Mechanistic basis for improved activity of Engineered AsCas12a

Abstract

CRISPR-associated proteins (Cas) are central to gene editing, forming nuclease complexes with guide RNA to enable precise genome modification. Among numerous Cas variants, Cas9 and Cas12a are the most extensively studied. While much is known about the genomic substrates for these enzymes, less is known about the determinants of the DNA cleavage activity. Wild-type Cas12a exhibits higher intrinsic specificity than Cas9, minimizing off-target activity, but lower overall potency. Recent protein engineering has sought to improve both parameters. Here, we shed light on the structural and mechanistic basis by which an engineered AsCas12a variant achieves high potency while retaining its hallmark specificity. We show that reduced protein–DNA interactions facilitate more rapid R-loop formation, thereby enhancing cleavage activity. These results provide mechanistic insight into Cas12a function and highlight strategies for designing genome-editing nucleases with optimal balance between efficiency and specificity.

Engineered AsCas12a increases gene editing potency, retaining high specificity. Potency increases by reducing protein-DNA interactions, mediating faster R-loop formation. This provides insight into design of efficient and precise CRISPR nucleases.

Introduction

Gene editing is a promising new therapeutic approach to treating genetic diseases by altering genomic target sites in patients’ own cells to restore, reduce, or repair gene expression or protein function¹. In particular, clustered regularly interspaced short palindromic repeats (CRISPR)–associated protein (Cas) gene editing has emerged as an exciting therapeutic modality due to its inherent ability to target and modify specific sites within the genome. The two most common Cas enzymes evaluated for therapeutic purposes are Cas9 and Cas12a, specifically Streptococcus pyogenes Cas9 (SpCas9) and Acidaminococcus sp. Cas12a (AsCas12a), respectively. There are distinct differences between Cas9 and Cas12a that impact their roles as gene-editing enzymes. Importantly, Cas12a recognizes a TTTV protospacer adjacent motif (PAM) while Cas9 recognizes an NGG PAM^2,3. Moreover, while Cas12a has only one unique nuclease domain, Cas9 has two unique nuclease domains^4–6. While two distinct nuclease domains facilitate simultaneous blunt end double-stranded DNA (dsDNA) cuts with Cas9, Cas12a generates staggered DNA cuts due to the sequential positioning of each DNA strand in the nuclease active site⁴. Additionally, Cas12a is less tolerant of mismatches within the RNA/DNA hybrid within the enzyme active site (R-loop) compared with Cas9 because of its requirement for both PAM proximal and PAM distal Watson–Crick base pairing within the 20 nt–long RNA/DNA duplex, which contributes to Cas12a’s superior specificity. In contrast, Cas9 is cleavage competent after formation of the first 8 to 12 RNA/DNA base pairs within the R-loop, allowing for cleavage even in the presence of mismatches⁷. Finally, Cas12a is guided by a single CRISPR RNA (crRNA) (~40 nt) while Cas9 requires a second trans-activating CRISPR RNA (tracrRNA) for activity (although they can be combined into a single long guide RNA [gRNA] at around 100 nt)⁸. The ability to use a short crRNA is particularly relevant for therapeutic purposes because of the facilitated chemical synthesis of the short gRNA.

Cas12a undergoes specific, previously characterized steps during catalysis⁹. First, Cas12a recognizes and binds to a TTTV PAM which distorts the DNA duplex and allows for strand invasion of the gRNA to initiate R-loop propagation between the template strand (TS) and the RNA. After formation of a full 20 bp R-loop, Cas12a undergoes conformational rearrangements that result in a cleavage-competent complex¹⁰. Non-template strand (NTS) DNA cleavage occurs immediately after this conformational change and is followed by a slower conformational change that positions the TS into the active site, leading to TS cleavage and eventually to dissociation⁹.

AsCas12a has been the active focus of many gene-editing efforts because of its enhanced intrinsic specificity compared with Cas9¹¹. However, while the specificity is higher for wild-type (WT) AsCas12a compared with WT Cas9, the potency is lower. To address this limitation, recent work has been performed to improve its editing efficiency as well as enhance the targeting range through the relaxation of PAM requirements^11–15. One of these efforts has resulted in an engineered variant formerly termed ‘AsCas12a Ultra’, now referred to as “engineered AsCas12a”¹¹. Interestingly, this variant of AsCas12a contains only two mutations that distinguish it from WT: M537R and F870L. These mutations result in several observable changes in AsCas12a behavior¹¹, particularly accelerated in vitro DNA cleavage, increased performance at lower temperatures, and increased potency in the context of genome editing. The mechanistic basis for how these two mutations contribute to an enzyme with such improved attributes remains unclear.

Renizgamglogene autogedtemcel (reni-cel), an investigational gene-edited autologous hematopoietic stem cell medicine, is manufactured with an AsCas12a variant that incorporates the mutations from engineered AsCas12a¹¹. To our knowledge, reni-cel is the only autologous AsCas12a-edited genomic medicine that has been evaluated for clinical use to date. Early data from clinical studies of reni-cel demonstrated improved clinical outcomes in patients living with sickle cell disease and transfusion-dependent beta thalassemia^16,17.

Here, we describe recent structural and biochemical work performed to elucidate the mechanism driving the heightened efficiency of this engineered AsCas12a compared with the WT enzyme, as well as specifically highlight why this variant has promising therapeutic potential.

