Casπ cleaves DNA targets using 5′ C-rich PAM distinct from other Cas12 variants

To further determine the biochemical characteristics of Casπ, we started with identifying the PAM preference of both orthologs using a plasmid library containing five randomized DNA nucleotides upstream of the protospacer (Fig. 2a; Supplementary information, Fig. S2a). Deep sequencing analysis suggests that both Casπ effectors recognize the 5′-CCN-3′ PAM (Fig. 2a; Supplementary information, Fig. S2b, c and Table S3). Specifically, for Casπ-1 effector, the strictness of PAM requirement increases when increasing the salt concentration in the cleavage buffer (Supplementary information, Fig. S2b). Notably, this C-rich PAM preference for Casπ is different from the T-rich PAM preference for all reported type V nucleases (Supplementary information, Fig. S2d), which will help expand the targeting scope for type V-based technologies. Using the most favorable CCC PAM determined by plasmid screening assay, we observed efficient cleavage activity for both Casπ effectors on the double-stranded DNA (dsDNA) target even compared to the large Lachnospiraceae bacterium Cas12a (LbCas12a, 1228 aa) effector (Fig. 2b; Supplementary information, Table S3). A further screening showed that both Casπ effectors can only robustly cleave the dsDNA target with CCC or CCT (CCY) PAM, indicating a more stringent PAM requirement on dsDNA target (linearized substrate) compared to plasmid target (negative supercoiled substrate) (Supplementary information, Fig. S2e, f). Gel analysis of the cleavage products from the DNA non-target strand (NTS) and target strand (TS) showed that both effectors generate a staggered cut on the dsDNA (Fig. 2c). Consistent with the deep sequencing analysis result for plasmid cleavage (Supplementary information, Fig. S2a, g, h), the exact cleavage sites locate at 11–14nt downstream of the PAM on the NTS and 2–4nt downstream of the protospacer on the TS, thus leaving a 5′ single strand overhang of 6–12nt on the products (Fig. 2d, e). Moreover, we observed the single-stranded DNA (ssDNA) TS cleavage (cis-cleavage) by both effectors, and the cleavage efficacy and pattern are comparable to the TS cleavage within dsDNA (Supplementary information, Fig. S2i).

Open in a separate window

Fig. 2

Casπ effector cleaves dsDNA with 5′ C-rich PAM.

a Graphical representation of the in vitro PAM depletion assay and the resulting PAMs for both Casπ effectors. b Top, in vitro dsDNA cleavage comparison between LbCas12a, Casπ-1 and Casπ-2 revealed by denaturing PAGE. NTS denotes the non-target strand DNA which is cy5-labeled at 5′ end. P denotes the cleavage products. Bottom, the plot of NTS dsDNA cleavage efficiency by LbCas12a, Casπ-1 and Casπ-2 (n=3 each; mean±SD). c Cleavage site mapping of Casπ-1 and Casπ-2 revealed by denaturing PAGE. Lane M shows cy5-labeled marker. d, e The cleavage sites for NTS and TS of Casπ-1 (d) and Casπ-2 (e) (marked in red arrows, large arrows indicate high probability of cleavage, PAM in blue) suggested by both gel analysis (n=3 each) and NGS analysis.

Casπ exhibits substantial tolerance of biochemical conditions with efficient trans-activity

To explore the application potential of Casπ, we performed a general screening for DNA cleavage by both effectors under various biochemical conditions in vitro. For RuvC-containing nucleases, divalent ions are typically important to coordinate the catalytic core for DNA hydrolysis. The ion screening suggested that either Mg²⁺ or Mn²⁺ can robustly activate the nuclease activity in Casπ (Fig. 3a; Supplementary information, Fig. S3a). Further experiments also showed that Casπ overcomes several disadvantages reported in other Cas nucleases. Normally, one common drawback of most compact CRISPR effectors (< 1000aa) is their limited tolerance range of salt concentration in vitro. For example, the compact AsCas12f and CasPhi (Cas12j) prefer low salt concentration (< 150mM NaCl) for detectable dsDNA cleavage, due to their limited dsDNA unwinding ability.^12,15 Meanwhile, PlmCasX (Cas12e) robustly unwinds the dsDNA for cleavage in high salt concentration condition (300–450mM NaCl), but gets denatured and precipitated in low-salt buffer (< 300mM NaCl) as seen.¹¹ In contrast, the compact Casπ persists a stable effector status for dsDNA cleavage in a wide range of salt concentrations from 50mM to 300mM NaCl (Fig. 3b; Supplementary information, Fig. S3b). Furthermore, unlike many Cas nucleases which get denatured and precipitated in solution when being concentrated to a high protein concentration (50–100μM), both Casπ nucleases behave robustly upon physical enrichment (30kD molecular weight cut-off centrifugal filters; see Materials and methods).¹¹ Therefore, we often stock the Casπ nucleases at the ultra-high protein concentration of 300μM for the following convenient use. Moreover, a huge limitation of employing biomolecular tools in different exogenous scenarios is that they only work efficiently in the temperatures that their source bacterial hosts prefer. To our surprise, although discovered in mesophilic environment, Casπ tolerates temperatures from 25°C even to 65°C (Fig. 3c; Supplementary information, Fig. S3c).

Open in a separate window

Fig. 3

Biochemical screening of Casπ nuclease activity.

a In vitro dsDNA cleavage by Casπ effectors using different divalent ions revealed by denaturing PAGE. NTS denotes the non-target strand DNA which is cy5 labeled at 5′ end. Bottom, P means cleavage products (n=3 each). b In vitro dsDNA cleavage by Casπ effectors in the buffers with different salt concentrations (n=3 each). c In vitro dsDNA cleavage by Casπ effectors at different temperatures (n=3 each).

