ABSTRACT
Filament or run-on oligomer formation by enzymes is increasingly recognized as an important phenomenon with potentially unique regulatory properties and biological roles. SgrAI is an allosterically regulated type II restriction endonuclease that forms run-on oligomeric (ROO) filaments with enhanced DNA cleavage activity and altered sequence specificity. Here, we present the 3.5 Å cryo-electron microscopy structure of the ROO filament of SgrAI bound to a mimic of cleaved primary site DNA and Mg2+. Large conformational changes stabilize a second metal ion cofactor binding site within the catalytic pocket and facilitate assembling a higher-order enzyme form that is competent for rapid DNA cleavage. The structural changes illuminate the mechanistic origin of hyper-accelerated DNA cleavage activity within the filamentous SgrAI form. An analysis of the protein-DNA interface and the stacking of individual nucleotides reveals how indirect DNA readout within filamentous SgrAI enables recognition of substantially more nucleotide sequences than its low-activity form, thereby expanding DNA sequence specificity. Together, substrate DNA binding, indirect readout, and filamentation simultaneously enhance SgrAI’s catalytic activity and modulate substrate preference. This unusual enzyme mechanism may have evolved to perform the specialized functions of bacterial innate immunity in rapid defense against invading phage DNA without causing damage to the host DNA.
INTRODUCTION
Filament formation by non-cytoskeletal enzymes is a newly appreciated phenomenon. Although first shown in vitro for the metabolic enzymes acetyl-CoA carboxylase and phosphofructokinase in the 1970s1–3, it was not until 40 years later that filament formation was demonstrated to modulate enzyme activity under physiological conditions4,5. At around the same time, filament formation was surprisingly discovered for the unfolded protein response RNase/kinase Ire16 and for numerous other enzymes previously unknown to form such assemblies7–9. Independently, filament formation was proposed to explain the unusual biophysical and biochemical activities seen in the type II restriction endonuclease SgrAI10.
SgrAI is a sequence-specific type II DNA restriction endonuclease that is allosterically activated by the same substrate DNA. A primary recognition sequence (CR|CCGGYG, where R=A or G and Y=C or T, | denotes cleavage site) activates the enzyme by over 200-fold, which in turn also relaxes the enzyme’s DNA sequence specificity, resulting in the cleavage of an additional 14 secondary sequences (CCCCGGYG or XRCCGGY G, where X=A,G, or T) up to 1000-fold faster10–14. In the absence of DNA, SgrAI is a homodimer composed of two 37 kDa chains, each with a single active site15–18. The enzyme binds primary or secondary site DNA in 1:1 ratios, and the ensuing complex is referred to as a DNA-bound dimer (DBD). Since the primary site DNA acts both as an allosteric effector as well as a substrate for enzymatic cleavage, it is expected that activated SgrAI possesses at least two DNA binding sites, one for enzymatic cleavage and one for the allosteric effector. Early studies suggested the formation of an activated SgrAI/DNA complex composed of at least two DBDs16,17, thereby providing a functional complex with two DNA binding sites. However, assemblies containing many more DBDs were shown by analytical ultracentrifugation and ionmobility mass spectrometry10,19. Negative stain EM revealed filaments of varied lengths with left-handed helical symmetry that we call run-on oligomers (ROO)20.
The ROO filament structure was previously resolved to ~nanometer resolution by cryo-EM and helical reconstruction, that revealed left-handed helical symmetry with approximately 4 DBDs per turn20. This structure inspired a low-resolution mechanistic model for the enzymatic behavior of SgrAI, wherein binding to primary site DNA induces a conformational change that favors ROO filament formation, which in turn stabilizes the activated enzyme state capable of accelerated DNA cleavage (Fig. 1A). SgrAI DBD bound to a secondary site DNA disfavors the activated conformation, and hence ROO filament formation by SgrAI bound to secondary site does not appreciably occur (Fig. 1B). This explains why cleavage of secondary site sequences is negligible unless a primary site is present – in sufficient concentration or on the same contiguous DNA – to promote ROO filament assembly10,12,14. Filaments formed by SgrAI bound to primary site DNA will drive ROO assembly, incorporating DBD-containing secondary sites, activating SgrAI for DNA cleavage on both primary and secondary sites, and thereby expanding the enzyme’s DNA sequence specificity (Fig. 1C).
