A gate and clamp regulate sequential DNA strand cleavage by CRISPR-Cas12a

CRISPR-Cas12a has been widely used for genome editing and diagnostic applications, yet it is not fully understood how RNA-guided DNA recognition activates the sequential cleavage of the non-target strand (NTS) followed by the target strand (TS). Here we used single-molecule magnetic tweezers microscopy, ensemble gel-based assays and nanopore sequencing to explore the coupling of DNA unwinding and cleavage. In addition to dynamic R-loop formation, we also directly observed transient dsDNA unwinding downstream of the 20 bp DNA:RNA hybrid and, following NTS cleavage and prior to TS cleavage, formation of a hyperstable “clamped” Cas12a-DNA intermediate resistant to DNA twisting. Alanine substitution of a conserved aromatic amino acid “gate” in the REC2 domain that normally caps the heteroduplex produced more frequent and extended downstream DNA breathing, a longer-lived twist-resistant state, and a 16-fold faster rate of TS cleavage. We suggest that both breathing and clamping events, regulated by the gate and by NTS cleavage, deliver the unwound TS to the RuvC nuclease and result from previously described REC2 and NUC domain motions.

Following R-loop formation the liberated NTS docks into the RuvC active site and nicking occurs (Supplementary Figure S1). The DNA ends then disengage and an uncut NTS phosphodiester can re-enter the active site and rounds of cleavage-release-rebinding lead to DNA end trimming (30). This gap formation may prevent steric occlusion of the open active site to allow subsequent TS binding and allow binding of non-specific ssDNA in trans ("bystander" cleavage), a key activity for nucleic acid detection diagnostics (31). A geometric limitation for TS cleavage is that the target site is displaced (32), in a region of dsDNA downstream of the R-loop, ~25 A˚ from the RuvC active site and in an incorrect orientation for in-line nucleophilic attack without rotation to match the polarity of the NTS (8,11). A suggested model is that downstream dsDNA unwinds, and the released single stranded TS is bent towards RuvC (Supplementary Figure S1) (13). Structures consistent with these molecular gymnastics have been observed for the related Cas12b and Cas12f complexes (33,34). Cryo-EM structures, single molecule FRET measurements and molecular dynamics of Cas12a additionally support an inward closing motion of the REC2 and Nuc domains following NTS nicking (13,24). Slower TS cleavage may reflect both the kinetics of TS gap formation and these domain transitions.
More recently, Cofsky et al (30) demonstrated that the downstream DNA that encompasses the TS cleavage site is subject to DNA breathing. They propose this is a fundamental property of the RNA 3′ end of an R-loop that is exploited by Type V enzymes due to their R-loop geometry. Extending the region of downstream ssDNA accelerated DNA cleavage and altered the TS cleavage loci, confirming the important role of the downstream unwinding and the kinetic constraint it places on the TS cleavage rate. Here we were able to corroborate these findings in the absence of DNA cleavage by directly observing the dynamics of transient and reversible downstream unwinding events by Lachnospiraceae bacterium ND2006 (Lb) Cas12a using a magnetic tweezers assay that can measure strand separation in real time (15,35). We also observed that the LbCas12a R-loop was highly dynamic and heterogeneous, forming multiple intermediate stable states and occasionally fully unzipping (rupture events).
It has been suggested that REC2 dynamics could control conformational changes in Nuc, regulating a functional role in loading the TS into RuvC (12,33). Cas12a ternary X-ray structures reveal that the R-loop is terminated by a stacking interaction with a conserved aromatic amino acid in REC2 (e.g., W355 in LbCas12a, Figure 1a, (21)). Although mutation of this residue had moderate reduced INDEL formation activity (20), we speculated that this "Gate" may control the downstream DNA breathing by transitioning between closed (stacked) and open states (Figure 1b). Indeed, mutation of the W355 Gate to alanine resulted in more frequent and extended downstream DNA breathing but also greater occupancy of longer Rloop states in general and less state heterogeneity, ruling out a role for the stacking interaction in stabilising the 20 bp R-loop. Using ensemble endonuclease assays, we demonstrated that the more frequent DNA breathing with W355A accelerated the rate of endonucleolytic TS cleavage by ~16-fold. Subsequent 5′-3′ processing of the cleaved end was also accelerated as mapped by nanopore sequencing. By following nuclease activity in the single molecule tweezers assay, we also observed a torque-resistant clamping of the downstream DNA after NTS cleavage, which we suggest corresponds to Nuc interactions that guide the TS to the RuvC active site. This clamped state was hyperstable in the W355 mutant and supported more rapid TS cleavage. We therefore propose that the aromatic Gate residue acts to control both downstream DNA breathing and the clamping that is necessary for TS cleavage.

