Structural basis of σ appropriation

Abstract Bacteriophage T4 middle promoters are activated through a process called σ appropriation, which requires the concerted effort of two T4-encoded transcription factors: AsiA and MotA. Despite extensive biochemical and genetic analyses, puzzle remains, in part, because of a lack of precise structural information for σ appropriation complex. Here, we report a single-particle cryo-electron microscopy (cryo-EM) structure of an intact σ appropriation complex, comprising AsiA, MotA, Escherichia coli RNA polymerase (RNAP), σ70 and a T4 middle promoter. As expected, AsiA binds to and remodels σ region 4 to prevent its contact with host promoters. Unexpectedly, AsiA undergoes a large conformational change, takes over the job of σ region 4 and provides an anchor point for the upstream double-stranded DNA. Because σ region 4 is conserved among bacteria, other transcription factors may use the same strategy to alter the landscape of transcription immediately. Together, the structure provides a foundation for understanding σ appropriation and transcription activation.


INTRODUCTION
Bacterial transcription is catalyzed by RNA polymerase (RNAP), which consists of ␣ I , ␣ II , ␤, ␤ and subunits. To initiate transcription without transcription activators (referred to as canonical transcription initiation in this paper), bacterial RNAP needs to form holoenzyme with factor (1-10). The principle factor, 70 in Escherichia coli (E. coli), contacts RNAP extensively and mediates sequencespecific interactions with promoter DNA. In particular, conserved region R2 contacts a domain of ␤ subunit known as the clamp helices and mediates sequence-specific interactions with the promoter -10 element, while conserved region R4 contacts the ␤ flap tip helix (FTH) and the ␤ C-terminal region (␤CTR), and mediates sequencespecific interaction with the promoter -35 element.
Transcription is the step at which most regulation of gene expression occurs. For many years, bacteriophage T4 has provided a simple model system to investigate mechanisms underlying this process (11,12). Because bacteriophage T4 does not encode its own RNAP, it must redirect RNAP of its host to the correct genes at the correct time. T4 middle genes encode proteins for replication, recombination and nucleotide metabolism (11). Unlike host promoters, T4 middle promoters contain a promoter -10 element and a MotA box, which is centered around position -30 and replaces the promoter -35 element (13). Transcription of T4 middle genes proceeds by activation of T4 middle promoters through a process called appropriation, which requires the concerted effort of two T4-encoded transcription factors: AsiA and MotA (11,14). In this process, the co-activator AisA binds to and remodels R4, which then allows the activator MotA to interact with R4, as well. In addition, the conformational change of R4 prevents its normal contact with the promoter -35 element, so the transcription of host genes is inhibited.
The structural basis of appropriation has been studied extensively. Available structures include AsiA alone and in complex with R4 (15)(16)(17), C-terminal domain of MotA (MotA CTD ) alone and in complex with DNA (18,19), and N-terminal domain of MotA (MotA NTD ) (20,21). Even a holistic model for appropriation complex was developed using experimentally determined structures, biochemically observed restrains and molecular modeling (22). Nevertheless, puzzle remains, in part, because of a lack of precise structural information for appropriation complex. For example, although the significance of the N-terminal region of AsiA in appropriation is conclusive (15,23,24), the role of the C-terminal region remains controversial (22,23,(25)(26)(27). Furthermore, the G1249D substitution within the ␤ subunit specifically impairs middle promoter activation both in vivo and in vitro (22,28). However, molecular modeling suggests that the position of G1249 is far from AsiA, MotA, DNA and R4 (22). Last but not least, AsiA contains a helix-turn-helix (HTH) motif, suggesting the possibility of an interaction between AsiA and DNA (17). However, no such interaction has been confirmed yet.
To solve these puzzles conclusively, we determined a single-particle cryo-electron microscopy (cryo-EM) structure of an intact appropriation complex at 3.79Å resolution. The structure defines the protein-protein and protein-DNA interactions that mediate appropriation. Strikingly, AsiA alters its conformation to engage RNAP and provides an anchor point for the upstream double-stranded DNA (dsDNA).