A full kinetic analysis of SgrAI-mediated cleavage of primary site DNA enabled the estimation of individual rate constants for the major steps in the reaction pathway, such as DBD association into the ROO filament, cleavage of DNA within the filament, DBD dissociation from the ROO filament, and dissociation of cleaved DNA from SgrAI21,22. These studies show that assembly of DBDs into the ROO filament is rate limiting when recognition sites are on separate DNA molecules, but fast when on the same contiguous DNA. Kinetic simulations show that this rate limiting step underlies the enzyme’s ability to sequester potentially damaging secondary site cleavage events onto invading phage DNA and away from the host, leaving host sites largely untouched23. The study also shows that DNA cleavage is fast, followed by a relatively slower disassembly of DBDs from the ROO filament. Although slower than DNA cleavage, the rate of disassembly is still sufficiently fast to prevent trapping of cleaved DNA within the filament. Furthermore, simulations also show that the very nature of filament formation imparts beneficial properties to the enzyme, both in terms of faster enzyme kinetics and its ability to sequester DNA cleavage onto phage and away from host, providing protection against infection and host DNA damage, respectively23. Importantly, mutations disrupting the ROO filament (even moderately) abolish all protection provided by SgrAI against phage infection, indicating that the rapid speed provided by ROO filament formation is critical to anti-phage activity23.
The previous low-resolution cryo-EM structure shows the general architecture of the ROO filament, but left open two important questions: 1) what is the mechanism for activated DNA cleavage? and 2) how do primary vs. secondary site DNA sequences differentially stabilize the activated conformation of SgrAI and the ROO filament? To address these questions and clarify the molecular mechanisms of enzyme activation and allosteric regulation, we determined the structure of activated, filamentous SgrAI to near-atomic resolution. The 3.5 Å cryo-EM maps allowed detailed model building and refinement, and enabled us to compare the ROO filament structure with previously determined x-ray crystal structures of the inactive, non-filamentous forms. The data revealed how numerous conformational changes in protein and DNA stabilize the SgrAI ROO, activate DNA cleavage, and how a single base pair change in secondary sites may modulate enzyme activation.
RESULTS AND DISCUSSION
Cryo-EM helical analysis of SgrAI ROO filaments resolves the activated enzyme form
To elucidate the molecular determinants of SgrAI activation, we sought to resolve the ROO filament to near-atomic resolution. We prepared ROO filaments from purified, recombinant, his-tagged wild type SgrAI14 bound to a 40-bp oligonucleotide DNA containing a pre-cleaved primary site sequence (PC DNA, see Methods). Cryo-EM micrographs revealed a large variety of differently-sized filaments, as expected for a ROO (Fig. S1A). Because the filaments were typically limited to ~<10 DBDs in size, thereby making it challenging to precisely define their orientations on the cryo-EM micrographs through conventional manual filament tracing, we sought to process the data in a single-particle manner. Template-based particle detection using a small distance between neighboring picks allowed us to select virtually all filamentous regions on the micrographs. 2D and 3D classification – without imposing helical symmetry – allowed us to separate the particle picks into compositionally distinct classes (Fig. S1B-C). Subsequent imposition of helical symmetry facilitated obtaining a high-resolution cryo-EM electrostatic potential map, resolved globally to 3.5 Å, which in turn enabled deriving an atomic model (Fig. 2A, Fig. S1D-H). The model is consistent with the final cryo-EM map, and has good geometry and statistics (Fig. S1F, Fig. S2, and Table S1).
