DNA interference states of the hypercompact CRISPR-CasΦ effector.

Generation of casΦ-2 mutant expression vectors

Plasmids were cloned and mutagenized via Golden Gate assembly as previously described 8 . In brief, the pRSFDuet-1 derived casΦ-2 overexpression vector pPP085 8 (Addgene #158795) was amplified around the horn using primers containing the desired mutation and AarI Golden Gate cloning sites. The resulting fragment was circularized using the restriction enzyme AarI (Thermo Fisher Scientific) and T4 ligase (NEB). Plasmids were propagated in Escherichia coli MachI (QB3-Macrolab, UC Berkeley). Generated plasmids ( Supplementary Tab. 1 ) were sequenced across the coding sequence of casΦ-2 .

CasΦ-2 protein production and purification

C-terminally hexa-histidine-tagged CasΦ-2 was produced by heterologous expression in E. coli and purified as previously described 8 . In brief, overexpression plasmids were transformed into E. coli BL21(DE3)-Star (QB3-Macrolab, UC Berkeley). Expression cultures were grown shaking vigorously at 37 °C in 1.5 L TB-Kan (50 μg/mL Kanamycin) media to an OD 600 of 0.6. Subsequently, cultures were cooled down on ice for 15 min and gene expression was induced with 0.5 mM IPTG before incubation overnight at 16 °C. Cells were harvested by centrifugation and resuspended in wash buffer (50 mM HEPES-Na pH 7.5 RT, 1 M NaCl, 20 mM imidazole, 5 % glycerol and 0.5 mM TCEP), subsequently lysed by sonication, followed by lysate clarification by centrifugation. The soluble fraction was loaded on a 5 mL Ni-NTA Superflow Cartridge (Qiagen) pre-equilibrated in wash buffer. Bound proteins were washed with 20 column volumes (CV) wash buffer and subsequently eluted in 4 CV elution buffer (50 mM HEPES-Na pH 7.5 RT, 500 mM NaCl, 500 mM imidazole, 5% glycerol and 0.5 mM TCEP). The eluted proteins were concentrated to 1-2 mL before injection into a HiLoad 16/600 Superdex 200pg column (GE Healthcare) pre-equilibrated in size-exclusion chromatography (SEC) buffer (20 mM HEPES-Na pH 7.5 RT, 500 mM NaCl, 5 % glycerol and 0.5 mM TCEP). Peak fractions were concentrated to 1 mL and concentrations were determined based on the absorbance at 280 nm using a NanoDrop 8000 Spectrophotometer (Thermo Scientific). Proteins were purified at a constant temperature of 4 °C and concentrated proteins were kept on ice to prevent aggregation, snap frozen in liquid nitrogen and stored at −80 °C. Binary and ternary complex reconstitution for cryo-EM CasΦ-2 was produced as described above. crRNA (rPP012; ( Supplementary Tab. 2 )) was ordered as a synthetic RNA oligonucleotide from IDT (Integrated DNA Technologies) and dissolved in DEPC treated ddH 2 0 to a concentration of 0.5 mM. Subsequently, the crRNA was heated to 65 °C for 3 min and cooled down to RT to allow for hairpin formation. DNA oligonucleotides ( Supplementary Tab. 3 ) were designed to contain a non-complementary protospacer segment to produce “bubbled” substrates and facilitate rapid R-loop formation during ternary complex reconstitution. Oligonucleotides were ordered from and synthesized by IDT. DNA oligonucleotides were combined in a 1:1.2 molar ratio (Target strand:non-target strand) and annealed to a final DNA-duplex concentration of 0.5 mM in hybridization buffer (10 mM Hepes-Na pH 7.5 RT, 150 mM NaCl) by heating for 5 min at 95 °C and a subsequent slow cool down in a thermocycler. The CasΦ-2 binary complex was reconstituted by incubation of 20 μM CasΦ-2 and 24 μM crRNA for 10 min at RT in a total volume of 250 μL SEC buffer. For formation of the ternary complex, 40 μM DNA-duplex was added to the assembly reaction after formation of the binary complex and the sample was incubated for an additional 10 min at RT. Subsequently, assembly reactions were injected into a Superdex 200 10/300 GL column (GE Healthcare) pre-equilibrated in low salt (LS) buffer (10 mM Hepes-Na pH 7.5, 150 mM NaCl) at 4 °C to separate complexes from excess nucleic acids. Peak fractions were concentrated to 250 μL at 4 °C and concentrations were estimated using the absorbance at 280 nm as measured on a Nanodrop 8000 Spectrophotometer (Thermo Scientific), based on the absorbance at 280 nm of a sample containing the individual components mixed in the expected molecular ratio at a known concentration. Assembled complexes were kept on ice to prevent aggregation.

