Transposons and CRISPR: Rewiring Gene Editing

Abstract

CRISPR-Cas is driving a gene editing revolution because of its simple reprogramming. However, off-target effects and dependence on the double-strand break repair pathways impose important limitations. Because homology-directed repair acts primarily in actively dividing cells, many of the current gene correction/replacement approaches are restricted to a minority of cell types. Furthermore, current approaches display low efficiency upon insertion of large DNA cargos (e.g., sequences containing multiple gene circuits with tunable functionalities). Recent research has revealed new links between CRISPR-Cas systems and transposons providing new scaffolds that might overcome some of these limitations. Here, we comment on two new transposon-associated RNA-guided mechanisms considering their potential as new gene editing solutions. Initially, we focus on a group of small RNA-guided endonucleases of the IS200/IS605 family of transposons, which likely evolved into class 2 CRISPR effector nucleases (Cas9s and Cas12s). We explore the diversity of these nucleases (named OMEGA, obligate mobile element-guided activity) and analyze their similarities with class 2 gene editors. OMEGA nucleases can perform gene editing in human cells and constitute promising candidates for the design of new compact RNA-guided platforms. Then, we address the co-option of the RNA-guided activity of different CRISPR effector nucleases by a specialized group of Tn7-like transposons to target transposon integration. We describe the various mechanisms used by these RNA-guided transposons for target site selection and integration. Finally, we assess the potential of these new systems to circumvent some of the current gene editing challenges.

Gene editing refers to the modification of an organism’s genetic material (addition, deletion, or alteration). Since the discovery that nuclease-induced DNA breaks can be used to promote genome editing,¹ various nuclease-based technologies have been developed (ZFNs, meganucleases, TALENs, and CRISPR-Cas). Among them, CRISPR-Cas (clustered regularly interspaced short palindromic repeats and CRISPR-associated proteins) has led a revolution that has accelerated the applications of gene editing in different disciplines. CRISPR-Cas has been adapted from a microbial defense system that provides adaptive immunity against genetic parasites.^2,3 After the first encounter with the invader, short pieces of the foreign genetic material are incorporated into CRISPR arrays. Later, these arrays are transcribed into RNAs, which associate with effector nucleases (e.g., Cas9 or Cas12), acting as a guide to cleave complementary DNAs or RNAs, thereby digesting the invader’s genetic material and stopping a second infection in a process known as interference. An additional determinant for target DNA selection is the presence of a short motif adjacent to the target site or PAM (protospacer adjacent motif). PAM sequences are essential for avoiding autoimmunity and are thought to act as “labels” to minimize the search space along the genome.⁴⁻⁷

CRISPR-Cas systems are widespread among microorganisms, and their variety is categorized into two classes and multiple types on the basis of the presence of the different protein and RNA components and their architecture. Typically, class 1 systems (including types I, III, and IV) use effector multisubunit complexes. For instance, type I effector or Cascade complexes comprise a gRNA and Cas3 and Cas5–Cas8 proteins, wherein Cas3 has nuclease activity. Likewise, type III-A and -B systems encode RNA-targeting multiprotein complexes (known as Csm for type III-A or Cmr for type III-B) containing Cas5 (Csm4/Cmr5), Cas7 (Csm3/Cmr4), Cas7-like (Csm5/Cmr1 and Cmr6), Cas10 (Csm1 and Cmr2), and Cas11 homologues (Csm2 and Cmr3), with Cas10 frequently displaying nuclease activity. In turn, class 2 effectors (type II or Cas9s, type V or Cas12s, and type VI or Cas13s) are composed of single-multidomain protein nucleases that have been extensively used in gene editing applications.⁸⁻¹⁰

The strength of CRISPR-Cas is based on its ease of reprogramming simply by altering the sequence of the guide RNA. Therefore, Cas nucleases can be redirected to specific sites in a genome adjacent to a PAM sequence to generate double-strand breaks (DSBs). This procedure exploits the host DSB repair pathways (DRPs) to induce gene editing, creating knockouts [through nonhomologous end joining (NHEJ) or microhomology-mediated end joining] or precisely correcting a lesion [through homology-directed repair (HDR) by providing an exogenous homologous repair template].¹¹ Gene editing with CRISPR-Cas nucleases is currently used in biotechnology, agriculture, basic research, and even human medicine.¹² However, CRISPR-Cas editing still faces substantial challenges in terms of efficacy (mainly due to the efficiency of the enzymes) and safety (associated with their accuracy and our lack of control over the host DRPs).¹³ In the past several years, the development of more efficient and precise CRISPR-nucleases,^14,15 CRISPR-based platforms that bypass the host DRPs,¹⁶ and strategies for controlling the DRPs¹⁷ has alleviated some of these limitations. Autonomous DNA transposons have also been investigated as alternative gene editing tools. These transposons are mobile genetic elements (MGEs) comprising their own target DNA recognition system and a DNA integrase in charge of DRP-independent transposon insertion. However, transposon usage is either too specific (not easily programmable) as is the case for Tn7¹⁸ or too unspecific as in the case for Tn5¹⁹ and PiggyBac.²⁰

This review focuses on new findings regarding the relationship between CRISPR interference nucleases and transposons. The potential advantages of these new RNA-guided tools for gene editing are also assessed.