To explore the cleavage specificity by Casπ effectors, we first performed the single mismatch screening on the DNA protospacer. The single mismatches between sgRNA and nucleotides 1–8 of the target DNA at the PAM-proximal region largely abolished the nuclease activity of Casπ, which suggests a ‘seed region’ located in the position of nucleotides 1–8 of the target DNA (Fig. 4a, b).^24,25 Besides, single mismatches between nucleotides 13–16 at the PAM-distal region also significantly decreased the cleavage efficiency of Casπ (Fig. 4a, b). Additionally, many Cas12 nucleases cleave random ssDNA (trans-activity) when activated by ssDNA or dsDNA target (activator), which has been harnessed for nucleic acid diagnosis.^10,26 Noteworthily, though compact in size, Casπ effectors show comparable trans-activity to the widely used LbCas12a with either ssDNA or dsDNA activator (Fig. 4c, d), indicating Casπ’s potential as a nucleic acid diagnosis tool. In summary, compared to many reported Cas effectors, Casπ presents a substantial advantage of flexibility and robustness for in vitro applications.

Open in a separate window

Fig. 4

Specificity of DNA cleavage by Casπ.

a Top, cleavage assay using single mismatched dsDNA targets by Casπ-1 effector at 1h. Bottom, the bar plot of cleavage efficiency (WT indicates that sgRNA and dsDNA target are fully paired. Numbering means the position of the mismatches between sgRNA and dsDNA target used in this assay; n=3 each; mean±SD). b Top, cleavage assay using single mismatched dsDNA targets by Casπ-2 effector at 1h. Bottom, the bar plot of cleavage efficiency (n=3 each; mean±SD). c Top, the trans-DNA cleavage by LbCas12a, Casπ-1 and Casπ-2 on ssDNA substrate with dsDNA activator (S means substrates and P means products). Bottom, the plot of trans-ssDNA substrate cleavage efficiency by LbCas12a, Casπ-1 and Casπ-2 with dsDNA activator (n=3 each; mean±SD). d Top, the trans-DNA cleavage by LbCas12a, Casπ-1 and Casπ-2 on ssDNA with ssDNA activator (S means substrates and P means products). Bottom, the plot of trans-ssDNA substrate cleavage efficiency by LbCas12a, Casπ-1 and Casπ-2 with ssDNA activator (n=3 each; mean±SD).

Casπ orthologs are active for DNA manipulation both in prokaryotic and eukaryotic cells

To further explore whether the compact Casπ effectors can be employed for DNA cleavage in prokaryotes, we performed a plasmid interference assay using E. coli BW25141 strain carrying a ccdB toxin plasmid with arabinose-inducible promoter (Fig. 5a). While few survival clones were observed in the non-targeting control due to ccdB toxicity, expressing either Casπ-1 or -2 with the ccdB-targeting sgRNA led to significantly more survival clones (Fig. 5a, b; Supplementary information, Fig. S4a, b). This plasmid interference activity was further verified via PCR analysis (Supplementary information, Fig. S4c).

Open in a separate window

Fig. 5

Casπ facilitates DNA manipulation in bacterial and human cells.

a Schematic illustration of the plasmid interference assay. b Bacteria survival assay on culture plates containing 10mM arabinose. NT, plasmid with Casπ and non-target sgRNA; ccdB, plasmid with Casπ and sgRNA targeting ccdB gene. Dilution gradient is shown on the left. c Scheme of Casπ-mediated EGFP lighting up in HEK293A cells. d EGFP lighting up results. Transfection of plasmids carrying Casπ and sgRNA activated EGFP (frame restored) with detectable green fluorescence signal. Both the bright field (BF) and fluorescent images of cultured cells are shown. e Editing efficacies determined by NGS from 5 targets mediated by Casπ-1, Casπ-2, Cas12a and Cas9 (n=3 each, mean±SD). f Analysis of INDELs generated by Casπ-1 editing within all 15 editing experiments. Left, pie chart showing percentage of each INDEL within all 15 editing experiments analyzed by NGS (Mixed means mixed editing with both insertion and deletion). Right, INDEL size distributions within all 15 editing experiments. g Analysis of INDELs generated by Casπ-2 editing within all 15 editing experiments. h Analysis of INDELs generated by Cas12a editing within all 15 editing experiments. i Analysis of INDELs generated by Cas9 editing within all 15 editing experiments.

Next, to investigate the genome-editing ability of Casπ in eukaryotic cells, we constructed a HEK293A cell line with the genome-integrated ORF containing the MYH8 exon and the out-of-frame EGFP (Fig. 5c; see Materials and methods). Expression of either Casπ-1 or -2 with sgRNA targeting the MYH8 exon efficiently lit up the cells with in-frame EGFP signal, which indicates that the DNA insertions or deletions (INDELs) were generated by Casπ editing (Fig. 5d). To compare the editing activity between Casπ effectors and the well-developed LbCas12a and SpyCas9 effectors, we designed five parallel targeting sites across the MYH8 exon (Supplementary information, Fig. S4d and Table S5). The edited genomes were PCR amplified, and the editing efficacies were validated by T7 endonuclease I (T7E1) assays and quantified by targeted sequencing (Supplementary information, Fig. S4e). Next-generation sequencing (NGS) revealed that both Casπ effectors introduced INDELs nearby the cleavage sites in TS as observed in vitro (Fig. 2d, e; Supplementary information, Fig. S4f, g). Overall, SpyCas9 presents an average editing efficacy of 30.9% across the five sites and a maximum efficacy of 37.1% at site 4 (Fig. 5e). LbCas12a shows an average editing efficacy of 6.7% and a maximum efficacy of 16.8% at site 5 (Fig. 5e). Casπ-1 shows an average editing efficacy of 2.7% and a maximum efficacy of 8.0% at site 1 (Fig. 5e). Casπ-2 shows an average editing efficacy of 5.4% and a maximum efficacy of 15.4% at site 2 (Fig. 5e). The combined INDEL analysis on the five targets shows that SpyCas9, LbCas12a and Casπ effectors mainly generate deletions on the targeted genome (Fig. 5f–i; Supplementary information, Fig. S4h). Of note, SpyCas9 may generate long deletions of ~40nt, while Cas12a and Casπ editing dominantly contributes to shorter deletions of < 25nt (Fig. 5f–i). Further, three more endogenous targets on B2M and TP53 genes were edited by Casπ effectors and the editing efficacies were quantified by NGS (Supplementary information, Fig. S4i–k).

Therefore, even without any optimizations, the naive version of compact Casπ effectors works comparably to LbCas12a and maximumly reaches over half of the editing ability of the well-developed SpyCas9, supporting Casπ’s potential to be a competitive and compact DNA manipulation platform with further engineering.

