and used their CREATE (CRISPR-enabled trackable genome editing) selection protocol to find gRNA designs that could activate gene editing in E. coli only when incubated with theophylline. They identified two switchable gRNAs that had similar efficiency as regular gRNA at several different loci in the presence of theophylline. Background activity in the absence of theophlline was minimal (3–9%). Such selection protocols offer a promising route for discovery and optimization of aptamer-controlled gRNAs. However, while excellent off to on switching was observed in bacteria, this system was not tested in mammalian cells. Thus, aptamer- and aptazyme-based systems for gRNA control may not be readily transferable to other cell models or CRISPR technologies that it was not originally designed for. The wide-range of ligand specific aptamers that have already been designed and the ability to integrate them into gRNA sequences using traditional molecular biology methods makes aptamer-based conditional control an attractive option. They also allow for experiments requiring multiplexed regulation of gene editing or expression. The ability to generate aptamers against theoretically any protein target offers opportunities for the design of complex genetic circuits. While several gRNA-aptamer designs seemed to have minimal background activity, significant gRNA activation was seen in many of these designs in the absence of ligand. Rigorous selection and optimization are needed to develop functioning gRNA-aptamers with higher on/off ratios.

5. Oligonucleotide-control of gRNA function

Nucleic acid-triggered gRNA activation offers a wide range of opportunties to interface CRISPR-based gene control with endogenous inputs. Moreover, it holds promise for interfacing with upstream DNA computation circuits and thereby provides unique ways to translate the recognition of diveres nucleic acid input patterns into defined gene regulatory outputs.^[55] Since DNA computation approaches are commonly based on toe hold-mediated strand exchange reactions,^[56] designs of integrating a blocked gRNA into these circuits can be readily generated.

A simple and elegant strategy by the Fulga group was to attach a spacer-complementary strand and 14 nucleotide loop to the 5’ end of the gRNA. A ssDNA activating oligonucleotide binds the loop between the spacer and complement, inducing RNase H-mediated degradation and removing the inhibitory complement strand (Figure 4a).^[57] Mammalian cells were transfected with plasmids encoding dCas9-VP64 and the oligonucleotide-responsive gRNA. 24 hours later, the activating oligonucleotide was delivered by transfection and cells were analyzed another 24 hours later. This method had almost no background and excellent (113-fold) activation of ECFP expression upon 20 nucleotide oligonucleotide addition (Figure 4b). The success of their approach is likely due to the accessibility of the target loop to the activating ssDNA oligonucleotide, which allows for RNase H-mediated cleavage between the spacer and inhibitor strand, as compared to toehold-mediated strand-displacement methods discussed later. Whether the oligonucleotide-sensing gRNA design is still capable of complexing with Cas9 was not assessed.

Figure 4.

a) Oligonucleotide-based control of gRNA function through loop binding and RNase H-mediated degradation. The activating oligonucleotide and its binding site is shown in red. b) ECFP transcriptional activation assay using an oligonucleotide-responsive gRNA with dCas9-VP64. The 20 nucleotide oligonucleotide had the best activation of gRNA function. Adapted from [57]. c) gRNA-based miRNA sensor. Red represents the miRNA and its binding site. RNA-induced silencing complex (RISC) is recruited by the miRNA to cleave the target site and release the functional gRNA. Activation of RFP expression is seen only in the presence of miR-294 or the homologous miR-291a. Bars represent the mean and error bars represent s.d. in b) and c). Adapted from [58] with permission from Springer Nature.

A dCas9 transcriptional activator has been repurposed by the Wang group as a microRNA (miRNA) sensor by modifying the gRNA to be activated in the presence of an miRNA of interest and driving RFP expression in mammalian cells.^[58] The transcribed gRNA contains a 5’ cap and 3’ poly-A tail, with complementary miRNA sequences flanking the gRNA sequence. miRNA binding induces site-specific RNA cleavage via Ago2 in the RNA-induced silencing complex (RISC), removing the 5’ cap and the 3’ poly A tail and generating functional gRNA (Figure 4c). The mechanism for activation was not explored; however, increased gRNA stability and gene editing in mammalian cells when the gRNA was transcribed with a 5’ cap and 3’ poly-A tail was previously reported.^[59] Regardless, the approach showed no background activity and was specific to the miRNA of interest, only showing off-target gRNA activation with other miRNAs that had high sequence homology. Wang and co-workers applied this method to study miRNA heterogeneity in mouse embryonic stem cells, and to track miRNA expression during cell differentiation.