IS200/IS605 Transposons: The Evolutionary Origin of Cas9 and Cas12 Nucleases

Mechanism of TnpA-Based IS200/IS605 Transposition

Insertion sequence (IS) transposons were identified in the 1970s due to their capacity to produce mutations in model systems during genome translocation.²¹⁻²⁴ The IS200/IS605 family is an unusual group within the IS transposons that use obligatory single-strand (ss) DNA intermediates for transposition. The transposition mechanism of this family has been reviewed extensively.²⁵ In brief, transposon excision and integration are intricately coordinated with genome replication and have a strong bias toward the lagging strand template.²⁶ The transposon cargo is typically ∼600–2000 nucleotides and encodes only one or two genes, TnpA and TnpB (Figure 1A). Transposon excision and integration are catalyzed by TnpA (or Y1), a small transposase (∼200 amino acids) member of the “HUH” (histidine–bulky hydrophobic residue–histidine) enzyme superfamily.²⁷ TnpA binds to subterminal palindromic structures at the flanking ends of the transposon DNA called the left and right ends [isLE and isRE, respectively (Figure 1A)]. Upon binding to the isLE and isRE, TnpA first catalyzes the transposon’s excision and then its integration into a DNA adjacent to a target site, a conserved tetra- or pentanucleotide ssDNA that is not duplicated after insertion (Figure 1A). Structural and biochemical data of different steps of the transposition reaction are available for members of two families, TnpAs from Helicobacter pylori and Deinococcus radiodurans from the IS608 and ISDra2 transposons (HpyTnpA and DraTnpA), supporting a trans/cis rotational model.²⁸⁻³² Overall, transposon mobilization is thought to take place through a “peel and paste” mechanism, in which the transposon excises (“peel”) as a circular single-strand DNA from the lagging strand (donor DNA) and is then inserted (“paste”) into a new genomic position next to the target site (preferentially at the lagging strand)²⁵ (Figure 1A).

Figure 1.

IS200/IS605 family of transposons that encodes Cas9 and Cas12 evolutionary ancestors. (A) Architecture of the D. radiodurans IS608 transposon and the proposed TnpA-mediated “peel and paste” transposition mechanism. Modified from refs (40) and (25). (B) Domain organization and comparison between DraTnpB and FnoCas12a [Protein Data Bank (PDB) entry 5NG6] and between OgeuIscB (PDB entry 8CTL) and SpyCas9 (PDB entry 4ZT0). The WED domain, RuvC domain, recognition lobe (REC), bridge helices (BH), NUC domain, zinc finger domain (ZF), PAM or TAM interacting domain (PI/TI), HNH domain, P1 interacting domain (P1D), and C-terminal domain are depicted.

IscBs and TnpBs Are RNA-Guided Endonucleases

TnpB proteins are not required for transposition^28,31,33 and seem to play a regulatory function.³⁴⁻³⁶ On the basis of sequence homology analyses, TnpBs are likely the evolutionary ancestors of Cas12 proteins.^9,37−39 The predicted structural model of TnpB from D. radiodurans (DraTnpB) is most similar to Cas12f and consists of a recognition lobe (REC), a WED domain, a RuvC domain, two bridge helices (BHs), and C-terminal zinc finger motive (ZF).⁴⁰ A group of TnpB-related proteins also encode an HNH domain. Therefore, they are thought to have evolved into Cas9 proteins and are named IscBs (insertion sequences Cas9-like similar to TnpBs).³⁷ IscBs also display an N-terminal domain without clear homology to other known domains containing a “PLMP” motif (PLMP), while they lack the characteristic REC domain³⁸ (Figure 1B).

Noncoding RNAs (ncRNAs) have been linked to TnpB-related transposons.^41,42 Recently, transcriptomics data from different TnpB transposons have revealed the expression of a ncRNA overlapping with the TnpB gene that extends beyond the isRE of the transposon, denoted as right end RNA [reRNA (Figure 2A)].^38,40 Similarly, several IscB transposons transcribe ncRNAs associated with the isLE [leRNAs (Figure 2B)] containing predicted structural hairpins that partially match the HEARO (HNH Endonuclease-Associated RNA and ORF) RNA,³⁸ a noncoding RNA usually associated with HNH containing loci, together thought to form a mobile genetic element.⁴³ Both reRNAs and leRNAs are also denoted ωRNAs (from OMEGA-associated RNAs, see Diversity of TnpB and IscB Proteins).³⁸ Some IscB transposons are associated with CRISPR arrays. The similarities between the structural hairpins of these CRISPR-RNAs (crRNAs) and the ωRNA suggest an evolutionary relationship, in which the isLE/isRE derived into the handle of the crRNA.³⁸ ωRNAs also have a sequence that probably evolved to form the guide of the crRNA, corresponding to the flanking region of the transposon (Figure 2A,B). This sequence changes after transposition, showing conceptual parallelisms with the update of the guide sequences during CRISPR immunity. Nevertheless, the guide update mechanism is not conserved in both systems. In CRISPR, it is accomplished by Cas1–Cas2 adaptation complexes that integrate pieces of the invader genetic material into the CRISPR arrays,⁴⁴ while in ωRNAs, it is mediated by transposon mobilization (Figure 1A).