Unique structural domains in Casπ responsible for DNA interference

To understand the molecular details underlying the DNA targeting behavior by Casπ effector and provide structural information for editing optimization in future studies, we achieved the cryo-EM map of the R-loop complex containing the deactivated Casπ-1 (D537A, E643A), sgRNA and dsDNA at 3.4-Å resolution (Supplementary information, Figs. S5a–c, S6a–e). The EM density of Casπ R-loop complex is well resolved, which allows us to build the complete atomic model ab initio (Fig. 6a–c; Supplementary information, Fig. S6e, f and Video S1). Consistent with the primary sequence BLAST suggesting no significant similarity to reported proteins, Casπ also exhibits a unique 3D architecture compared to other CRISPR-Cas nucleases revealed by structural alignment with Dali server (Supplementary information, Fig. S7a, b).²⁷ Only moderate similarity was observed between Casπ and Cas12 nucleases, mainly within the RuvC domain and oligonucleotide binding domain (OBD) (Supplementary information, Fig. S7c, d). Then, referring to CasX (Cas12e) which shares the top structural similarity with Casπ and also uses a large RNA guide (Supplementary information, Fig. S7e), we further located the conserved bridge helix (BH) element and four unique structural domains within Casπ, including the ‘lock-catch’ (LC) domain, proline-rich string (PRS), Helical-I domain and NTSB (non-target strand binding domain) chimera (HNC), and Casπ (Pi) C-terminal (PCT) domain (Fig. 6a–c; Supplementary information, Video S1).

Open in a separate window

Fig. 6

The structure of Casπ nuclease.

a The domain organization aligned with primary sequence. LC domain is colored in dark orange, PRS in red, HNC in yellow, OBD in purple, RuvC in dark green, BH in dark blue, and PCT in pink. b The base pairing details for the R-loop region. The sequences for NTS (light orange color), TS (light green color) and sgRNA spacer (cyan color) are presented. The PAM region is marked with rectangle. c The atomic model for Casπ R-loop complex. The protein domains, DNA and sgRNA are colored referring to a and b. The front and top views are presented. d The structural details within Casπ RuvC domain (dark green color). The three catalytic residues were highlighted with dark yellow color. In this complex, D537 and E643 were mutated to alanine. e The molecular details for PAM recognition. The amino acids involved in dG(2) and dG(3) recognition are labeled. The key hydrogen bonds are shown by dashed lines. f The structural details within Casπ HNC domain (yellow color). g The structure of Casπ PCT domain (pink color) and TS DNA loading model. The 5′ end of TS DNA is hypothetically modeled as dashed line (light green) and loaded into RuvC nuclease pocket by PCT domain.

The RuvC domain in Casπ displays a canonical DNA cleavage pocket with the conserved triplet of catalytic residues D537, E643 and D796 (Fig. 6d). D537 and E643 are mutated to alanine in this study for stabilizing the complex (Supplementary information, Fig. S5a–c). Different from other type V CRISPR nucleases which prefer T-rich PAM, two unique residues in Casπ OBD domain, Arg390 and Arg392, were observed to recognize the two guanine nucleotides (dG(2) and dG(3) in the TS) complementary to the CCN PAM (in the NTS) (Fig. 6b, e). Both the single mutations (R390A or R392A) and double mutation (R390A/R392A) totally abrogated the nuclease activity of Casπ (Fig. 6e; Supplementary information, Fig. S8a, b). In addition, the side chain of Gln133 inserts into the downstream site of PAM duplex, which may lead to local dsDNA melting for sgRNA–spacer invading as discussed in other type V nucleases (Fig. 6e).²⁴

The HNC domain, which presents as a structural chimera of Helical-I domain and NTSB domain in CasX, interacts with both the ‘seed region’ of sgRNA–DNA heteroduplex at the PAM-proximal region and the backbone of DNA NTS to stabilize the R-loop conformation (Fig. 6f; Supplementary information, Fig. S8c). Meanwhile, neither primary sequence BLAST nor structural search for PCT domain (Trp703–Asp794 and Arg836–Ile867) reveals any suggestive similarity to annotated proteins, indicating that this unique feature is specific to Casπ nucleases (Supplementary information, Fig. S7b). Since the PCT domain sits at similar primary and spatial locations to the target-strand loading (TSL) domain of CasX (Supplementary information, Fig. S8d), we then hypothesize that the PCT domain may help with the target strand loading into RuvC nuclease domain (Fig. 6g), and this needs to be further explored in future studies.¹¹

Casπ presents a ‘bracelet’ architecture encircling the nucleic acid target

Strikingly, a long ‘proline-rich string’ (PRS) loop composed of 69 aa (Pro72–Trp140) is largely resolved in the EM map (Fig. 7a; Supplementary information, Fig. S6f and Video S1). There are 14 prolines and 17 charged residues within this ‘string’ which makes it adopt high structural accessibility and electrostatic capacity to tie up the whole complex via multi-interactions with other protein domains, sgRNA and also the DNA target (Fig. 7a; Supplementary information, Fig. S9a). Directly next to the PRS N-terminus, Casπ folds into a two-helix structure (Met1–Asp71) which serves as a ‘lock’ and tightly interacts with a three-helix ‘catch’ module (Val317–Ala375) through multiple interactions, such as the hydrogen bonds (E28 and Y61 interact with R339 and E337, respectively) (Fig. 7b), the charged interactions and van der Waals interactions (not shown in the figure). Via this unique structure never observed in other Cas nucleases, the ‘lock-catch’ (LC) domain further locks the ‘tie-up’ conformation mediated by the PRS (Fig. 7a; Supplementary information, Fig. S9a and Video S1). Moreover, similar to the Helical II domain in CasX (Cas12e),¹¹ the ‘lock’ part in LC domain also intensively interacts with the sgRNA stem to stabilize the assembly of R-loop complex (Fig. 7b; more details discussed in next section). Remarkably different from the canonical ‘two-lobe’ architecture for Class 2 Cas nucleases, the PRS and LC domains string all other protein domains together, and make the Casπ fold as a locked ‘bracelet’ encircling the nucleic acid target (Fig. 7c, d; Supplementary information, Fig. S9b, c).