Results

Crystal structure of engineered AsCas12a in complex with a 20 bp R-loop

In order to determine how the two mutations that constitute the engineered AsCas12a (M537R, F870L) increase its rate of cleavage, we solved a 2.7 Å resolution crystal structure of engineered AsCas12a in complex with a 43 nt gRNA, a 24 nt target DNA strand, and a 10 nt NTS containing a 5’-TTTT-3’ PAM (Fig. 1a–c) using molecular replacement with WT AsCas12a (PDB accession code: 5B43; Table 1, Table 2)¹⁸. The structure revealed minimal differences between the WT and engineered ribonucleoproteins (RNPs) (Fig. 1d). The M537R and F870L mutations do not induce any gross conformational changes at this stage of complexation with the DNA substrate. The one noticeable change between the two structures exists within a loop in the wedge (WED) domain. M537R causes steric clashes with Y575, displacing it by 16.7 Å and the entire loop by 8.4 Å (Fig. 2a–c, Supplementary Fig. 1). This displacement removes a hydrogen bond between Y575 and the NTS DNA and allows the NTS more flexibility by removing repulsive interactions between the hydroxyl group of the Y575 side chain and the NTS backbone. The M537R mutation also introduces a hydrogen bond with the backbone of K550 of the WED domain (Supplementary Fig. 2a). Together, these changes appear to reduce interactions of Cas12a with the DNA and stabilize an NTS cleavage-competent conformation. Additionally, to further interrogate the position of M537R, we performed Gaussian accelerated molecular dynamics (GaMD) simulations to observe alternative energetically accessible conformations not captured in the crystallography. We examined sections of M537R to assess potential electrostatic influence due to close proximity to the DNA backbone. We do indeed see that M537R samples conformational states that make hydrogen bonds with the DNA backbone, typically preferring conformations that bridge between the backbone carbonyl and DNA backbone phosphate, suggesting that the position of M537R in the engineered enzyme is controlled by interactions with both K550 and the DNA backbone. (Supplementary Fig. 2b–d).

Fig. 1. Structure of engineered AsCas12a.

a Domain outline of AsCas12a, mutations that result in the engineered variant are indicated in orange. b Schematic of DNA/RNA interactions within RNP complex. c Crystal structure of engineered AsCas12a. Protein (dark gray, PDB accession code: 9MK3) is bound to a crRNA (magenta and complexed with an NTS/TS DNA duplex in blue). Mutations that constitute the engineered variant are shown as spheres in orange. d Engineered AsCas12a (dark gray) aligned to WT AsCas12a (light gray, PDB accession code: 5B43). AsCas12a Acidaminococcus sp. CRISPR-associated protein 12a, CRISPR clustered regularly interspaced short palindromic repeats, crRNA CRISPR RNA, Nuc nuclease, NTS non-template strand, PAM protospacer adjacent motif, PDB Protein Data Bank, RNP ribonucleoprotein, TS template strand, WT wild-type.

Table 1.

X-ray diffraction statistics

	Cas12a/DNA/RNA
Sample ID	X26p10
X-ray source	NSLS-II (FMX)
X-ray beam wavelength (Å)	0.97934
Space group	P 21
Unit cell (a,b,c = Å; α,β,γ °)	a = 81.916, b = 198.585, c = 140.363; α = γ = 90, β = 95.272
Resolution bin	Overall	Highest
Resolution (Å)	40.47–2.70	2.86–2.70
Measured reflections	697,664	90,176
Unique reflections	121,265	18,529
Rmeas (%)	12.5	126.9
Completeness (%)	98.09	93.7
I/σ	10.80	1.11
CC(1/2)(%)	99.8	48.7
Redundancy	5.75	4.87

Cas12a CRISPR-associated protein 12a, CRISPR clustered regularly interspaced short palindromic repeats.

Table 2.

Structure refinement statistics

	Cas12a/DNA/RNA
Refinement range (Å)	41.13–2.70
Rcryst (%)	18.91
Rfree (%)	23.94
Protomer details (per asymmetric unit)
Cas12a	2
tDNA	2
ntDNA	2
crRNA	2
Bond lengths, RMSD (Å)	0.006
Bond angles, RMSD (°)	1.59
Average B-factors (Å2)	Main chain	Side chain
Cas12a (chain A)	71.85	73.06
crRNA (chain B)	64.24	N/A
tDNA (chain C)	75.61	N/A
ntDNA (chain D)	94.08	N/A
Cas12a (chain E)	70.78	71.61
crRNA (chain F)	56.90	N/A
tDNA (chain G)	71.78	N/A
ntDNA (chain H)	88.05	N/A
Waters (chain W)	55.00	N/A
Ramachandran plot (%)
Preferred	97.28
Allowed	2.64
Outliers	0.08

Cas12a CRISPR-associated protein 12a, CRISPR clustered regularly interspaced short palindromic repeats, crRNA CRISPR RNA, N/A not applicable, ntDNA non-target DNA, RMSD root mean square deviation, tDNA target DNA.

Fig. 2. Reduced DNA/protein interactions in engineered AsCas12a.

a Hydrogen bond and steric interactions between Y575 (light gray) and the NTS backbone (blue) in WT AsCas12a. b M537R (orange) in the engineered enzyme pushes Y575 (dark gray) out of the way, eliminating the protein/DNA interactions shown in (a). c Overlay of WT (light gray) and engineered AsCas12a (dark gray) WED loop structures. d Representative gel of in vitro cleavage assay for engineered and WT AsCas12a with a TTTV PAM (top) and TTTT PAM (bottom). Only the bottom two bands were used for data quantification as upper bands represent remaining duplexed DNA. Uncropped gel images are shown in Supplementary Fig. 6. e Quantification of in vitro cleavage assays. Data was quantified in ImageQuant and plotted in Prism. Each individual replicate is shown. AsCas12a Acidaminococcus sp. CRISPR-associated protein 12a, CRISPR clustered regularly interspaced short palindromic repeats, NTS non-template strand, PAM protospacer adjacent motif, WED wedge, WT wild-type.