Figure 2.

TnpBs and IscBs are RNA-guided endonucleases. (A) DraTnpB RNA-guided dsDNA cleavage. (B) OgeuIscB RNA-guided dsDNA cleavage.

IscBs and TnpBs are RNA-guided endonucleases that use ωRNAs as guides to target a guide RNA complementary dsDNA.^38,40 They also specify a target or transposon-associated motive (TAM) adjacent to the guide RNA target, which is analogous to the recognition of PAM sequences by CRISPR nucleases to initiate dsDNA unwinding. TAM discovery experiments have revealed that different members of the TnpB and IscB families recognize different TAMs.^38,40 Intriguingly, DraTnpB recognizes a TAM that partially matches the transposon’s target site, suggesting that the PAM might have originated from this sequence (Figure 2A). TnpBs cleave both strands of the DNA using a single catalytic RuvC domain, like Cas12 proteins^38,40 (Figure 2A). On the contrary, IscB proteins use their RuvC and HNH domains to cleave specifically the nontarget and target strand, resembling Cas9 cleavage (Figure 2B).³⁸ High-resolution cryo-electron microscopy structures of an IscB isolated from a human gut stool metagenome (OgeuIscB) in complex with the ωRNA and a nonscissile target DNA have revealed the molecular basis of TAM recognition, R-loop formation, and DNA cleavage, showing that IscB shares a similar interference mechanism with Cas9.⁴⁵ OgeouIscB structures show that the function of the REC domain is substituted by the ωRNA lobe formed by the HEARO RNA, which contacts the RNA/DNA heteroduplex. This supports an evolutionary trend in which protein domains (such as REC) have substituted ωRNA structural motifs.⁴⁵ Despite the structures of OgeuIscB showing unambiguous density for the PLMP domain, its function is still unclear. The PLMP domain interacts with the last loop of the ωRNA (P5), suggesting that it could play a role in RNP biogenesis by regulating the termination of transcription. The structure-assisted removal of this domain had only a slight effect on RNA-guided activity,⁴⁵ while its removal from other members has a more drastic effect.³⁸ Further research is needed to uncover the role of the PLMP domain.

Diversity of TnpB and IscB Proteins

Intriguingly, several groups of TnpB- and IscB-related proteins have been identified. The family of these nucleases, named OMEGA (obligate mobile element-guided activity), includes members that represent steps of the evolutionary history of Cas9 and Cas12 proteins. For instance, a group of smaller IscB-like proteins (∼350 amino acids) named IsrB (insertion sequence RuvC-like OrfB) contain PLMP and RuvC domains but lack the HNH domain and consequently are nickases that cleave the nontarget strand.³⁸ On the contrary, IshB nucleases (insertion sequence HNH-like OrfB, ∼180 amino acids) display PLMP and HNH domains but lack the RuvC domain and are predicted to nick the target strand.³⁸ Furthermore, a group of IscBs display putative ancestral REC domains and probably represent a later stage in the evolution closer to Cas9s.³⁸ Additionally, two new groups of Cas9 have been discovered, one of them containing relatively small Cas9s (II-D, ∼700 amino acids).³⁸

The TnpB family is more diverse than the IscB family. A bioinformatic analysis supports the idea that TnpB is one of the most common genes in prokaryotes, suggesting that RNA-guided mechanisms are more broadly spread than previously thought.³⁸ Remarkably, TnpBs are possibly ancestors of Fanzors, larger proteins (∼2000 amino acids) present in diverse eukaryotic transposons.⁴⁶ Further phylogenetic and functional investigation of the TnpB family will help to accurately establish the evolutionary history of Cas12 proteins.

Biological Function of TnpB and IscB Proteins

The role of TnpB and IscB proteins remains enigmatic. Initial observations show that TnpB is dispensable for transposition,^28,31,33 and genetic and functional experiments point to a possible regulatory role.³⁴⁻³⁶ On the basis of the current evidence and similarities with CRISPR immunity, the guide sequences of the OMEGA systems are updated after the migration of the transposon to a new place in the genome (Figure 1A). Therefore, the target site for TnpB and IscB proteins, defined by TAM sequence directly adjacent to the guide RNA complementary sequence, is present in only one of the daughter dsDNAs after transposon excision (Figure 3). In this framework, two main hypotheses for the function of TnpB and IscB proteins during the transposon life cycle have been suggested. A “toxin–antitoxin” model supports the idea that the cleavage of the transposon-less daughter DNA would be toxic for the cells, while DNAs containing the transposon would behave as an antitoxin at the population level.³⁸ An alternative model of “peel, paste, and copy” for transposon homing proposes that the cleavage of the transposon-less daughter dsDNA would stimulate its recombination with the sister dsDNA containing the transposon.⁴⁰ It is important to note that the “peel, paste, and copy” model might allow us to explain the increase in the transposon copy number observed in some D. radiodurans strains.⁴⁷ The outcome of both scenarios would be similar at the population level, ensuring the propagation of the transposon. Further research is needed to uncover the biological role of the RNA-guided activity of OMEGA nucleases during the transposon life cycle.

