Activation of Csm6 ribonuclease by cyclic nucleotide binding: in an emergency, twist to open

Abstract Type III CRISPR systems synthesize cyclic oligoadenylate (cOA) second messengers as part of a multi-faceted immune response against invading mobile genetic elements (MGEs). cOA activates non-specific CRISPR ancillary defence nucleases to create a hostile environment for MGE replication. Csm6 ribonucleases bind cOA using a CARF (CRISPR-associated Rossmann Fold) domain, resulting in activation of a fused HEPN (Higher Eukaryotes and Prokaryotes Nucleotide binding) ribonuclease domain. Csm6 enzymes are widely used in a new generation of diagnostic assays for the detection of specific nucleic acid species. However, the activation mechanism is not fully understood. Here we characterised the cyclic hexa-adenylate (cA6) activated Csm6’ ribonuclease from the industrially important bacterium Streptococcus thermophilus. Crystal structures of Csm6’ in the inactive and cA6 bound active states illuminate the conformational changes which trigger mRNA destruction. Upon binding of cA6, there is a close to 60° rotation between the CARF and HEPN domains, which causes the ‘jaws’ of the HEPN domain to open and reposition active site residues. Key to this transition is the 6H domain, a right-handed solenoid domain connecting the CARF and HEPN domains, which transmits the conformational changes for activation.


INTRODUCTION
Type III CRISPR systems eliminate invading nucleic acids via a multi-faceted immune response which, in addition to target RNA and non-specific ssDNA cleavage at the effector comple x, involv es the acti vation of ancillary defence enzymes.Target RNA binding and recognition licenses cyclic oligoadenylate (cOA) second messenger synthesis by the large Cas10 subunit of the effector complex ( 1 , 2 ).cOA, composed of 4 or 6 3 -5 linked AMP subunits (denoted cA 4 or cA 6 ), binds to CARF (CRISPR-associated Rossmann Fold) domains, activating CRISPR ancillary defence nucleases to abrogate viral replication.CARF-family proteins include the ribonucleases Csx1 / Csm6 ( 3-10 ), PD-D / ExK famil y DN A nucleases Can1 ( 11 ) and Can2 / Card1 ( 12 , 13 ) and the transcription regulator Csa3 ( 14 ).cA 4 or cA 6 stimulate Csx1 / Csm6 enzymes by binding to a dimeric CARF domain, w hich allostericall y activates the connected HEPN (Higher Eukaryotes and Prokaryotes Nucleotide-binding) RNase domains for cellular defence.cA 4 binding to dimeric Card1 results in a rotation of the two monomers with respect to one another, coupled with activation of the nuclease domains ( 12 ).In the past few years, the properties of Csx1 / Csm6 proteins as acti vatab le, non-specific RNases have been harnessed in a range of diagnostic assays to detect specific nucleic acid species (15)(16)(17)(18)(19).
Csx1 / Csm6 enzymes cleave nucleic acid non-selectively, ther efor e collateral damage to host transcripts during the immune response slows cell growth ( 5 ), posing a significant risk to long-term cell survi val.To remov e e xtant cOA, many type III CRISPR systems encode dedicated ring nucleases of the Crn1-3 families ( 20 , 21 ), which deactivate Csx1 / Csm6 enzymes by degrading the antiviral second messenger (re vie wed in ( 22 )).Some Csx1 / Csm6 enzymes also degrade their own cOA activators ( 4 , 23-27 ) and viruses utilize highly acti v e anti-CRISPR ring nucleases of the DUF1874 (AcrIII-1) family to subvert type III immunity ( 28 ).
For historical reasons, CARF-domain containing CRISPR ancillary ribonucleases associated with type III-A systems have been termed Csm6 and those associated with type III-B / D as Csx1.Howe v er, this classification does not adequately reflect the structural differences observed acr oss the family.Pr oteins that recognize a cA 4 activator are found in both the Csm6 and Csx1 groups ( 4 , 6 , 29 ).Although the enzymes are obligate dimers, some assemble into hexameric (trimer of dimer) conformations ( 6 ).The activation of this sub-class of enzymes on cA 4 binding appears to be quite subtle, with minimal changes in the ar chitectur e of the proteins observed when the apo and cA 4 -bound structures are compared ( 4 , 6 ).
The subset of Csm6 enzymes that recognize cA 6 , with r epr esentati v es in Streptococcus thermophilus , Mycobacterium tuberculosis and Enterococcus italicus , have a significantly different domain organization, with an alpha-helical '6H' domain linking the CARF and HEPN domains ( 30 ).The structure of E. italicus Csm6 (EiCsm6) bound to a fluorinated analogue of the cA 6 activator re v ealed the activated form of this family ( 24 ).cA 6 binding was predicted to cause conformational changes that activate the HEPN RNase site, while deactivation upon cA 6 cleavage at the CARF domain would re v erse these changes.Unfortunately, the absence of a Csm6 apo structure has pre v ented a direct test of this hypothesis.
Here we carried out a biochemical characterization of S. thermophilus Csm6' (StCsm6') and solved X-ray crystal structures in the presence and absence of cA 6 .This revealed a dramatic conformational change upon cA 6 binding, with a pronounced rotation of the dimeric CARF domains (through 58 • ), transmitted via the 6H domain, to prise open the HEPN domains, like the opening of a jaw (see Video Abstract).Based on these structures and supporting biochemical and EPR (electron paramagnetic resonance) data, we propose a model for activation of Csm6 proteins b y cA 6 , whereb y a large rotation of the CARF domains results in activation of the HEPN domains.