Meanwhile, inhibition of gRNA function upon oligonucleotide addition was accomplished by two labs. Both the Moon lab and the Pierce lab independently used an RNA oligonucleotide to bind an exogenous sequence inserted in the middle of the gRNA, thus turning off its function in bacteria.^[60] The Moon lab used an Hfq recruiting sequence in the inhibiting RNA oligonucleotide that stabilized the oligonucleotide as well as the dsRNA hybrid. The Pierce lab also designed an off-to-on switch analogous to the 5’ loop approach mentioned first in this section. However, their designed used a hairpin with an additional toehold 5’ to the spacer complement sequence. The RNA-activating oligonucleotide bound the toehold and the spacer complement and resulted in a 3-fold off-to-on response, considerably weaker than the previously discussed design that targeted the 14 nt loop.^[57] The results were similar to the Chen group’s toehold-mediated strand displacement approach, with a median 6.6-fold off-to-on response.^[61] The Simmel group also conducted strand-displacement reaction-based activation of gRNA function with Cas12a.^[62] By inserting sequence complementary to the Cas12a binding region of the gRNA at the 5’ end with a 12 nt toehold, they were able to construct AND and NAND logic gates using 2 and 3 RNA oligonucleotide inputs. They demonstrated 100-fold suppression of fluorescent protein transcription 4 hours after induction of dCas12a expression in E. coli that constitutively expressed the gRNA and trigger RNA oligonucleotide.

Oligonucleotide-mediated activation of gRNA function has many properties that make it an appealing method for conditional control of CRISPR/Cas9 activity. First, similar to aptamer-based strategies, it is relatively simple to apply. Second, most approaches discussed above had excellent response to trigger, with low background activity. Third, this method provides an exciting tool for studying miRNA function or wiring the presence of an endogenous nucleic acid molecule to gene editing or transcriptional events. Lastly, as oligonucleotides are utilized as inputs, they can be readily generated by DNA computing devices, which have been shown to exhibit functionality in biological systems, including mammalian cells.^[63] With that said, this approach lacks the diversity of activating signals that is possible with aptamers, and the ease and temporal (and spatial) precision that small molecules and light provide. Introduction of triggering-oligonucleotides necessitates transformation or transfection of the oligonucleotide, or expression from a plasmid already delivered into the cell, which is a major weakness of this approach compared to activation with small molecules or light.

6. Conclusion

A multitude of strategies for placing CRISPR/Cas9 under conditional control have been developed.^[15] While earlier approaches focused on protein engineering, more recently, modification of the gRNA has received increasing attention, as it provides distinct opportunities. By themselves much smaller biomolecules than the Cas9 proteins, gRNAs are more easily synthesized, modified, randomized, and screened for desired biochemical functions.^[64] gRNAs can either be expressed from a range of vectors or delivered directly by cationic lipids or through injection either alone or in complex with the Cas9 protein,^{[34a, 65]} alleviating limitations on transducing large genetic elements^[66] (such as split-Cas9-estrogen receptor fusion proteins).