Figure 3.

Possible functions of TnpB and IscB’s RNA-guided activity in the transposon life cycle. The proposed model is depicted for the D. radiodurans IS608 transposon. After transposon excision and replication of the DNA, one of the resulting dsDNAs presents the target site for DraTnpB, defined by the presence of the TAM next to the guide complementary sequence. DraTnpB RNA-guided cleavage of this strand would stimulate the HDR using the sister DNA, containing the transposon, as a template (transposon homing by the “peel, paste, and copy” mechanism). Alternatively, the generated DSB might have toxic effects on the cell, the transposon working as an antitoxin. Schematic inspired by ref (40).

Potential of OMEGA Systems in Gene Editing

OMEGA constitutes a broad collection of architecturally and functionally diverse RNA-guided endonucleases.^38,40 Two OMEGA nucleases have been proven to induce genome edits in mammalian cells, establishing a proof of principle of OMEGA system’s potential in gene editing. OgeuIscB-ωRNA (496 amino acids) produced insertions and deletions (indels) at the target site, next to a 3′ CTAGAA TAM, in HEK293FT with a frequency of ∼5% when co-expressed with 16-nucleotide guides. Similarly, transient transfection of DraTnpB (408 amino acids) together with reRNAs containing 20-nucleotide guides produced indels with a frequency of ∼10–20% next to a 5′ TTGAT TAM, and the observed mutational profile resembles that of Cas12a-based editing.⁴⁰

The development of compact CRISPR enzymes is of current interest due to their advantage when they are packed into mature delivery vehicles, such as adenoviral-associated vectors (AAVs).^13,48−51 OMEGA nucleases are substantially smaller (∼400 amino acids on average) than type II (Cas9s, 900–1300 amino acids) and type V (Cas12s, 500–1300 amino acids) enzymes, and still smaller than the most compact CRISPR-Cas gene editors (Cas12fs, ∼500 amino acids, and Cas12js, ∼800 amino acids).⁴⁸ The miniaturization of enzymes usually comes at a cost of efficiency or specificity, because of the simplification of accessory regulatory protein domains. For example, Cas12js constitute a minimal protein scaffold that can perform RNA-guided DNA cleavage.^49,50 However, their efficiency is lower than those of other gene editors,⁴⁸ and attempts to improve their efficiency have resulted in a parallel increase in RNA/DNA mismatch tolerance and reduced RNA/DNA specificity.⁵⁰ A recent study with Cas9 suggests that the tolerance to RNA/DNA mismatches inversely correlates with the number of structural contacts with the RNA/DNA hybrid.¹⁵ In this regard, OgeuIscB intertwines an extensive protein and ωRNA contact network with the DNA/RNA hybrid, suggesting a possible high DNA/RNA substrate specificity despite its small size.⁴⁵

OgeuIscB and DraTnpB are less efficacious than Cas9 or Cas12a when used in mammalian cells. However, OMEGA nucleases represent diverse new starting points for the design of compact enzymes for gene editing. Further mechanistic and functional studies of the different IscB and TnpB families will be essential for fully uncovering the similarities between OMEGA and CRISPR nucleases and for the efficient development of TnpB- and IscB-based gene editing tools.

CRISPR-Cas Recruitment by TN7-like Transposons

The intricate evolutionary history of transposons and both CRISPR-Cas classes is further intertwined by the appearance of specialized transposons that have recruited an inactivated CRISPR-Cas interference effector to direct the transposase to its target (Figure 4).⁵²⁻⁵⁵ These CRISPR-Cas-associated transposons (CAST) have lost their target cleavage activity by eliminating Cas3 in class 1 systems and mutating Cas12’s catalytic amino acids in class 2 systems. Both complexes retain the ability to recognize DNA sequences in a PAM-dependent manner, targeting transposon integration. Convergent evolution has produced a great diversity of these elements, resulting in different combinations of CRISPR-Cas subtypes (e.g., I-F, I-B, and V-K) and transposon families [e.g., Tn7 and Tn5053 (Figure 4)].⁵⁶⁻⁵⁹

Figure 4.

Gene architectures of Tn7, CAST, and Tn5053 systems. Each system contains different combinations of transposon (tnsA, tnsB, tnsC, or tnsD/tniQ) and cas (Cascade or cas12k) proteins. The general structure of the CRISPR arrays (shown on the right of each locus) and the RE and LE sequences of each transposon (shown beneath each transposon end) also vary between systems. Canonical CRISPR array repeats are colored pale red, while divergent repeats are colored dark red. Subtype I-F and I-B CASTs resemble the Tn7 family of transposons, including tnsA, tnsB, tnsC, and tnsD/tniQ, while type V-K resembles the Tn5053 family that includes only tniA (tnsB homologue), tniB (tnsC homologue), and tniQ.

