Structure 14

Supplemental Data

Molecular Architecture and Conformational

Flexibility of Human RNA Polymerase II Seth A. Kostek, Patricia Grob, Sacha De Carlo, J. Slaton Lipscomb, Florian Garczarek, and Eva Nogales

Figure S1. Purification and Activity of hRNAPII (A) SDS-Page of inmuno-purified hRNAPII from HeLa cell nuclei. (B) In vitro

transcriptional activity of our purified hRNAPII using the Kashlev method of transcription

initiation.

Figure S2. FSC and Angular Distribution (A) Fourier shell correlation curve indicating a resolution of 22 Å at the end of the

FREALIGN refinement. (B) Final angular distribution plot of the particles showing an

isotropic distribution of orientations.

Figure S3. 3D Variance and Docked Crystal Structure, Insertions and Deletions

The 3D variance (yellow mesh) is superimposed onto the yeast crystal structure docked

into the EM density map (PDB entry 1Y1V). Deletions in the hRNAPII sequence are

represented in white with green outline. Residues not present in the yeast crystal

structure and insertions in hRNAPII are shown as red dashes. Left and right show front

and side views, respectively.

Figure S4. 3D Variance and Docking Left and right show front (as in Fig. 4B) and back views, respectively, of the yRNAPII

crystal structure docked into the EM hRNAPII map. The latter highlights the position of

pore 1, the Rpb6 N-terminus and the CTD linker. The 3D variance map is shown as a

yellow density.

Figure S5. Rigid Body Docking of the hRNAPII Homology Model into the

Cryo-EM Density Map

The same views of the EM density map from human RNAPII are shown as in Fig.

2 B (mesh). The homology model of hRNAPII (based on its sequence alignment

with yRNAPII —see supplementary Materials and Methods—) was docked

independently into the cryo-EM density using rigid body docking. The resulting

position for the core complex is identical. The only noticeable dissimilarity is in

the stalk, with a different size (smaller) and forming a different angle with the core

of hRNAPII than in yeast. Also visible are the models of most of the flexible loops

and some short termini.

Table S1

Location on the yRNAPII sequence of the insertions and deletions in the

hRNAPII amino-acid sequence (see Fig. 2B). The number of residues is

indicated (with h for human, y for yeast if the numbers are different) as well as

the location relative to the secondary structure of the different subunits, as

defined in Cramer et al. 2001 (core complex) and Meka et al. 2005 (Rpb4/7

complex). The regions closest to the main high variance areas determined by our

analysis are also mentioned, with their corresponding number as in Fig. 4. Yellow

text correspond to regions highlighted in the text concerning the contact of the

stalk with the core enzyme.

Subunit Deletions in hRNAPII

(number of residues)

Insertions in hRNAPII

(number of residues)

Not localized in

yRNAPII

(number of residues)

Regions of high

variance (variance

region #)

Rpb1

44-45 (2) Zipper

1187-1188 (1, 4 total,

but 3 of them are not

localized in yeast Xtal

structure) loop a40-b29

3-4 (4), N-ter

34-35 (2), Zipper

129-130 (2) loop a3-a4

155-156 (11) loop a4-b3

583-584 (8) loop b20-

b21

1286-1287 (5) loop b32-

b33

186-195 (8) loop b4-b5

1176-1187 (10) loop

a40-b29

1243-1254 (10 y + 6 h)

loop b31-a43

1456-1979 - part of linker

and CTD, C-ter

Clamp head (1)

Lid (2)

Rudder (2)

Switch 2 (3)

a38-a39 (4)

a43-b32 (close to

interaction with Rpb9

and Rpb2) (4)

b30-b31 (4)

a41-a42 (4)

Linker and CTD (6)

Rpb2

231 (1) loop b7-b8

642-647 (6) loop b21-

a16

668 (10 total, but not all

localized) loop a16-a17

268-269 (6) loop b9-b10

20 (19) N-ter

Protrusion 70-90 (19) loop a2-b1

335-345 (9) a8

437-446 (8) a11-a12

Fork loop 1 (2)

Protrusion (2)

lobe

external 1

716-733 (17) loop a19-

b24

1178 (1) b44-b45

1220-1224 (5) C-ter

134-164 (19 y, 5 h)

loop b2-b3

External 1 668-678 (9 y, 0 h)

715-722 (6 y, 1 h) loop

a19-b24

Wall flap loop 919-933 (13)

b7-b8 (1)

b10-b11 (1)