Single molecule observation of dynamic R-loops and downstream DNA breathing
To observe Cas12a-dependent DNA unwinding events in real time, we used a single molecule magnetic tweezers assay used previously with cas9 and Cas12a (Figure 2a) (15,35). A 2 kbp linear DNA target was tethered between a glass coverslip and a magnetic particle (500 nm diameter) in a flow cell. The apparent length of the DNA above the surface was monitored by video microscopy of the bead image (36). A pair of permanent magnets above the flow cell stretched the DNA with varying force depending on the vertical position and could be rotated to introduce positive or negative turns into the DNA, inducing supercoiling at low forces. We initially used an EDTA-based buffer so that the dynamics of R-loop formation could be monitored without DNA cleavage (15).
At the 0.3 pN stretching force used here, introducing negative turns formed negative supercoils that shortened the apparent DNA length (Figure 2b). Negative torque supports Rloop formation (15), which resulted in a reduction in DNA supercoiling to balance the altered DNA linking difference, observed as an increase in apparent DNA extension (IN events in Figures 2a,b). To force the R-loop out, positive turns were introduced to generate positive supercoiling. The positive torque causes R-loop dissociation (OUT events in Figures 2a,b). The process was repeated by cycling between negative and positive turns using the magnets.
Hereafter we used a linear relationship to convert DNA extension to DNA turns (Supplementary Figure S2) so that we could compare between different DNA molecules where relative attachment points can cause variations in apparent DNA length.
An example R-loop formation profile at -7 pN nm using 1 nM WT Lb Cas12a is shown in Figure   2c. The improved signal-to-noise using the smaller diameter (500 nm) magnetic particles (37), revealed a much more dynamic R-loop than was detectable previously. Each trace was fitted with a Hidden Markov model (HMM) to identify the minimal number of discrete states that could describe the data and to calculate the state positions in turns and their relative probabilities. The example data in Figure 2c could be described by 3 discrete states (S2, S3, and S4) showing sequential hopping transitions. We noted that there were frequent "rupture" events where the turns returned to zero, suggesting a complete reversible unzipping of the R-loop (S0).
A second example profile (Figure 2d) illustrates the heterogeneity in R-loop dynamics, since the S3 state is more occupied and we could identify an additional state, S5, with a turns value of 2.0 ± 1.0. We interpret the transient S5 events as PAM distal DNA breathing, as observed by Cofsky et al (see below) (30). A third example profile (Figure 2e) reveals another discrete state (S1) representing a much shorter R-loop. The full length S4 R-loop state does not appear to be accessed during this event, although there is a transient increase that could be accessing S4 and/or S5, but this could not be identified from the HMM analysis because of its infrequency during the measurement window. In general, there was great heterogeneity between events, both in the total turn sizes, number of states, and dynamics of the states (Supplementary We previously measured that the 20 bp R-loop of Cas12a corresponds to a change of 1.8 turns (15). This matches the average turns for state S4 here. We therefore mapped the other states onto possible R-loop sizes (Figure 2g). From the occupation probabilities, we can also suggest a simple free energy diagram for the different states ( Figure 2h). The S1, S2 and S3 states could correspond to the previously identified conformation checkpoints that couple R-loop propagation to nuclease activation [8]. The linker and lid interactions may produce the occupancy of the S1 and S2 states, respectively. HMM modelling suggests that the rupture events can occur directly from these states. The "finger" interaction may produce the occupancy of the S3 state.
The S5 state (2.0±1.0 turns) can be estimated to correspond to an additional unwinding of ~2 bp. Although we cannot rule out that the change in turns partly corresponds to a change in writhe, this additional unwinding is consistent with the observations of Cofsky et al (30). We suggest there is a free energy penalty to accessing this state from the full R-loop S4 state due to the W355 Gate (see Discussion). The transient downstream unwinding may be due to movement ("opening") of the Gate residue that otherwise caps the R-loop. This idea is tested in the next section.
In summary, our data reveals a more dynamic R-loop than was presumed previously, rapidly switching between S2-S4 states and even a fully ruptured R-loop state, possibly controlled by reversibility of the conformational checkpoints. An additional S5 state is consistent with a transient and reversible unwinding of a few additional base pairs downstream of the R-loop.   (Figure 3e). The S2 state was less frequently observed than with WT, while the S1 state was not occupied to a measurable degree ( Figure   3d). The most occupied state was 2.0±0.3 turns (S4) which we suggest corresponds to a 20 bp R-loop plus 2 further unwound base pairs, equivalent to the transient S5 state of the WT enzyme ( Figure 3f). In other words, the Gate mutant seems to lock in the additional downstream breathing. The additional occupied S5 state corresponds to 2.3±0.9 turns, which we interpret as additional downstream unwinding by 2-3 bp not observed as a long-lived state with WT. Rupture events were observed, but less frequently that with WT (Figure 3g), which may reflect that the shorter R-loop states were less occupied and so there was less chance for R-loop collapse.