A general model explaining the new system’s transposition mechanism has been proposed on the basis of previous knowledge of Tn7^18,60,61 and the recent molecular, biochemical, and structural data generated for class 1 (subtypes I-B and I-F)^54,62−65 and class 2 (subtype V-K)⁶⁶⁻⁶⁸ transposons (Figure 5). Tn7’s locus comprises five genes, namely, TnsA–TnsE, which are used in two different combinations (Figures 4 and 5). The first combination, TnsABCD, serves as a homing mechanism to insert Tn7 into the host’s chromosome. The second combination, TnsABCE, is used to insert Tn7 into replicating plasmids, opening the possibility of horizontal gene transfer to other hosts by conjugation.

Figure 5.

Tn7 and CAST transposition mechanisms. Each system relies on different machinery to recognize its target and catalyze transposition. Tn7 uses TnsD. Subtype I-B CAST uses both TnsD and Cascade. Subtype I-F CAST uses only Cascade. Subtype V-K CAST uses Cas12k. Transposon excision and integration are catalyzed by TnsA and TnsB in Tn7, I-B, and I-F CAST, while V-K CAST uses only TnsB. TS, target strand; NTS, nontarget strand.

During the homing phase or TnsABCD transposition, TnsD recognizes the target, the glucosamine 6-phosphate synthase gene (glmS), creating a distortion in the DNA double helix¹⁸ (Figure 5). TnsC, a transposition regulator belonging to the AAA+ family, recognizes both TnsD and this DNA distortion and forms a ring around the DNA.^60,69 This process leads to the activation of the transposase complex comprising TnsA and TnsB.⁶⁹⁻⁷² TnsA is an endonuclease that cleaves the 5′ ends of the transposon. TnsB belongs to the DDE tranposase family and binds to a series of DNA repeats present at the left and right ends (LE and RE, respectively) of the transposon (Figures 4 and 5). Upon activation, TnsB catalyzes the cleavage of the 3′ ends of the transposon leaving free hydroxyl groups, later used to perform a nucleophilic attack downstream of the target sequence. Together, TnsA and TnsB excise the transposon from the donor DNA and TnsB inserts it into the attachment site located 25 bp downstream of the glmS transcriptional terminator (Figure 5).¹⁸ In TnsABCE transposition, TnsE is responsible for sequence-independent insertion into mobile plasmids by recognizing 3′ recessed ends and the sliding clamp protein during DNA replication.⁷³

With respect to CAST, target recognition relies on the CRISPR-Cas-inactivated effector, instead of TnsD or TnsE, except in subtype I-B, in which both CRISPR-Cas and TnsD target recognition are used.⁵²⁻⁵⁴ Generally, the inactivated CRISPR-Cas effector recognizes the target sequence based on the presence of a PAM and complementarity to the guide RNA. Subsequently, TniQ and TnsC are recruited by the DNA-CRISPR-Cas effector complex. Activation of TnsA and TnsB by TnsC is predicted to occur in a manner similar to that of Tn7. This results in transposon insertion a certain distance from the target sequence previously recognized by the effector complex (Figure 5).

Even though the three systems, CAST subtypes I-B, I-F, and V-K, share similar components, the molecular mechanisms leading to DNA integration differ. Target recognition is the most evident difference due not only to the divergence between the CRISPR-Cas subtypes but also the distinct functions of the TnsD/TniQ homologues and the usage of gRNA variants.^54,65 Moreover, donor DNA integration and resolution also differ depending on the presence or absence of TnsA in class 1 and class 2 systems.^74,75

CAST Targeting Mechanisms

Homing transposition (transposon integration in the host’s chromosome) in the subtype I-B system is supposed to work in the same way as in Tn7. In this system, the targeting function of TnsD has not been completely replaced by the CRISPR-Cas effector. Instead, the tnsD gene was duplicated and each copy performs a specialized function. Homing transposition uses one TnsD copy to recruit TnsC, TnsA, and TnsB to the target site in the bacterial genome. The other copy, TniQ, is used together with Cascade to target mobile genetic elements⁵⁴ (Figure 5).

In subtype I-F, the TnsD/TnsE targeting function has been completely replaced by the CRISPR-Cas effector. Homing transposition and mobile genetic element targeting are differentially achieved by categorizing the gRNA sequences encoded in the array⁶⁵ (Figure 4). New spacer acquisition results in new spacers being added to the leader-proximal end of the CRISPR array and leader-proximal repeat duplication. Given errors in replication, leader-distal repeats (“older repeats”) are increasingly more divergent. As a result, the crRNA sequences can be divided between a highly divergent crRNA encoded in the leader-distal end and the more recently acquired sequences with less degenerate repeats. Petassi et al.⁶⁵ found that the spacers from the divergent repeats were used for homing while the spacers from the less degenerate repeats were used for targeting MGEs. This categorization system is especially advantageous because I-F CAST gRNAs meant for transposon homing could potentially be used by the host’s canonical I-F CRISR-Cas system (i.e., a I-F Cascade complex comprising a catalytically active Cas3 nuclease). Because a gRNA coming from I-F CAST would target a DNA sequence in the host’s genome, its presence in a catalytically active Cascade complex would result in a DSB on the host’s target DNA. However, repeat divergence in I-F CAST renders the processed gRNA unusable by catalytically active I-F systems and therefore prevents target DNA cleavage.⁶⁵ As an additional advantage of gRNA categorization, repeat divergence can even increase transposition efficiency in specific cases with a divergent homing gRNA having integration frequencies higher than those of gRNAs containing typical repeats.⁶⁵