Expression and purification of Csm6' and variants
The pEV5HisTEV-csm6' wild-type and mutant constructs were transformed into C43 (DE3) E. coli cells.StCsm6' protein was expressed according to the standard protocol described previously (Rouillon et al., 2019).Briefly, 2-4 l of cell culture (in LB broth) containing the pEV5HisTEV-csm6' plasmid was induced with 0.4 mM isopropyl-␤-D-1-thiogalactoside (IPTG) at an OD 600 of ∼0.6 and grown overnight at 16 • C. Cells were harvested (4000 rpm; Beckman Coulter JLA-8.1 rotor) and resuspended in lysis buffer containing 50 mM Tris-HCl pH 7.5, 0.5 M NaCl, 10 mM imidazole and 10% glycerol.Cells were lysed by sonicating six times, with cycles of 1 min on ice and 1 min rest intervals and the cell debris removed by centrifugation.StCsm6' was purified with a 5 ml HisTrapFF column (GE Healthcare); following loading of the supernatant following cell lysis, the column was washed with 20 column volumes (CV) of buffer containing 50 mM Tris-HCl pH 7.5, 0.5 M NaCl, 30 mM imidazole and 10% glycerol, and the protein was eluted with a linear gradient of buffer containing 50 mM Tris-HCl pH 7.5, 0.5 M NaCl, 0.5 M imidazole and 10% glycer ol acr oss 15 CV.Pr otein containing fractions were concentrated and the 8-His affinity tag was removed by incubating protein with Tobacco Etch Virus (TEV) protease (10:1 StCsm6':TEV protease) overnight at room temperature.
Cleaved StCsm6' was further purified by repeating the immobilized metal affinity chromato gra phy step and collecting the unbound fraction.Size exclusion chromatography (HiLoad 16 / 60 Super de x 200 pg, GE Healthcare) was used to complete purification, with pure StCsm6' protein eluted isocratically in buffer containing 20 mM Tris-HCl pH 7.5, 150 mM NaCl.The protein was concentrated using a centrifugal concentrator, aliquoted and frozen at -70 • C.
For seleno-methionine labelled expression, the plasmid containing the csm6' gene was transformed into E. coli B834 (DE3) cells.Cells were grown in M9 minimal medium supplemented with Selenomethionine Nutrient Mix (Molecular Dimensions, Ne wmar ket, Suffolk, UK) and 50 mg l −1 ( L )-selenomethionine (Acros Organics).The protein was purified by the same method described for nati v e StCsm6'.StCsm6' variants were expressed and purified as described for the WT protein.

Radiolabelled RNA cleavage assays
To determine RNA cleavage by StCsm6' and variants, 50 nM radiolabelled RNA oligonucleotide A1 (5 AGGGU AUUAUUUGUUUGUUUCUUCU AAA CUAUAAGCUAGUUCUGGAGA) was incubated with protein (1 M dimer) and cOA activator (cA 3 , cA 4 or cA 6 ) (BIOLOG, Life Sciences Institute, Bremen) in buffer containing 20 mM HEPES pH 7.0, 150 mM NaCl, 1 mM dithiothreitol (DTT), 1 mM EDTA and 3 units SUPERase •In RNase inhibitor a t 45 • C .Control reactions incubating RNA in buffer without protein, and RNA with protein in the absence of cOA activator were also carried out.Reactions were stopped by adding phenol-chloroform and vortexing to remove protein, and 5 l of reaction product was extracted into 5 l 100% formamide xylene-cyanol loading dye.All experiments were carried out in triplicate and RNA cleavage was visualized by phosphor imaging following denaturing polyacrylamide gel electrophoresis (PAGE).
For deactivation assays, cA 6 (10,20,50, 100, 200 nM) was preincuba ted with buf fer alone (control) or with StCsm6', S105W or H336A variants (1 M dimer) for 60 min at 45 • C prior to adding 50 nM radiolabelled RNA and StCsm6' (1 M dimer) and incubating for a further 30 min at 45 • C. Control reactions, with cOA not preincubated or treated with StCsm6' (c1), and no cOA but RNA and StCsm6' added after preincubation of buffer (c2), were carried out.Reactions were stopped by adding phenol-chloroform as detailed above and RNA cleavage was visualized by phosphor imaging after denaturing PAGE.RNA cleavage was quantified using the Bio-Formats plugin ( 32 ) of ImageJ as distributed in the Fiji package ( 33 ).RNA protected was calculated using the no cOA control reaction as 100% protection.All experiments were carried out in triplicate.

Crystallization
StCsm6' was mixed with cA 6 (hereafter StCsm6'-cA 6 ) in a 1:1.5 molar ratio of StCsm6':cA 6 and incuba ted a t room temperature for 30 min prior to crystallization.Crystallization trials were performed using sitting drop vapour diffusion, with JCSG and PACT 96-well commer cial scr eens (Jena Bioscience), set up with an Art Robbins Gryphon robot.Non-labelled StCsm6'-cA 6 crystallized at a concentration of 12.5 mg / ml and the selenomethionine-labelled StCsm6'-cA 6 (hereafter SeMet-StCsm6'-cA 6 ) at 10 mg / ml.Following optimization using hanging drops in a 24 well plate, crystals were obtained from 1.85 M sodium malonate and 5% ethanol, over a reserv oir v olume of 600 l, for StCsm6'-cA 6 , and 1.6 M sodium malonate, over a reservoir volume of 1 ml, for SeMet-StCsm6'-cA 6 .3 l drops in a 2:1 or 1:1 protein:mother liquor ratio were added to a silanized cover slip and sealed with high-vacuum grease (DOW Corning, USA) and incuba ted a t room tempera ture.Crystals were harvested and cryoprotected with the addition of 20% PEG 1000 and 10% glycerol to the mother liquor, mounted on loops and vitrified in liquid nitrogen.
No diffraction quality crystals were immediately forthcoming for apo StCsm6'; a crystal was observed after 6-12 months incubation in the initial 96 well crystallization screens, with the drop noticeably dehydrated.Apo-StCsm6'was crystallized from 0.95 M sodium citrate, 0.18 M sodium bromide and 0.1 M HEPES, pH 8. Crystals were harvested and cryoprotected with the addition of 20% PEG 1000 to the mother liquor, mounted on loops and vitrified in liquid nitrogen.
X-ray data processing, structure solution and refinement X-ray data for SeMet-StCsm6'-cA 6 were collected at Diamond Light Source (DLS) on beamline I04, at a wavelength of 0.9795 Å , to 2.61 Å resolution.Dif fraction da ta were automatically processed through the Xia2 pipeline ( 34 ), using XDS and XSCALE ( 35 ), and showed a strong anomalous signal.The data were phased using the automated experimental phasing pipeline SHELX ( 36 ) in CCP4 Online ( 37 ), and an initial model of StCsm6' was built using the ARP / wARP w e bservice ( 38 ).No further refinement was carried out on this model.X-ray data for StCsm6'-cA 6 were collected at DLS on beamline I04 and processed to 1.96 Å resolution using autoPROC ( 39 ) and STARANISO (40; http://star aniso.globalphasing.org/cgi-bin/staraniso.cgi).MOLREP ( 41 ) was used to phase the data for StCsm6'-cA 6 using molecular replacement with the model generated by ARP / wARP (from the SeMet-StCsm6'-cA 6 data) as the search model.Iterati v e cy cles of REFMAC5 ( 42 ) or PHENIX ( 43 ) and COOT ( 44 ) were used for automated and manual refinement of the model respecti v ely, including addition of water molecules.Electron density for cA 6 was clearly visible in the maximum likelihood / A weighted F obs -F calc electron density map at 3 .cA 6 was drawn using Chemdraw (Perkin Elmer), restraints generated in JLigand ( 45 ), and positioned using COOT ( 44 ).
X-ra y data f or apo-StCsm6' were collected at DLS on beamline I03 and processed to 3.54 Å resolution using FAST DP ( 34 ), which incorporates XDS ( 35 ), CCP4 ( 37 ) and CCTBX ( 46 ).Molecular replacement using PHASER ( 47 ) was used to phase the apo-StCsm6' data, using the StCsm6'-cA 6 structure as the model (with cA 6 r emoved).Successful phasing r equir ed the N-and Cterminal domains of StCsm6' to be separated into two distinct search models.The model was refined as described above.
Throughout refinement, the quality of both models was assessed and validated using PDB-Redo ( 48 ) and Molprobity ( 49 ), and corrected as required.For StCsm6'-cA 6 , the Molprobity score is 1.35; 99th centile.Ramachandran statistics are 97% fav oured; 2% allo wed; 1% outliers.For apo-StCsm6', the Molprobity Clashscore is 3.14; 84th centile.Ramachandran statistics are 76% favoured; 18% allowed; 6% outliers.Data processing and refinement statistics are shown in Supplementary Table 1.Structural figures and movies were created using CCP4MG ( 50 ) and PyMol (Schr ödinger Inc.).Alignments were performed using the DALI server ( 51 ), and all values calculated are shown in Supplementary Tables 2 and 3.The model coordinates and structure factors have been deposited in the Protein Data Bank with accession codes 8PE3 (StCsm6' in complex with cA 6 ) and 8PCW (apo-StCsm6').
Successful spin labelling was confirmed via electrospray ionization (ESI) mass spectrometry using the in-house mass spectrometry facility.ESI mass spectrometry was performed on samples before (control) and after spin labelling.Samples were diluted to 1 M in 1% formic acid (FA).10 l per sample was injected onto the liquid chromatography (LC) system (Waters Xevo G2 TOF MS with Acquity HPLC) using a MassPrep cartridge column (Waters), a ppl ying a 5 minute gradient from 95% water, 5% acetonitrile to 5% water, 95% acetonitrile (eluents supplemented with 1% FA).Data were collected in positi v e mode from 500-2500 m / z and charged ion series deconvolution to 0.1 Da resolution was performed using the MaxEnt I algorithm utilizing a peak width at half height of 0.4 m / z .Expected masses were obtained for both mutants before and after labelling (see Supplementary Figure S2).Labelling efficiencies were obtained from continuous wave (CW) EPR spectra as described below (Supplementary Figure S2).