To further test the importance of the interaction with K550 as well as the displacement of Y575, we generated Y575A and K550A single mutants of engineered AsCas12a. We used a dual fluorescently labeled DNA template to separately examine the kinetics of various steps in the Cas12a mechanism, namely R-loop formation, the subsequent metal ion–dependent conformational rearrangement that poises the enzyme for NTS cleavage¹⁹, and NTS cleavage and product release. Probing R-loop formation with all variants revealed significantly impaired R-loop formation in the K550A mutant and wild-type relative to the engineered Cas12a and Y575A mutant (Supplementary Fig. 3a). This impaired activity further supports the hypothesis that hydrogen bonding between K550 and M537R helps stabilize the WED domain alpha helix that K550 is positioned in, and that this helix is critical for activity. Additionally, the Y575A mutant showed indistinguishable R-loop formation dynamics relative to the engineered variant, implicating Y575 displacement as beneficial for R-loop formation and generating a more active enzyme. The dramatic differences in R-loop formation rates between the Cas12a variants prompted a further examination of R-loop formation, further detailed in later sections. To separate metal ion–dependent structural rearrangements from strand cleavage events, Ca²⁺ was added to initiate the conformational changes while maintaining the Cas12a active site in a cleavage incompetent state (Supplementary Fig. 3b). All variants showed fast structural rearrangement limited by sample mixing time with no apparent impairment observed in any variant. NTS and TS cleavage steps were then initiated via addition of excess Mg²⁺ to displace the previously added Ca²⁺. This step was probed by fluorescence anisotropy of the carboxy-rhodamine (ROXN) dye on the 5’ end of the TS (Supplementary Fig. 3c). Interestingly, we observed a small and rapid increase in anisotropy from ~0.25 to ~0.27 upon adding Mg²⁺ followed by a larger and slower decrease in anisotropy to around ~0.044, consistent with TS strand cleavage and separation from the RNP–DNA complex²⁰. The initial rapid increase in anisotropy is indicative of a transient restriction in the mobility of the ROXN dye prior to TS cleavage. We attribute this to the structural rearrangements preceding TS cleavage observed previously in which the TS is bent into the RuvC active site¹⁰, bringing the 5’ end of the TS in towards the enzyme and leading to the restriction of dye motion. We observe no difference in TS cleavage dynamics between the WT, engineered Cas12a, and Y575A mutant. The K550A did show a smaller magnitude of anisotropy decrease in the slow phase compared with the other variants but very similar kinetics, likely due to the presence of some inactive enzyme in that preparation.

On the other hand, the F870L mutation in the engineered AsCas12a variant has no obvious effect on the complex (Supplementary Fig. 4). However, this mutation may play a bigger role during earlier stages of complex formation, such as during PAM recognition or early R-loop formation. The exact mechanism for how F870L contributes to enhanced activity of this engineered AsCas12a remains to be determined. It should also be noted that in the original study, the F870L single mutant had a lower overall effect on potency compared with the single M537R mutant¹¹. However, the double mutant had overall better activity than either single mutant, albeit in a sequence-dependent context. Therefore, it could also be that with this particular DNA target, the F870L mutation has a minimal effect.

M537R allows for relaxed PAM recognition

NTS and TS DNA cleavage by engineered and WT AsCas12a was determined by in vitro gel-based cleavage assays using carboxyfluorescein (FAM)-labeled DNA substrates. As expected, based on previous data, engineered AsCas12a cleaves a FAM-labelled NTS substrate faster than WT (Supplementary Fig. 5a, Supplementary Fig. 6)⁹. In contrast, TS cleavage is not significantly different between the two enzymes, highlighting the slow, rate-determining, structural rearrangements required for the TS to be positioned in the active site (Supplementary Fig. 5b). Together, these results indicate that the increased efficiency of engineered AsCas12a is the result of mechanistic steps that occur upstream of DNA cleavage.

One possible contributor to the increased efficiency is the reduced DNA interaction caused by M537R which could allow for recognition of a more relaxed PAM sequence. To test this hypothesis, we performed NTS cleavage assays on DNA substrates with different PAMs. According to previous literature, this engineered AsCas12a is active on TTTN PAM sequences, as opposed to TTTV for WT¹¹. Using a gRNA targeting the beta-2-microglobulin (B2M) locus, we performed cleavage assays for engineered and WT AsCas12a on identical DNA sequences with either a TTTA or TTTT PAM. As expected, engineered AsCas12a maintains cleavage activity on a TTTT PAM substrate, albeit with a reduced cleavage rate, while WT has very little activity (Fig. 2d, e, Supplementary Fig. 6). Together, these data support the hypothesis obtained from the crystal structure that the displacement of Y575 allows for more promiscuous DNA interactions.

To determine if this promiscuity also allows for the cleavage of mismatched templates, we performed NTS cleavage assays using gRNAs with single mismatches in the spacer (Supplementary Fig. 7). Both WT and engineered AsCas12a were severely impaired by both PAM-proximal and PAM-distal mismatches within the spacer. A mismatch at position 10 did not affect either WT or engineered AsCas12a under these cleavage conditions, consistent with earlier findings of a reduction in specificity in the middle of the R-loop compared to the edges^9,15. Interestingly, while significantly impaired compared against a perfectly matched spacer, PAM-proximal mismatches still retain some cleavage activity with the engineered AsCas12a but not with WT, arguing that the increased PAM promiscuity with the engineered enzyme does indeed lower the target specificity. However, it is important to note that the increased potency of this variant allows for reduced concentrations of enzyme while maintaining high levels of activity, which inherently reduces the risk for off-target cleavage. Additionally, any genomic location that includes even one mismatch would not be a viable target site, further reducing the possibility of off-target cleavage by either WT or engineered AsCas12a.

Engineered AsCas12a has increased protein interactions in the WED domain

To determine if the M537R and F870L mutations contribute to any differences in protein flexibility, we compared the temperature factors (B-factors) for both the WT (5B43) and engineered AsCas12a crystal structures. B-factors were plotted for each residue and compared without application of a normalization procedure since the B-factors averaged over the whole protein complex differed by only 0.24% between the WT and engineered Cas12a structures (Supplementary Fig. 8)²¹. Interestingly, many regions within engineered AsCas12a appear to exist in a more rigid state, particularly in the WED domain surrounding the M537R mutation (Fig. 3, Table 3). Introduction of additional protein–protein interactions appear to stabilize multiple WED domain residues that are close to the NTS, although not directly interacting with the DNA within this ternary complex (Fig. 3a). However, it is possible that this more rigid protein interaction network may promote a more stable, cleavage-competent conformation, essentially lowering and stabilizing the transition state for NTS cleavage. In contrast, in WT AsCas12a, many residues within this region appear to be very flexible, potentially preventing efficient formation of an NTS-cleavage-ready complex (Fig. 3b).