AsiA protein
Gene encoding AsiA was synthesized and subcloned to pET28a by GENEWIZ, Inc. Escherichia coli strain BL21(DE3) (Invitrogen, Inc.) was transformed with plasmid pET28a-NH-AsiA (GENEWIZ, Inc.) encoding Nhexahistidine-tagged AsiA under the control of the bacteriophage T7 gene 10 promoter. Single colonies of the resulting transformants were used to inoculate 1 L LB broth containing 50 g/ml kanamycin, cultures were incubated at 37 • C with shaking until OD 600 = 0.6, cultures were induced by addition of isopropyl ␤-D-1thiogalactopyranoside (IPTG) to 0.4 mM, and cultures were incubated 75 min at 30 • C. Then, cells were harvested by centrifugation (5000 rpm; 10 min at 4 • C), resuspended in 20 ml buffer A (20 mM Tris-HCl, pH 8.0, 0.1 M NaCl, 5% glycerol) and lysed using a JN-02C cell disrupter (JN-BIO, Inc.). The lysate was centrifuged (12 000 rpm; 45 min at 4 • C), and the supernatant was loaded onto a 2 ml column of Ni-NTA agarose (Qiagen, Inc.) equilibrated with buffer A. The column was washed with 10 ml buffer A containing 0.16 M imidazole and eluted with 10 ml buffer A containing 0.5 M imidazole. The sample was further purified by anion-exchange chromatography on a Mono Q 10/100 GL column (GE Healthcare, Inc.; 160 ml linear gradient of 0.1-1 M NaCl in buffer A). Fractions containing AisA were pooled and stored at −80 • C. AsiA derivative was expressed and purified in the same way as wild-type protein. Yields were ∼2 mg/l, and purities were >95%.

MotA protein
Gene encoding MotA was synthesized and subcloned to pET21a by GENEWIZ, Inc. Escherichia coli strain BL21(DE3) (Invitrogen, Inc.) was transformed with plasmid pET21a-MotA (GENEWIZ, Inc.) encoding MotA under the control of the bacteriophage T7 gene 10 promoter. Single colonies of the resulting transformants were used to inoculate 1 L LB broth containing 50 g/ml ampicillin, cultures were incubated at 37 • C with shaking until OD 600 = 0.6, cultures were induced by the addition of IPTG to 1 mM, and cultures were incubated 3 h at 37 • C. Then, cells were harvested by centrifugation (5000 rpm; 10 min at 4 • C), resuspended in 20 ml buffer B (10 mM Tris-HCl, pH 7.5, 0.2 M NaCl, 5% glycerol, 1 mM EDTA and 1 mM dithiothreitol (DTT)) and lysed using a JN-02C cell disrupter (JN-BIO, Inc.). The lysate was centrifuged (12 000 rpm; 45 min at 4 • C), and the supernatant was loaded onto a 5 ml column of HiTrap Heparin HP (GE Healthcare, Inc.) equilibrated in buffer B and eluted with a 100 ml linear gradient of 0.2-1 M NaCl in buffer B. The sample was further purified by cation-exchange chromatography on a Mono S 10/100 GL column (GE Healthcare, Inc.; 160 ml linear gradient of 0.2-1 M NaCl in buffer B). Fractions containing MotA were pooled and stored at −80 • C. Yields were ∼2 mg/l, and purities were >95%.

Assembly of appropriation complex
DNA oligonucleotides (sequences in Figure 1A) (Sangon Biotech, Inc.) were dissolved in nuclease-free water to ∼1 mM and stored at −80 • C. Template strand DNA and nontemplate strand DNA were annealed at a 1:1 ratio in 10 mM Tris-HCl, pH 7.9, 0.2 M NaCl and stored at −80 • C. appropriation complex was prepared in reaction mixtures containing (500 l): 9 M 70 , 18 M AsiA, 4.5 M E. coli RNAP core enzyme, 5 M DNA scaffold and 18 M MotA. 70 was incubated with AsiA for 10 min at 37 • C, incubated with core for 10 min at 37 • C and incubated with MotA and DNA scaffold for 10 min at 37 • C. The mixture was applied to a Superose 6 Increase 10/300 GL column (GE Healthcare, Inc.) equilibrated in 10 mM HEPES, pH 7.5, 50 mM KCl, and the column was eluted with 24 ml of the same buffer. Fractions were checked by SDS-PAGE and the peak containing appropriation complex was concentrated to 20 M using an Amicon Ultra-0.5mL centrifugal filter (10 kDa MWCO; Merck Millipore, Inc.).