Sample pr epar ation f or pulse dipolar EPR spectroscop y and cryogenic CW EPR
Exchange of the spin labelled protein into deuterated buffer was performed by means of repeated steps of dilution followed by centrifugal concentration until a theoretical deutera tion of a t least 99.5% was reached.Samples f or cry ogenic CW and pulse EPR measur ements wer e pr epar ed at a final protein concentration of 50 M (100 M spin label).Both StCsm6' mutants were prepared with and without 25 M cA 6 .50% (v / v) deuterated glycerol (CortecNet) was used for cryoprotection.The samples with a final volume of 65 l were transferred to 3 mm quartz EPR tubes which were immediately frozen in liquid nitrogen.

Pulse dipolar EPR spectroscopy (PDS)
PDS experiments were performed at 50 K on a Bruker ELEXSYS E580 spectrometer with an overcoupled 3 mm cylindrical resona tor (ER 5106QT-2w), opera ting a t Qband frequency (34 GHz), using a second frequency option (E580-400U).Pulses were amplified by a pulse travelling wave tube (TWT) amplifier (Applied Systems Engineering) with nominal output of 150 W. Temperature was controlled using a cryogen-free variable tempera ture cryosta t (Cryogenic Ltd) operating in the 3.5 to 300 K temperature range.
Pulse electr on-electr on doub le resonance (PELDOR) e xperiments were performed with the 4-pulse DEER ( 52-54 ) pulse sequence ( / 2( A ) -1 - ( 55 ), with a frequency offset (pump -detection frequency) of +80 MHz ( ∼3 mT).Shot repetition times (SRT) were set to 2 or 3 ms; 1 was set to 380 ns, and 2 was set to 8000 ns for the samples without addition of cA 6 and to 6500 ns for those with cA 6 .The echo decays as function of available dipolar evolution time were assessed from refocused echo decays by incrementing 2 in the 4 pulse DEER sequence from a start value of 760 ns and omitting the B inversion pulse.Pulse lengths were 16 and 32 ns for / 2 and detection, and 12 or 14 ns for the ELDOR pump pulse.The pump pulse was placed on the resonance frequency of the resonator and applied to the maximum of the nitroxide field-swept spectrum.
PDS experiments were analyzed using DeerAnalysis2015 ( 56 ).PDS data were first background-corrected using a 3D homogeneous background function and ghost suppression (power-scaling) ( 57 ) for a four-spin system, before Tikhonov regularization followed by statistical analysis using the validation tool in DeerAnalysis2015, varying backgr ound start fr om 5% to 80% of the trace length in 16 trials.Resulting background start time for the best fit was then used as starting point for a second round of Tikhonov regularization followed by a second round of statistical analysis, this time including the addition of 50% random noise in 50 trials, resulting in a total of 800 trials.
Traces recorded with a 2 of 8000 ns were cut at 7300 for processing to remove artifacts at the end of the tr ace.Tr aces recorded with a 2 of 6500 ns resulted in a best fit where the background fit had a positi v e and thus unphysical slope after the first validation.Ther efor e, data wer e cut iterati v ely by 10% of the initial trace length and Tikhonov regularization and first validation rounds were repeated until the best fit had a decaying background function.This resulted in a 10% cut for the D88R1 / N209R1 + cA 6 data and a 30% cut for the R85R1 / N209R1 + cA 6 da ta.Valida tion trials from the second validation round were pruned with a prune le v el of 1.15, where trials exceeding the root mean square deviation of the best fit by at least 15% were discarded.In all cases the regularization parameter ␣ was chosen according to the L-curve criterion ( 58 ) and the goodness-of-fit.
For comparison, raw PDS data were subjected to the Comparati v eDEERAnalyzer (CDA) v ersion 2.0 within DeerAnalysis2022 (DEERNet ( 59 ) Spinach SVN Rev 5662 and DeerLab ( 60 ) 0.9.1 Tikhonov regularization) for userindependent data processing and analysis, in line with curr ent r ecommendations ( 61 ).CDA r eports ar e provided as shown in Supplementary Table 4.

Continuous wave (CW) EPR
CW EPR measur ements wer e performed using a Bruker EMX 10 / 12 spectrometer equipped with an ELEXSYS Super Hi-Q resonator at an operating frequency of ∼9.9 GHz (X-band) with 100 kHz modulation.