Open in a separate window

Fig. 7

The structure of ‘bracelet’ architecture of Casπ.

a The structural distribution of PRS domain across the Casπ R-loop complex. For clear presentation, the sgRNA and DNA are shown in cyan. ‘Lock’ part in LC domain is shown in light orange and the ‘catch’ part in deep orange. PRS is shown in red and all other protein domains in gray. The primary sequence for PRS loop is colored in red and shown at the bottom. The prolines in PRS are specially depicted in blue. The positively charged amino acids are labeled with ‘+’ and negatively charged ones with ‘–’. b The structural details of LC domain. The elements are colored referring to a. Side chains of the amino acids involved in the interactions between ‘lock’ and ‘catch’ are shown, and the formed hydrogen bonds are presented as dashed lines. The charged interactions and van der Waals interactions are not shown. The cartoon shapes for the ‘lock (L)’, ‘catch (C)’, PRS and sgRNA stem are outlined and presented at the top left. The domain organization of ‘lock (L)’, ‘catch (C)’ and PRS is presented at the bottom. The interactions between ‘lock (L)’ and ‘catch (C)’ are connected with orange curves. c A cartoon model for Casπ ‘bracelet’. The PRS is modeled as a half-ring colored in red. The LC is modeled referring to b, and all other domains are modeled as a half-ring colored in gray. The nucleic acid target is also modeled and labeled, accordingly. The RuvC active sites are indicated in green color. d The cartoon model of ‘two-lobe’ architecture for canonical Class 2 Cas nuclease. The protein part was colored in gray and nucleic acid part in cyan. The RuvC active sites are indicated in green color.

Strictness for PAM preference varies in different scenarios

PAM sequence is essential for dsDNA targeting by Class 2 Cas nucleases, and it is often determined by the cleavage of plasmid library containing randomized PAM either in vitro or in vivo. In our experience, Cas effectors usually show more robust cleavage on the plasmid target than linearized dsDNA,⁸ as plasmids contain melting bubbles in the supercoil conformation.²⁸ Compared to the plasmid, a more stringent PAM requirement was observed on the linearized dsDNA target (Supplementary information, Fig. S2e, f). Moreover, we also found that the dC gradually dominated the third position of the PAM in the depletion analysis for Casπ-1 while increasing the salt concentration in the cleavage buffer, which indicates a more stringent PAM preference for Casπ-1 effectors in high-salt buffer (Supplementary information, Fig. S2b). Similar patterns were observed in CasX enzymes (unpublished data). Referring to previous biophysical studies, either linearizing the plasmid (relax the supercoil and re-anneal the bubbled strands in plasmids) or increasing the salt concentration (stabilize the dsDNA conformation) may contribute to ‘tougher’ targets for Cas effectors to unwind.²⁸ Therefore, we would suggest that a stringent PAM sequence determined in the ‘tough’ condition (linearized dsDNA target in the buffer with the highest salt concentration that Cas effectors can tolerate) may be the prioritized choice for gene editing application.

Casπ detection and phylogenic analysis of type V CRISPR systems

The assembled contigs were scanned for Cas nucleases using HMM profiles, which were built using the HMMER,⁴⁰ based on Cas nuclease sequence alignments from Clustal Omega (1.2.4).⁴¹ CRISPR arrays were identified using local version of the CRISPRCasFinder (4.2.20) and CRISPRidentify (v1.1.0).^42,43 Loci that contained both cas1 and the CRISPR array were further analyzed to identify the proteins located within the range from 20,000nt upstream to 20,000nt downstream of the CRISPR array. Potential functions of these proteins were annotated by HMMs and the local version of eggNOG mapper (2.1.6, eggNOG DB version: 5.0.2, MMseqs2 version: 13.45111).^44,45 Proteins larger than 600aa were selected as potential Class 2 Cas nucleases with nucleic-acid interference activity, and were further clustered by phylogenetic analysis.

For phylogenetic analysis, sequences of reported Cas nucleases were collected from UniProt database by searching keywords of each nucleases, like Cas9 and Cas12a.^{10,12,13,46,47} Sequence alignment of Casπ with the selected type V Cas nucleases was generated using Clustal Omega (1.2.4).⁴¹ Phylogenic reconstruction was performed using IQ-TREE2 (2.0.7) with VT+F+R7 as the substitution model and 1500 bootstrap sampling.⁴⁸ Reconstruction result was visualized and edited using iTOL v6.5.8.⁴⁹

Protein sequence and CRISPR repeat analysis

The protein and CRISPR repeat sequences of four Casπ orthologs were analyzed by Clustal Omega server with default parameters,⁴¹ and the two heatmaps illustrating the sequence similarity were built using the similarity score matrix (Sequences shown in Supplementary information, Table S1). For protein alignment with other type V CRISPR, the protein sequences of four Casπ orthologs were aligned with LbCas12a, AsCas12a, AaCas12b and DpbCas12e proteins using NCBI COBALT program,²² and the key amino acids in RuvC domains of Casπ were inferred from the alignment results.^7,11,22,50

tracrRNA identification and PAM prediction

For CRISPR-Casπ system, tracrRNA 3′-region was determined by anti-repeat identification, transcriptome mapping and promoter prediction. Anti-repeats were searched against a 5kb window upstream of the CRISPR locus using blastn with (E-value<0.2).¹⁸ Subsequently, the meta-transcriptomic reads of the sludge sample were extracted and mapped to their native genome locus around the anti-repeat region to analyze the tracrRNA expression. The transcript coverage was calculated by log₁₀ formula. Finally, the 5′-boundary of tracrRNA was determined by promoter prediction using BDGP-Promoter Prediction program.⁵¹ All tracrRNAs were determined in this manner as shown in Fig. 1c and the sequences were shown in Supplementary information, Table S1.

To predict the PAM sequence for Casπ-1 and Casπ-2, all the spacers present in both CRISPR arrays were manually extracted and aligned against the default databases using CRISPRTarget to search the potential protospacer sequences.²³ Sequences 3bp upstream of the identified protospacers were extracted and aligned to predict the PAM sequences. The PAMs ranking at the top for both Casπs were further used for plasmid cleavage in vitro.