Subtype V-K also categorizes crRNA sequences albeit not in the same manner as subtype I-F.⁵⁴ This system comprises crRNAs encoded in a regular CRISPR array and an ectopic crRNA coding sequence not included in it (Figure 4). The difference between both types of crRNA lies in not only its position within the transposon locus but also the length of the repeat. The ectopic crRNA presents a shortened repeat and is used for homing transposition targeting tRNAs encoding sequences, while the regular array is used for transposition into MGEs. The evolutionary cause for subtype V-K gRNA categorization is, however, still not clear. According to Saito et al.,⁵⁴ avoidance of autoimmunity by another type V system is unlikely because no other type V systems co-occur in the current available genomes. Nonetheless, this might have been the case assuming the presence of an ancestral Cas12 nuclease during the first steps of Cas12k’s recruitment by the transposon.⁵⁴

Transposon Excision and Integration

Subtypes I-B and I-F rely on TnsA and TnsB to carry out transposon integration (Figure 5).^53,54 It is expected that the reaction mechanism for excision and insertion proceeds in the same way as in Tn7 because this system also comprises TnsA and TnsB as explained above. In contrast to subtypes I-B and I-F, TnsA is conspicuously absent in subtype V-K. This fact has implications for integration intermediate resolution given that no enzyme would cleave the 5′ end of the transposon DNA, leading to cointegrate transposition products.^74,75

Transposons encoding only TnsB or TnsB homologues have different strategies for dealing with this problem.⁷⁶⁻⁷⁸ In the case of Tn5053, the transposon contains resolvase TniR that uses res sites, also present in the transposon, to resolve the cointegrate^79,80 (Figure 4). Other transposases can generate hairpins to excise the remaining DNA and later insert it into the target DNA.⁸¹ Moreover, DNA replication and RecA-dependent recombination can also solve the cointegrate in the case of replicative transposition, where the transposon increases its copy number as opposed to “cut-and-paste” transposition in which there is only one copy of the transposon.^78,81

The cointegrate resolution mechanism in the subtype V-K system is still unknown. However, there are two points that suggest the V-K system might involve a specific resolvase present only in the natural host. First, even though the V-K transposon is present as a single copy in its natural host, when transposition is carried out in a heterologous system, 20% of insertion products are present as cointegrates.⁷⁵ Second, upon comparison of the V-K system components (TnsB, TnsC, and TniQ) to other transposons, this combination is most similar to Tn5053, which contains TniA, TniB (TnsB and TnsC homologues), and TniQ.⁷⁵ This could point to the existence of a resolvase not encoded in the V-K transposon but present in the natural host that would fill in the function of TniR. So far, this putative resolvase is unknown. Obviously, the understanding of this important stage during integration is key for the possible development of subtype V-K genome editors.

CAST Gene Editing

Notwithstanding the colossal advances in genome editing, most insertion techniques are still limited by sequence length, efficiency, DNA targeting/programmability, and homologous recombination dependency. For instance, current CRISPR-Cas techniques are programmable but depend on homologous recombination so they can be used on only actively dividing cells.⁸² Recombineering using the RecET or gamma-red system is also programmable but has a low efficiency.⁸³ Transposase usage is either too specific (not easily programmable) as is the case for Tn7¹⁸ or too unspecific as is the case for Tn5¹⁹ and PiggyBac.²⁰ CAST has been proposed as a promising strategy for overcoming the limitations of these systems. Because targeting depends on naturally inactivated RNA-guided nucleases, it can be programmed by changing the spacer in the crRNA/gRNA. Moreover, given that CAST has its own DNA integration system, it does not rely on homologous recombination and could therefore be used in virtually any cell type.^52,53

To date, the capabilities of CAST to perform gene editing have been tested only in different species of bacteria. In the first of these studies, Vo et al.⁸⁴ optimized subtype I-F achieving highly accurate and efficient DNA integration in bacteria, as well as multiplexed insertions and deletions. They called this system INTEGRATE (insertion of transposable elements by guide RNA-assisted targeting). INTEGRATE is composed of a single vector comprising all genes necessary for transposition (cas6–cas8, tnsA–tnsC, and tniQ), a crRNA array, and a donor DNA. INTEGRATE can transpose up to 10 kb cargo sequences with ∼100% efficiency, although multiple insertions occurring 5 kb apart have an only ∼20% efficiency due to transposition immunity (see CAST Gene Editing Challenges). Multiple insertions were also explored by means of an orthogonal system comprising subtype I-F and V-K transposons. The results of these experiments showed that these transposon subtypes do not display cross-reactivity; i.e., the transposase from subtype I-F does not mobilize the donor DNA from subtype V-K and vice versa. However, subtype V-K displayed a larger number of off-target integration events.

