Structure, Volume 23

Supplemental Information

Near-Atomic Resolution for One State of F-Actin

Vitold E. Galkin, Albina Orlova, Matthijn R. Vos, Gunnar F. Schröder, and Edward H. Egelman

Supplementary Figures

Supplementary Figure 1

Supplementary Figure 2

A" B"

C" D"

Supplementary Figure 3

Supplementary Figure 4

Supplementary Figure 5

Supplementary Figure 6

Supplementary Figure 7

0 2 4 6 8 10 12 14 16 18Refinement cycle #

0.525

0.53

0.535

0.54R

free

(4.4

-4.5

Å)

Supplementary Figure 1, related to Figure 1. Preparation of grids for cryo-EM involves large

forces on actin filaments. A) An image of a field of filaments showing a common orientation,

almost certainly due to fluid flow in one direction during the blotting that occurs prior to freezing.

B) Breaks (black arrows) can occasionally be observed in actin filaments due to the mechanical

forces that are placed on these filaments from the fluid flow. The sharp bend (white arrow) is

inconsistent with the known rigidity of the actin filament and thermal flexing, and must also arise

from mechanical forces.

Supplementary Figure 2, related to Figure 1. Out-of-plane tilt of segments correlates with the

perceived thickness of the ice. Typical images of actin in what were perceived as thin ice (A) and

thick ice (C), both using lacey carbon grids imaged with a Titan Krios. In the thin ice, actin

filaments (A, arrows) can be frequently found that are anomalously straight (Galkin et al., 2012)

due to the large mechanical forces that are present. Histograms for out-of-plane tilt from thin ice

(B) and thick ice (D) have been fitted with Gaussians, and the distributions show that the amount

of out-of-plane tilt is increased when the ice is thicker, consistent with the fact that filaments

have a greater freedom to deviate away from the perpendicular to the beam when the film

containing the filaments is thicker prior to freezing.

Supplementary Figure 3, related to Figure1. Comparison of twist distributions of F-actin in

thinner ice (A) and thicker ice (B) were calculated using cross-correlation of images with

projections from a set of models having a range of twist from 162° to 170° with a step of 0.5°.

The reliability of the sorting was established by computing averaged power spectra from

segments having smaller (red bins) or larger (blue bins) angles. A) Twist distribution based on a

sorting of 55,052 segments in thinner ice, with the two classes used to calculate power spectra

marked in red (n=8,727) and blue (n=11,900). Comparison of the two power spectra is shown in

insert where positions of the n=4 layer line are marked with white arrows. B) Twist distribution

based on a sorting of 89,160 segments in thicker ice, with the two classes used to calculate power

spectra in red (n=6,770) and blue (n=8,995). Comparison of the two power spectra is shown in

insert where positions of the n=4 layer line are marked with white arrows.

Supplementary Figure 4, related to Figure 2. (A) The Fourier Shell Correlation (FSC)

between reconstructions from half data sets, with each half set containing from 6,250 to 25,000

segments. Completely independent data sets were generated, so that there was no overlap of

boxes (as opposed to the extensive overlap that would exist if one took odd/even images to create

the two sets). (B) The resolution (using the FSC=0.143) found in (A) is plotted versus the

number of segments, using a log10 scale. This can be fit nicely by a straight line (dotted red line)

as observed by others (Stagg et al., 2014). The estimate for the resolution of 50,000 segments

(approximately the number of segments in the full data set) is 0.223 (vertical black line) or 1/(4.5

Å).

Supplementary Figure 5, related to Figure 2. The FSC between the atomic model for the

canonical state of F-actin and the actual reconstruction. The FSC is 0.5 at 4.66 Å.

Supplementary Figure 6, related to Figure 4. The FSC for the T2 reconstruction. The data

were split into two halves with almost no overlap of boxes. Each subset was iterated

independently for five cycles, and then aligned to generate the FSC curve.

Supplementary Figure 7, related to Experimental Procedures. Rfree value during the final

refinement and model building cycles. Simulated annealing refinements were performed with

CNS using DEN restraints. Structure factors between 4.4-4.5 Å were chosen for the calculation

of Rfree.

Oda et al., Nature (2009) Fujii et al., Nature (2010)

Current manuscript

Longitudinal SD3-up→SD4-low

200-208→283-294 241-247→283-294

205→286 241→324 244→290 245→322

241→324 244→325 245→322

Longitudinal SD3-up→SD2-low

61-65→283-294 40→169 61→167 62→288 64→166

38→169 44→168 61→167 62→288 64→166

Longitudinal SD1-up→SD2-low

38-49→139,140,143 38-49→346,351,374 43→346 44→375

45→143

44→143 47→352

Lateral HP plug→SD2-across the strand

265-271→39-42 265→40 268→40 270→39

N/A

Lateral HP plug→SD3-across the strand

265-271→170-174 267→173

267→173

Lateral HP plug→SD4-across the strand

265-271→201-203 265-271→285-286

none 271→201/202

Lateral SD4→SD1

191-199→110-115 191→110 194/195→110 195→113

194→111 195→113

Lateral SD4→SD3

none 194→177

none

Supplementary Table I, related to Figure 3 The residues predicted to be involved in the subunit-subunit interface are shown for the fiber diffraction model of Oda et al. (Oda et al., 2009), the cryo-EM model of Fujii et al. (Fujii et al., 2010) and this paper. Differences between our model and that of Fujii et al. are shown in red.