Room-temper atur e CW EPR
Room-temperature CW EPR measurements were performed to assess labelling efficiency.Samples were recorded using a 100 G field sweep centred at 3455 G, a time constant of 20.48 ms, a conversion time of 20.12 ms, and 1707 points resolution.An a ttenua tion of 20.0 dB (2 mW power) and a modulation amplitude of 0.7 G were used.Protonated StCsm6' samples were measured in 50 l capillaries at 50 M protein (100 M spin) concentration and double integrals wer e compar ed to MTSL as a standard.Labelling efficiency was ≥97% for D88R1 / N209R1 and ≥87% for R85R1 / N209R1, and samples showed negligible free spin label contribution.

Cryogenic CW EPR
Cryogenic CW EPR spectra were obtained at 120 K.The temperature was controlled with an ER4141 VTM Nitrogen VT unit (Bruker) operated with liquid nitrogen and a quartz Dewar insert.Samples wer e r ecorded using a 200 G field sweep centred at 3256 G, a time constant of 20.48 ms, a conversion time of 20.00 ms, and 2000 points resolution.An a ttenua tion of 40.0 dB (20 mW power) and a modulation amplitude of 1 G were used.
To estimate the dipolar broadening by short inter-spin distances in CW EPR, the spectra of cA 6 bound constructs were simulated as dipolarly broadened spectra obtained by convoluting the apo spectrum with the dipolar broadening functions corresponding to a Gaussian distribution centred at 1.1 and 0.9 nm with widths of 0.2 and 0.35 nm for StCsm6' R85R1 / N209R1 and D88R1 / N209R1, respecti v ely.

Modelling for PDS measurements
Distance distributions were modelled based on the crystal structures obtained in this study for StCsm6' in the presence and absence of cA 6 .R1 moieties were introduced at residues 85 or 88 and 209 of both chains of the StCsm6' dimer using mtsslWizard ( 62 ) within the mtsslSuite ( 63 ) server-based modelling software.Cartoon structural representations of spin-labelled StCsm6' constructs were generated using Pymol (Schr ödinger Inc.).

Statistical analyses
For kinetic analyses, all assays were repeated in triplicate.Means and standard deviations were calculated and shown along with original data points.

StCsm6' is a ribonuclease activated by cA 6
The type III-A CRISPR system of S. thermophilus encodes two Csm6 proteins, denoted Csm6 (StCsm6) and Csm6' (StCsm6') (Figure 1 A).The kinetic properties of StCsm6 have been studied extensively ( 25 ), while StCsm6' has only been confirmed as a cOA dependent ribonuclease ( 1 ).As StCsm6' was the subject of this study, we first wished to examine its enzymatic properties.We incubated StCsm6' and radiolabelled RNA with different cOA molecules to confirm that, as expected, StCsm6' is specifically activated by cA 6 (Figure 1 B).Extending this analysis, we utilized a realtime fluorescence assay to determine the sensitivity to the cA 6 acti vator, re v ealing that StCsm6' is activated by as little as 60 pM and fully activ ated b y ∼100 nM cA 6 (Figure 1 C).

Structure of the active form of StCsm6'
We proceeded to solve the structure of StCsm6' bound to its cA 6 activator (Figure 2 A) using X-ray crystallo gra phy with data to 1.96 Å resolution.StCsm6' is symmetrical dimer, with each monomer displaying a CARF domain (residues 1-173), a central 6H domain (residues 174-239) ( 30 ), and a HEPN domain (residues 240-386).In the dimer, each domain interacts with the equivalent domain in the other monomer, but they cross over each other at the 6H domains to gi v e an 'X' arrangement (Supplementary Figure S3; Supplementary Movie 1).The CARF domain comprises fiv e ␣helices and fiv e ␤-strands which alternate to form a central parallel ␤-sheet sandwiched between a three helix bundle and the other two helices.The CARF domain is linked to the HEPN domain via the 6H domain, comprising three ␣helices (ther e ar e a total of six ␣-helices in the dimer, hence 6H).The HEPN domain comprises a total of six ␣-helices; the C-terminal helix packs against the ␣-helices in the 6H domain.
A DALI ( 51 ) search shows the closest structural match to StCsm6' is EiCsm6 in complex with a fluorinated cA 6 mimic (cFA 6 , where each C2' hydr oxyl gr oup is replaced with a fluorine atom) (PDB: 6TUG) ( 24 ), with a root mean square deviation (RMSD) of 6.6 Å over 369 C ␣ atoms for the monomer; the two proteins have a sequence identity of 26%.Like StCsm6', EiCsm6 is a dimer, with each monomer possessing a CARF, a 6H and HEPN domain, and displays a similar arrangement where the monomers cross over at the 6H domain.Although the overall RMSD between the two structures is high (and higher still for the dimer, see Supplementary Table 2), this is largely reflective of a different orientation of the domains in the monomer relati v e to each other.This suggests the possibility of some rotational motion of the CARF domains with respect to the HEPN domains in the dimeric protein.
The monomeric CARF domain of StCsm6' and EiCsm6 in the cA 6 / cFA 6 complexes are largely superimposable (RMSD of 2.1 Å over 162 C ␣ atoms for the monomer) (Figure 2 B).The core ␤-sheet in the CARF domains, as well as the two ␣-helices closest to the 6H domain, which form the majority of the interactions at the dimer interface, show the highest structural similarity.The region with the three ␣helices in the CARF domains of StCsm6', which are linked by long loops, will r equir e significant movement to 'open' in order to allow cA 6 to bind, and to 'close' once cA 6 is bound.The equivalent region in the CARF domains of EiCsm6, which has four ␣-helices, one of which is distorted, has a different arrangement of these elements compared to StCsm6', and thus overall lower structural similarity.
The 6H domain, first predicted bioinformatically ( 30 ), links the CARF and HEPN domains.The most structurally similar protein to the StCsm6' 6H domain is EiCsm6 (RMSD of 1.8 Å over 65 C ␣ atoms for the monomer), despite only 18% sequence identity.The high structural similarity between these domains in the two proteins, over a relati v ely small number of residues comprising just three ␣helices, suggests this domain plays an important functional role rather than being present solely to link the CARF and HEPN domains.
The closest structural homologues to the HEPN domain of StCsm6' are EiCsm6 and Csm6 from Staphylococcus epi-dermidis (SeCsm6; PDB code 5YJC) ( 64 ), with RMSDs of 1.9 and 1.7 Å , both over 141 C ␣ atoms of the monomer, respecti v ely.The ov erall domain structur e between the thr ee proteins is very similar (Figure 2 C; Supplementary Figure S3B).Howe v er, both EiCsm6 and SeCsm6 have an additional ∼40 residues in the HEPN domain compared to StCsm6'.This comprises two short ␤-str ands arr anged in an anti-parallel ␤-sheet, and a 15 residue ␣-helix close to the tip of the HEPN domain.