U C G A G C U -3 ʹ C A A U U C G A G C U -3 ʹ U U U C A A U U C G A G C U -3 ʹ
We previously used rapid switching to positive supercoiling to measure the R-loop lifetime under varying positive torque. Our observations of dynamic states here suggest that it would be difficult to reliably capture a defined S-state and that there may also be dynamic changes during the magnet rotation period (~1 s). Furthermore, although stable states at positive torque were noted for WT (Figure 2b), as observed previously (15), the majority of W355A events were immediately dissociated so that lifetime distributions could not be measured (Supplementary Figure S4). This result was surprising given that rupture events at negative torque were infrequent and the average R-loop states were longer in length. However, if Rloop rewinding under positive torque is driven from the PAM-distal end (the PAM end being clamped by the PID), the difference in stability may reflect that in the WT enzyme, the Gate can resist the rewinding force.
In summary, the removal of the Gate residue resulted in more stable R-loops and additional downstream unwinding at negative torque, but unstable R-loop formation at positive torque.
The change in turns captured was greater than that with WT, indicating that the downstream DNA was breathing more readily (  The steps leading to engagement of the TS by RuvC may be triggered by the breakage of the NTS strand or may have a requirement for RuvC to go through its cleavage chemistry in a more ordered conformational exchange. To test these alternatives, we took advantage of the slow TS cleavage rate of Cas12a to produce a pre-nicked DNA substrate by 10 s treatment with WT enzyme (Figure 4d). Since the cleavage rate of nicked DNA is rate-limited by the slower R-loop formation compared to supercoiled DNA, a control pre-nicked DNA was produced using Nt.BspQI cleavage at a distal site (Figure 4d). The cleavage of the nicked substrates after 30 s using either WT or W355A Cas12a was compared with cleavage of supercoiled DNA (Figure 4d). For WT enzyme, pre-nicking of the NTS resulted in more cleavage of the TS, indicating that a barrier had been lifted. In contrast, W355A produced similar levels of linear DNA on all substrates, consistent with the removal of the Gate having a larger effect on accelerating the cleavage process.

C G A G C U -3 ʹ C A A U U C G A G C U -3 ʹ
Given that the locations of NTS and TS cleavage can be altered by mismatches within the Rloop (29) and that RuvC-dependent processing of the DNA ends can also be affected by downstream DNA unwinding (30), we reasoned that the differences in R-loop dynamics and DNA breathing between WT and W355A Cas12a might also result in measurable differences in cleavage loci. To explore this, we used ENDO-Pore, a nanopore-based method that allows

A stable downstream DNA Clamp state following non-target strand cleavage
To further explore the nuclease mechanism and role of the W355 Gate, we adapted the magnetic tweezers assay by using a Mg 2+ -based buffer that supports DNA cleavage but kept all other conditions the same. A DNA was positively supercoiled before enzyme was introduced into the flow cell; under positive torque, R-loop formation is inhibited so the DNA is not cut. The DNA was then rapidly unwound (<1 s) to produce negatively supercoiled DNA (event 1 in Figure 5a), facilitating R-loop formation recorded as a change in apparent bead height (event 2). We expected that R-loop formation would then support NTS cleavage and that free rotation at the NTS nick would release the negative supercoiling strain, producing a further increase in apparent bead height to full length (39). Such an event was observed The slower TS cleavage rate for WT observed in the ensemble cleavage assays (Figure 4c)  The times for NTS cleavage estimated from the single molecule assays were similar for WT and W355A (Figure 5h). This suggests that the observed difference in clamping occurs after TS cleavage; i.e., initial R-loop lifetimes reflect the chemical cleavage kinetics. As observed in the ensemble reactions, TS cleavage measured from bead release was faster for W355A, although there were far fewer measurable events for WT Cas12a over a reasonable experimental timescale (see below). Bead loss with WT Cas12a was only observed in 57% of molecules over the experimental timescale while for W355A this value was significantly higher at 94 % (p < 0.05) (Figure 5h). The less frequent cleavage by WT could be due to the less stable clamped state (Figure 5e), or to TS cleavage being inhibited by applied force, possibly by preventing conformation changes needed for gate opening.
In summary, single molecule cleavage experiments revealed that Cas12a can form a clamped state following NTS cleavage that is substantially more torque stable than the R-loop formed in EDTA. The stability of the clamped state is promoted by opening of the Gate and appears to be necessary for TS cleavage. Consequently, the slow TS cleavage of WT Cas12a can be accelerated by removing the Gate in the W355A mutant.