Cryo-EM grid preparation
Immediately before freezing, 8 mM CHAPSO was added to the sample. C-flat grids (CF-1.2/1.3-4C; Protochips, Inc.) were glow-discharged for 60 s at 15 mA prior to the application of 3 l of the complex, then plunge-frozen in liquid ethane using a Vitrobot (FEI, Inc.) with 95% chamber humidity at 10 • C.

Cryo-EM data acquisition and processing
The grids were imaged using a 300 kV Titan Krios (FEI, Inc.) equipped with a K2 Summit direct electron detector (Gatan, Inc.). Images were recorded with Serial EM (31) in counting mode with a physical pixel size of 1.307 A and a defocus range of 1.5-2.5 m. Data were collected with a dose of 8 e/pixel/s. Images were recorded with a 12 s exposure and 0.25 s subframes to give a total dose of 56 e/Å 2 . Subframes were aligned and summed using MotionCor2 (32). The contrast transfer function was estimated for each summed image using CTFFIND4 (33). From the summed images, ∼10 000 particles were manually picked and subjected to 2D classification in RELION (34). 2D averages of the best classes were used as templates for auto-picking in RELION. Auto-picked particles were manually inspected, then subjected to 2D classification in RE-LION. Poorly populated classes were removed, resulting in a dataset of 654 794 particles. These particles were 3D classified in RELION using a map of E. coli elongation complex (EMD-8585) (35) low-pass filtered to 40Å resolution as a reference. 3D classification resulted in 4 classes. Particles in Class 3 and Class 4 were 3D auto-refined, then subjected to 3D classification focused on the MotA box. From this classification, the best-resolved class containing 105 108 particles was 3D auto-refined and post-processed in RELION.

Cryo-EM model building and refinement
The models of RNAP core enzyme, R2 and R3 from the structure of E. coli RPo (PDB 6CA0) (9), the NMR structure of AsiA-R4, the crystal structure of MotA NTD and the crystal structure of MotA CTD -MotA box were fitted into the cryo-EM density map using Chimera (36). The model of nucleic acids was built manually in Coot (37). The coordinates were real-space refined with secondary structure restraints in Phenix (38).  Figure 4B) were annealed at a 1:1 ratio in 10 mM HEPES, pH 7.5, 50 mM KCl. 5-Hydroxymethylated cytosines within the DNA scaffold were glucosylated by treatment with ␤-glucosyltransferase (NEB, Inc.) according to the company protocol and purified using standard phenol-chloroform extraction followed by ethanol precipitation. Mass spectrometry confirmed that all cytosines within the DNA scaffold are hydroxymethylated and glucosylated. Equilibrium fluorescence polarization assays were performed in a 96-well microplate format.

Electrophoretic mobility shift assay to quantify appropriation complex
where I VV and I VH are fluorescence intensities with the excitation polarizer at the vertical position and the emission polarizer at, respectively, the vertical position and the horizontal position. Equilibrium dissociation constant, K D , was extracted by non-linear regression using the equation: where P is the fluorescence polarization at a given concentration of AsiA, P f is the fluorescence polarization for free 6-FAM-labelled DNA scaffold, P b is the fluorescence polarization for bound 6-FAM labeled DNA scaffold and [A] is the concentration of AsiA or AsiA derivative.

Overall structure of appropriation complex
To obtain a structure of appropriation complex, we used a nucleic-acid scaffold corresponding to positions −45 to +15 of a T4 middle promoter (P uvsX , positions numbered relative to the transcription start site; Figure 1A). The scaffold contains a MotA box centered between positions −30 and −31, a consensus promoter -10 element and a 13-bp transcription bubble maintained in the unwound state by having non-complementary sequences on non-template and template strands. First, 70 was incubated with excessive AsiA to ensure that there was no free 70 . Then, RNAP core enzyme was added to form AsiA-associated RNAP. Finally, MotA and the nucleic-acid scaffold were added to form appropriation complex, which was purified by gel filtration (Supplementary Figure S1A). SDS-PAGE confirmed that all protein components are present in the complex with stoichiometric levels (Supplementary Figure S1B). Nucleicacid staining confirmed that the complex contains DNA scaffold (Supplementary Figure S1C). In vitro transcription assay with ATP, CTP and UTP confirmed that the complex is capable of transcription initiation (Supplementary Figure  S1D).
We then froze the complex, collected data on Titan Krios and determined the structure at a nominal resolution of 3.79 A ( Figure 1B Table S1). Local resolution calculation indicates that the central core of the structure was determined to 3.4-4.0Å resolution (Supplementary Figure S3C).
Nucleic Acids Research, 2019, Vol. 47, No. 17 9427 The RNAP, R2 and R3 of the structure are very similar to the previously reported E. coli RNAP-promoter open complex (RPo) structure (9) with root-mean-square deviation (RMSD) of 1.284Å (3551 C␣s aligned). Moreover, R2 and R3 interact with the promoter -10 element and extended -10 element in the same way as in RPo, in accordance with the observation that the protein-DNA interactions downstream of position -20 are not affected by the presence of AsiA and MotA (14). It is worth noting that nearly all T4 middle promoters have a stringent requirement for an excellent match to the consensus promoter -10 element and about half of the middle promoters also contain a consensus extended -10 element (11).