Electron microscopy grid preparation and data collection

CasΦ-2 complexes were frozen using an FEI Vitrobot Mark IV cooled to 8 °C at 100% humidity. The ternary complex was frozen at a concentration of 9 μM on carbon 1.2/1.3 400 mesh C-flat grids (Electron Microscopy Sciences #CF413-50), which were glow discharged at 15 mA for 25 sec prior to sample application using a PELCO easyGLOW. 4 μL of sample was applied to the grid and immediately blotted for 5 sec with blot force 8. The ternary complex in the presence of magnesium was prepared by addition of MgCl 2 to a final magnesium concentration of 5 mM and subsequently incubated for 15 min at RT before application onto 1.2/1.3 300 mesh UltrAuFoil gold grids (Electron Microscopy Sciences #Q350AR13A). Grids were glow discharged with 15 mA for 40 sec, 4 μL of sample was applied and blotted for 5 sec with blot force 8. The binary complex was frozen at a concentration of 3.7 μM on 1.2/1.3 300 mesh UltrAuFoil gold grids (Electron Microscopy Sciences #Q350AR13A), which were glow discharged with 15 mA for 25 sec. 4 μL of sample were applied to grids and grids were immediately blotted for 4 sec with blot force 8. Micrographs for all data sets were collected on a Talos Arctica operated at 200 kV and 36,000x magnification (1.115 Å pixel size), using the super resolution camera setting (0.5575 Å pixel size) on a K3 Direct Electron Detector. Cryo-EM data was collected using SerialEM v. 3.8.7 software. Images were collected using beam shift.