Each DBD constitutes the basic building block of activated and filamentous SgrAI. The enzyme can oligomerize in a run-on manner, from either side along the helical axis (Fig. 2B). In previous experiments, this property has been shown to specifically benefit the enzyme’s ability to quickly sequester phage DNA within the oligomeric form for rapid cleavage24. Within the filament, SgrAI makes several inter-DBD contacts between asymmetric units. The loop spanning residues 126-134 interfaces with flanking DNA. Electrostatic interactions from charged amino acids within this patch of the protein help maintain the oligomeric enzyme form (Fig. 2C). This patch also makes up a portion of the allosteric interface of the enzyme. Immediately alongside, loop 55-62 also packs against the flanking DNA, although this interface is primarily maintained by weaker van der Waals interactions with the DNA backbone. Several regions of the NTD make protein-protein interactions between neighboring DBDs within the central filament axis, including clusters of salt bridges that help to facilitate helical packing. Arg84 from DBDn makes a salt bridge with Asp85 from DBDn+1 (Fig. 2D). Several other charged residues, including Arg11 and Glu8, reside immediately nearby, although their exact contributions cannot be accurately determined at the current resolution. These charged regions from multiple neighboring DBDs, both immediately alongside and at the opposing side of the helix, generally contribute to interface packing along the central helical axis. These data suggest that the interactions made within the central filament axis are specific, and are designed to facilitate oligomerization. The structural organization, combined with previous biochemical and functional data, implies that filament formation is an inherent property of the enzyme.
The activated DNA-bound SgrAI enzyme exhibits multiple conformational rearrangements that stabilize the filamentous ROO form
To determine the molecular changes underlying enzyme activation, we compared the structure of the high activity ROO filament to a previously published X-ray structure of a low-activity, non-filamentous SgrAI DBD (PDB code 3DVO18). Superposition of the two DBDs reveals a ~9° rotation of one chain relative to the other (Fig. 3A and Movie S1), with shifts also occurring in the dimeric interface to accommodate this rotation (Fig. 3B). Such conformational changes allow for favorable interactions to occur within the ROO filament. Specifically, conformational changes place residues 84-87 and 22-34 into position for filament formation (Fig. 3C). Without such structural changes, the subunits would clash sterically at the interface marked by a red “X” in Fig. 3C, and thus these residues must facilitate proper helical packing in the filament. Changes in conformation also occur in the bound DNA, and result in favorable interactions between the base pairs flanking the recognition sequence and SgrAI residues (56-57 and 127-134) of a neighboring DBD in the ROO filament. Figure 3D show these interactions, comparing the ROO filament structure (in magenta, neighboring SgrAI in green), the non-filamentous low-activity DBD (gray), and an idealized B-form DNA (tan). The ROO DNA is bent towards neighboring SgrAI in the ROO filament relative to the idealized B-form DNA, but takes on a different path compared to the DNA in the non-filamentous enzyme assembly. Furthermore, DNA bending allows for interactions with residues 56-57 and 127-134 of the neighboring SgrAI, without steric conflicts. These residues have been shown to be important for enzyme activation, presumably by stabilizing the ROO filament14. Overall, the structural rearrangements observed within the filamentous SgrAI ROO (Movie S1) prevent unfavorable clashes and promote interactions between neighboring DBDs.
Mechanism of SgrAI hyper-activation for DNA cleavage through stabilization of a second Mg2+ binding site
Many nucleases, as well as other phosphoryl transfer enzymes, are dependent on divalent metal ion cofactors such as Mg2+ for their activity. SgrAI is no exception, and cleaves nucleic acids in the presence of Mg2+, and will also utilize Mn2+ or Co2+ for catalysis18,25. The now classical two-metal ion mechanism was first proposed for divalent ion-dependent DNA hydrolytic cleavage based on structures of a 3’-5’ DNA exonuclease26, and has since been used as a mechanistic model for many metal ion-dependent phosphoryl transfer enzymes27–29. Based on the activated cryo-EM reconstruction of filamentous SgrAI, the two-metal ion mechanism, adapted for the enzyme, is shown schematically in Figure 4A with experimental structural data supporting the mechanism in Figure 4B. In describing both panels, we have utilized the “tense” T and “relaxed” R states – terms adopted based on early enzymology work with hemoglobin30 – to describe the low- and high-activity forms of the enzyme, respectively.
