Cryo-EM structure of a mammalian RNA polymerase II elongation complex inhibited by α-amanitin

RNA polymerase II (Pol II) is the central enzyme that transcribes eukaryotic protein-coding genes to produce mRNA. The mushroom toxin α-amanitin binds Pol II and inhibits transcription at the step of RNA chain elongation. Pol II from yeast binds α-amanitin with micromolar affinity, whereas metazoan Pol II enzymes exhibit nanomolar affinities. Here, we present the high-resolution cryo-EM structure of α-amanitin bound to and inhibited by its natural target, the mammalian Pol II elongation complex. The structure revealed that the toxin is located in a pocket previously identified in yeast Pol II but forms additional contacts with metazoan-specific residues, which explains why its affinity to mammalian Pol II is ∼3000 times higher than for yeast Pol II. Our work provides the structural basis for the inhibition of mammalian Pol II by the natural toxin α-amanitin and highlights that cryo-EM is well suited to studying interactions of a small molecule with its macromolecular target.

RNA polymerase II (Pol II) is the central enzyme that transcribes eukaryotic protein-coding genes to produce mRNA. The mushroom toxin ␣-amanitin binds Pol II and inhibits transcription at the step of RNA chain elongation. Pol II from yeast binds ␣-amanitin with micromolar affinity, whereas metazoan Pol II enzymes exhibit nanomolar affinities. Here, we present the high-resolution cryo-EM structure of ␣-amanitin bound to and inhibited by its natural target, the mammalian Pol II elongation complex. The structure revealed that the toxin is located in a pocket previously identified in yeast Pol II but forms additional contacts with metazoan-specific residues, which explains why its affinity to mammalian Pol II is ϳ3000 times higher than for yeast Pol II. Our work provides the structural basis for the inhibition of mammalian Pol II by the natural toxin ␣-amanitin and highlights that cryo-EM is well suited to studying interactions of a small molecule with its macromolecular target.
The toxin ␣-amanitin occurs in poisonous amanita mushrooms and inhibits Pol II, 2 the enzyme that transcribes proteincoding genes in eukaryotes to produce mRNA (1). The toxin ␣-amanitin is a modified peptide that comprises eight amino acids forming two ring systems (1). It contains the noncanonical amino acid residues dihydroxy isoleucine (Ile(OH) 2 ), hydroxyl proline (Hyp), and hydroxyl tryptophan (Trp(OH)), which contribute to its toxicity and its affinity for the Pol II enzyme (2).
Previous structural work used Pol II from the yeast Saccharomyces cerevisiae to reveal that ␣-amanitin binds in a pocket of the enzyme formed by the polymerase bridge helix, an element of the active center, and the RPB1 funnel domain helices ␣21 and ␣23, and loop ␣23-␣24 (3). The subsequent structure of a yeast Pol II elongation complex with bound DNA template and RNA transcript further showed that ␣-amanitin contacts two elements of the polymerase active center: the bridge helix and the trigger loop (4). The toxin in particular traps the trigger loop, which is a mobile element that undergoes folding for catalyzing extension of the RNA chain and is also important for the translocation of nucleic acids to the next DNA template position after catalysis (4).
Different organisms vary strongly in their sensitivity to ␣-amanitin (5), and their Pol II enzymes bind ␣-amanitin with different affinities (2). Whereas Pol II from yeasts such as S. cerevisiae binds ␣-amanitin with micromolar affinity (2), metazoan enzymes show much higher affinity, with mammalian Pol II binding with nanomolar affinity (6). The reasons for this dramatic variation in eukaryotic species remain unknown. Here we provide the cryo-EM structure of ␣-amanitin bound to its natural target, the mammalian Pol II elongation complex, and describe contacts of the toxin with Pol II that explain why its affinity is much higher for the mammalian enzyme.

Results
To investigate the binding affinity variation of ␣-amanitin in eukaryotic species, we determined the cryo-EM structure of ␣-amanitin bound to its natural target, the mammalian Pol II elongation complex (EC). We purified Pol II from pig thymus ("Experimental procedures") and added human Gdown1 (hGdown1) (Fig. 1A), which is often associated with Pol II in metazoan cells (7). The EC was formed with a DNA-RNA scaffold that was highly similar to a previously used one (8). The EC was active in RNA synthesis and was inhibited after ␣-amanitin addition (Fig. 1B). The EC sample was cross-linked with BS3, incubated with ␣-amanitin, and immediately applied to EM grids before flash freezing. Cryo-EM analysis revealed a homogeneous distribution of particles that could be classified easily (supporting Figs. S1 and S2D). 134,512 particle images were extracted and used for 3D reconstruction, resulting in a cryo-EM density map at a nominal resolution of 3.4 Å (supporting Fig. S1).
To obtain an atomic model of the mammalian Pol II EC-␣amanitin complex, we placed the previously refined bovine Pol II structure (8) into the density and adjusted it locally. There was no density for hGdown1, which apparently dissociated from the complex. The region around the Pol II active center, including ␣-amanitin and its binding pocket, was well resolved, with an estimated local resolution of ϳ3.0 Å (supporting Fig.  S2, A and B). There were no other significant additional densities observed.