Single particle cryo-EM data processing and 3D volume reconstruction

For the binary complex, 2,934 movies in super resolution were collected with defocus ranging from −0.7 to −1.8 μm. Movies were processed in cryoSPARC (Structura Biotechnology Inc.) and resulted in a map with a preferred orientation. In order to overcome this problem a tilted data set with a 20° tilt was collected. The tilted data set was processed in cryoSPARC v3.1, independently from the untitled dataset and resulted in the final map and structure presented in this study ( Extended Data Fig. 1 ). For the tilted data set, 1,641 movies were collected in super resolution with defocus ranging from −1.0 to −2.2 μm. Movies were corrected for beam induced motion with patch motion and manually curated. After curation, 1,233 micrographs were used for CTF parameter calculation with patch CTF. Then, particles were picked with a blob picker from all 1,233 micrographs and 2,349,114 particles were extracted with bin factor 2. After one round of 2D classification, 435,648 particles were selected and re-extracted with re-centering and then used for ab initio reconstruction with 2 classes. Particles (298,679) from the best class were again re-extracted with re-centering. After local motion correction in cryoSPARC, the corrected particles were downsampled using bin factor 2 and the data sign was flipped. These particles were used to refine the best ab initio class using non-uniform refinement 45 with per group CTF optimization including beam tilt and trefoil fitting. The resulting map reached 3.54 Å resolution. Half maps from this refinement were used for LocSpiral 46 enhancement of weaker density regions. The LocSpiral map was used for model building. For the ternary complex ( Extended Data Fig. 4 ), 4,374 movies were collected with defocus ranging from −0.8 to −2 μm. Data processing was performed with cryoSPARC v3.0.1. Movies were corrected for beam induced motion using patch motion, and CTF parameters were calculated using patch CTF. After micrograph curation, 4,090 micrographs were included for further data analysis. 20 micrographs were used for particle picking with blob picker for preliminary 2D classification and template selection for Topaz particle training 47 . The resulting Topaz model was used to pick particles from all micrographs. 2,666,893 particles were extracted from micrographs with binning factor 2 and used for further analysis. Two rounds of 2D classification were performed, leading to a set of 1,733,254 selected particles. These particles were re-extracted with re-centering and used for ab initio reconstruction with 5 classes. The classes were evaluated visually and the class with the best defined densities for all domains and the largest number of particles (819,071 particles) was chosen and used as a seed for 3 classes in heterogeneous refinement performed on a set of particles from this class. Upon visual inspection, a class from heterogeneous refinement with the highest number of particles (419,818) and 3.64 Å resolution was chosen for map refinement. Particles from this class were again re-extracted with re-centering and duplicate particles were removed (23,287 particles removed). We performed a homogeneous refinement and non-uniform refinement with per group CTF correction without spherical aberration fitting and without tetrafoil fit and obtained volumes with resolutions of 3.13 Å, and 3.05 Å, respectively. We chose the 3.05 Å resolution map as our final map. Half maps from the final volume were used in LocSpiral in order to enhance weaker map regions. The LocSpiral volume was used for model building. For 3D variability analysis with cryoSPARC v3.1, the best class from ab initio reconstruction (819,071 particles) was used. First, the map from this class was homogeneously refined with particles from this class. The output from this refinement was used for 3DVA 48 with 3 modes of motion. The 3 modes of motion were visualized by 3DVA simple display with 20 frames each (two frames from two modes of motion are included in Extended Data Fig. 10 ). For the ternary complex in the presence of magnesium ( Extended Data Fig. 8 ), 4,827 micrographs were collected with −0.5 to −1.8 μm defocus range. Micrographs in super resolution were corrected for beam induced motion with patch motion, and CFT was estimated using patch CTF in cryoSPARC v2.15. After curation, 1,962 micrographs were retained for further analysis. An initial set of particles was picked using Blob picker from a small subset of micrographs and used for initial 2D classification and template selection for Topaz particle picking training. The optimized Topaz picking model was used to pick particles from all micrographs. A total of 1,059,466 particles were extracted from micrographs with pixel size binning factor 2 and used for 2D classification. Promising classes with 679,252 particles were selected for ab initio reconstruction using 3 classes. One class volume (410,553 particles) appeared to correspond to the full-length complex and was used for homogeneous refinement, resulting in a map at 2.87 Å resolution. The homogeneous refinement output was used for non-uniform refinement, which improved the resolution to 2.84 Å. Half maps from non-uniform refinement reconstruction were used for LocSpiral map enhancement in order to improve regions with weaker density. The LocSpiral map was used for model building. Outputs from homogenous refinement were used for 3D variability analysis (3DVA) with 5 motion modes. The 5 modes of motion were visualized by 3DVA simple display with 20 frames each. Two frames from each mode of motion are shown in the supplement ( Extended Data Fig. 10 ).

Model building and refinement

The model of the CasΦ-2 ternary complex was built de novo in Coot 49 using LocSpiral improved and unfiltered cryoSPARC maps, and refined using the Real-space refinement tool 50 as implemented in Phenix (Version 1.18, 51 ). The CasΦ-2 binary complex structure was modelled into the respective LocSpiral map, based on the ternary structure of CasΦ using rigid body fit and Real-space refinement as implemented in Coot and refined using the Real-space refinement tool as implemented in Phenix. The model of CasΦ-2 bound to phosphorothioate-modified DNA in the presence of magnesium was built based on the CasΦ-2 ternary model into the respective LocSpiral map, and refined using the Real-space refinement tool as implemented in Phenix. 2-deoxy-cytidine-5'-thiophosphorate (SC ligand) restraints were generated using Elbow 52 as implemented in Phenix. 3'-oxygen to 5'-phosphate bond angles involving phosphorothioate-modified DNA, as well as distances between the two magnesium cofactors and D394/D695, were restrained using custom geometry restraints in the Real-space refinement tool as implemented in Phenix. Custom angle geometry restraints were estimated by measuring the angles of 3'-oxygen to 5'-phosphate bonds observed for unmodified DNA, as refined in the corresponding cryoEM map using the Phenix implemented Real-space refinement tool and PyMol ( pymol.org ). For the binary model refinement, distances between the zinc cofactor and the surrounding cysteine residues were restrained based on distances measured for the ternary model. Models were refined using atomic displacement parameters, global minimization and local grid search refinement strategies as implemented in the Phenix Real-space refinement tool. The models were additionally refined using secondary structure, rotamer and ramachandran restraints as implemented in the Phenix Real-space refinement tool. Prior to the final refinement, hydrogens were added to the models using ReadySet as implemented in Phenix and local_grid_search refinement was disabled.