In the first panel, SgrAI is in the low activity T state, and only metal ion Site A is occupied. This metal ion ligates both a nonesterified oxygen of the scissile phosphate (SP, Fig. 4A, panel 1), as well as a water molecule, inducing its deprotonation to hydroxide (blue). The previous X-ray crystal structure of SgrAI bound to uncleaved primary site DNA and Ca2+ (PDB file 3DVO18), shows this state (Fig. 4B, panel 1). The next panel of Figure 4A, panel 2, shows the shift in conformation to the activated R state, which brings the backbone carbonyl of Thr 186 closer to the metal ion binding Site B. The current cryo-EM ROO filament structure of SgrAI bound to Mg2+ and primary site DNA (but missing the SP) shows this state (Fig. 4B, panel 2). In this structure, the Site B Mg2+ is unoccupied, presumably due to the absence of the SP. The shift in residues 184-187 is shown by the black arrow. Panel 3 of Figure 4A shows the active site immediately prior to the reaction. Once Mg2+ binds Site B, it ligates the SP via the same nonesterified oxygen as the Site A Mg2+, as well as the leaving group, the O3’. Although currently there is no direct evidence for ligation between the Site B Mg2+ and the O3’, it is commonly proposed to occur in two-metal ion mechanisms of other enzymes28. The Site B metal ion also ligates a water molecule positioned to donate a proton to the O3’ leaving group (cyan, Fig. 4A, panel 3), thereby stabilizing the otherwise unfavorable negative charge that forms on the O3’ as the bond with phosphorus (red, Fig. 4A, panel 3) breaks. Panel 3 of Figure 4B shows the cryo-EM ROO filament structure in magenta, overlaid onto the Site B Mg2+ from a post-catalytic product structure of SgrAI bound to cleaved DNA and residing in the T state (PDB code 3MQY), in blue. The shift in residues 183-188 in the R state brings the backbone carbonyl of Thr 186 1.5 Å closer to the Site B metal ion (blue, Fig. 4B, panel 3), which would allow for stabilization of the ion either through direct or second shell ligation. Finally, panel 4 of Figure 4A shows the predicted product structure after the reaction has occurred. Panel 4 of Figure 4B shows the same product structure of 3MQY, with all regions displayed, superimposed on the cryo-EM ROO. In the T state conformation of 3MQY, the backbone carbonyl of Thr 186 is too far from the site B Mg2+ to directly coordinate the metal ion (Fig. 4B, panel 4). Presumably, this previously observed low-activity conformation included other structural differences in comparison to the expected enzyme configuration in the high-activity state. In contrast, the current cryo-EM reconstruction provides the first glimpse into the structural configuration of the enzyme active site in the hyper-activated R state that is competent for rapid DNA cleavage and an improved model for the post-catalytic product state of the hyper-activated form. Together with previous low-activity crystal structures, and building upon prior knowledge of the two-metal ion hydrolytic cleavage mechanism, these snapshots provide the most comprehensive insight into the full mechanism of catalytic cleavage activity for SgrAI.
DNA distortions suggest that ROO-mediated sequence specificity expansion is influenced by indirect readout
To understand how DNA sequence affects SgrAI preferences for primary and secondary binding sites, we investigated the protein-DNA interactions and DNA structure in the inactive and active enzyme conformational states. The inactive state is represented by PDB 3DVO, and the active state is represented by the current cryo-EM ROO model. We could find no significant differences in direct readout or in other direct contacts to the 8-bp recognition sequence (Fig. S3). However, analysis of the DNA structure showed a large rearrangement at the center (4th) base step, between the C4 and G5 nucleotides in CACCGGTG (Fig. 5A). The 4 Å rise at this base step is unusually high in both conformations (Fig. 5A right, the van der Waals rise in B form DNA is 3.4 Å). Such a large rise will likely weaken the stacking energy of these bases. Analysis of RMSD of the 8-bp recognition sequence between the two structures reveals better alignments when superimposing each 4-bp half-site, rather than the full 8-bp sequence (Table S2). Furthermore, calculation of the base overlap areas between neighboring dinucleotide pairs confirmed that the 4th base step undergoes the largest structural change upon transitioning from the low to the high activity state (Table S3). Hence, it appears that the DNA accommodates the large conformational change in the SgrAI dimer, without disrupting important direct readout contacts between SgrAI and the DNA, by allowing each 4-bp half-site to move independently. The weakened stacking and large rise at the center base step allows for this independent movement.