Molecular recognition and cleavage of cA 6 by StCsm6'
The F obs -F calc map re v ealed electron density consistent with a molecule of cA 6 bound to the StCsm6' dimer, which was fitted into the model (Figure 2 A, D).Upon refinement, it became clear there was actually a mix of species in the binding site, which corresponded to intact cA 6 and the two products of cA 6 cleavage, linear tri-adenylate molecules containing 2 ,3 -cyclic phosphates (A 3 > P), with occupancies estimated as 0.50 (cA 6 ) and 0.50 (both A 3 > P) (Supplementary Figure S3C).All interactions made with StCsm6' were the same for both the intact cA 6 and the hydrolysed products, so will only be described for the former (Figure 2 D; Supplementary Figure S4A).
All interactions between cA 6 and StCsm6' were made e xclusi v ely to residues in the CARF domains, and were symmetrical with respect to each monomer in the dimer.cA 6 formed hydrogen bonds with main chain atoms in D10, T11, R15, D19, A107, Q110, S133, H136, A137 and N139, and with side chain atoms in T11, S39, D73, K112, N139 and R167, as well as a stacking interaction between H77 and one of its adenosine moieties (Supplementary Figure S4A).cA 6 is fully enclosed within the binding site of the CARF domains (Supplementary Figure S4B), suggesting there is significant mobility in the region around this site.This is consistent with higher temperature factors displayed by the residues in the ␣-helix and loops that enclose cA 6 (Supplementary Figure S4C), w hich notabl y is the region least structurally conserved with the CARF domains of EiCsm6.
The presence of both intact cA 6 and the cleaved A 3 > P products provides unique insights into catalysis by residues in the CARF domains.Mechanistic studies on cyclic oligoadenylate cleavage propose that an activated 2 -OH group on a ribose is responsible for in-line nucleophilic attack on the adjacent scissile phosphodiester bond ( 21 , 24 , 28 ).Gi v en the symmetrical nature of the StCsm6' interactions with cA 6 , cleavage takes place at the identical (but opposite) sides of the cA 6 ring to produce two A 3 > P molecules.The observed position of cleavage in cA 6 is consistent with the notion that the angle formed between the 2 -OH, P and O should be close to 180 • .The mean of this angle at the two positions in cA 6 where cleavage occurs is 172 • , compared to 100 • and 159 • at the other equivalent positions in the cA 6 ring (Supplementary Figure S5A).Interestingly, the 2 -OH group of ribose and the phosphate involved in cleavage form minimal interactions with StCsm6'; the 2 -OH forms a hydrogen bond with the backbone amide of N10, and an oxygen atom in the phosphate with T11 (Supplementary Figure S5B).
The binding environment for cA 6 in StCsm6' is structurally similar to that of EiCsm6 bound to fluorinated cA 6 (Supplementary Figure S6).cA 6 in StCsm6' superimposes with cFA 6 in EiCsm6 with an RMSD of 1.3 Å over 126 atoms, and visually is conformationally identical (Figure 2 D).The conservation of interactions made between the ligands and StCsm6' or EiCsm6 vary a t dif ferent positions around the ring.The residues (N10 and T11 in both structures) and hydrogen bond interactions with the 2 -OH of the ribose and phosphate involved in cleavage are absolutely conserved, as are the interactions of the adenine of the same adenylate moiety with main chain atoms of R15 and N19 and side chain of R167.This suggests the arrangement of these residues is critical to getting the cA 6 into a conforma tion commensura te with ca talysis.The other two pairs of adenylate moieties display very few conserved interactions, with the exception of S39, suggesting a greater plasticity in these regions of the binding site.It is worth not-ing that overall EiCsm6 makes fewer interactions with cFA 6 than StCsm6' with cA 6 .

Characterization of RNA and cA 6 cleavage by StCsm6'
The mechanism of RNA degradation by StCsm6' was investigated by site-directed mutagenesis of the arginine and histidine residues within the R-X 4 -H HEPN catalytic motif.We replaced R331 and H336 with glutamate and alanine, respecti v ely (Supplementary Figure S7).Each of these mutations abolished RNA cleavage, as expected (Figure 3 A).To verify that cA 6 binding to the CARF domain was essential for activation of RNA cleavage at the HEPN acti v e site, we generated a S105W variant to disrupt cA 6 binding at the CARF domain (Supplementary Figure S7).The S105W muta tion abolished cA 6 -activa ted RNA cleavage, demonstra ting tha t replacement of S105 by a bulky tryptophan residue most likely pre v ents cA 6 binding and subsequent activation of RNA cleavage by StCsm6'.
Gi v en the observation of a mixed population of cleaved and intact cA 6 in the crystals, we carried out a two-stage ribonuclease deactivation assay ( 28 ) to investigate this biochemically.By pre-incubating StCsm6' with different concentrations of cA 6 before adding radiolabelled A1 RNA and fresh wild type (WT) StCsm6', we could evaluate the extent to which StCsm6' and variants degraded cA 6 at stage 1, and thus protected RNA from enzymatic degradation in the second stage of the assay (Figure 3 B).We incubated the same range of cA 6 concentrations with either reaction buffer (control), WT StCsm6', S105W StCsm6' (CARF variant) or H336A StCsm6' (HEPN variant).The control reaction demonstrated that cA 6 remaining in the reaction mixture after stage 1 could activate RNA degradation by StCsm6' in stage 2. In contrast, when cA 6 was pre-incubated with WT StCsm6', RNA was pr otected fr om degrada tion a t stage 2, across the different concentrations of cA 6 tested.When cA 6 was preincubated with the H336A variant at stage 1, RNA degradation was greatly reduced at stage 2, suggesting that this variant can still degrade cA 6 similarly to the WT protein.In contrast, the S105W variant degraded only very small amounts of cA 6 at stage 1.Thus, the CARF domain seems to be the major site for cA 6 degradation in StCsm6' under the conditions tested, which had low (nM) cA 6 and an excess of enzyme.These data fit well with previous findings for StCsm6 where cA 6 degradation was measur ed dir ectly ( 25 ).In that study, increasing the cA 6 concentration to M le v els led to a much greater contribution by the HEPN domain to cA 6 degr adation.Over all, it appears for both enzymes that at low cA 6 concentrations the activator is degraded by ring nuclease activity in the high affinity CARF domain binding site.This is consistent with our observation of cleaved cA 6 in the crystal structure.