Data deposition and figure preparation

Cryo-EM maps and model coordinates were deposited to the EMDB (codes EMD-23600, EMD-23601 and EMD-23678) and PDB (codes 7LYS, 7LYT and 7M5O). Figures were prepared in UCSF ChimeraX 53 , UCSF Chimera 54 and Coot 49 . Cryo-EM map σ-levels were calculated as: map level/root-mean-square deviation from zero (RMS). Orientation distribution plot was generated using pyem csparc2star.py and star2bild.py programs 55 .

Map vs. model

FSC graphs were calculated in Mtriage, as implemented in Phenix 56 .

Gold standard FSC and map vs model

FSC were replotted using Prism 8 (GraphPad).

RNP complex reconstitution for DNA cleavage and binding assays

CasΦ-2 was produced as described above. crRNA guides ( Supplementary Tab. 2 ) were ordered as synthetic RNA oligonucleotides from IDT (Integrated DNA Technologies) and dissolved in DEPC treated ddH 2 0 to a concentration of 0.5 mM. Subsequently, the crRNA was heated to 65 °C for 3 min and cooled down to RT to allow for hairpin formation. CasΦ-2 RNP complexes were reconstituted at a concentration of 10 μM by incubation of 10 μM CasΦ-2 and 12 μM crRNA for 10 min at RT in 2x cleavage buffer (2xCB) (20 mM Hepes-Na pH 7.5, 300 mM KCl, 10 mM MgCl 2 , 20 % glycerol, 1 mM TCEP). Formed RNPs were aliquoted to a volume of 10 μL, flash frozen in liquid nitrogen and stored at −80 °C. Before usage, RNP aliquots were thawed on ice.

DNA cleavage and mismatch tolerance assay

DNA targets ( Supplementary Tab. 4 ) were cloned as previously described 8 . Mutations (A->C, T->G, C->A, G->T) were introduced in the target region by Golden Gate mutagenesis as described above. DNA targets were produced by PCR as previously described 8 and diluted to a final concentration of 20 nM in CB buffer (10 mM Hepes-Na pH 7.5, 150 mM KCl, 5 mM MgCl 2 , 10 % glycerol, 0.5 mM TCEP). CasΦ-2 RNPs were prepared as described above with either oligonucleotide rPP012 (20 nt spacer) for variant protein cleavage assays, or rPP013 (22 nt spacer) for mismatch tolerance assays. RNPs were diluted on ice in CB buffer to a final concentration of 2 μM. Reactions were initiated by addition of the RNP (1 μM final) to the target DNA (10 nM final) and incubation at 37 °C in a thermocycler for 1 h. Subsequently 0.08 units/μL Proteinase K (NEB) were added and the reaction was incubated for 30 min at 37 °C in a thermocycler. Loading dye (Gel Loading Dye Purple 6X, NEB) was added and samples (10 μL) were analyzed by electrophoresis on an 1% agarose gel, stained with SYBR Safe (Thermo Fisher Scientific). For the cleavage assay involving alternative PAMs and supercoiled DNA, 2 μL samples were separated on an 1% agarose gel, stained with SYBR Gold (Thermo Fisher Scientific). Gels were imaged using a ChemiDoc MP and Image Lab v5.2.1 software (Biorad). Substrate and product intensities were quantified using ImageJ (Version 1.51, https://imagej.nih.gov/ij/ ) 57 and the cleaved fraction was calculated as the product intensity sum divided by the combined substrate and product intensity sum. Data was plotted using Prism 8 (Graphpad). Filter binding assay Non-target DNA strands ( Supplementary Tab. 3 ) were 5′-end-labelled using T4-PNK (NEB) in the presence of 32 P-γ-ATP to track binding of the PAM-proximal DNA segment. Oligo duplex targets were generated by combining 32 P-labelled and unlabelled complementary oligonucleotides in a 1:1.5 molar ratio and annealed to a final duplex concentration of 50 nM in hybridization buffer (10 mM Hepes-Na pH 7.5 RT, 150 mM NaCl) by heating for 5 min at 95 °C and a subsequent slow cool down in a heating block. Binding reactions were initiated by combining CasΦ-2 with 0.1 nM DNA duplex in CB buffer. Reactions were subsequently incubated for 1 h at 37 °C in a thermocycler. For the binding experiment in the absence of magnesium, 12.5 mM EDTA was supplemented in CB buffer (w/o magnesium) throughout the experiment. Binding reactions were subsequently blotted at RT using a 96-well dot blot apparatus (Minifold I, S&S) assembled from below to top with two layers Whatman 1 filter paper (Whatman), one layer Amersham Hibond-N (GE Healthcare), one layer Amersham Protran 0.1 μm NC (GE Healthcare) and one layer HT-450 Tuffryn (PALL Life Sciences). Membranes and filter paper were pre-equilibrated for 1 h in CB buffer prior to assembly of the dot blot apparatus. Membranes were washed with 50 μL CB buffer per well prior to application of 30 μL CasΦ-2 binding reaction at an approximate flow rate of 15 μL/min . Bound proteins and nucleic acids were subsequently washed with 100 μL CB buffer. Membranes were air dried before phosphor-imaging visualization using an Amersham Typhoon scanner, v2.0.0.6 firmware version 208 (GE Healthcare).