Fig. 3. Introduction of new hydrogen bonds in WED domain.

a Protein interaction network surrounding M537R mutation in engineered AsCas12a. Residues are colored by B-factor. b Protein interaction network surrounding M537R mutation in WT AsCas12a. Color scale for B-factor colors is shown below both panels. Yellow dotted lines indicate a hydrogen bond. AsCas12a Acidaminococcus sp. CRISPR-associated protein 12a, CRISPR clustered regularly interspaced short palindromic repeats, WED wedge, WT wild-type.

Table 3.

Differences in distance between M537-adjacent residues in engineered vs. WT AsCas12a

Donor	Acceptor	Distance (Å) engineered	Distance (Å) WT	Delta (Å)
Arg537 Amide NH	Asn534 Amide CO	2.80	3.00	−0.20
Asn534 SC NH	Arg537 Amide CO	3.40	3.30	0.10
Phe535 Amide NH	Asp40 SC COO‒	3.10	3.70	−0.60
Asn534 Amide NH	Asp40 SC COO‒	2.60	3.20	−0.60
Lys570 Amide NH	Phe535 Amide CO	3.50	4.90	−1.40
Gln34 SC NH	Gln571 SC CO	3.00	6.20	−3.20
Lys572 Amide NH	Gln34 SC CO	3.10	6.90	−3.80
Leu540 Amide NH	Arg537 Amide CO	3.10	3.30	−0.20
Arg43 SC NH	Gln536 SC CO	2.90	7.10	−4.20
Arg43 SC NH	Phe165 Ring	3.65	5.07	−1.42

AsCas12a Acidaminococcus sp. CRISPR-associated protein 12a, CRISPR clustered regularly interspaced short palindromic repeats, SC side chain, WT wild-type.

Engineered AsCas12a exhibits accelerated R-loop formation kinetics

Previous work on Cas12a has established that R-loop formation represents the rate-limiting step in Cas12a catalysis⁹. Similarly, for Cas9, R-loop formation has been found to be rate limiting for target cleavage²². To determine if the engineered AsCas12a mutations impact the kinetics of R-loop formation, we performed temperature-dependent stopped-flow fluorescence spectroscopy. A DNA template containing a fluorescent tricyclic cytidine analog (tC^o) in the NTS was used to probe R-loop formation²². These modified cytidine analogs represent ideal probes for exploring Cas protein dynamics because of their compact size and minimal perturbation of nucleic acid properties when substituting for cytidine DNA or RNA bases^23,24. The quantum yield of tC^o increases from ~0.22 in dsDNA to ~0.30 when contained in single-stranded DNA (ssDNA). Thus, formation of the TS–gRNA hybrid R-loop and displacement of the NTS converts the NTS to ssDNA form and leads to a fluorescence increase, reporting R-loop formation in real time²⁴.

We first verified the increase in fluorescence upon reaction of RNP with fluorescent DNA substrate due to R-loop formation via steady-state fluorescence spectroscopy (Supplementary Fig. 9). Next, we performed temperature-dependent stopped-flow fluorescence spectroscopy to examine the kinetics and activation thermodynamics of R-loop formation in WT AsCas12a and the engineered variant. A large excess of RNP (final concentration: 100 nM) was shot against a limiting amount of DNA target duplex (final concentration: 5 nM), substituted with tC^o at position 9 in the spacer sequence on the NTS (Fig. 4a, b). Under these conditions, the bimolecular association and PAM recognition steps are predicted to occur in a single exponential phase with a 92.6 ms time constant as shown by numerical simulation (Supplementary Fig. 10)⁹, enabling observation of slower dynamics. A strong increasing fluorescence signal was observed immediately on RNP–DNA target mixing for both WT and engineered AsCas12a between 37 °C and 20 °C which was well fit by bi-exponential functions. In the WT, the kinetics became too sluggish to observe below 20 °C (Fig. 4a), while for the engineered variant, an R-loop formation signal was evident down to 13 °C (Fig. 4b).

Fig. 4. Time-resolved fluorescence kinetics of R-loop formation.

a WT and b engineered AsCas12a R-loop formation dynamics monitored by tC^o fluorescence. Samples excited at 360 nm and fluorescence intensity measured at 450 nm. Data were fit to the sum of two exponentials (black traces) and τ_R-loop, calculated as the amplitude weighted lifetime: ${\tau }_{obs}={\sum }_i\frac{a_i{\tau }_i}{a_i}$, is reported in c. Reported error is the propagated standard deviations from fitting the kinetic traces. AsCas12a Acidaminococcus sp. CRISPR-associated protein 12a, cps counts per second, CRISPR clustered regularly interspaced short palindromic repeats, nd not determined, tC^o tricyclic cytidine analog, WT wild-type.

To quantify the differences in the R-loop formation dynamics between the WT and engineered variants, we treated the bi-exponential kinetics as being representative of the overall R-loop formation process and extracted the amplitude-weighted lifetimes (Fig. 4c) from the bi-exponential fits. Between 37 °C and 20 °C, engineered AsCas12a shows a remarkable acceleration of R-loop formation by nearly an order of magnitude compared with the WT enzyme. Together, these results demonstrate that R-loop formation is greatly accelerated in engineered AsCas12a compared with WT.

Heterogenous R-loop formation kinetics likely arise from multiple initial conformations

Previous observations⁹ of AsCas12a R-loop formation dynamics probed by 2-amino purine (2-AP) fluorescence of target DNA showed biphasic kinetics with an initial slow phase attributed to R-loop formation and a second faster phase attributed to a conformational rearrangement (Scheme S1). Similarly, biphasic R-loop formation kinetics probed by fluorophore-labeled target DNA in Cas9 were attributed to a sequential 2-step unwinding process which gated each strand cleavage event²². Other work has highlighted conformational heterogeneity in Cas12a²⁵ with apo, gRNA-bound, and DNA-bound forms all sampling multiple distinct conformations in solution. The various observations^9,22 (see above) of heterogeneity in R-loop formation kinetics suggest the possibility that the observed heterogenous kinetics could arise from conformational equilibria between states of gRNA-bound Cas12a, which are competent to initiate DNA recognition and R-loop formation but exhibit distinct R-loop formation dynamics (Scheme S2a, b).