Plasmid construction

Bacterial and human codon-optimized casπ-1 and casπ-2 genes were ordered from Sangon Biotech. For Casπ protein expression in E. coli, casπ genes were cloned into pET28a-based vector with an N-terminal hexa-histidine tag and a SUMO tag by homologous recombination (One Step Seamless Cloning Mix, CWBIO). For the D537A and E643A mutations in RuvC domain, R390A and R392A mutations in OBD domain of Casπ-1, mutated fragments were PCR amplified via mutagenetic PCR primers containing mutated sequences and inserted into pET28a-based vector by homologous recombination. For PAM depletion assay, the plasmid library containing five randomized nucleotides upstream of the target sequence was constructed as previously described.⁵² For in vitro plasmid cleavage, pUC19-based plasmids containing target sequence with different PAMs were constructed via homologous recombination. For bacterial plasmid interference, pBAD-driven arabinose inducible ccdB toxin plasmid (p11-LacY-wtx1) was requested from Prof. Wei Li group in the Institute of Zoology, Chinese Academy of Sciences.⁵³ casπ genes were cloned into MCSI of pCDFDuet vector by Gibson assembly with a sgRNA region, containing 2 SapI sites for target spacer exchange by Golden Gate, inserting into MCSII of pCDFDuet (sgRNA spacer sequences were listed in Supplementary information, Table S5).

For constructing the EGFP report cell line, the CMV-driven fusion fragment of MYH8 (270bp), a flanking sequence (32bp) and EGFP (1436bp) was cloned into psi-LVRU6MP vector by Gibson assembly. For cell editing assay, plasmid vector was obtained from circular PCR amplification of pBLO62.5 (Addgene plasmid# 123124) with two primers respectively pairing to N-terminal and C-terminal NLS sequence.⁸ Subsequently, Casπ (SpyCas9 or LbCas12a) genes were inserted into the region downstream of the CMV promoter and N-terminal NLS by homologous recombination. Then, sgRNAs (containing 2 SapI sites for spacer insertion) were inserted into the circular PCR-amplified vector containing Casπ (SpyCas9 or LbCas12a) genes with a U6 promoter and a poly-T terminal signal by homologous recombination. Primers containing the target spacer sequences were annealed and phosphorylated prior to Golden Gate assembly (SapI restriction sites) for stuffer–spacer exchange insertion (target protospacer sequences were listed in Supplementary information, Table S5).

A list of plasmids and a brief description are summarized in Supplementary information, Table S4.

Protein expression and purification

Casπ expression plasmids were transformed into E. coli BL21(DE3) (TIANGEN) and incubated overnight at 37°C on LB-Kan⁺ agar plates (50μg/mL Kanamycin). Single colony was overnight cultured as seed in LB-Kan⁺ medium (50μg/mL Kanamycin) at 37°C. Each 1L of LB-Kan⁺ medium (50μg/mL Kanamycin) was then inoculated with 100mL seed culture and incubated at 37°C. As the culture OD reached 1.0, the protein expression was induced with 0.2mM IPTG for 20h at 16°C. Bacterial cells were collected and resuspended in lysis buffer (800mM NaCl, 20mM HEPES-Na, pH 7.5, 10% glycerol, 40mM imidazole, 1mM TCEP and 1mM PMSF) and lysed by sonication. The lysate was centrifuged at 15,000×g for 80min at 4°C and applied to Ni-NTA gravity column. The resin was then washed with 20 column volumes (CVs) of wash buffer (500mM NaCl, 20mM HEPES-Na, pH 7.5, 10% glycerol, 40mM imidazole, 1mM TCEP), and resuspended in 5 CVs of tag-removal buffer (500mM NaCl, 20mM HEPES-Na, pH 7.5, 10% glycerol, 40mM imidazole, 1mM TCEP and 0.6μg/mL ulp1 protease) for 1h incubation at 4°C. Next, the supernatant was loaded into 5mL HiTrap Heparin HP column (GE Healthcare) and eluted with a linear gradient of heparin elution buffer (buffer A: 20mM HEPES-Na, pH 7.5, 10% glycerol, 1mM TCEP; buffer B: 2M NaCl, 20mM HEPES-Na, pH 7.5, 10% glycerol, 1mM TCEP). Elution fractions with Casπ were pooled together and concentrated using 30 kD molecular weight cut-off centrifugal filters (Merck Millipore), and further purified by size exclusion chromatography (SEC) column (Superdex 200 Increase 10/300, GE Healthcare) with S200 buffer (400mM NaCl, 20mM HEPES-Na, pH 7.5, 10% glycerol, 1mM TCEP). Protein concentrations were measured by NanoDrop One (Thermo Scientific) and protein samples were stocked at –80°C after flash-frozen in liquid nitrogen. The Casπ protein samples are usually stocked at the concentration of 300μM. LbCas12a was expressed as previously described.²⁶

In vitro transcription of CRISPR RNA

DNA sequences containing T7 RNA polymerase promoter upstream of the Casπ tracrRNA, crRNA and sgRNA were assembled by overlap PCR and validated by Sanger sequencing. The validated sequences were then PCR amplified as the template for in vitro transcription (IVT). All reactions were performed in IVT buffer (30mM Tris, pH 8.1, 25mM MgCl₂, 0.01% Triton, 2mM spermidine) with 4mM NTP mix and 0.4mg/mL T7 RNA polymerase. The transcribed product was loaded into 10% Urea-PAGE with 2× formamide loading buffer (95% formamide, 0.02% SDS, 0.02% BPB, 0.01% xylene cyanole FF, 1mM EDTA) for electrophoresis. The gel region containing the target RNA band was extracted, smashed and soaked in soaking buffer (0.38M NaAc, pH 5.2, 0.8mM EDTA, 0.8% SDS) for 8h at 4°C. The dissolved RNA was then concentrated using 3 kD molecular weight cut-off centrifugal filters (Merck Millipore) and stocked at –80°C. The RNA samples are usually stocked at the concentration of 50μM. The RNA sequences and related description are listed in Supplementary information, Table S5.