cro ARTICLE
We could build an atomic model for ␣-amanitin and define its chemical interactions with Pol II (Figs. 2 and 3 and Table 1). The structure was completed by classification and refinement focused on the flexible Pol II stalk subcomplex RPB4 -RPB7, upstream DNA, and the mobile trigger loop, followed by manual adjustments and real-space refinement (supporting Fig. S1; see also "Experimental procedures").
The structure of the Pol II EC is highly similar to the previously determined structure of the bovine counterpart (8). Pig Pol II differs from bovine Pol II in only five residues: RPB1 Glu 1968 , RPB5 Glu 32 and Asp 46 , RPB6 Ser 126 , and RPB9 Phe 11 . The EC adopts the post-translocation state with a straight bridge helix, different from the slightly bent bridge helix observed in the yeast Pol II-␣-amanitin crystal structure (4), which is thought to reflect a translocation intermediate. The trigger loop adopts a conformation that most closely resembles the "wedged" conformation previously observed in the yeast EC bound by ␣-amanitin (4). However, residue Leu 1104 (Leu 1081 in yeast), which forms a wedge behind the bridge helix in the yeast structure (4), protrudes ϳ2 Å less in between the bridge helix and the polymerase cleft module, essentially not forming a wedge anymore, and consistent with the observed straight bridge helix. We refer to this slightly altered trigger loop conformation as "unwedged" because it is likely that it is adopted after the wedged conformation and before the addition of the next nucleotide.
The position and binding pocket of ␣-amanitin is as observed in the yeast EC (4) (Figs. 2B and 3B and Table 1). Most contacts between ␣-amanitin and yeast Pol II observed in the EC are conserved in the mammalian complex, as expected by the high conservation of residues involved in binding the toxin ( ). This contact is not present in the yeast Pol II-amanitin complex, because the yeast counterpart of mammalian Asn 792 is Ser 769 , and the observed hydrogen bond is thus not possible. There is a third residue in the amanitin-binding pocket that differs, Asn 742 (Fig.  3, A and B), which corresponds to Val 719 in yeast, but this is unlikely to contribute strongly to the difference in affinity because in both structures these residues form van der Waals contacts with the side chain of isoleucine in ␣-amanitin.
Thus, compared with the yeast structure, two additional hydrogen bonds are formed between ␣-amanitin and the mammalian EC. It is known that two additional hydrogen bonds can give rise to enthalpy changes that account for changes in dissociation constants by 3 orders of magnitude (9,10). We therefore suggest that the two additional hydrogen bonds account for the much higher affinity of mammalian Pol II for the toxin. This interpretation is supported by known biochemical data obtained with amanitin derivatives that lack certain functional groups (1,11). In particular, alkylation of the hydroxyl group in the indole ring is predicted to prevent hydrogen bond formation and is known to decrease toxicity and inhibitory potential of amanitin (1).
The structure also suggests the molecular basis for ␣-amanitin resistance arising from mutations in the binding pocket in Pol II enzymes from mice (12) and Drosophila (13). Modeling shows that mutation I779F in mouse RPB1 leads to a steric clash that likely prevents ␣-amanitin from binding (Fig. 4B). The additional mouse mutations L745P and R749P likely destabilize helix ␣21, which forms part of the binding pocket (Fig. 4B). The Drosophila melanogaster Rpb1 mutations N792D and N793D (13) are predicted to disrupt hydrogen bonds between Pol II and amanitin, thereby decreasing affinity.

Discussion
More than one century after the discovery of ␣-amanitin (14), we now provide an atomic model of its structure in com-

Mammalian Pol II inhibited by ␣-amanitin
plex with its natural target, the mammalian Pol II EC. This work provides the structural basis of mammalian Pol II inhibition by ␣-amanitin. Whereas insights into the mechanism of transcription inhibition by ␣-amanitin were already derived from structures of the yeast Pol II (3) and the yeast EC (4), our current work additionally provides a molecular explanation for the long-standing observation that ␣-amanitin has a much higher affinity for mammalian Pol II, compared with the yeast enzyme. Most notably, we observe two additional, well defined hydrogen bonds that are possible in mammalian Pol II enzymes, but not in yeast Pol II, explaining the tighter binding of the toxin to the former.
Together with recent studies (15,16), our work also shows that cryo-EM can now be used to study the detailed interactions of small molecules with proteins, as required for drug design. We note that such applications of cryo-EM still often require that the target molecule or complex has a critical size. In the future, further developments of cryo-EM will, however, likely remove this limitation such that the inhibition of target molecules and complexes of lower molecular weight by small molecules can also be studied.