Assessment of the HT-450

Tuffryn membrane showed no aggregated fraction, containing the radiolabeled substrate. Spot intensities were quantified using ImageQuant TL 8.1 (Cytivia) and the bound DNA fraction for each spot was calculated relative to the mean (n = 3) of the intensities observed at the maximum concentration of CasΦ-2 WT RNP. Curves were fitted using a sigmoidal four parameter logistic curve model in Prism 8 (Graphpad). Of note, “apparent K d ’s” were not derived from the data, since the assay was performed under non-equilibrium conditions using a final 100 μL wash step with CB buffer w/o DNA. In vitro cleavage assays - radiolabeled nucleic acids CasΦ RNP were assembled as described above. Substrates were 5′-end-labelled using T4-PNK (NEB) in the presence of 32 P-γ-ATP (Substrate sequences are given in Supplementary Tab. 3 ). Oligonucleotide-duplex targets were generated by combining 32 P-labelled and unlabelled complementary oligonucleotides in a 1:1.5 molar ratio. Oligos were hybridized to a DNA-duplex concentration of 50 nM in hybridization buffer (10 mM Hepes-Na pH 7.5 RT, 150 mM NaCl), by heating for 5 min to 95 °C and a slow cool down to RT in a heating block. Cleavage reactions were initiated by combining 200 nM RNP with 2 nM substrate in CB buffer and subsequently incubated at 37 °C. Reactions were stopped by addition of two volumes formamide loading buffer (96 % formamide, 100 μg/mL bromophenol blue, 50 μg/mL xylene cyanol, 10 mM EDTA, 50 μg/mL heparin), heated to 95 °C for 5 min, and cooled down on ice before separation on a 12.5 % denaturing urea-PAGE. Gels were dried for 4 h at 80 °C before phosphor-imaging visualization using an Amersham Typhoon scanner, v2.0.0.6 firmware version 208 (GE Healthcare). Bands were quantified using ImageQuant TL 8.1 (Cytivia) and the cleaved fraction was calculated as the product intensity sum divided by the combined substrate and product intensity sum. Curves were fitted to a One-Phase-Decay model in Prism 8 (graphpad) to derive the rate of cleavage. Fluorophore quencher assay CasΦ RNP were assembled as described above. Reactions were initiated by combining 100 nM RNP (100 nM CasΦ, 120 nM rPP012 crRNA), 100 nM DNase Alert (IDT) FQ probe, with and without activator ssDNA ( Supplementary Tab. 3 ) in cleavage buffer (10 mM Hepes-Na pH 7.5, 150 mM KCl, 5 mM MgCl 2 , 10 % glycerol, 0.5 mM TCEP) in a 384 well flat bottom black polystyrene assay plate (#3820, Corning). Three replicates for each reaction were monitored (λ ex : 530 nm; λ ex : 590 nm) in a Cytation 5 plate reader (BioTek, software Gen v3.04) at 37 °C every 1.5 min for the activator titration experiment. For the FQ-mismatch-assay, 2 nM activator oligonucleotides were used in singlicates. The data were background-subtracted using the mean values of the measurements taken for three no-activator controls at the respective time point. To derive the % relative fluorescence, the values were calculated for each variant as the ratio of the intensity measured for the respective mismatched activator divided by the intensity measured for the no-mismatch activator. Data were plotted in Prism 8 (graphpad).

Protein sequence and structure topology analysis

Protein sequences were aligned using the MUSCLE multiple sequence alignment tool58 and visualized in Jalview (version 2.11.1.4)59. The CasΦ secondary structure topology diagram was generated by PDBsum60.