External 1

Wall flap loop (5 and 2)

Rpb3

196-199 (4) b10-b11

216 (1) b10-b11

268 (1 localized, 44 total,

but most were not

localized) Tail, C-ter

3 (1) N-ter

76-77 (1) a2-b5

123-124 (4) b7-b8

3 (1) N-ter

268 (44) C-ter Tail

b7-b8

Tail

Rpb4

4-7 (4) N-ter

42-77 (6) a1-a2 loop

118-136 (9) a1-a2 loop

Tip

And/or interface with

Rpb7

Rpb5 7-8 (2) N-ter

68-73 (6) a4, loop a4-b2

48-49 (2) a3-b1

122-123 (1) a6

a8-b6 – contact with

Rpb1 (foot)

b7-b8 – contact with

Rpb1 (foot)

Rpb6 72-74 Tail (N-ter) …72 (71 y, 44 h) Tail (N-

ter)

Tail (5)

Possibly contacts with

Rpb1 and Rpb5 (5)

Rpb7 57-58 (2) b3 141-142 (2) b2-b3

171 (1 localized) C-ter

b3-b4 loop – interface

with Rpb4 ?

Rpb8 32 (1) b3-b4 loop

82-87 (6) b5-b6 loop

2 (1) N-Ter

18-19 (2) b3-b4

105-106 (5) b7-b8

139-140 (1) b9-b10

146 (3 ) C-ter

63-76 (12 ) b5-b6 loop b5-b6 loop

Rpb9 113-120 (10) C-ter 2 (10) N-ter

70-71 (1) b5-b6 loop C-ter (2) C terminal

105-106 (2 ) b7-b8 loop

Rpb10 65 (5y 3h) Cter N terminal (contacting

Rpb3)

Rpb11 114 (6y 3h) C-ter

Rpb12 25 (24y 12h) b3-b4 (interaction with

Rpb2)

Supplemental Experimental Procedures

hRNAPII Purification and Activity

Human RNAPII was purified as previously described [1, 2]. In brief RNAPII was

extracted from HeLa nuclear pellets through sonication and precipitated with a 42%

ammonium sulfate cut. The pellets were resuspended and dialyzed to 0.15 M

ammonium sulfate and placed on a DEAE52 anion exchange column. Subsequent to

thorough washing, RNAPII was eluted with a buffer containing 0.4 M ammonium sulfate

and assayed by western blot. The fractions containing RNAPII were pooled and this

eluate was dialyzed into a buffer containing 0.2 M ammonium sulfate. This was

subsequently placed over a protein G affinity column containing 8WG16 antibodies from

NeoClone. After several high-salt washes the RNAPII was eluted four times with buffer

containing a tri-heptapeptide repeat of the CTD. These fractions were then pooled and

dialyzed against a buffer containing 0.15 M ammonium sulfate and then placed on a

DEAE-5PW ion exchange column to separate the different phosphorylation states of the

protein complex. Fractions were immediately dispensed into 5 µl aliquots and frozen in

liquid nitrogen for later use. Transcriptional activity was measured using a promoter-

less RNA polymerase initiation system prepared as described by Kashlev and

coworkers [3]

EM Sample Preparation and Data Collection

Fractions of 50 ug/mL hRNAPII were thawed just prior to EM use, diluted 5 fold

and 5 µL of this dilution was applied for 30 seconds to a continuous carbon-coated, 400-

mesh copper grid that had been glow discharged. The grid was then placed on a 100 µL

drop of stain that consisted of a saturated solution of ammonium molybdate neutralized

to a pH of 7.2. with 10 N NaOH [4]. After 30 seconds of exposure to the stain the grid

was mounted on a plunger, blotted to a thin layer, air-dried for 1-2 sec and finally

vitrified in liquid ethane. EM data was collected on an FEI CM200-FEG transmission

electron microscope at an acceleration voltage of 200 kV, with a calibrated

magnification of x50280, and using a Gatan 626 cryo-specimen holder (Gatan Inc.,

Warrendale, PA, USA) while keeping the specimen temperature at approximately –

180˚C. Images were taken under low-dose conditions (17 e-/Å2) with a defocus range of

1.2 µm to 3.6 µm and recorded on Kodak SO163 plate films. The quality of the

micrographs was checked by visual inspection for astigmatism and drift. The best 19

micrographs were digitized on a Nikon Super CoolScan 8000 with a 6.353 µm raster

size, resulting in a pixel size of 1.25 Å. Transmission values from the scanner were

converted to optical density with tm2od, an in-house convenience script that employs

the proc2d module of EMAN (see

http://cryoem.berkeley.edu/~slaton/bash/tm2od.shtml). Particles were then decimated to

2.5 Å/pixel.