Both R3.2 and R4 are remodeled by AsiA
Compared with the structure of RPo, the conformation of the C-terminal helix of R3.2 and R4 in appropriation complex is dramatically different (Figure 2A and Supplementary Figure S4A). The C-terminal helix of R3.2 (530-543) and the first helix of R4 (S1) form a continuous helix in RPo. However, most residues of the C-terminal helix of R3.2 (532-539) become disordered in appropriation complex probably due to clash with the C-terminal helix of AsiA (A6; see below). R4 adopts an elongated conformation with all helices completely reorganized. For example, the third and forth helices (S3 and S4) of R4 and the loop connecting them, which constitute the HTH motif and participate in binding the -35 element in RPo, are converted into one continuous helix in appropriation complex.
Early work indicated that AsiA binds very tightly to R4 and a NMR structure of AsiA in complex with R4 has been determined (15). The structure of AsiA and R4 in appropriation complex is superimposable on the NMR structure (neglecting A6 of AsiA and S5 of R4, which do not contribute to AsiA-R4 interaction; Figure 2B), indicating that AsiA makes a similar set of interactions in both structures. Substitutions of some of the interacting residues ( Figure 2C) have been demonstrated to impair AsiA's function (23,24,(39)(40)(41), verifying that the cryo-EM structure is biologically relevant. The AsiA-R4 interface area is large (1194Å 2 ), which explains the tight binding between AsiA and and emphasizes the critical role of AsiA in the remodeling of R4.

AsiA alters its conformation to engage RNAP
Strikingly, whereas in the NMR structure of AsiA in complex with R4, A6 binds to a pocket on the remainder of AsiA, in appropriation complex, A6 is rotated by ∼90 • , about a 'pivot' formed by the short loop between A5 and A6, inserted into the RNA exit channel in the same orientation as the C-terminal helix of R3.2 in RPo, and contacted by ␤CTR (Figure 3A and B; Supplementary Figure S4B). Specifically, AsiA-conserved residue R82 forms a salt bridge with ␤ residue D1310 ( Figure 3C and Supplementary Figure S4C), consistent with the previous report that MotA-dependent transcription activation was severely compromised by charge-reversal substitution of AsiA residue R82 (26). In addition, AsiA conserved residue M86 makes van der Waals interactions with ␤ residue Y1305, consistent with the observation that AsiA substitution M86T also compromises MotA-dependent transcription activation (26). The G1249D substitution within the ␤ subunit specifically impairs middle promoter activation both in vivo and in vitro (22,28). In our structure, the distance between the ␣-carbon of ␤ residue G1249 and the carbonyl oxygen of AsiA residue N89 is 3.9Å (Figures 3C and Supplementary figure S4C). We infer that the G1249D substitution would introduce steric clash between the side chain of the aspartic acid residue and the carbonyl oxygen of N89 and be incompatible with the current con- formation of appropriation complex. In accordance with our inference, benzoyl-phenylalanine (BpA) photocrosslinking experiments showed that even though RNAP bearing G1249D substitution within the ␤ subunit formed an AsiA-associated complex, the conformation of the complex differed from that formed by wild-type RNAP (22). In appropriation complex, FTH binds the hydrophobic pocket of AsiA, which is occupied by A6 in the NMR structure ( Figure 3D and Supplementary figure S4D). In particular, three hydrophobic residues of FTH (L901, I905 and F906) insert deeply into the pocket, which is consistent with the previous report that amino acid substitutions I905A and F906A in combination strongly disrupt the AsiA-␤ flap interaction (27). Additionally, the N74D substitution of AsiA and the G907K substitution of FTH that were reported to jeopardize AsiA-␤ flap interaction (27) can be interpreted by our structure. In our structure, AsiA conserved residue N74 forms an H-bond with FTH residue E898, and FTH residue G907 is close to a patch of positively charged residues on AsiA. The N74D substitution of AsiA would disrupt the H-bond and cause electrostatic repulsion between the aspartic acid residue and FTH residue E898. Similarly, the G907K substitution of FTH would cause electrostatic repulsion between the lysine residue and the patch of positively charged residues on AsiA.