Therefore, we further examined the temperature dependence of the biphasic kinetic data observed with tC^o fluorophore to distinguish between these two possibilities as well as gain further insight into differences in the R-loop formation mechanism between the engineered and WT variants. We considered the fast and slow phases of R-loop formation separately (Fig. 5a). The lifetimes of the fast and slow phases for each variant show that both processes are substantially faster in the engineered enzyme compared with the WT. The temperature dependence of the relative amplitude of the fast phase (Fig. 5b, Supplementary Fig. 11) reveals that the relative contributions of the fast and slow processes to the total signal also show a dramatic temperature dependence, with the fast phase contribution predominating at 37 °C and the slow phase contribution predominating at lower temperatures. Furthermore, both variants show very similar temperature-dependent contributions of the fast and slow phases to the total signal amplitude. The engineered variant does show an unexpected increase in the contribution of the fast phase at 13 °C, however we were not able to observe an increasing fluorescence signal in the WT enzyme at this temperature. Further, the dramatic slowing of R-loop formation at this temperature precluded observation of the full extent of R-loop formation, potentially impacting the accuracy of the amplitudes derived from fits to the data at that temperature. With this analysis, it becomes possible to consider the possible origin of the multiphasic kinetics. While several kinetic models could explain this biphasic behavior (Supplementary Section 1), the data are well described by a model in which R-loop formation initiates from at least two distinct populations of RNP conformations that each have different intrinsic reaction rates, which interconvert via a temperature dependent equilibrium external to the R-loop formation process as depicted in Scheme S2a. This working model allows for a direct comparison of the R-loop formation dynamics between WT and engineered enzymes.

Fig. 5. Fast and slow phases of R-loop formation.

a Temperature dependence of the lifetimes of the fast and slow phases of R-loop formation in WT and engineered AsCas12a. b Temperature dependence of the fraction of the total amplitude of the fluorescence signal contained in the fast phase for WT AsCas12a (orange) and engineered AsCas12a (blue). Reported error is the standard deviation from Igor Pro 9 curve the fitting procedure. AsCas12a Acidaminococcus sp. CRISPR-associated protein 12a, CRISPR clustered regularly interspaced short palindromic repeats, nd not determined, WT wild-type.

Engineered AsCas12a mutations lower activation barrier for R-loop formation

To gain further insight into the effect of the M537R and F870L mutations on the mechanism of R-loop formation, we examined the R-loop formation dynamics from the perspective of transition state theory. We first considered the combined biphasic kinetics as representative of the overall or average R-loop formation process. Eyring plots were constructed with the amplitude-weighted lifetimes (Fig. 6a, b) of R-loop formation. In both variants, linear regions were evident between 13 °C and 31 °C with the rate at 37 °C noticeably deviating from the linear trend coinciding with the temperature where the fast process contributes most of the signal amplitude. Such non-linear Eyring plots commonly arise in enzyme kinetics. This downward curvature is often indicative of a change in rate limiting step with temperature. For instance, the intrinsic chemical steps of R-loop formation may be rate limiting at lower temperature while a less temperature-sensitive upstream step such as DNA binding or PAM interrogation could become a kinetic bottleneck at higher temperatures. Alternatively, significant non-linearity can imply a significant change in the heat capacity of activation, $\mathrm{\Delta }{C}_p^{‡}$^26,27, which is sometimes invoked in processes involving large scale conformational changes or significant solvent reorganization or even enzyme denaturation at higher experimental temperatures, though it should be noted that the temperature range is comfortably below the denaturation temperature of the enzyme–gRNA complex (46 °C)²⁸ and attempts to fit the data with a non-zero $\mathrm{\Delta }{C}_p^{‡}$ using macromolecular rate theory²⁶ yielded extremely poor fits. Given this complexity we considered the linear region of the data between 13 °C and 31 °C. Data were fit to the Eyring Eq. (1) in the linear region for both WT and engineered AsCas12a, revealing substantial differences in apparent activation parameters between the two enzyme variants. The apparent activation enthalpy of R-loop formation, ΔH^‡, dropped from 60.8 kcal/mol to 52.4 kcal/mol in the engineered variant compared with the WT. The apparent activation enthalpies for the WT enzyme value of 60.8 kcal/mol and in the engineered enzyme are exceptionally high for a single biochemical step, underscoring the complexity of the kinetic barrier. Even so, the comparison is informative since the engineered enzyme shows a substantial 14% reduction in this apparent enthalpic barrier. Perhaps more revealingly, WT shows a positive entropy of activation, ΔS^‡, of 137.6 cal/(mol*K) which is reduced to 114.4 cal/(mol*K) in the engineered variant, suggesting that an increased organization is retained in the transition state of engineered AsCas12a while the overall large entropic gain in both cases is consistent with a transition state that involves the melting of the DNA duplex and the displacement of ordered solvent molecules.

$$ln\frac{k}{T}=ln\frac{k_B}{h}-\frac{\mathrm{\Delta }{H}^{‡}}{RT}+\frac{\mathrm{\Delta }{S}^{‡}}{R}$$ 1

Fig. 6. Eyring analysis of R-loop formation kinetics.

a Kinetic data for R-loop formation was plotted ln(k_R-loop/T) against 1/T for WT (orange) and engineered AsCas12a (blue) where k_R-loop = τ_R-loop^-1. Data was fit to the linear Eyring equation between T = 31°C and 13°C (black dashed lines). Solid colored lines depict the 68% confidence bands for the linear fits. b Table of extracted parameters for fits to the linear Eyring equation. The reported error represents ± 1 standard deviation reported by Igor Pro 9 curve fitting procedure and is propagated for the % difference. AsCas12a Acidaminococcus sp. CRISPR-associated protein 12a, CRISPR clustered regularly interspaced short palindromic repeats, WT wild-type.