PAM depletion assay and analysis

PAM depletion assay was performed as previously described with modifications (Supplementary information, Fig. S2a).⁵² Plasmids containing a PAM library were transformed into E. coli DH5α (TIANGEN) and incubated overnight at 37°C on LB-Amp⁺ agar plates (100μg/mL Ampicillin), and then all colonies were harvested to extract the plasmids using HighPure Maxi Plasmid Kit (TIANGEN). For cleavage reaction, sgRNA was diluted to the concentration of 30μM in refolding buffer (50mM KCl, 5mM MgCl₂) and refolded at 72°C for 5min, and then slowly cooled down to room temperature (RT). Subsequently, active RNP complexes were assembled by incubating 1μM Casπ protein with 1.2μM sgRNA in assembly buffer (100mM NaCl, 10mM HEPES-Na, pH 7.5, 1mM TCEP, 5mM MgCl₂) at RT for 30min. The reaction was initiated by adding 20nM plasmid and performed as three individual replicates in cleavage buffers (50–300mM NaCl, 10mM HEPES-Na, pH 7.5, 1mM TCEP, 10mM MgCl₂) at 37°C for 1h, and then quenched with loading buffer (Gel Loading Dye Purple 6×, NEB) supplemented with 20mM EDTA and 25μg/mL heparin. The cleaved products were analyzed and purified by electrophoresis on the 1.2% agarose gel with GelRed staining (Vazyme). Then, the end of linearized products was repaired by T4 DNA polymerase (Thermo Fisher Scientific) with 1mM dNTP (Sangon Biotech). dA oligo was further added to the 3′ end of the products by Dreamtaq polymerase (Thermo Fisher Scientific) with 1mM dATP (Sangon Biotech). Adapters with 3′ dT overhang were ligated with the products containing 3′ dA overhang by fast T4 DNA ligase (Beyotime). The DNA fragments containing the recognized PAM sequence were PCR amplified using a primer pairing to the adapter and the other primer pairing to the 120bp upstream region of the PAM. Next, the PCR-amplified PAM-containing products were purified by VAHTS DNA Clean Beads (Vazyme) and further amplified by TIANSeq Fast DNA Library Prep Kit (TIANGEN) for Illumina Novaseq PE150 sequencing. In control groups, the plasmids were treated with blank buffer instead of Casπ effectors, and DNA fragments containing PAM library were directly amplified by two primers covering the PAM region for the following process as described above. The depletion fold-change for each PAM was analyzed using the number of matched reads in Casπ and control groups normalized with total reads.

A list of depleted PAMs and related fold-change values are summarized in Supplementary information, Table S3.

In vitro cleavage assays

For cleavage assays with labeled NTS, the dsDNA substrate was prepared by PCR extension using a 65nt ssDNA template and a 5′-cy5-labeled 16nt primer (ordered from Sangon Biotech). Then the extended dsDNA was purified by DNA Clean & Concentrator-25 (Zymo Research) and diluted to 1μM in nuclease-free water (Invitrogen). The sgRNA was diluted to the concentration of 30μM in refolding buffer (50mM KCl, 5mM MgCl₂) and refolded as described above. Subsequently, Casπ effectors were assembled in a 1:1.2 protein to sgRNA ratio (1μM Casπ protein and 1.2μM refolded sgRNA) in assembly buffer (100mM NaCl, 10mM HEPES-Na, pH 7.5, 1mM TCEP, 5mM MgCl₂) at RT for 30min. The reaction was started by mixing 1μM RNP with 20nM dsDNA substrate in cleavage buffer (150mM NaCl, 10mM HEPES-Na, pH 7.5, 1mM TCEP, 10mM MgCl₂) at 37°C and aliquots were collected at the following time points: 0 mim, 2min, 5min, 15min, 30min, 60min, 90min and 120min. For biochemical screenings, only the reaction buffers were modified accordingly, such as the salt concentration (50mM, 150mM, 300mM or 450mM NaCl with 10mM HEPES-Na, pH 7.5, 1mM TCEP, 10mM MgCl₂), type of divalent ions (10mM Mg²⁺, Mn²⁺, Ca²⁺ or Co²⁺ with 150mM NaCl, 10mM HEPES-Na, pH 7.5, 1mM TCEP) and temperatures (25°C, 30°C, 37°C, 45°C, 55°C or 65°C with 150mM NaCl, 10mM HEPES-Na, pH 7.5, 1mM TCEP, 10mM MgCl₂). The products were analyzed as described above.

For cleavage assays with labeled TS, the 5′-cy5-labeled TS ssDNA was synthesized by Sangon Biotech and diluted to 10μM in nuclease-free water (Invitrogen). dsDNA was prepared by mixing 5′-cy5-labeled TS and unlabeled complementary oligo at the molar ratio of 1:1.2 in annealing buffer (10mM HEPES-Na, pH 7.5, 150mM KCl), followed by heating for 5min at 95°C and slow cooling down to RT. Cleavage reactions were initiated by mixing 1μM RNP with 20nM ssDNA or dsDNA substrate in cleavage buffer (150mM NaCl, 10mM HEPES-Na, pH 7.5, 1mM TCEP, 10mM MgCl₂) at 37°C and the product aliquots were collected at the following time points: 0min, 2min, 5min, 15min, and 60min.

For mismatched cleavage assay, the dsDNA substrates with single mismatches were prepared by PCR extension using a 65nt ssDNA template with single mismatch and a 5′-cy5-labeled 16-nt primer (ordered from Sangon Biotech). Then the extended dsDNA was purified by DNA Clean & Concentrator-25 (Zymo Research) and diluted to 1μM in nuclease-free water (Invitrogen). Cleavage reactions were initiated by mixing 1μM RNP with 20nM dsDNA substrate in cleavage buffer (150mM NaCl, 10mM HEPES-Na, pH 7.5, 1mM TCEP, 10mM MgCl₂) at 37°C and the product aliquots were collected at 1h.

For trans-cleavage assay, 1μM Casπ or LbCas12a RNP was first incubated with 1.5μM dsDNA or ssDNA activator at 37°C for 30min. Then 20nM 5′-cy5-labeled random 60nt ssDNA was mixed into the reaction. The product aliquots were collected at the following time points: 0min, 2min, 5min, 15min, 30min, 60min, 90min and 120min.