To investigate changes in DNA structure further, the stacking overlap areas were calculated for all base steps of the 8-bp recognition sequence (Table S3) in both conformations. Similar DNA structure and stacking areas were found at the first and third base steps (Table S3, Fig. S4A-B). However, the bases in the second base step, between nucleotides A2 and C3, showed a large rearrangement (Table S3, Fig. 5B and Movie S2), corresponding to a change in stacking area of ~2 Å2 (a greater than 2-fold difference). This base step is affected by substitutions in primary site sequences (i.e. CRCCGGYG) found in two of the fourteen secondary site sequences (i.e. CCCCGGYG, Y=C or T). Differences in base stacking are likely to result in differences in stacking energy, and hence different degrees of stabilization of the low and high activity conformations.
Mechanistic model of SgrAI activation via filament formation and secondary site cleavage activity
One important issue yet to be fully understood is the mechanism by which primary and secondary site sequences influence SgrAI activation. Secondary sites differ from primary sites by a single base pair, either in the first or second position. Some discrimination occurs at the DNA binding step, however SgrAI binds to both types of sites with nanomolar or tigher affinity10. Yet cleavage of secondary sites is nearly undetectable unless primary sites are present on the same DNA, or at sufficient concentrations. A working model to rationalize this behavior postulates that SgrAI bound to DNA resides in an equilibrium between a low activity “tense” T state and a high activity “relaxed” R state (Fig. 5C). Only the high activity R state forms ROO filaments. When the bound DNA has the primary site sequence, the R state is sufficiently populated to influence ROO filament formation (Fig. 5C and Fig. 1). However, when the bound DNA contains the secondary site sequence, the T state is favored to a greater extent, thereby lowering the occupancy of the R state conformation and hence the propensity to form ROO filaments (Fig. 5C and Fig. 1). Occupancy of the R state is not completely disallowed, since SgrAI bound to secondary site DNA will join filaments formed by SgrAI bound to primary site DNAs (Fig. 1), stabilizing the R state conformation and activating the enzyme for broad cleavage of all bound DNAs.
In the context of the T and R state model, the X-ray crystal structures of non-filamentous SgrAI bound to DNA (primary or secondary site) represent the T state, and the current cryo-EM ROO filament structure of SgrAI bound to primary site DNA represents the R state. We found that the R state is characterized by a reconfiguration of Thr 186 in the enzyme active site, which is predicted to stabilize the site B metal ion (Fig. 4). Previous data has shown that SgrAI in the R state exhibits rapid DNA cleavage kinetics, and will form the ROO filament (Fig. 5C). The T state has much lower DNA cleavage kinetics, or perhaps, is completely inactive.
The equilibrium between the two states can be estimated by comparing the single turnover DNA cleavage rate constants of SgrAI with primary and secondary site DNAs in the absence of activation and without significant formation of ROO filaments. If only the R state is capable of DNA cleavage (estimated at 0.8 s-1)22, then the observed cleavage rate constant of the unactivated, non-ROO form of the enzyme bound to primary site DNA (0.0017 s-1)10 can be used to estimate the proportion of SgrAI/DNA complexes in the T state. This gives an equilibrium constant that shows that the T state is favored over the R state by 470-fold (0.8 s-1/0.0017 s-1), corresponding to a free energy difference of −3.6 kcal/mol (at 25°C). Since SgrAI cleaves secondary site DNA much more slowly14, the T state is favored by this complex to a greater extent, giving −4.9 kcal/mol (Fig. 5D). SgrAI bound to either type of site favors the T state conformation, however when the secondary site sequence is bound, the complex favors the T state conformation by an estimated 1.3 kcal/mol more than when the primary site is bound (Fig. 5D).