We then examined the two phases independently in Eyring plots (Supplementary Fig. 12a, b) which again gave non-linear curves for each component. The fast and slow phases clearly respond differently to the mutations in the engineered variant. It is also clear that the average parameters calculated (Supplementary Fig. 12c) above are primarily due to the contribution of the slow phase, for which ΔH^‡ and ΔS^‡ are both lowered by similar percentages to the averaged case (Fig. 6b). Interestingly, for the fast component, the slope of the high temperature region of the Eyring plot begins to deviate from the lower temperature slope at a lower temperature breakpoint than all the other phases (at 31 °C). At this point, it is possible that the intrinsic rate of the fast phase of R-loop formation becomes gated by a preceding step in the mechanism, such as DNA binding or PAM recognition.

Discussion

In this study we suggested a mechanistic model where the increased potency of engineered AsCas12a compared with WT AsCas12a arises from accelerated R-loop formation. Previous comparisons of WT Cas12a and WT Cas9 have shown that while Cas12a shows significantly improved specificity, it is slightly less potent compared with Cas9²⁹. The engineered variant discussed in this report is not only much more potent than its WT counterpart but maintains the intrinsic specificity that is unique to Cas12a¹¹. The enhanced rate of RNA/DNA hybrid formation in engineered AsCas12a is likely accomplished by reduced protein interference, which allows NTS/TS unwinding at a faster rate than it does in the WT enzyme. The displacement of Y575 by M537R releases the DNA from a sterically confined position (Fig. 2b, c). This altered protein/DNA interaction likely affects the ability of the DNA to move more independently during catalysis by reduction of steric barriers. While the crystal structure highlighted in this study represents an enzyme in complex with a fully formed R-loop, the structural details of how this altered protein/DNA interaction affects initial DNA binding and R-loop propagation remain to be elucidated. Although, as evidenced by a more relaxed PAM requirement (from TTTV to TTTN), one can assume that the initial DNA binding is influenced by a reduced interaction with the 4^th PAM nucleotide. Additionally, the increased rigidity seen in the WED domain around the M537R mutation may facilitate quicker DNA binding through stabilization of a more cleavage-competent complex, which in turn would further enhance the efficiency of the enzyme.

Understanding the mechanistic differences between engineered and WT AsCas12a may allow further manipulation of Cas12a to generate even more potent and efficient versions. Previous protein engineering efforts have generated a variety of optimized Cas12a enzymes^12,14,15,30. Most variants harbor mutations in the DNA-binding domains, hinting at a shared engineering feature. By comparing all known Cas12a engineered variants, additional combinations of mutations or underexplored regions to perturb may become apparent.

In addition to uncovering the mechanistic differences between WT and engineered AsCas12a, we have also demonstrated a robust method for elucidating the energetics of rate-limiting R-loop formation. This study uses readily accessible stopped flow fluorescence spectroscopy methodology, and a commercially available fluorescent modified nucleoside, making it widely applicable for the study of other CRISPR nucleases. In our engineered variant, acceleration of R-loop formation kinetics is primarily due to a reduction in the enthalpic component of the activation barrier, which is somewhat compensated by a decrease in the entropy of activation (Fig. 7). Some hints for the structural basis of these altered energetics are apparent in the structural data. The altered conformation of the WED domain loop containing Y575 shows reduced interaction with the non-template DNA strand in the engineered Cas12a structure. While this structure reflects the state of the complex at the point of R-loop completion, it can be inferred that this loop conformation likely contributes to reducing target DNA steric clashes and polar interactions with the WED domain loop as the gRNA invades the target DNA along the R-loop formation coordinate. Our data suggest that laboratory evolution of CRISPR nucleases may leverage tuning of activation enthalpy to control enzyme reactivity when selection pressure to increase enzyme activity is applied. This motivates a general investigation of other CRISPR enzymes and engineered variants to uncover if this is a general strategy exploited by natural systems to tune the reactivity of Cas systems for optimal activity as viral defence mechanisms.

Fig. 7. Model of transition state difference between wild-type and engineered AsCas12a.

Proposed differences in the energy landscape for WT (orange) and engineered (blue) AsCas12a during the course of DNA binding, R-loop formation, and DNA cleavage. AsCas12a, Acidaminococcus sp. CRISPR-associated protein 12a, CRISPR clustered regularly interspaced short palindromic repeats, WT wild-type.

This mechanism also provides a framework for understanding the physical basis of Cas enzyme specificity and its relation to increased reactivity. A global lowering of the activation barrier is expected to increase reactivity towards both on-target cleavage and potential off-target as well (Supplementary Fig. 7). Mismatches, however, are discriminated against through an increase in the activation barrier to R-loop formation and a corresponding slowing of the R-loop formation kinetics⁹. Therefore, in the context of therapeutic applications, the main effect of examined engineered variants is to increase editing potency. This enables delivery of a smaller dose of effector complex to achieve a desired clinical effect. Delivery of a lower dose of effector complex also serves to counterbalance any increased propensity for cleavage of off-target sites resulting in a similar specificity of WT and engineered AsCas12a variants¹⁵. Increasing potency also contributes to mitigating potential dose-limiting toxicities of delivery modalities such as messenger RNA-lipid nanoparticles or viral vectors, widening their therapeutic window. This also highlights the importance of ensuring that, when comparing specificity and off-target activity between different gene-editing systems, conditions are carefully controlled to compare between similar (and non-saturating) levels of on-target activity¹⁵.

Here we have provided insight into the mechanism of the increased potency of an engineered Cas12a. An AsCas12a that incorporates the mutations from this highly efficient variant has been tested in clinical trials^16,17. Similarly, Cas9 engineering efforts are focused on improving specificity while maintaining potency³¹. The inherent differences between Cas9 and Cas12a make them complementary tools in the gene-editing toolbox. With their different PAM requirements and distinct DNA cutting mechanisms, a wider array of DNA sequences can be successfully targeted. However, the enhanced specificity and resulting reduced off-target DNA cleavage of Cas12a compared with that of Cas9, as well as the shorter gRNA advantage of Cas12a, make it a particularly promising enzyme to put further focus on, especially as more gene editing therapeutics progress towards the clinic.

Methods

AsCas12a protein expression and purification

The DNA / RNA / protein complex (Cas12 + gRNA + target DNA) was purified by size-exclusion liquid chromatography. This purification is based on Yamano et al. ¹⁸. In short, AsCas12a was expressed in E. coli Rosetta2 cells at 20 °C. After purification, AsCas12a, gRNA, target DNA, and non-target DNA were complexed in a 1:1.5:2.3:3.4 ratio in 10 mM HEPES pH 7.5 and 150 mM NaCl. gRNA, target DNA, and non-target DNA were synthesized in-house.