Structure of apo StCsm6' reveals dramatic conformational changes on cA 6 binding
Apo StCsm6' proved to be extremely difficult to crystallize, and only a single crystal in a dehydrated drop of mother liquor following ∼1 year of incubation was obtained.Consistent with its incalcitrant nature towards crystallization, the resolution of diffraction was low.Ne v ertheless, data on this crystal were collected to 3.54 Å resolution, and the structure solved by molecular replacement with the separate HEPN and CARF domains of the StCsm6' complex structure used as the search models.Aside from a couple of disorder ed r egions, the peptide backbone of apo StCsm6' could be built with confidence, but far fewer side chains were modelled due to the low resolution of the data.The structure revealed that apo StCsm6', like in the complex with cA 6 , is a symmetrical dimer in an 'X' arrangement.Strikingly, howe v er, the ov erall structure for apo StCsm6' and in comple x with cA 6 were markedly different, with an RMSD of 7.7 Å over 327 C ␣ atoms for the monomer and 4.9 Å over 441 C ␣ atoms for the dimer (note, howe v er, ther e ar e 386 r esidues in the monomer and 772 residues in the dimer, meaning the RMSDs over the full length dimeric protein would be much higher) (Figure 4 A).This was evident from the movement of some secondary structure elements in all domains, but in particular the relati v e orientation of the domains with respect to each other (Supplementary Movie 1).
The monomeric form of the individual CARF, 6H and HEPN domains of StCsm6' are very similar in the apo structure and in complex with cA 6 , with RMSD values of 1.4 Å or lower over the entirety of each domain (Supplementary Table 3).The only difference to note is that one of the loops in the CARF domain (residues 133-149), which encloses the acti v e site in the complex with cA 6 , could not be modelled in the apo structur e, pr esumably due to conformational flexibility in this region which may not be stabilized until cA 6 is bound.
Howe v er, although the intrinsic secondary structure of individual monomeric domains did not change significantly upon binding of cA 6 , the RMSD values for superimposition of the dimers of each domain were considerably higher (Supplementary Table 3; Supplementary Movie 1).Visual inspection of the superimposition of the apo StCsm6' and cA 6 complex structures highlights there was considerable movement in all domains.The most apparent difference in the CARF domains involves an ␣-helix adjacent to the loops that encloses the cA 6 binding site, which is outside of the canonical CARF domain core (Figure 4 A, B).This ␣helix moves up and inwards when in complex with cA 6 .In doing so, ther e ar e also some smaller movements in other secondary structure elements in the CARF domains, as a result of the 'tightening up' movement towards the dimer interface to both interact with, and enclose, cA 6 .The 6H domain comprises six ␣-helices (three in each monomer), which are positioned at a similar angle in the two structures, but those in complex with cA 6 are pulled inwards towards the dimer interface (Figure 4 A,C).Whilst upon binding of cA 6 the CARF and 6H domains appear to tighten with mov ement towar ds the dimer interface to enclose cA 6 , at the opposite side of the protein, in the HEPN domain, a number of the secondary structure elements swing outwards to open up the cavity.This is accompanied by a significant rotation of the CARF domains relati v e to the HEPN domains, of around 58 • (Figure 4 A, C; Supplementary Figure S8), making the overall effect akin to opening a pair of jaws upon binding of cA 6 (Supplementary Movie 1, Video Abstract).
The dramatic movement of StCsm6' upon binding cA 6 leads to significant changes at the HEPN dimer interface w here the catal ytic cleavage of RN A takes place.The displacement of the secondary structure elements in the HEPN domain causes an increase in distance of ∼6.5 Å at the opening to the acti v e site (Figure 4 C).In turn, this movement brings the catalytic residues closer together in order to form a functional acti v e site.For e xample, the acti v e site R331 residues of each monomer move closer together by ∼2 Å (Figure 4 D; Supplementary Movie 1).Unfortunately, a number of the acti v e site residues could not be modelled in the apo StCsm6' structure; this could reflect the flexibility in this region prior to cA 6 binding which forms the functional acti v e site, but also could be a consequence of the lower resolution data.
Unsurprisingly, DALI searches showed that EiCsm6' was the closest homologue to apo StCsm6', both over the monomeric full length protein and individual domains (although SeCsm6' was also a good hit for the HEPN domain) (Supplementary Table 2).As seen with the structure of StCsm6' in complex with cA 6 , searches with the dimeric units of StCsm6' generally showed poor alignment with high RMSDs over a limited number of r esidues.Inter estingly, howe v er, the apo StCsm6' monomer showed a lower RMSD with EiCsm6 in complex with cFA 6 compared to StCsm6' in complex with cA 6 against EiCsm6 in complex with cFA 6 .The RMSDs for the individual domains for both StCsm6' structur es compar ed to EiCsm6 in complex with cFA 6 were similar, suggesting that the key difference is the orientation of the domains in relation to each other in each of the full length proteins (Supplementary Figure S9).

Conformational change probed by pulse EPR
To further validate the large-scale conformational transition between the apo and cA 6 bound states of StCsm6' observed in the crystal structures, pulse dipolar electron paramagnetic r esonance spectroscop y (PDS) was employed ( 61 ).PDS gi v es access to ensemble distance distributions between paramagnetic spin labels in frozen solution and has been successfully utilized to probe conformational flexibility and ligand-induced structural changes in soluble and membr ane proteins.A str aightforward appr oach intr oduces two cysteines at the desired sites via site directed mutagenesis and post-translationally modifies them by site-specific spin-labelling with thiol specific nitroxide reagents yielding labelled side-chains (i.e. the spin-bearing residue R1 in the present case).As StCsm6' is a homo-dimer, introduction of a single cysteine would yield a protein complex containing two spin labels.Howe v er, modelling re v ealed that in the dimer only minimal distance changes between any two identical amino acid residues in the two monomers are expected upon cA 6 binding.Thus, double cysteine mutants were designed for which significant distance changes could be predicted (Figure 5 A).While these result in multi-spin systems that are known to complicate analysis, this can be dealt with entirely on a post-processing le v el for up to four spins under our experimental conditions ( 65 ).
A double cysteine mutant of StCsm6' at positions 85 and 209 was produced and spin-labelled with good yields ( > 85%) (Supplementary Figure S2).For R85R1 / N209R1 in the apo structure of StCsm6', a multimodal distance distribution was predicted with the shortest population cen- A key ␣-helix in the 6H domain (burgundy) is important in facilitating the movement between the CARF and HEPN domains.The cA 6 molecule bound to StCsm6' is shown in stick r epr esenta tion; carbon a toms are shown as orange or pink for StCsm6' and EiCsm6, respecti v ely, phosphate as dark orange, nitrogen as blue and oxygen as red.( B ) Superposition of the CARF domains of apo StCsm6' (grey) and in complex with cA 6 (orange).The most significant difference is the movement towards the dimer interface of two ␣-helices which enclose cA 6 , as illustrated by the arrows.( C ) Superposition of the HEPN domains of apo StCsm6' (grey) and in complex with cA 6 (blue).Like the CARF domains, the 6H domains are pulled inwards towards the dimer interface, illustrated by the top arro ws.Ho we v er, the opposite is true for the HEPN domains which swing outwards like the opening of a jaw, illustrated by the bottom arrows.As an example, the distance between F303 (last residue of the ␣-helix indicated) in each monomer is shown with a dashed line in the colour corresponding to each structure; this shows a movement of 6.5 Å outwards for StCsm6' in complex with cA 6 relati v e to the apo enzyme.( D ) Superposition of the HEPN domains of apo StCsm6' (grey) and in complex with cA 6 (blue), with R331 (first residue in the R-X 4 -H catalytic motif) in each monomer shown in sticks of the same colour.The dashed line, in the colour corresponding to each structure, shows the distance between R331 in each monomer; this highlights a movement of 2.0 Å to bring the arginine residues closer together in the complex with cA 6 , to form a functional active site, relative to the apo enzyme.The other active site residues could not be modelled in the apo structure.tred around 3.0 nm and further intensity from 4.0 to 5.5 nm (Figure 5 B, C; Supplementary Figure S10).While this latter population remains unchanged upon addition of cA 6 , the shorter distance is predicted to significantly reduce to 1.0-2.0nm, which is well into the lower limit of PDS.Frozen solution CW EPR displayed dipolar broadening and the PDS echo decay was significantly accelerated, both consistent with the presence of a short distance (generally the lower limit of our PDS methodology lies at 1.5-1.8nm).Despite the spin label distances in both states exceeding the models and hinting at a more expanded state in solution, the increase of short distance distributions was very clear in the PDS data.Results on an additional double cysteine mutant of StCsm6', D88R1 / N209R1, confirm this finding (Supplementary Figure S11).These data thus support the observation of the cA 6 -mediated conformational change observed in the crystallo gra phic stud y, indica ting the structures obtained in crystallo are consistent with those in solution.