Comparison of the R and T state SgrAI structures bound to primary and to secondary DNA should reveal the origin of the differential stabilization. Only a single structure of SgrAI bound to a secondary site has been determined, that with CCCCGGTG, and in the T state25. Previously, no differences in interactions between enzyme and DNA, or in DNA structure, between this and the T state structure with primary site were observed. For this reason, it was necessary to derive a structure of the activated enzyme form in the R state and bound to primary site DNA, which is now represented by our cryo-EM ROO model. The current structure – in particular the configurations of the 1st and 2nd base pairs that account for differences in primary vs. secondary site sequences – was therefore used to investigate possible new interactions or DNA structure that could explain differential activity between primary and secondary site cleavage. The only change in base stacking was identified at the second base step, (CACCGGTG). Importantly, this subtle change represents a 50% increase in base stacking overlap area in the R state relative to the T (Fig. 5E). While the exact energies of base stacking are difficult to calculate, they generally correlate with stacking area. Therefore, this increase in stacking area may be expected to be stabilizing to the R state conformation, relative to the T. Stacking energies measured for different dinucleotide sequences show that the stacking energy of primary sequences at this base step (AC or GC, blue bars, Fig. 5F) provide more favorable energy (up to 0.5 kcal/mol) than the secondary site sequence (CC, gold bar, Fig. 5F)31. Hence, the difference in base stacking and its associated energy may explain at least part of the differential stabilization of the R (vs. T) state by primary and this class of secondary site sequences. A better understanding of the origin in the other class of secondary site sequences (XRCCGGY G, X=G, T, or A) awaits their structure determination in the relevant conformational states.
Biological role and relationship to other fìlament forming enzymes
The unusual, allosteric, filament forming mechanism of SgrAI may have evolved due to evolutionary pressure imposed by the relatively large genome of its host, Streptomyces griseus10. The larger genome results in a greater number of recognition sites, which must be protected from SgrAI-mediated cleavage via methylation by the cognate methyltransferase SgrAI.M. Such pressure to protect these sites from damaging cleavage would favor an increase in the activity of the methyltransferase, and/or a decrease in the activity of the SgrAI endonuclease. We find that the activity of SgrAI is in fact reduced compared to that of similar endonucleases, in that its 8-bp recognition sequence is longer than in typical endonucleases, making it rarer in genomes, and its DNA cleavage rate is remarkably slow in the absence of activation10. Activation occurs only under particular conditions, namely the assembly of SgrAI into ROO filaments when bound to unmethylated primary site DNA. This is less likely to occur within the host genome, because most primary sites are methylated, but predicted to occur on invading phage DNA containing unmethylated primary sites. In addition to becoming activated upon ROO filament formation, the specificity of SgrAI is also expanded to include cleavage of a second class of recognition sequences, called secondary sites, which increases the total number of recognition sequences from 3 to 17. A greater number of potential cleavage sites increases the probability of phage DNA cleavage, and may also aid in neutralization by preventing transcription, replication, repair, as well as methylation by the SgrAI.M enzyme, also present in the cell.
Beyond the well-known cytoskeletal NTPases, filament formation by enzymes is a newly appreciated phenomenon, with the advantages and evolutionary driving forces yet to be fully understood. Proposed purposes of filament formation vary among enzymes and include rapid enzyme activation, sequestration of activity, buffering of activity, and even in functioning as cytoskeletal structures5,23,32–35. In the case of SgrAI, an extensive kinetic study has recently been performed showing that association of SgrAI/DNA complexes into the ROO filament is the rate determining step under most conditions, and is overcome only through high enzyme and DNA concentrations (obtainable in vitro). In a biological context, and owing to local concentration effects, filament formation is expected to occur when two cleavage sites reside on the same contiguous DNA21–23. This effect is significant and is responsible for sequestering SgrAI activity on phage DNA and away from the host genome23. Sequestration of DNA cleavage activity to the activating (i.e. phage) DNA is critical, since most secondary sites on the host DNA are not methylated36, and can be cleaved by activated SgrAI.
Simulations using the same kinetic parameters determined for SgrAI, but with a theoretical model where only binary oligomeric states are allowed to form, shows that the filament forming mechanism is superior in both speed and sequestration23. Both the binary and the ROO filament model effectively sequester activated DNA cleavage on phage DNA, with minimal predicted host DNA cleavage, as a result of the slow, rate-limiting enzyme association into higher-order complexes (i.e. the binary complex or ROO filaments). However, the ROO filament mechanism is 2-fold faster in DNA cleavage, owing to the multiple ways enzymes can assemble into filaments (i.e. at either end), compared to the binary mechanism with only a single manner of association. The binary mechanism can be “sped up” to achieve the same rate of DNA cleavage as the ROO filament mechanism by increasing the association rate constant for binary assembly formation, however, it must be increased by 4.5-fold to achieve the same fast DNA cleavage kinetics as the ROO filament mechanism23. This increased association rate constant results in more predicted DNA cleavage of secondary sites on the host DNA, and hence a loss of sequestration of DNA cleavage activity. Speed is likely to be critical to protecting against phage infection by SgrAI, as is also supported by a recent study23. Therefore, the ROO filament mechanism may have evolved to meet the opposing requirements of rapid activation and sequestration of activity.