Optical control methods demonstrate simple and generalizable design of caged gRNAs, excellent off-to-on switching, and precise spatiotemporal manipulation due to the distinct advantages of light as an activator. While light penetration is not an issue in cell culture, tissue slices, and many aquatic embryos due to their optical transparency, application in non-transparent animal models like mice is more difficult. Small molecule decaging methods capitalize on some of the weaknesses of optical triggering. Whole organism exposure to the small molecule stimulus can be accomplished through feeding or injection, and the caging reaction can be performed on in vitro transcribed gRNA, bypassing the need for oligonucleotide synthesis. However, this method does no offer spatial specificity, unless the expression of the Cas9 protein is restricted to specific tissue. Also, the introduction of the small molecule-responsive caging groups into the gRNA sequence was non-specific, modifying random nucleotides, which might cause batch-to-batch variation in efficacy. The chemically acylated gRNAs did not form an RNP with the Cas9 protein, which can be important as delivery becomes more difficult and the gRNA is more prone to degradation when not in a protein complex.^[31] There are a plethora of aptamer-based designs for conditional control of CRISPR/Cas9. This method is attractive because of the versatility of aptamers, as they have been engineered to bind a wide range of proteins or small molecules. Embedding the aptamer into a self-cleaving ribozyme has allowed for control of gRNA function by small molecule addition. However, aptamer- and aptazyme-based approaches can display significant CRISPR/Cas9 background activity. Lastly, oligonucleotide-responsive gRNAs have been engineered and hold promise for sensing endogenous DNA/RNA. The most effective off-to-on switch utilized a simple 5’ loop and spacer-complement strand to block the spacer, and a DNA input strand that binds the loop and induces RNase-H-mediated degradation to restore gRNA function. However, these approaches require inserting additional RNA sequence into the gRNA, meaning some rational design and optimization is needed. Detection of miRNAs is an important application of oligonucleotide-mediated gRNA regulation, demonstrating the possibility of wiring specific cell responses to endogenous nucleic acid triggers. In addition, triggering of aptamer- and aptazyme-based gRNAs and oligonucleotide-responsive gRNAs is much slower than light activation or small molecule-induced decaging.

Overall, an ideal system for conditional gRNA regulation can be defined by the following criteria: 1) a wide dynamic range with minimal background activity and function comparable to the non-modified gRNA when activated, 2) temporal control with fast activation kinetics for interrogation of fast and dynamic biological events, 3) the potential for spatial control, 4) reduced off-target effects, 5) the potential for conditional control interfaced with endogenous triggers, e.g., porteins and nucleic acids, and 6) ease-of-use across different experimental settings and model organisms, e.g., use of RNPs facilitating delivery and stability against RNases. Each currently available method offers unique benefits. For example, light activation of caged gRNA enjoys the advantages of a large dynamic range, ultrafast activation, potential for spatial control, and the possibility for the delivery of RNPs. Acylation of gRNAs for small molecule activation uniquely provides protection against RNases and good activation; hower phosphine-triggered decaging can be slower than optical control and lacks localized targeting. Oligonucleotide- and protein-controlled gRNAs allow activation by endogenous signals and can shed light on cellular decision-making logic. We envision that future endeavors in the development of conditionally controlled gRNAs will provide broader access of these diverse approaches to the general biological community.

Development of conditional gRNA function with optical, small molecule, protein, or nucleic acid stimuli opens the door to many possibilities of perturbing and sensing of biological systems. This could be useful in the field of synthetic biology to design sensors for RNA and proteins.^[67] Optical control over multiple genes could be possible with the use of different photocages with red-shifted absorbances,^[68] allowing for gene editing in specific tissues and timepoints in model organisms. Several examples of gRNA functional control discussed in this review, such as by the Cai and Simmel groups, demonstrate the feasibility of generating synthetic gene circuits capable of computing protein or oligonucleotide inputs, respectively.^[55] Coupling conditional control with epigenetic modifying CRISPR/dCas9 will be fruitful for study of epigenetic roles during development at specific loci.^[69] Many of the designs discussed here may also work with other Cas9 proteins, such as Cas13 for targeting RNA in cells, permiting transcript editing.^[70] In a therapeutic context, conditionally controlled CRISPR technologies can be useful for limiting off-target effects by constraining activity of gene editing to specific cells or tissues and limiting the duration of activity.^[16]

Biographies

Alexander Deiters received his PhD from the University of Münster in 2000 and subsequently conducted postdoctoral studies at the University of Texas and The Scripps Research Institute. He currently is a professor at the University of Pittsburgh, where his group conducts research at the interface of chemistry and biology, with an emphasis on optical control of oligonucleotide and protein function.

Wes Brown received his B.S. in Bioresource Research and B.A. in International Studies from Oregon State University. Since 2016, Wes has conducted research on genetic code expansion and optical control of protein function in zebrafish under the mentorship of Prof. Deiters at the University of Pittsburgh.

Wenyuan Zhou received his B.S. in Chemical Biology from Peking University. Since 2014, he has been conducting research for his PhD on optical control of cell signalling and gene editing under the guidance of Prof. Deiters at the University of Pittsburgh.