Image Processing

The EMAN software package [5] was used to manually establish particle

coordinates (boxer) and window 9225 images at a size of 120 x 120 pixels (batchboxer).

Two methodologies were employed to estimate the contrast transfer function (CTF) for

each particle. The first method used the ctfit module of EMAN [5] to estimate the

defocus for an entire micrograph with subsequent assignment of CTF parameters to the

corresponding particles using IMAGIC [6]. This procedure corrected the individual

image CTFs by flipping the phase only, without regard for the amplitudes. The second

method relied on the ctftilt program of FREALIGN to estimate defocus and astigmatism

for each particle [7]. The program determines the specimen tilt parameters by

measuring the defocus at a series of locations on the image while constraining them to

a single plane. This information was stored and used for refinement in FREALIGN.

The individual images were arithmetically normalized, iteratively centered, and

then multiplied by a Gaussian blurred circular mask, with a radius 90% of the total

image radius. 2-D analysis was performed using the program IMAGIC [6]. Particles

were subjected to multivariate statistical analysis (MSA) and hierarchical ascendant

classification (HAC).

Initial Euler angle assignment was performed with SPIDER [8] using a projection

matching strategy. The particles used for 2-D analysis were first converted to SPIDER

format using IMAGIC (em2em). The initial 2-D analysis indicated that the

crystallographic model of the yeast homologue of human RNAPII (PDB coordinates

1Y1V) could be used as an initial reference. The yRNAPII crystal structure was low-

pass Fermi-filtered to a spatial frequency of 1/50Å-1 thus allowing alignment of the

overall shape without bias from high-resolution frequencies. A gallery of 2-D references

was generated from the filtered yeast model by reprojecting it with an angular step

(delta theta) of 15 degrees. Then the individual experimental particles images were

cross-correlated to these references for alignment and Euler angular assignment [9].

After translational and rotational alignment of the particles a 3-D volume was calculated

by back projection using the assigned Euler angles. This procedure was iterated using

the volume derived from the previous round to generate reprojections. The theta angle

step size was decreased after several rounds of projection matching were performed for

each theta step.

Refinement and full CTF correction of the SPIDER model was performed with the

image processing software package FREALIGN [10]. This program refines the x, y

shifts and the three Euler angles for each particle, and performs a CTF correction on the

generated volume in Fourier space. Previously determined particle parameters are

used as initial input parameters. fwrap, an in-house developed package, was used to

divide the dataset across the processors of a Linux cluster and automatically manage

the multiple rounds of FREALIGN refinement (see

http://cryoem.berkeley.edu/~slaton/emperl/fwrap/index.shtml). CTF correction

refinement was performed with an initial resolution range of 200 to 40 Å (10 rounds)

with all 9225 particles. Subsequently the phase residual cut-off parameter was lowered

from 90 to 60 to obtain the best matching 6238 particles. Further refinement using

these particles, up to round 63, was performed with a resolution range of 200 to 10 Å.

The resolution of the final reconstruction was estimated from the Fourier shell

correlation function obtained by comparing two independent reconstructions, which

were generated by splitting randomly the data set in half [11]. The two data sets were

independently reconstructed and the resolution was given according to the 0.143 cut-off

in the Fourier shell correlation curve [12] (Supplementary Figure 1). The density

threshold was calculated with IMAGIC (threed-sexy). This program calculates the

protein mass covered by the above threshold voxels, assuming a protein density of

0.844 Daltons/ Å3.

To assess the potential conformational heterogeneity, 2-D analysis was again

performed with IMAGIC on particles from angular distribution groups generated during

projection matching. Class averages were generated by averaging particles in angular

groups at a theta step of 15˚. Particles in each class were subjected to MSA and MRA

generating initially up to 5 sub-classes. However visual inspection indicated that the

variability in the dataset was represented by two main sub-classes, with the others

being underrepresented.