The HTH motif of AsiA contacts the upstream dsDNA
The NMR structure of AsiA shows that A3, A4 and the loop connecting them constitute a HTH motif (15)(16)(17). In our structure, the HTH motif contacts the upstream portion of the MotA box with A4 inserting into the major groove of DNA between positions -32 and -37 ( Figure 4A and Supplementary figure S4E). In particular, two conserved positively charged residues (R55 and K56) on A4 and one conserved positively charged residue (R47) on the loop form three salt bridges with the DNA backbone phosphates. In accordance, alanine substitution of these residues in combination (HTH − ) eliminates AsiA-DNA interaction ( Figures  4B).
Previous studies have shown that P uvsX can be transcribed by RNAP alone, which is inhibited by AsiA in the absence of MotA and activated by the presence of both AsiA and MotA (23). In accordance with the previous studies, gel shift assay shows that RPo is formed by RNAP alone (Figure 4C, lane 1). AsiA inhibits the formation of RPo ( Figure  4C, lane 2), while AsiA and MotA in combination promotes the formation of appropriation complex, which migrates slower than RPo on native gel ( Figure 4C, lane 3). AsiA (HTH − ) inhibits the formation of RPo as efficiently as the wild-type protein ( Figure 4C, lane 4), indicating that AsiA (HTH − ) is well folded and the AsiA-DNA interaction is not required for its transcription inhibition activity. AsiA (HTH − ) forms less appropriation complex than the wildtype protein ( Figure 4C, lane 5), indicating that the AsiA-DNA interaction is important for its transcription activation activity.
For canonical transcription initiation, S4 of R4 inserts into the major groove of the promoter -35 element to read out sequence ( Figure 4D). Compared with S4, A4 does not insert so deeply and can not read out the bases in the major groove, which is consistent with the observation that the sequences between positions -32 and -37 are not conserved in T4 middle promoters (42)(43)(44)(45). To further confirm that the AsiA-DNA interaction is non-sequence-specific, we tested the interaction between AsiA and an irrelevant DNA scaffold using fluorescence polarization assay. It turned out that AsiA binds the irrelevant DNA scaffold with comparable affinity ( Figure 4B). Each cytosine in T4 DNA is hydroxymethylated and glucosylated, which places a bulky group within the major groove of T4 DNA (11,19). We infer that A4 of AsiA inserts shallowly into the major groove of T4 DNA, so it will not clash with the glucosylated hydroxymethyl moiety. In order to test this hypothesis experimentally, we prepared hydroxymethylated and glucosylated DNA scaffold using ␤-glucosyltransferase and confirmed its binding with AsiA using fluorescence polarization assay ( Figure 4B).