METHODS
Protein preparation
SgrAI enzyme used in assays contains 13 additional C-terminal residues (ENLYFQSHHHHHH) which include 6 histidine residues to be used for SgrAI purification, as well as a cleavage site for TEV protease, and was purified using previously described methods14. Briefly, SgrAI was expressed in BL21 (DE3) E. coli (which also contain a constitutive expression system for the methyltransferase MspI.M) overnight at 17°C. Cells were sonicated, centrifuged to remove cell debris, and SgrAI was isolated using Talon resin chromatography (Clonetech, Inc.), followed by further purification using heparin resin chromatography (GE, Inc.). Purified SgrAI was concentrated and stored in single use aliquots at −80°C in buffer containing 50% glycerol. Enzyme purity was assessed using Coomassie blue staining of SDS-PAGE and assessed to at least 99% purity.
DNA preparation
The oligonucleotides were prepared synthetically by a commercial source and purified using C18 reverse phase HPLC. The concentration was measured spectrophotometrically, with an extinction coefficient calculated from standard values for the nucleotides 37. Equimolar quantities of complementary DNA were annealed by heating to 90°C for 10 minutes at a concentration of 1 mM, followed by slow cooling to room temperature. The sequence of the DNA used in SgrAI/DNA preparations is shown below (red indicates the SgrAI primary recognition sequence, and | indicates cleavage site):
PC-DNA-top 5’-GATGCGTGGGTCTTCACA −3’
PC-DNA-bottom 3’-CTACGCACCCAGAAGTGTGGCC-5’
Two copies of PC DNA (the duplex formed by annealing of PC-top and PC-bot) self-assemble via annealing of their 5’ “overhanging” CCGG sequences to simulate a 40 bp DNA duplexcontaining a single primary site sequence (shown in red above) after cleavage by SgrAI, however with the exception that it is missing the 5’phosphate at the cleavage site.
Sample preparation
SgrAI/DNA samples were prepared using: 30 μl of 6.4 μM SgrAI (in 10 mM Tris-HCl pH 7.8, 300 mM NaCl, 0.1 mM EDTA, 1 mM 2-mercaptoethanol), 2 μl of 420 μM stock PC DNA in H2O, 1.5 μl of 100 mM Mg(OAc)2 and incubation at room temperature for 50 min. The final concentrations are 5.8 μM SgrAI, 25 μM PC DNA (4.3:1 ratio of PC DNA:SgrAI), and 4.5 mM MgCl2.
Data collection
Images were recorded on a Titan Krios electron microscope (FEI) equipped with a K2 summit direct detector (Gatan) at 1.31 Å per pixel in counting mode using the Leginon software package38. Data was acquired using a dose of ~55 e-/Å2 across 60 frames (50 msec per frame) at a dose rate of ~7.8 e-/pix/sec. A total of 216 micrographs were recorded over a single session. All imaging parameters are summarized in Table S1.