All-Atom Contacts

Clashscore, all atoms: 3.24 97th percentile* (N=1784, all resolutions) Clashscore is the number of serious steric overlaps (> 0.4 Å) per 1000 atoms.

Protein Geometry

Poor rotamers 65 20.44% Goal: <1% Ramachandran outliers 23 6.17% Goal: <0.05% Ramachandran favored 314 84.18% Goal: >98% MolProbity score^ 2.78 32nd percentile* (N=27675, 0Å - 99Å) Cβ deviations >0.25Å 23 6.63% Goal: 0

Supplementary Table II, related to Supplementary Experimental Procedures

Supplemental Experimental Procedures

Model building of F-actin

The starting structure for the model building was based on the X-ray structure 2BTF.PDB

(Schutt et al., 1993), which we refined against an earlier EM reconstruction of lower resolution

using the program DireX (Schröder et al., 2007). A short six monomer filament was then

extracted from this model which we used to start a 88 ns MD simulation at a temperature of 300

K in explicit water using Gromacs (Hess et al., 2008) with the Amber-SB99-ILDN force field.

From the resulting trajectory 230 equally spaced frames were extracted. One of the two central

protomers in the short 6-mer filament was extracted from each frame yielding 230 protomer

structures. These protomer structures were then refined against a masked protomer density from

our 4.7 Å EM reconstruction using DireX with cross-validation (free interval 4.0–4.5 Å). The

best 100 structures with the lowest Cfree values were averaged. The geometry of this averaged

structure was then optimized by energy minimization with CNS (Brunger et al., 1998) and again

refined against the EM density with DireX.

A minimal filament model was built by surrounding one protomer with four others using

the helical symmetry of the filament such that this single protomer makes all possible

interactions within the filament. Several rounds of manual model building and correction with

Coot (Emsley et al., 2010), followed by simulated annealing structure refinement with CNS with

DEN restraints, were performed. For the structure refinement in CNS, the EM density for the

five protomers was masked with a soft mask and then converted to an HKL structure factor file

using the CNS task em_map_to_hkl.inp.

NCS restraints were used during the refinement to keep the protomers similar to each

other. The MLHL target function was used while the standard X-ray scattering factors were

replaced by electron scattering factors in CNS. The final refinement yielded an R-value of 36.1%

(Supplementary Fig. 7) and the grouped and restrained B-factor refinement yielded an average

atomic B-factor of 111 Å2. The Molprobity analysis is shown in Supp. Table II.

Supplemental References Brunger, A.T., Adams, P.D., Clore, G.M., DeLano, W.L., Gros, P., Grosse-Kunstleve, R.W.,

Jiang, J.S., Kuszewski, J., Nilges, M., Pannu, N.S., et al. (1998). Crystallography & NMR

system: A new software suite for macromolecular structure determination. Acta

CrystallogrDBiolCrystallogr 54, 905-921.

Emsley, P., Lohkamp, B., Scott, W.G., and Cowtan, K. (2010). Features and development of

Coot. Acta crystallographica Section D, Biological crystallography 66, 486-501.

Fujii, T., Iwane, A.H., Yanagida, T., and Namba, K. (2010). Direct visualization of secondary

structures of F-actin by electron cryomicroscopy. Nature 467, 724-728.

Galkin, V.E., Orlova, A., and Egelman, E.H. (2012). Actin filaments as tension sensors. Current

Biology 22, R96-R101.

Hess, B., Kutzner, C., van der Spoel, D., and Lindahl, E. (2008). GROMACS 4:   Algorithms for

Highly Efficient, Load-Balanced, and Scalable Molecular Simulation. Journal of Chemical

Theory and Computation 4, 435-447.

Oda, T., Iwasa, M., Aihara, T., Maeda, Y., and Narita, A. (2009). The nature of the globular- to

fibrous-actin transition. Nature 457, 441-445.

Schröder, G.F., Brunger, A.T., and Levitt, M. (2007). Combining efficient conformational

sampling with a deformable elastic network model facilitates structure refinement at low

resolution. Structure 15, 1630-1641.

Schutt, C.E., Myslik, J.C., Rozycki, M.D., Goonesekere, N.C.W., and Lindberg, U. (1993). The

structure of crystalline profilin:·-actin. Nature 365, 810-816.

Stagg, S.M., Noble, A.J., Spilman, M., and Chapman, M.S. (2014). ResLog plots as an empirical

metric of the quality of cryo-EM reconstructions. J Struct Biol 185, 418-426.