Crystallization of Cas12a/DNA/RNA ternary complex

The Cas12a/DNA/RNA complex was crystallized as described by Yamano et al. ¹⁸. A custom screen was designed based on the published crystallization conditions, then further optimized after initial crystals were obtained. The best crystals grew in hanging-drop trays with 3 μL protein complex and 2 μL reservoir solution (0.1 M Na acetate, pH 4.3, 6% polyethylene glycol (PEG) 3350, 10% hexanediol) incubated over 500 μL of reservoir solution at 20 °C within a few days. These crystals were cryo-protected with reservoir solution supplemented with 10% xylitol/10% erythritol and flash-frozen for shipment to a synchrotron X-ray source.

Data collection, processing, and structure determination

X-ray diffraction data were measured at beamline 17-ID-2 (FMX) at the NSLS-II synchrotron, Brookhaven National Laboratory, NY, USA. All data were collected at a single wavelength of 0.97934 Å using a Dectris Pilatus 6 M, on August 2, 2022.

X-ray diffraction data from selected data sets were indexed and integrated with X-ray Detector Software (XDS)³², merged, scaled, and averaged using AIMLESS (CCP4 Suite)³³. The structure of the ternary complex was determined by molecular replacement using PDB accession code 5B43. The protein models were optimized using repeated rounds of model building and real-space model refinement performed with the program Crystallographic Object-Oriented Toolkit (COOT)³⁴, alternating with restrained refinement with the Phenix program Phenix.Refine³⁵. The DNA and RNA molecules were mutated according to their sequences using COOT after a few cycles of refinement.

In vitro cleavage assays

For all gel-based cleavage assays, 1 uM DNA duplex was first annealed in 1X anneal buffer (50 mM HEPES pH 7.5, 100 mM NaCl). TS and NTS were annealed in 1:1.2 ratios of FAM-labeled:unlabeled strand. The DNA duplex was heated at 95 °C for 5 min followed by slow cooling to room temperature on the bench top. 1 uM AsCas12a RNP was assembled at room temperature in 1X assembly buffer (50 mM HEPES pH 7.5, 120 mM NaCl, 5 mM MgCl2, 2 mM TCEP, 0.2 mg/ml BSA) with 2-fold excess gRNA (2 uM). RNP was incubated at room temperature for 20 minutes before proceeding with the cleavage reaction. Cleavage reactions were performed with 50 nM AsCas12a RNP and 10 nM FAM-labeled DNA duplex in 20 ul total of 1X assembly buffer. Cleavage was initiated by addition of 2 ul 100 nM FAM-labelled DNA duplex into the reaction. 2 ul aliquots were taken at indicated timepoints and quenched immediately into a 4 ul denaturing quench solution (90% formamide, 20 mM ethylenediaminetetraacetic acid [EDTA], 300 nM non-competing DNA oligo). Samples were heated at 90 °C for 3 minutes and then immediately run on a 20% denaturing PAGE gel. Gel was run at 100 V for 2 hours and imaged on a Typhoon scanner. Gel images were analyzed in ImageQuant and cleavage data was quantified in Prism.

Time-resolved fluorescence kinetics

Fluorescence data for the dual-dye labeled DNA template (Oligo table: DNA FRET template) was collected on a Horiba Fluorolog-3 instrument using quartz cuvettes under rapid magnetic stirring at 37 °C. For R-loop formation and subsequent structural changes, the template was excited at 532 nm and emission was observed at 550 nm to be sensitive to changes in the distance between the two PAM distal dyes. A duplexed dual-dye labeled template was prepared in tris-buffered saline (TBS) plus 1 mM EDTA at 10 uM and the sample was prepared by dilution to 2 nM in the same buffer plus 10% superblock buffer (Fisher Scientific). R-loop formation reaction was initiated by addition of a 20 uL aliquot of RNP for a final concentration of 50 nM. The subsequent structural change was initiated by addition of an aliquot of 500 mM Ca²⁺ to a final concentration of 3 mM. To measure fluorescence anisotropy within the same run, optical collection parameters were adjusted on the fly to 590 nm excitation and emission collected at 620 nm with a 15 nm slit width along with 0 ° and 90° polarization for the detectors in a T-shaped collection geometry and 0 ° polarization for the excitation. The strand cleavage reaction was then initiated by addition of an aliquot of 1 M Mg²⁺ to a final concentration of 6 mM. The G factor was calculated by rotation of the excitation polarization to 90 ° and collecting a short kinetic trace after completion of the reaction and stabilization of the fluorescence signal. Each condition was measured in duplicate or triplicate, and representative data is shown.

Stopped flow data was collected on a Horiba Fluorolog-3 instrument equipped with an SFA-20 stopped flow accessory with temperature control via a recirculating chiller. The two sample syringes were loaded with (a) RNP at 2x final concentration and (b) tC^o-labeled DNA template at 2x final concentration. Samples were prepared in 5 mM Mg²⁺, 1 mM tris (2-carboxyethyl) phosphine (TCEP) containing TBS buffer. Sample excitation was set to 350 nm and emission was collected in a 90 ° geometry at 450 nm. Excitation and emission slits were set to 5 nm and 8 nm bandpass respectively. Data was collected between 25 and 50 ms integration times and each curve represents the average of at least three individual shots. Between temperatures the system was allowed to equilibrate for ~10 minutes.

Data plotting and analysis was performed in Igor Pro 9.0. Curves were fit using Igor Pro’s built in curve fitting routine and reported errors are the standard deviations of the fitting parameters. Time-resolved fluorescence data was subjected to a binomial smoothing routine in Igor spanning 5 data points.