Mutagenesis of key interface residue D80 blocks activation of StCsm6'
To test the hypothesis that the conformational change observed upon cA 6 binding to StCsm6' is r equir ed for activation, we mutated residue D80 to a bulkier tyrosine residue.In the apo StCsm6' structure, D80 sits in a solvent exposed position in the CARF domain (Figure 6 A; Supplementary Figure S7), but upon binding of cA 6 the rotation of the CARF domain brings D80 into close contact with the residues at the top of the 6H domain (residues 172-175) (Figure 6 B).We hypothesized there would be insufficient room to accommodate a bulkier tyrosine side chain in place of the aspartate, and thus would block the conformational change r equir ed for activ ation.The D80Y v ariant was expressed and purified as described for the wild type protein, but was almost completely inacti v e in a cA 6 dependent RNase assay (Figure 3 A).This is consistent with a model whereby cA 6 -media ted conforma tional changes, transmitted from the CARF domains to the HEPN domains, occurs through precise movement of the 6H domains, confirming their role in activation of the enzyme.

DISCUSSION
Csm6 family proteins have attr acted consider able interest as an exemplar of acti vatab le, non-specific nucleases linked to type III CRISPR systems.They display minimal activity in the absence of the cOA effector ( 29 ), but once activated function as efficient ribonucleases that degrade both viral and host RNA ( 5 , 9 , 66 ).These properties have been utilized to de v elop a range of sensiti v e ne w diagnostic assays ( 19 , 67 ).The key question of how cOA binding at the CARF domain results in activation of the ribonuclease acti v e site in the HEPN domain has remained unresolved, with only modest changes detected by crystallo gra phy w hen the two states are accessible to crystallization.Here, we have focussed on the cA 6 activated family of Csm6 enzymes, which are broadly distributed in bacteria ( 1 , 2 , 8 ).These comprise a distinct sub-family of the structurally di v erse Csm6 / Csx1 famil y, w hose members are most commonly activated by cA 4 .We succeeded in crystallizing apo-and cA 6 bound S. thermophilus Csm6', allowing a detailed comparison of the inacti v e and activa ted sta tes.This re v ealed a remar kab le structur al tr ansition, where binding of cA 6 to the CARF domains results in a tightening of the secondary structure elements concomitant with a 58 • rotation of the CARF domains relati v e to the HEPN domains.This mov ement, transmitted through the 6H domain, allows the 'jaws' of the HEPN domains to open and causes a rearrangement of the catalytic residues to form a functional acti v e site prior to catalysis.To check the structural rearrangements observed were not artefacts of crystallization, EPR spectroscopy studies were conducted on StCsm6' in the absence and presence of cA 6 in solution, which showed changes in distances between key atoms were consistent with the X-ray crystallo gra phic structures.
The mechanics of opening the jaws of the HEPN domains upon cA 6 binding suggests this plays an important regulatory role to control ssRNA binding and nuclease activity.Although there is no structure of RNA bound to the HEPN domain of a member of the Csm6 / Csx1 family for direct comparison, it is plausible that there is insufficient space for ssRNA to bind in the 'closed' form of the HEPN domains.Upon binding of cA 6 to the CARF domains, the structural transition opens the HEPN domains providing ssRNA easier access to the acti v e site, as well as rearranging residues to form a functional acti v e site poised for catalysis.The relati v e importance of these two factors for Csm6 activation is unclear, and it remains to be determined whether the HEPN dimer binds and cleaves a single molecule of ssRNA in a composite acti v e site.
The observation of both cA 6 and the cleaved A 3 > P products in the acti v e site of the CARF domains of StCsm6' provides unique insights into catalysis.The cleavage presumably occurred during the co-crystallization of StCsm6' with cA 6 , but the restrictions of crystal formation meant catalysis was slow, thus allowing both substrate and product to be observed.It is proposed that an activated 2 -OH group on a ribose in a cyclic oligoadenylate molecule undergoes in-line nucleophilic attack on the adjacent scissile phosphodiester bond, and in order for this to happen the angle formed between the 2 -OH, P and O should be close to 180 • ( 21 , 24 , 28 ), in this case 172 • .Like EiCsm6 ( 24 ), the 2 -OH group of ribose and the phosphate involved in cleavage form only two interactions with StCsm6', and neither of them are sufficiently reacti v e to play a role in catalysis.This is highly suggesti v e that StCsm6' does not assist cleav age b y deprotonation of the nucleophile or protonation of the leav-

Figure 1 .
Figure 1.StCsm6' is a cyclic hexa-adenylate activated ribonuclease.( A ) Type III CRISPR locus of Streptococcus thermophilus .Two cOA activated CARFfamily ribonucleases, denoted Csm6 and Csm6', are found adjacent to cas genes encoding the type III CRISPR complex.( B ) Denaturing PAGE visualizing cleavage of a radiolabelled RNA oligonucleotide A1 by StCsm6' (1 M dimer) with cA 3 , cA 4 or cA 6 activator (20 M).StCsm6' is specifically activated by cA 6 .The image is r epr esentati v e of three technical replicates.Control c1 is RNA alone and c2 is RNA incubated with StCsm6' in the absence of activator.( C ) Fluorogenic RNase activity assay; cleavage of a FAM ™ reporter substrate (RNaseAlert, IDT) by StCsm6' (1 M dimer) across a range of cA 6 concentrations.Columns depict the mean of three technical replicates (individual data points shown) and error bars show the standard error of the mean.