MotA NTD contacts both S4 and S5 of R4
MotA contains two domains connected by a flexible linker. In the structure of appropriation complex, the transactivation domain, MotA NTD , interacts with the remodeled R4, while, the DNA binding domain, MotA CTD binds the MotA box ( Figure 5A and Supplementary Figure S4F). In other words, R4 is sandwiched between MotA NTD and A B C AsiA, while the upstream dsDNA is sandwiched between MotA CTD and AsiA. The structure of MotA NTD in appropriation complex is very similar to the previously reported crystal structure and NMR structure of MotA NTD (20,21), with root-meansquare deviation (RMSD) of 0.879Å (89 C␣s aligned) and 1.719Å (89 C␣s aligned), respectively. S5 of R4 binds to a basic/hydrophobic groove on MotA NTD ( Figure 5B), which is consistent with the previous report that mutation of S5 residues F610 and L611 impairs both MotA binding to 70 and MotA-dependent transcription activation (46). The structure is also consistent with the observation that mutation of MotA residues K3, K28 and Q76, which lie in the basic/hydrophobic groove, impairs the ability of MotA to interact with S5 and to activate transcription (47). Besides the interaction of S5, basic residues on S4 interact with an acidic/hydrophobic surface on MotA NTD and mutation of MotA residues D30, F31 and D67, which lie on the acidic/hydrophobic surface, also have deleterious effects on the interaction with 70 , MotA-dependent transcription activation and phage viability (20,47).
Because of the glucosylated hydroxymethyl moiety in the major groove, a HTH motif cannot be used to read out sequence from the major groove. Therefore, a novel strat-egy is utilized by T4. MotA CTD adopts a saddle-shaped, 'double wing' conformation, straddles the major groove and leaves enough space to accommodate a glucosylated hydroxymethyl moiety, while the linker engages the minor groove in the upstream portion of the MotA box ( Figure  5A). Structural modeling based on biochemical data has predicted that MotA linker engages the minor groove in the upstream portion of the MotA box (22), which was further confirmed by the crystal structure of MotA in complex with MotA box (19). The structure of MotA CTD , the linker and the MotA box in appropriation complex is superimposable on the crystal structure of MotA in complex with MotA box (Figure 5C), indicating that MotA makes a similar set of interactions in both structures.

DISCUSSION
Previous work has shown that AsiA containing either a deletion of the C-terminal half of AsiA (23) or a deletion of A6 (22,23) is still active in vivo and in vitro. Other work has shown that the AsiA ortholog within the bacteriophage KVP40 can substitute for T4 AsiA in transcription, even though there is practically no homology between these AsiA proteins after residue 60 (25). According to our structure, Nucleic Acids Research, 2019, Vol. 47,No. 17 9431 AsiA engages RNAP through two sets of interactions. First, 70 , which is anchored to RNAP through R2 and R3, grabs the N-terminal half of AsiA through R4. Second, the C-terminal half of AsiA interacts with the ␤CTR and FTH of RNAP directly. If the first set of interactions are strong enough to hold AsiA and RNAP together, disruption of the second set of interactions will not necessarily eliminate the formation of appropriation complex. To test this possibility, we truncated the C-terminus of AsiA ( A6) and tested appropriation complex formation with the mutant protein (Supplementary Figure S5). As expected, truncation of the C-terminus of AsiA affects the formation of appropriation complex, but there is still a notable level of appropriation complex. Therefore, we infer that the notable level of appropriation complex is enough for the transcription of T4 middle genes in vivo and in vitro.
Despite the absence of sequence homology or structural similarity except for the HTH motif, the function of AsiA is analogous to that of R4 during canonical transcription initiation. During canonical transcription initiation, R4 contacts the FTH and the ␤CTR, and mediates sequencespecific interaction with the promoter -35 element (Supplementary Figure S6A). As a mimic of R4, AsiA also contacts the FTH and the ␤CTR, and provides an anchor point for the upstream dsDNA (Supplementary Figure S6B). The anchoring of the upstream dsDNA by AsiA may compensate for the weak binding between MotA and MotA box (11).
Extensive biochemical studies with T4 middle promoter P rIIB2 have been conducted in order to unveil the mechanism of MotA-dependent transcription activation (48,49). Consistent with the cryo-EM structure, surface plasmon resonance experiments proved that MotA promotes the formation of RNAP-promoter closed complex (RPc), but not the transition from RPc to RPo (48). Further consistent with the cryo-EM structure, in vitro transcription experiments indicated that MotA and AsiA work together to facilitate promoter escape (48), which can be attributed to the conformational change of the C-terminal helix of R3.2 and R4.
Compared with classical transcription activation, appropriation is more efficient. Typical transcription activators, such as class I and class II transcription activators, function by improving target promoters, but they have little effect on other promoters (7,(50)(51)(52)(53). On the contrary, AsiA and MotA shut down the majority of cellular genes, whose transcription requires 70 , and devote most resources to one set of promoters. This system potentially could be engineered to change the landscape of transcription with high efficiency in synthetic biology.

DATA AVAILABILITY
The accession number for the cryo-EM density map reported in this paper is Electron Microscopy Data Bank: EMD-9916. The accession numbers for the atomic coordinates reported in this paper are Protein Date Bank: 6K4Y.