DISCUSSION
We explored the R-loop dynamics and activation of NTS and TS cleavage by Cas12a using a combination of single molecule R-loop assays, ensemble DNA cleavage assays, single cleavage event mapping by nanopore sequencing (ENDO-Pore) and single molecule cleavage assays.
Our results show that an aromatic residue in the REC2 domain that terminally caps the full 20 bp heteroduplex, which we term the Gate (Figure 1a), does not stabilise the R-loop directly through stacking interactions but controls downstream DNA breathing (Figure 1b), and following NTS cleavage, conformational activation of a clamped state that is necessary for primary TS cleavage and subsequent end processing (Figure 5b).
By using smaller diameter magnetic particles in our magnetic tweezers assays, we could observe dynamic interchange between defined R-loops states in the absence of DNA cleavage that could be mapped to approximate locations of previously identified conformational checkpoints (Figures 2g & 3f). We also captured reversible R  (14).
A key difference between Type II-A and Type V-A CRISPR-Cas effectors is that the former use two separate nuclease domains to cut the NTS and TS while the latter use a single RuvC domain that must transition sequentially between the NTS and TS (8). It has been suggested that R-loop asymmetry is exploited by Cas12a to allow downstream DNA breathing that is a key step in providing the single strand TS that can be delivered to the RuvC nuclease active site (30). We directly corroborated such reversible DNA breathing events for WT Cas12a as transient increases in DNA unwinding measured from changes in supercoiling (Figure 2 & 3).
Additionally, we found that in the W355A mutant there was an increase in frequency and size We interpret the clamped state as downstream protein-DNA interactions that are necessary for delivering the TS to the RuvC active site (13,24,30). Hence TS cleavage is produced in a clamped state. Our data is consistent with clamping being controlled by the Gate since its removal favoured the clamped state and hence accelerated the observed TS cleavage rate.
The reversibility of the clamping seen with WT Cas12a could be due to reversible openingclosing of the Gate even following NTS cleavage. The 5′-3′ processing of the TS (Figure 4e Single molecule, structural and molecular dynamics studies of Cas12a have indicated that the REC2 and Nuc domains move towards each other upon NTS cleavage, contracting the groove between the TS and RuvC active site (12,13,24). We interpret the clamped state observed after NTS cleavage ( Figure 5), as resulting from this motion. Since a combined path created by residues from both RuvC and Nuc domains is essential for TS cleavage (12,33), we further suggest that TS interactions by Nuc, possibly following DNA breathing, are responsible for the torque stable clamped state (Figure 5b)

Single Molecule Magnetic Tweezers Experiments
The magnetic tweezers experiments were performed using a commercial PicoTwist microscope (Fleurieux sur L'Arbresle, France) equipped with a 60 Hz Jai CV-A10 GE camera (36) . For flow cell preparation, glass coverslips (Menzel Gläser No.1, 24 x 60 mm x160 µm) were cleaned in 3 repeated cycles of 1 hr sonication in 1M KOH and then Acetone and were subsequently cleaned with milliQ water and dried using compressed air. The coverslips were kept enclosed in glass jars to prevent any moisture until needed. Flow channels were prepared as previously. DNA molecules (a 2 kbp section of pSP1) were tethered to 500 nm paramagnetic beads (Adamtech) (37), and the glass coverslip of the flow cell as previously described (15,35). Topologically-constrained DNA were identified from rotation curves at 0.3 pN and the rotational zero reference (Rot0) set. 1 nM RNP was used for all measurements. To make pre-nicked DNA, 5 nM tritiated pSP1 was nicked either using 50 nM Cas12a RNP (WT protein with crRNA 24) at 25°C in Buffer RB for 10 s or using 10 units Nt.BspQI (New England Biolabs) per µg DNA, at 50°C for 1 hr. Nicked DNA was purified using Qiagen PCR purification kit. Cleavage assays were then performed as described above and stopped after 30 s by addition of 0.5 volumes of STEB followed by incubation at 67 ᵒC for 10 minutes. Debranched products were purified using AMPure XP beads followed by size selection using the Short Read Elimination XS kit (Circulomics). Samples were prepared for Nanopore sequencing using the Ligation Sequencing Kit (SQK-LSK109) combined with the Native Barcoding Expansion kit and sequenced using R9.4.1 cells (Oxford Nanopore Technologies).