Purification of Sus scrofa RNA polymerase II and recombinant hGdown1
S. scrofa Pol II was purified essentially as described for the bovine Pol II preparation (8), except that pig thymus instead of bovine thymus was used. Briefly, thymus was homogenized, and the supernatant was filtered. After polyethyleneimine precipitation, Pol II was purified with a MacroPrepQ column, followed by ammonium sulfate precipitation and an affinity column with 8WG16 (␣RPB1 CTD) antibody-coupled Sepharose, a UNO-Q anion exchange column, and finally a Sephacryl S-300 HiLoad sizing column. The typical yield was 2-4 mg from ϳ500 g of thymus.
Recombinant expression and purification of hGdown1 was performed as described previously (8). After elution from the UnoQ column (Bio-Rad), Pol II was combined with a 3-fold molar excess of hGdown1 and incubated for 2-3 h before applying the complex to a HiPrep 16/60 Sephacryl S-300 HR column (GE Healthcare). Fractions containing the Pol II-hGdown1 complex were collected and concentrated to a

Mammalian Pol II inhibited by ␣-amanitin
concentration of 2-3 mg/ml using an Amicon Ultra-15 centrifugal filter unit (100-kDa molecular mass cut-off) (Merck). Sample aliquots were snap frozen in liquid nitrogen and stored at Ϫ80°C prior to use.

Elongation complex preparation
The DNA scaffold used for the EC is the same as the one used for the bovine RNA polymerase II with the transcription elongation factor 5,6-dichloro-1-␤-D-ribofurano-

EM
4 l of the protein solution was applied to glow-discharged Quantifoil R2/2 gold grids (Quantifoil) and plunged into liquid ethane after blotting with a FEI Vitrobot Mark IV (FEI, Hillsboro, OR). The data were acquired on a FEI Titan Krios, operated at 300 keV, and equipped with a Gatan K2 Summit direct electron detector and a Quantum GIF. Micrographs were collected automatically with the software package EPU (FEI) at a nominal magnification of 130K (1.07 Å per pixel) in counting mode. The dose rate was 3.8 e Ϫ /pixel/s. Three images were acquired per foil hole. Each micrograph was collected with a total dose of 35 electrons per square angstrom over a 10-s exposure, fractionated into 40 frames (0.25 s each). Defocus values ranged from 1 to 3 m. Micrograph frames were aligned and corrected with MotionCor2 (19). Unless otherwise noted, data processing was performed using RELION 2.1 (20). Contrast transfer function parameters were estimated using Gctf (21). Initial 2D classes were calculated from 2,909 manually selected particles from 37 micrographs. The initial 2D classes were used as templates for autopicking. After manual inspection of all 2,049 micrographs, a total of 207,410 particles were obtained. Two rounds of 2D classification were performed, and bad particles were removed. The resulting data set of 134,512 particles was used for further refinement and focused classification refinement in 3D. The Bos taurus Pol II structure (8) (EMDa-taBank accession code EMD-3219) was low-pass filtered to 40 Å as an initial model for 3D refinement. Initial 3D refinement followed by movie processing and particle polishing yielded a reconstruction at an overall resolution of 3.4 Å (gold-standard Fourier shell correlation criterion 0.143, RELION 2.1). Focused 3D classification without image alignment was performed on the ␣-amanitinbinding pocket, the Pol II stalk (RPB4 -RPB7), and upstream DNA, followed by global 3D refinement.

Model building and refinement
Model building was based on the previously published bovine Pol II structure (8) (Protein Data Bank accession code 5FLM). The model was manually fitted in COOT (22). The ␣-amanitin molecule was taken from a S. cerevisiae ␣-amanitinbound Pol II structure (4) (Protein Data Bank accession code 2VUM). The ␣-amanitin molecule was rigid body fitted into the density. The structure was refined in real space with special restraints to the nucleic acids and ␣-amanitin using PHENIX (23).

Transcription assay
Template DNA and RNA were mixed at a molar ratio of 1:1 and annealed as described (18). The annealed DNA-RNA was mixed with Pol II-hGdown complex at a molar ratio of 1:2 and incubated at 28°C for 10 min. Nontemplate DNA was added and incubated at 28°C for an additional 10 min. The elongation complex was mixed with ␣-amanitin or buffer (control) at the same molar ratio used for the complex formation. The sample was subsequently incubated on ice for 20 min. 100 M UTP was added to both control and experimental reactions. The reaction was incubated in transcription buffer (20 mM Na-HEPES, pH 7.5, 60 mM (NH 4

Mammalian Pol II inhibited by ␣-amanitin
to the reaction. The product RNA was separated using a 20% denaturing urea polyacrylamide gel (300 V) and visualized using a GE Typhoon FLA 9500 (GE Healthcare).