All cleavage products collected above were quenched with 2× Urea-loading buffer (8M urea and 2mM Tris-Cl, pH 7.5) supplemented with 20mM EDTA and 25μg/mL heparin, and then analyzed in 15% urea-PAGE and visualized using Amersham Typhoon 5 (GE Healthcare). Product bands were quantified using ImageJ and cleaved fraction was calculated using the intensity of product bands divided by input intensity.⁵⁴ Curves of cleavage efficiency were plotted using a One-Phase-Decay model in Prism 8 (GraphPad).

For plasmid cleavage assay, 1μM Casπ RNP effectors were incubated with 20nM target plasmids at 37°C for 30min and then quenched with loading buffer (Gel Loading Dye Purple 6×, NEB) supplemented with 20mM EDTA and 25μg/mL heparin. The samples were analyzed by electrophoresis on a 1.2% agarose gel with GelRed staining (Vazyme). For non-labeled dsDNA cleavage assay, the dsDNA target was PCR amplified from the plasmid containing the protospacer and purified by DNA Clean & Concentrator-25 (Zymo Research). The reaction was initiated by incubating 1μM Casπ RNP effectors with 20nM dsDNA target at 37°C for 30min and then quenched with loading buffer (Gel Loading Dye Purple 6×, NEB) supplemented with 20mM EDTA and 25μg/mL heparin. The samples were analyzed by electrophoresis on the 1.2% agarose gel with GelRed staining.

All experiments were performed at least three times for replicability. A list of oligonucleotides used in this study and related description are summarized in Supplementary information, Table S5.

Determination of cleavage sites

The cleavage products and sites on dsDNA were analyzed by electrophoresis using 15% urea-PAGE as described above. To determine the cleavage sites on plasmids, linearized plasmids were purified and subjected to NGS library construction for Illumina Novaseq PE150 sequencing as described in PAM depletion assay. Paired-end reads were mapped to the target sequence using BWA and 3′-ends were selected to determine the cleavage sites. The abundance of each site was normalized to the total reads and plotted using Prism 8 (GraphPad).

Plasmid interference in bacteria

E. coli BW25141 cells were requested from Prof. Guangdong Shang group in College of Life Sciences, Nanjing Normal University. E. coli BW25141 competent cells carrying the ccdB toxin plasmid (p11-LacY-wtx1) was prepared following the protocol previously described.⁵³ For each group, 200ng plasmid expressing Casπ and sgRNA (ccdB-targeting or non-targeting) was electroporated into 50μL competent cells with 0.2cm cuvette (BIO-RAD) under 2.5kV using Eppendorf eporator. After 1.5h of recovering in 5mL SOC medium (Sangon Biotech) under 37°C, the bacterial cells were enriched by centrifugation and resuspended in 5mL liquid LB-Strep⁺ medium (50μg/mL streptomycin), and cultured for an extra 8h. Subsequently, to investigate the effects on bacterial survival by Casπ editing, 5µL of culture with gradient dilutions from 10⁰ to 10^–7 was spotted onto the LB-Amp⁺ agar plates (100μg/mL ampicillin) or LB-Strep⁺-Ara⁺ agar plates (50μg/mL streptomycin, 10mM arabinose), respectively, and incubated overnight at 37°C. In the meantime, to validate the transformation efficiency of Casπ–sgRNA expression plasmids, 10μL of culture was spreaded on LB-Strep⁺ agar plates (50μg/mL streptomycin) for overnight incubation at 37°C, and colony number on each plate was manually counted. 5μL of edited bacterial cells was used for PCR validation of the plasmid interference with Phanta Max Super-Fidelity DNA Polymerase Mastermix (Vazyme).

Construction of EGFP report cell line

To obtain a natural target sequence with diverse targeting windows (different GC contents and PAMs), a sequence survey was performed in mouse genome. Via screening by 20nt window, we allocated a 270bp fragment within the Mus musculus myosin heavy polypeptide 8 (MYH8) exon (NCBI accession: NM_177369.3 (3650-3919)) which presents a well distribution of targeting windows with various GC contents (30%–85%) and PAMs (Supplementary information, Table S3). This region shows low sequence similarity to human genome. Frameshifting EGFP (3n+2) was created by fusing the MYH8 fragment, a 32bp random flanking sequence and EGFP ORF (1436bp). The MYH8-EGFP was further inserted into lentiviral packaging plasmid. The LV-MAX lentiviral production system (Thermo Fisher Scientific) was used to produce the lentivirus for inserting the MYH8-EGFP (3n+2) fragment into HEK293A cell genome via infection. The selection and enrichment of genome-modified cells were performed according to the manufacturer’s protocol (Thermo Fisher Scientific).

Gene editing assay in human cells

For EGFP activation editing assay in human cells, the EGFP HEK293A reporter cells were cultured in DMEM (Gibco) supplemented with 10% (v/v) FBS (Gemini) and 1% (v/v) penicillin streptomycin (Gibco) at 37°C in 5% CO₂. About 8.0×10⁴ cells were seeded onto the each well of 48-well plate for ~16h incubation. When the cell confluency reached 60%–70%, 300ng plasmid expressing NLS-Casπ- or Cas9-P2A-PuroR-NLS with sgRNA (MYH8-targeting and non-targeting) was transfected into the cells within each well using Lipofectamine 3000 (Life Technologies) according to the manufacturer’s protocols. One day after transfection, the old medium was replaced by fresh DMEM-Puro⁺ medium (1.5μg/mL puromycin, Sigma) for 3-day culturing. Then the enriched cells were further cultured for another 3 days using fresh DMEM medium without puromycin for gene editing analysis. The EGFP signal was observed with fluorescent microscopy (Nikon Eclipse TS2FL fluorescence microscope). Edited cells were also collected and stored at –80°C. For more endogenous gene editing assay, the HEK293T cells were treated the same as mentioned above, but transfected with NLS-Casπ-P2A-PuroR-NLS with sgRNA targeting other endogenous genes.

A list of targeting sequences is summarized in Supplementary information, Table S5.