3-D variance was calculated as previously described [13, 14]. Briefly, 500

bootstrap versions of the dataset were picked randomly with replacement from our

original dataset, leading to as many 3D reconstructions using the same 3D alignment

parameters, and low-pass filtered at 1/30Å resolution. Some background “noise

particles” were extracted from the area around each particle of the dataset and treated

in the same way. An estimate of the 3D structure variance can be calculated using the

bootstrap variance σB2 between the B=500 reconstruction obtained from the resampled

data particles and the average of the “noise” variance σ Back2 calculated with the same

method (command VA 3R in SPIDER):

σ Struct2 =K(σ B

2 −σ Back2 ) ,

where K is the number of particles in the original data set.

Targeted Classification and Reconstruction

The different regions of the 3D variance were selected using spherical masks

centered at the highest variance peaks in SPIDER. They were then individually

projected in quasi-evenly distributed directions (15˚ spacing), identical to the directions

of the particle “angular groups”, to generate a series of 2D masks. 2D classification of

the particle data belonging to each angular group was performed within the

corresponding masks as in [14, 15]. Classification for the high variance region between

the clamp and the lobe (region 1, Fig 4) gave the clearest result, with two main classes

corresponding to a higher or lower density within the mask. The data was partitioned

accordingly for each view into two groups, “closed” or “open” clamp. Two 3D

reconstructions were obtained from the particle groups, which were then refined

simultaneously against the entire dataset; the final particle assignment to either

conformation was determined by the highest cross-correlation coefficient. An additional

refinement of the alignment parameters was performed for each separate group in

FREALIGN [10]. The 3D variance was calculated again for each group of particles,

showing a notably reduced variance level after partition of the data.

Docking of the Crystallographic Data

The atomic coordinates of the 12-subunit yeast RNAPII (PDB 1Y1V,

Kettenberger et al., 2004) were initially docked manually into the EM density map of the

human polymerase in Chimera (see below). The initial fit was refined using the rigid-

body docking program colores from the SITUS package [16, 17]. This program

performed an initial extensive search, with an angular step size of 15˚, followed by an

off-lattice Powell optimization of the fit. Rigid-body docking of the crystallographic data

into the two conformations obtained from targeted classification was also attempted.

The resulting best fit was obtained in approximately the same position as before

partition of the data, with lower cross-correlation coefficients. While most of the crystal

structure matched closely the envelope from the cryo-EM data, some domains seemed

to adopt different positions. Manual docking of the clamp and jaw-lobe domains gave a

better fit in those areas of the structure (Fig. 5, bottom row). The docking of the TFIIB-

yRNAPII core complex gave a slightly better fit for conformation 1, but this structure

lacks the stalk domain.

Homology Model of RNA Polymerase II from Human

A three-dimensional homology model of the human RNAPII was produced using

MODELLER 8v2 [18]. The sequence of the target protein was aligned to the

yeast RNAPII sequence using the program ClustalW [19] and the known X-ray

structure 1Y1V [20] (without TFIIS) was used as a template for the determination

of the tertiary structure. Prior to determination of the homology model, all amino

acids which were not present in the X-ray structure (loops and termini) were

removed from the yeast sequence. The MODELLER default script (model-

default.py) was used and extended with the “env.io.hetatm = True” command to

keep all ions (1 x Mg2+, 9 x Zn2+) within the homology model. Ten models were

created and the one with the lowest value of the MODELLER objective function

was chosen. Termini of the homology RNAPII structure corresponding to regions

which are not resolved within the yeast X-ray structure and which are longer than

15 residues are removed from the homology model (chain A 1486-1970(1486-

1970), chain B 1971-1986 (1-15), chain F 3772-3818 (1-46)).

The resulting human model (available upon request) was docked into the cryo-

EM density map using the SITUS rigid body docking command colores and

represented in the same position of the EM map as the docked yeast model in

figure S5 (1Y1V, Fig. 2B).

Volume Rendering

All the volumes represented were filtered to 22 Å resolution, as determined by

the FSC criterion at 0.143, using a cosine type filter. The density threshold was

calculated to include a protein volume corresponding to 517kDa, the molecular mass of

holo-hRNAPII. The 3D density maps and atomic structures were rendered with the

UCSF Chimera package from the Computer Graphics Laboratory, University of

California, San Francisco ([21] , supported by NIH P41 RR-01081).

Supplemental References 1. Thompson, N.E., Aronson, D.B., and Burgess, R.R. (1990). Purification of

eukaryotic RNA polymerase II by immunoaffinity chromatography. Elution of

active enzyme with protein stabilizing agents from a polyol-responsive

monoclonal antibody. J Biol Chem 265, 7069-7077.