Figure 2 .
Figure 2. Comparison of StCsm6' in complex with cA 6 and EiCsm6 in complex with cFA 6 structures.( A ) Cartoon representation of the secondary structure elements of StCsm6' (left) and EiCsm6 (right); the structures are orientated to gi v e the same view of their HEPN domains.The CARF domains are shown in orange and pink, the 6H and HEPN domains in blue and yellow, and a key ␣-helix in the 6H domain in burgundy and white for StCsm6' and EiCsm6, respecti v ely.The cA 6 molecule bound to StCsm6' and the cFA 6 molecule bound to EiCsm6 are shown in stick r epr esenta tion; carbon a toms are shown as orange or pink for StCsm6' and EiCsm6, respecti v ely, phosphate as dar k orange, fluorine as cyan, nitrogen as blue and oxygen as red.( B ) Structural superimposition of the CARF domains from StCsm6' and EiCsm6; colours as described in (A).( C ) Structural superimposition of the 6H and HEPN domains; colours as described in (A).( D ) Superimposition of the cA 6 molecule bound to StCsm6' and cFA 6 molecule bound to EiCsm6; colours as described in (A).

Figure 3 .
Figure 3. StCsm6' is a self-limiting ribonuclease which cleaves cA 6 at the CARF domains.( A ) Denaturing PAGE gel visualizing RNA A1 cleav age b y WT StCsm6' and variants (all at 1 M dimer) in the absence and presence (+) of 20 M cA 6 .Control 'c' refers to RN A onl y; 'wt' refers to WT StCsm6'; H336A and R331E are HEPN domain variants; S105W is a CARF domain variant designed to block cA 6 binding; and D80Y is discussed later.All are catal yticall y inacti v e.The image is r epr esentati v e of three technical replica tes.( B ) Deactiva tion assays where a range of cA 6 concentrations (10, 20, 50, 100, 200 nM) was incubated with either buffer, WT StCsm6' (wt), S105W variant or H336A variant (all at 1 M dimer) at stage 1, prior to addition of radiolabelled RNA A1 and WT StCsm6' at stage 2, analysed by denaturing PAGE and phosphorimaging.Control 'c1' refers to RN A onl y, not subjected to pre-incubation or enzyme; 'c2' refers to no cA 6 but StCsm6' and RNA added after buf fer incuba tion.The images ar e r epr esentati v e of three technical replica tes.( C ) Quantifica tion of the densiometric signals from B to determine the le v el of RNA protected from WT StCsm6' activity upon incubating with either buffer, WT StCsm6', S105W or H336A variants.Bars depict the average from three technical replicates (individual data points shown) and error bars are the standard deviation of the mean.

Figure 4 .
Figure 4. Comparison of the structure of apo StCsm6' and in complex with cA 6 .( A ) Cartoon r epr esentation of the secondary structure elements of apo StCsm6' (left) and in complex with cA 6 (right); the structur es ar e orientated to gi v e the same view of their HEPN domains.The side-by-side view highlights the movement of the CARF domains (orange) relative to the 6H and HEPN domains (blue).A key ␣-helix in the 6H domain (burgundy) is important in facilitating the movement between the CARF and HEPN domains.The cA 6 molecule bound to StCsm6' is shown in stick r epr esenta tion; carbon a toms are shown as orange or pink for StCsm6' and EiCsm6, respecti v ely, phosphate as dark orange, nitrogen as blue and oxygen as red.( B ) Superposition of the CARF domains of apo StCsm6' (grey) and in complex with cA 6 (orange).The most significant difference is the movement towards the dimer interface of two ␣-helices which enclose cA 6 , as illustrated by the arrows.( C ) Superposition of the HEPN domains of apo StCsm6' (grey) and in complex with cA 6 (blue).Like the CARF domains, the 6H domains are pulled inwards towards the dimer interface, illustrated by the top arro ws.Ho we v er, the opposite is true for the HEPN domains which swing outwards like the opening of a jaw, illustrated by the bottom arrows.As an example, the distance between F303 (last residue of the ␣-helix indicated) in each monomer is shown with a dashed line in the colour corresponding to each structure; this shows a movement of 6.5 Å outwards for StCsm6' in complex with cA 6 relati v e to the apo enzyme.( D ) Superposition of the HEPN domains of apo StCsm6' (grey) and in complex with cA 6 (blue), with R331 (first residue in the R-X 4 -H catalytic motif) in each monomer shown in sticks of the same colour.The dashed line, in the colour corresponding to each structure, shows the distance between R331 in each monomer; this highlights a movement of 2.0 Å to bring the arginine residues closer together in the complex with cA 6 , to form a functional active site, relative to the apo enzyme.The other active site residues could not be modelled in the apo structure.

Figure 5 .
Figure 5. EPR data for StCsm6' variant R85R1 / N209R1 in the presence and absence of cA 6 .( A ) Cartoon r epr esentation of the secondary structure elements of apo StCsm6' (left) and in complex with cA 6 (right), with colouring as described in Figure 4 , showing the position and predicted MTSL rotamers for StCsm6' D85R1 / N209R1.( B ) Background-corrected traces without (black) or with (red) addition of cA 6 with fits (grey).( C ) Distance distributions (shown as 95% confidence bands without (black) or with (red) addition of cA 6 ) and modelled distributions based on the corresponding crystal structures (grey).Colour bars indicate reliability ranges (green: shape reliable; yellow: mean and width reliable; orange: mean reliable).Peak * corresponds to the distance between D85R1 and N209R1 in different monomers of the StCsm6' dimer, which changes in the absence or presence of cA 6 .Peak ** corresponds to the overlapping distances between D85R1 in each monomer, N209R1 in each monomer, and D85R1 and N209R1 in the same monomer of StCsm6', which are the same in the absence or presence of cA 6 .

Figure 6 .
Figure 6.The role played by D80 in StCsm6' upon cA 6 binding.Cartoon r epr esentation of ( A ) apo StCsm6' and ( B ) StCsm6' in complex with cA 6 , showing key interactions between the CARF domain (grey) and an ␣-helix in the 6H domain (burgundy) is important for the conformational change brought about upon binding of cA 6 .In particular, D80 (shown as spheres, with carbon atoms in yellow and oxygen in red) in the CARF domain moves significantly upon binding of cA 6 to come into close proximity with residues 172-175 (modelled in surface representation) at the top of the key ␣-helix in the 6H domain.