INTEGRATE can be also used to create deletions in combination with Cre recombinase. Briefly, Vo et al.⁸⁴ used two target sites to perform a double insertion of a transposon with a LoxP site within its cargo. Afterward, the sequence comprised between the two LoxP sites was excised by Cre recombinase. Between 2.4 and 20 kb could be excised using this methodology proving INTEGRATE’s versatility as a genome editing tool.

Genome Editing in Bacterial Communities

Functionality in a broad range of species is a desirable feature for a genome editing tool. However, many traditional techniques fail in this regard because most microorganisms cannot be cultured using current protocols. Vo et al.⁸⁴ and Rubin et al.⁸⁵ tried to circumvent this problem by using INTEGRATE and DART (DNA-editing all-in-one RNA-guided CRISPR-Cas transposase), a system similar to INTEGRATE but comprising barcodes for environmental-transformation sequencing (ET-seq), in bacterial communities. Using either electroporation or conjugation, INTEGRATE and DART successfully inserted a cargo DNA into a bacterial consortium with no off-targets when using subtype I-F. However, when using subtype V-K, a vast majority of insertions were off-target. According to these results, INTEGRATE and DART have the potential to shed light on the function and fitness of specific genes in microbial communities while eliminating the limitations of isolating unculturable bacteria.^84,85

Protein Engineering

Protein engineering has been used in two ways to optimize subtype V-K. First, additional components have been added to the excision and integration system to avoid cointegrates, and second, protein fusions have been designed to simplify the targeting system.⁸⁶

In the first approach, given that subtype V-K lacks TnsA, a high percentage of its transposition products are cointegrates. In an effort to supply TnsA’s function, Tou et al.⁸⁶ fused TnsB to a homing endonuclease mutant with specific ssDNA nickase activity (nAniI). These combinations led to a dramatic decrease in the level of cointegrate products from 20% to <1% when using target sequences present in plasmids and 3% when using genomic target sequences. However, only fusions of nAniI to the N-terminus of TnsB supported transposition while C-terminal fusions did not. This could be explained by pull-down assays performed by Querques et al.⁶⁷ in which interaction between TnsC and TnsB was mediated by the C-terminus of the latter, meaning that nAniI could block TnsB:TnsC interaction.

In the second approach, Cas12k has been fused to either TniQ or TnsC and the insertion efficiency was measured. However, none of the constructs tested yielded an efficiency higher than that of the separate components.⁸⁶ This could be due to the wrong orientation of the proteins imposed by the design of the construct, misfolding, steric impediment, etc. Further experimentation is needed to design simpler systems with efficiencies equal to or higher than those of the natural ones. Structural data are particularly important to guide CAST engineering efforts, as artificial intelligence programs have not yet developed reliable modules for predicting protein–nucleic acid interactions.

So far, the structures of the individual components of subtype V-K as well as the Cas12k-TnsC-TniQ complex and the I-F Cascade-TniQ complex are known.^{62,66−68,87} These complexes reflect different steps in transposon integration particularly related to target recognition. Undoubtedly, this information will be fundamental to the redesign of CAST protein interactions and DNA binding. For instance, complexes comprising Cas and transposon proteins with affinities higher than those of the natural ones could be obtained by mutating the amino acids present in the interfaces between the proteins. This could potentially lead to complexes with higher stabilities or to simplified complexes where regions not necessary for protein interactions are removed. Obtaining the structure of complexes comprising TnsA and TnsB reflecting the different states in transposon excision and integration is the next step to understanding CAST’s molecular mechanism and consequently facilitating the design of optimized CAST variants suitable for genome editing.

CAST Gene Editing Challenges

Despite the recent advances in the determination of the integration mechanism of CASTs, a number of challenges are yet to be surpassed if this system is to be used for genome editing. Because TnsB activity relies on the sequence of LE and RE, CAST genome editing is not a scarless editing method. Transposon integration both leads to insertion of six to eight DNA repeats corresponding to the LE and RE and causes target site duplications in the insertion site (Figures 4 and 5).^52,53 Additionally, integration directionality needs to be addressed. Despite certain insertion bias in the subtype V-K system, i.e., LE being inserted proximal to the target site and RE being inserted distal to the target site,⁵² transposition in subtype I-F still presents products with LE and RE being inserted either way.⁵³ This phenomenon represents a problem for applications such as fixing exons or regulatory sequences. Generating a CAST system with integration unidirectionality would therefore be ideal. Accordingly, further studies of the TnsB insertion mechanism are still needed to be able to reduce the donor sequence requirements while ensuring the integration of the transposition product in one sense only.

With respect to TnsC, it constitutes an interesting challenge when CAST simplification is the aim. In the case of Tn7, it controls transposition immunity; i.e., no further transposon copies are inserted into the target locus after one copy has already been inserted.¹⁸ This mechanism involves TnsC binding and unbinding to the target DNA. Prior to transposition, TnsC binds to the target DNA. However, TnsB promotes TnsC’s ATP hydrolytic activity, which in turn leads to the disassembly of the TnsC ring/filament.⁶⁷ Because TnsB remains bound to the transposon LE and RE in the post-transposition complex, TnsC cannot bind the DNA in the vicinity of the integrated transposon and consequently cannot recruit additional transposase complexes carrying another copy of the transposon.⁶⁰ This means that trying to eliminate TnsC from the system by engineering a Cas12k or Cascade complex that interacts directly with TnsB might lead to multiple insertions of the donor DNA given that no component would regulate the transposase activity, only its targeting. However, if TnsC is to be kept, it might cause gene expression dysregulation because in the case of the V-K system, it wraps around DNA not only downstream of the targe sequence but also nonspecifically to any other DNA sequence.^67,68 This might not be the case for the class 1 systems that are more similar to Tn7 and whose TnsC requires the presence of TnsD or a D-loop to interact with DNA.