2. Maldonado, E., Drapkin, R., and Reinberg, D. (1996). Purification of human RNA

polymerase II and general transcription factors. Methods Enzymol 274, 72-100.

3. Kireeva, M.L., Komissarova, N., Waugh, D.S., and Kashlev, M. (2000). The 8-

nucleotide-long RNA:DNA hybrid is a primary stability determinant of the RNA

polymerase II elongation complex. J Biol Chem 275, 6530-6536.

4. Adrian, M., Dubochet, J., Fuller, S.D., and Harris, J.R. (1998). Cryo-negative

staining. Micron 29, 145-160.

5. Ludtke, S.J., Baldwin, P.R., and Chiu, W. (1999). EMAN: semiautomated

software for high-resolution single-particle reconstructions. Journal of Structural

Biology 128, 82-97.

6. van Heel, M., Harauz, G., Orlova, E.V., Schmidt, R., and Schatz, M. (1996). A

new generation of the IMAGIC image processing system. J Struct Biol 116, 17-

24.

7. Mindell, J.A., and Grigorieff, N. (2003). Accurate determination of local defocus

and specimen tilt in electron microscopy. J Struct Biol 142, 334-347.

8. Frank, J., Radermacher, M., Penczek, P., Zhu, J., Li, Y.H., Ladjadj, M., and Leith,

A. (1996). SPIDER and WEB - Processing and visualization of images in 3D

microscopy and related fields. J. Struc. Biol. 116, 190-199.

9. Penczek, P.A., Grassucci, R.A., and Frank, J. (1994). The ribosome at improved

resolution: new techniques for merging and orientation refinement in 3D cryo-

electron microscopy of biological particles. Ultramicroscopy 53, 251-270.

10. Grigorieff, N. (1998). Three-dimensional structure of bovine NADH: ubiquinone

oxidoreductase (complex I) at 22 Å in ice. J. Mol. Biol. 277, 1033-1046.

11. Frank, J. (1996). Three-dimensional electron microscopy of macromolecular

assemblies (San Diego: Academic Press).

12. Rosenthal, P.B., and Henderson, R. (2003). Optimal determination of particle

orientation, absolute hand, and contrast loss in single-particle electron

cryomicroscopy. J Mol Biol 333, 721-745.

13. Penczek, P.A., Yang, C., Frank, J., and Spahn, C.M. (2006). Estimation of

variance in single-particle reconstruction using the bootstrap technique. J Struct

Biol.

14. Grob, P., Cruse, M.J., Inouye, C., Peris, M., Penczek, P.A., Tjian, R., and

Nogales, E. (2006). Cryo-electron microscopy studies of human TFIID:

conformational breathing in the integration of gene regulatory cues. Structure 14,

511-520.

15. Penczek, P.A., Frank, J., and Spahn, C.M. (2006). A method of focused

classification, based on the bootstrap 3D variance analysis, and its application to

EF-G-dependent translocation. J Struct Biol 154, 184-194.

16. Wriggers, W., and Birmanns, S. (2001). Using situs for flexible and rigid-body

fitting of multiresolution single-molecule data. J Struct Biol 133, 193-202.

17. Chacon, P., and Wriggers, W. (2002). Multi-resolution contour-based fitting of

macromolecular structures. J Mol Biol 317, 375-384.

18. Sali, A., and Blundell, T.L. (1993). Comparative protein modelling by satisfaction

of spatial restraints. J Mol Biol 234, 779-815.

19. Thompson, J.D., Higgins, D.G., and Gibson, T.J. (1994). CLUSTAL W: improving

the sensitivity of progressive multiple sequence alignment through sequence

weighting, position-specific gap penalties and weight matrix choice. Nucleic Acids

Research 22, 4673-4680.

20. Kettenberger, H., Armache, K.J., and Cramer, P. (2003). Architecture of the RNA

polymerase II-TFIIS complex and implications for mRNA cleavage. Cell 114, 347-

357.

21. Pettersen, E.F., Goddard, T.D., Huang, C.C., Couch, G.S., Greenblatt, D.M.,

Meng, E.C., and Ferrin, T.E. (2004). UCSF Chimera--a visualization system for

exploratory research and analysis. J Comput Chem 25, 1605-1612.