GaMD simulations

GaMD simulations were carried out using OpenMM 8.2^36,37. Simulations were performed using the Amber ff19SB protein force field along with the OL3 force field for RNA, the OL24 force field for DNA³⁸, and the TIP3P model for water. The OL24 forcefield for DNA was implemented to better model the A-form of the DNA/RNA hybrid R-loop which can be distorted by other DNA force fields which bias towards B-form DNA structures³⁹. An integration time of 2 fs was used and temperature control implemented via Langevin dynamics with collision frequence, γ = 1 ps^-1. All hydrogen bond lengths were constrained via the SHAKE algorithm. GaMD parameters were selected based on previous work⁴⁰ on Cas9 and applied the dual boost potential to both the total energy and dihedral energy. Conventional and GaMD equilibration prep stages were carried out for 1 ns. The conventional MD stage was run for 96 ns and GaMD equilibration for 48 ns. GaMD production was carried out for 618 ns and 699 ns for engineered Cas12a and WT Cas12a, respectively.

Starting structures for simulations were based on the reported crystal structure for WT AsCas12a in complex with gRNA and its DNA target (PDB accession code: 5B43) and the reported structure of engineered AsCas12a (PDB accession code: 9MK3). Coordinates of missing loop regions were obtained modeled based on other AsCas12a structural data (PDB accession code: 8SFO)¹⁰. The structures were embedded in an octahedral box of TIP3P water with a minimum 12 Å distance from the protein to the box edge along with 524 Cl^‒ ions and sufficient NA²⁺ ions to neutralize excess charge to model a typical experimental ionic strength of 0.12 M. Coordinate and topology files were constructed using the LEaP module in AmberTools 23.6. Reported PMF were calculated using the PyReweighting⁴¹ python scripts using the “noweight” option. As such the reported PMF are not corrected for the biasing of the GaMD boost potential due to insufficient sampling to supress the energetic noise in this large protein system.

Oligo table

1) CTCAGACACAGTCAGTTTATCACAGCCtC^OATGATAGTTAACCTGTC R-loop 9 NTS

2) GACAGGTTAACTATCATGGGCTGTGATAAACTGACTGTGTCTGAG R-loop 9 TS

3) NTS for gel-based cleavage CTCAGACACAGTCAGTTTATCACAGCCCATGATAGTTAACGCTACGACATTACAG

4) TS for gel-based cleavage CTGTAATGTCGTAGCGTTAACTATCATGGGCTGTGATAAACTGACTGTGTCTGAG

5) FAM-labeled NTS for gel-based cleavage /56-FAM/CTCAGACACAGTCAGTTTATCACAGCCCATGATAGTTAACGCTACGACATTACAG

6) FAM-labeled TS for gel-based cleavage /56-FAM/CTGTAATGTCGTAGCGTTAACTATCATGGGCTGTGATAAACTGACTGTGTCTGAG

7) Gel TTTT PAM NTS - /56-FAM/CTCAGACACAGTCAGTTTTTCACAGCCCATGATAGTTAACGCTACGACATTACAG

8) Gel TTTT PAM TS – CTGTAATGTCGTAGCGTTAACTATCATGGGCTGTGAAAAACTGACTGTGTCTGAG

9) Crystallization NTS fragment-

CAGTCCTTTT

10) Crystallization TS

GGTTGCCAAGCGCACCTAATTTCCAAAAGGACTG

11) Crystallization gRNA

UAAUUUCUACUCUUGUAGAUGGAAAUUAGGUGCGCUUGGCAACC

12) gRNA for B2M target DNA UAAUUUCUACUCUUGUAGAUUCACAGCCCAUGAUAGUUAAC

13) DNA FRET template NTS

5’- CTC AGA CAC AGT CAG TTT ATC ACA GCC CAT GAT AGT TAA CCT TGC /3ATTO532N/ 3’

14) DNA FRET template TS

5’- /56-ROXN/GCA AGG TTA ACT ATC ATG GGC TGT GAT AAA CTG ACT GTG TCT GAG - 3’

Statistics and reproducibility

For fluorescence kinetics experiments: Data plotting and analysis was performed in Igor Pro 9.0. Curves were fit using Igor Pro’s built-in curve fitting routine and reported errors are the standard deviations of the fitting parameters. Standard deviations were propagated by standard error propagation to report errors for parameters calculated from the fitting parameters. Time-resolved fluorescence data was subjected to a binomial smoothing routine in Igor spanning 5 data points. Experiment was carried out in duplicate or triplicate for the combined FRET-anisotropy experiment and for stopped-flow fluorescence each kinetics trace is the average of at least 3 separate kinetics traces.

For gel-based assays: All gel-based assays were performed in triplicate with each replicate being performed on a separate day. All cleavage data was analyzed using ImageQuant TL 1D v8.2.0. Background was subtracted using the rolling ball method. Areas under the curve of cleaved and uncleaved bands were automatically calculated and exported into GraphPad Prism 10.4.0. Nonlinear fits were calculated for each data set using the Specific binding with Hill slope function with B_max = 1. Data were plotted with each replicate shown.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

L.J-F., J.-N.L., and B.C. performed all experiments and wrote the manuscript. C.L., B.C., and S.W. conceived of the study. S.K. formed and purified the ternary complex, A.D. designed the crystallization experiments, optimized the crystal growth, process the data, solved crystal structure, and analyzed the data, A.O. and A.W. analyzed the structural data. All authors contributed to editing and reviewing the manuscript.

Peer review

Peer review information

Communications Biology thanks Jianhang Yin and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editors: Mo Li, Aylin Bircan and Mengtan Xing.

Data availability

The data generated and/or analyzed for this study are available from the corresponding author on reasonable request. The crystal structure of engineered AsCas12a has been deposited in the Protein Data Bank repository (PDB: https://www.rcsb.org/) under the four-digit PDB accession code: 9MK3. The source data can be found in the Supplemental Data File.

Competing interests

L.J-F., J.-N.L., and B.C. declare the following competing interests: current or former employees and shareholders at Editas Medicine, Inc. C.L. and S.W. declare the following competing interests: shareholders at Editas Medicine, Inc. A.O., A.D., A.W., and S.K. declare the following competing interests: current or former employees of Schrödinger and were contracted by Editas Medicine to support crystallographic experiments.

Supplementary information

The online version contains supplementary material available at 10.1038/s42003-026-09799-1.