Evaluation of gene editing efficacy

For T7E1 assay, the genome of edited cells was extracted using Ezup Column Animal Genomic DNA Purification Kit (Sangon Biotech). The edited genome was used as the template for PCR amplification of target region using Phanta Max Super-Fidelity DNA Polymerase Mastermix (Vazyme) (primers listed in Supplementary information, Table S4). The PCR product was gel-purified, and ~200ng purified DNA was re-annealed for T7E1 cleavage assay according to the manufacturer’s protocol (Vazyme). Cleavage products were analyzed by electrophoresis using 2% agarose gel with GelRed staining (Vazyme).

For NGS, ~210bp regions nearby the target protospacers were amplified via PCR with Q5 polymerase (NEB) and primers containing Illumina adaptor sequences. Amplicons were verified by electrophoresis using 2% agarose gel with GelRed staining (Vazyme), purified by VAHTS DNA Clean Beads according to the manufacturer’s protocol (Vazyme) and further loaded onto Illumina Novaseq PE150 sequencing by Tianjin Novogene Bioinformatic Technology Co., Ltd. Sequencing reads were analyzed by CRISPResso2 with the following parameters: quantification window centered at 3bp for Casπ-1 (2bp for Casπ-2, 1bp for Cas12a and –3bp for Cas9) according to cleavage sites of both Casπs (Supplementary information, Fig. S2g, h), quantification window size of 14bp for both Casπs (8bp for Cas9), and plot window size of 40bp (to visualize large indels).⁵⁵ Cells treated with plasmids carrying codon-optimized Cas genes with a non-targeting sgRNA were evaluated at every spacer sequence within every read as a negative control. Percentage of each indel plotted (regardless of substitution) was based on the results of modified reads from the CRISPResso2 output. For the indel size distribution plots, unmodified reads (indel length of 0bp) were plotted as 0% of the total reads for clarify and the remaining reads were grouped and plotted based on the modified results.

Reconstitution of Casπ R-loop complex

Deactivated Casπ-1 (dCasπ-1, D537A, E643A) was purified as described above. The sgRNA was diluted to 40μM in refolding buffer (50mM KCl, 5mM MgCl₂) and refolded as described above. The dCasπ-1–sgRNA binary was reconstituted by incubating 20μM dCasπ-1 and 25μM sgRNA for 30min at RT in a total volume of 150μL assembly buffer (100mM NaCl, 10mM HEPES-Na, pH 7.5, 1mM TCEP, 5mM MgCl₂). To facilitate the R-loop formation, the bubbled dsDNA substrate with 10nt mismatch in the protospacer was used for R-loop ternary complex assembly. The bubbled dsDNA was diluted to 30μM in 150μL assembly buffer, and mixed with 150μL binary complex at RT for 30min incubation. Subsequently, the assembled sample was purified by size exclusion column (Superdex 200 Increase 10/300, GE Healthcare) in SEC buffer (150mM NaCl, 10mM HEPES-Na, pH 7.5, 1mM TCEP, 0.1% glycerol, 5mM MgCl₂) at 4°C. After flash freezing by liquid nitrogen, the aliquots of purified sample were stocked at –80°C. The reconstituted complex was usually stocked at the concentration of 3μM. A list of DNA oligonucleotides and sgRNA sequences with brief descriptions are presented in Supplementary information, Table S5.

Cryo-EM sample preparation and data collection

4μL of purified Casπ R-loop complex (~1.5μM) was crosslinked by BS3 (Sigma-Aldrich) and applied to the graphene oxide grid from Shuimu Biosciences Ltd. (Quantifoil Au 1.2/1.3, 300 mesh), which was glow-discharged (in a HARRICK PLASMA) for 10s at middle level after 2min evacuation. The grid was then blotted by a pair of 55mm filter papers (Ted Pella) for 0.5s at 22°C with 100% humidity, and flash-frozen in liquid ethane using FEI Vitrobot Marke IV. Cryo-EM data were collected on a Titan Krios electron microscope operated at 300kV equipped with a Cs-corrector and Gatan K3 direct electron detector with Gatan Quantum energy filter using EPU. Micrographs were recorded in counting mode at a nominal magnification of 105,000×, resulting in a physical pixel size of 0.856Å per pixel. The defocus was set between –1.5μm and –2.5μm. The total exposure time of each movie stack led to a total accumulated dose of 50 electrons per Ų which fractionated into 32 frames. More parameters for data collection are shown in Supplementary information, Table S6.

Image processing and 3D reconstruction

The raw dose-fractionated image stacks were 2× Fourier binned, aligned, dose-weighted, and summed using MotionCor2.⁵⁶ CTF-estimation, blob particle picking, 2D reference-free classification, initial model generation, final 3D refinement and local resolution estimation were performed in cryoSPARC.⁵⁷ Two rounds of 3D reference-based classification were performed in RELION.⁵⁸ The details of data processing were summarized in Supplementary information, Fig. S5 and Table S6.

Model building and refinement

The initial protein model was generated using AlphaFold2 and manually revised in UCSF-Chimera and Coot.^20,59,60 The DNA substrates and sgRNA were manually built in Coot based on the cryo-EM density. The complete model was refined against the EM map by PHENIX in real space with secondary structure and geometry restraints.⁶¹ The final model was validated in PHENIX software package. The structural validation details for the final model are summarized in Supplementary information, Table S6.

EM data were collected at the Tsinghua Cryo-EM facility and Shuimu Bioscience. The data were analyzed using the Bio-Computation platform at the Tsinghua University Branch of the Chinese National Center for Protein Sciences (Beijing). We thank the supports from the Tsinghua University Technology Center for Protein Research, Genome Sequencing and Analysis. We thank J.L. Lei, X.M. Li, and X.D. Li for expert electron microscopy assistance. We thank T. Yang, Y.K. Wang, A.B. Jia for computational support. We thank D. Chia, Y. Lin, and N. Liu for their kind advice on the manuscript. The work was supported by the National Key R&D Program of China (2022YFF1002801 to J.J.G.L.), the Ministry of Agriculture and Rural Affairs of China (J.J.G.L.), the National Natural Science Foundation of China (32150018 to J.J.G.L and 32101195 to S.Z.), and start-up funds from Tsinghua University, Beijing (J.J.G.L.).