Image analysis
Preprocessing steps, including frame alignment, CTF estimation, and particle selection, were performed within the Appion pipeline39. Movie frames were aligned using MotionCor240 on 5 by 5 patch squares and using a b-factor of 100. Micrograph CTF estimation was performed using CTFFind441. Particles were selected using DoG picker, with an overlap that did not allow any two picks to be closer than ~150 Å in distance. This provided sufficient separation within the picks to subsequently enable helical averaging of multiple asymmetric units within extracted particle boxes in the 3D classification and refinement stages. 31,988 particles were extracted at this point. Reference-free 2D classification was performed to remove any non-filamentous particles. For 3D classification, initially, an asymmetric (C1) classification was employed to separate helical filaments of distinct compositional heterogeneity and to remove bad particles from the data, which did not give rise to an interpretable reconstruction. At the next step, we imposed helical symmetry during 3D classification to classify the particles. After a round of 2D classification, 3D classification without imposition of helical symmetry, and 3D classification with imposition of helical symmetry, 6,894 particles remained. Finally, 3D refinement was performed using a soft-edged mask, and the resulting map was subjected to B-factor sharpening that yielded a final map of the ROO resolved to ~3.5 Å. The final helical symmetry parameters that were used for refinement were 21.6 Å and −86.2° for the rise and twist, respectively. During refinement, we also determined the optimal number of asymmetric units within a box by iteratively varying the number of asymmetric units (and thus how much helical averaging is performed within each box). Based on examination of the FSC curve and visual inspection of the resulting map, this number was determined to be ~7. Thus, 7-fold helical averaging was performed for each windowed particle. Using this scheme, it is possible that a few of the asymmetric units would appear in the reconstruction more than once, and conversely, that some not at all. However, this was the simplest manner by which to deal with the highly heterogeneous nature of the short, helical fragments that are found in the cryo-EM experiment, while still providing a high-resolution map. The final map was evaluated using Fourier Shell Correlation (FSC) analysis to calculate global and local map resolution (sx_locres.py, implemented within Sparx42, provided local FSC maps) and the 3DFSC program suite43 to calculate degree of directional resolution anisotropy.
Atomic model refinement of 3.5 Å resolution cryo-electron microscopy model
Figure S2 shows the quality of the 3.5 Å cryo-EM envelope. Model building proceeded with real-space refinement after placement of the SgrAI/DNA model derived from X-ray crystallography (PDB 3DVO18) using Phenix44. Model adjustment and refinement were performed iteratively in Coot45 and Phenix46, and the statistics were examined using Molprobity47 until no further improvements were observed. The final model was also evaluated using FSC analysis against the map (Figure S1) and using EMRinger48 to compare the fit of the model backbone into the cryo-EM map. The final model statistics showed good geometry and matched the cryo-EM reconstruction (Figure S1F, Figure S2, and Table S1).
Structural comparisons-RMSD and alignments
RMSD calculations were performed with the UCSF Chimera package49. Chimera is developed by the Resource for Biocomputing, Visualization, and Informatics at the University of California, San Francisco (supported by NIGMS P41-GM103311). The “Matchmaker” tool (with Needleman-Wunsch matrix for protein, and “nucleic” for DNA) was used with structure-based alignment using alpha carbons, phosphorus atoms, backbone atoms only, or all atoms of selected residues of selected chains. Because the scissile phosphate is missing in the EM structure determined here, the three atoms of the scissile phosphate (the phosphorus atom and two non-esterified oxygen atoms) were deleted from the other structures to allow for superposition with all atoms. PyMOL software (The PyMOL Molecular Graphics System, Version 2.0 Schrödinger, LLC.) was used for figures and some alignments, where indicated, using the align command, or the alignment wizard and selected atoms. Analysis of subunit rotation was performed with UCSF Chimera, after first superimposing both chains A of 3DVO and the cryo-EM model using Matchmaker. The match command was used to calculate the rotation angle of chains B relative to each other.
Contributions
N.H. prepared and assembled SgrAI and DNA. D.L. collected the cryo-EM data. S.P. processed the cryo-EM data. S.P., N.H., and D.L. built and refined the atomic model; N.H. and D.L. wrote the manuscript; all authors contributed to manuscript editing.
Conflict of Interest
The authors declare that they have no conflicts of interest with the contents of this article.
Acknowledgments
We thank Yong Zi Tan and Youngmin Jeon for establishing conditions for SgrAI vitrification and Bill Anderson and Jean-Christophe Ducom at The Scripps Research Institute for help with EM data collection and network infrastructure. Molecular graphics and analyses were performed with the USCF Chimera package (supported by NIH P41 GM103331). Research reported in this publication was supported by the National Science Foundation under Grant No. MCB-1410355 (to N.H.) and by the National Institutes of Health Grant No. DP5 OD021396 (to D.L.). The EM map and atomic model of activated filamentous SgrAI will be deposited into the EMDB and PDB under accession codes X and Y respectively.