Finally, even though the natural hosts of the V-K systems display only single on-target copies of the transposon, when they are used in other species, most insertions are off-target, in contrast to the I-F system.^84,85 This could indicate that the I-F system would be more amenable to be used for biotechnological purposes.

Future Directions

The natural diversity of nucleoproteins is an unlimited source of new systems waiting to be harnessed for biotechnological applications. Chief among them are CRISPR-Cas and transposons, which have developed sophisticated mechanisms for DNA excision and integration during their evolutionary history, and from which countless gene editing applications have been produced. The recently described OMEGA and CAST transposons constitute new avenues for circumventing current gene editing challenges. The former has the potential to minimize the size of programmable nucleases, while the latter facilitates targeted DNA insertions of large payloads without relying on homologous recombination.

However, as prokaryotic systems, both face the challenge of accessing the chromatin-packed DNA when used in eukaryotic cells. Experimental evidence has shown that nucleosomes directly impede DNA binding and cleavage, while chromatin remodeling can restore editing by Cas9 and Cas12a.⁸⁸⁻⁹⁰ How the new editors reviewed here could behave when facing chromatin is unknown. In the case of the OMEGA systems, while they have been shown to be able to perform editing in eukaryotic cells, their editing levels are low when compared to those of the classical Cas9 and Cas12a tools. This is similar for other RNA-guided nucleases, such as Cas12j,⁴⁸ which share with the OMEGA family members a reduced protein moiety. A possible explanation for this lower editing level is that the force produced by conformational changes in these small protein subunits might not be enough to displace DNA-bound proteins preventing access to the target sequence in a chromatin environment.

In the case of CAST, their behavior in a chromatin context is unknown. It is possible that CAST might encounter difficulties in accessing close chromatin based on data from other Cas proteins showing how chromatin accessibility decreases gene editing efficiency.⁹¹ A similar phenomenon is observed with Tn7, which can use gfpt-1 and gfpt-2 sequences (the human analogues of glmS) as target sequences in vitro but displays impaired transposition when target DNA is reconstituted in nucleosomes.⁹² In contrast, an engineered Tn7 element has been used to generate a mutant library in 45% of the yeast genome in vivo.⁹³ Taken together, these results suggest that CAST systems could have the potential to overcome the chromatin context, albeit with variable efficiency.

Nevertheless, further investigations addressing the behavior of OMEGA and CAST systems in a chromatin context need to be addressed in depth to evaluate their properties in genome editing. We expect that the current research on both systems will certainly accelerate their redesign as new gene editing tools in basic research and in biomedical applications.

Glossary

Abbreviations

CRISPR

clustered regularly interspaced short palindromic repeats

Cas

CRISPR-associated proteins

OMEGA

obligate mobile element-guided activity

CAST

CRISPR-Cas-associated transposon

ZFN

zinc finger nuclease

TALEN

transcription activator-like effector nuclease

gRNA

guide RNA

DSB

double-strand break

DRP

DSB repair pathways

NHEJ

nonhomologous end joining

HDR

homology-directed repair

isLE

left elements from IS transposons

isRE

right elements from IS transposons

LE

left end

RE

right end

IscB

insertion sequences Cas9-like similar to TnpBs

ncRNA

noncoding RNA

crRNA

CRISPR RNA

leRNA

left element RNA

reRNA

right element RNA

IsrB

insertion sequence RuvC-like OrfB

IshB

insertion sequence HNH-like OrfB

glms

glucosamine 6-phosphate synthase gene

INTEGRATE

insertion of transposable elements by guide RNA-assisted targeting

DART

DNA-editing all-in-one RNA-guided CRISPR-Cas transposase

Author Contributions

F.T.-C. and A.C. contributed equally to this work. The manuscript was written through contributions of all authors. A.C. wrote the draft of the first part (OMEGA systems). F.T.-C. wrote the draft of the second part (CAST systems). A.C. and F.T.-C. prepared the figures. G.M. provided feedback on the initial drafts and contributed to the writing of the final manuscript. All authors read and corrected the manuscript and approved the final version of the manuscript.

The Novo Nordisk Foundation Center for Protein Research is supported financially by the Novo Nordisk Foundation (NNF14CC0001), a Distinguished Investigator award (NNF18OC0055061 to G.M.), and the Copenhagen Bioscience PhD Programme (NNF19SA0035440 to F.T.-C.).

The authors declare the following competing financial interest(s): G.M. is co-founder and member of the advisory board of TwelveBio. The University of Copenhagen has patents issued for the development of Cas12j CRISPR tools in which G.M. and A.C. are named as inventors. F.T.-C. declares no competing interests.