Materials and methods
Micrographs from the 80SeEF2sordarin complex (Gomez-Lorenzo et al., 2000)
were recorded under low-dose conditions in a defocus range between 1.4 µm and 4.9 µm
on a Philips F20 (FEI/Philips, Eindhoven) at a magnification of about 52,000 and on a
Philips F30 (FEI/Philips, Eindhoven) at a magnification of about 38,000. The
micrographs were scanned on a Zeiss flatbed scanner (Z/I Imaging Corporation,
Huntsville, Alabama), with a step size of 14 μm and 7 μm, respectively.
The data were analyzed using the SPIDER system (Frank et al., 1996). After
automated particle selection and visual verification, the newly obtained particles were
used to increase the previous data set, which had been collected on a Philips 420 EM
equipped with a LaB6 filament. The data set was subdivided into defocus groups keeping
particles from different microscopes in different defocus groups. The data set for the final
reconstruction consisted of 63,898 particles in 43 defocus groups (Supplemental Table).
A refined CTF-corrected 3D reconstruction was calculated as described (Gabashvili et
During the refinement procedure, the relative size of the particles derived from the
three different EMs was adjusted by interpolation. In order to establish the relative size,
volumes were calculated from particles that were derived from the respective EM. The
density maps were re-scaled to different sizes and compared pair-wise using the cross-
correlation coefficient as a measure of similarity. The highest cross-correlation
coefficient was taken as an indication of matching size and the magnification of the
corresponding particles was adjusted accordingly. With the re-scaled particles the
refinement procedure was repeated. The discrepancy between the calculated
magnification and the nominal magnification of the electron microscopes was in the
range of the error of the magnification calibration with TMV virus pictures; i.e., about ±2
%. The final pixel size was estimated as 2.74 Å on the object scale. The corresponding
scale agreed best with the scale of our previous reconstructions of the yeast 80S ribosome
(Beckmann et al., 2001; Gomez-Lorenzo et al., 2000; Spahn et al., 2001a) and also
showed good agreement with the atomic models.
During the standard projection matching procedure, an angular increment of 2.0°
was used. The reconstruction was further improved by performing a local search with
smaller angular increments (0.5°). During this local search, an inverse B-factor of
110 Å-2 was applied to the reference structure to account for the decay of Fourier
amplitudes at high special frequencies due to the envelope function (Gabashvili et al.,
2000; Saad et al., 2001).
The resolution was estimated by the Fourier shell correlation with a 0.5 threshold.
Visualization and interpretation of the maps and the docked models was done using IRIS
Explorer (Numerical Algorithms Group, Inc., Downers Grove, IL), and Ribbons (Carson,
1991). Docking was done manually using O (Jones et al., 1991). The cross-correlation-
based 3D orientation search of certain parts of the cryo-EM maps in order to achieve
optimal alignment and detect conformational changes was done in SPIDER using the
commands OR 3Q and PK 3 (Spahn et al., 2001b)
We further validated the eEF2-cryo model which was derived from docking by
comparing it to eEF2-sor and apo-eEF2 with respect to volume and buried interface
between the two blocks (domains I/G’/II and domains III-V). Because the rigid body
docking should not lead to a substantial overlap between the two parts of eEF2, these
values should not change dramatically for a realistic model. In order to compare volumes
of eEF2 from the two crystal structures and the atomic model of eEF2 fitted into the cryo-
EM map (eEF2-cryo), we calculated the volume of a mask generated with a solvent
radius of 1.4 Å. The buried interface area was determined between residues 2-480 of
domains I/II, while residues 489-560, 728-759, and 763-800 were used for domains III
and V. This excludes residues 481-488, which cannot be modeled in eEF2-cryo, and
residues 760-762 which is only modeled in the crystal structure of apo-eEF2. All
calculations were done with CNS (Brunger et al., 1998).
The volumes of all three structures are 150 x 103 Å3. The buried interface area for
eEF2-cryo is 1761 Å2, compared to 1534 and 432 Å2 for apo-eEF2 and eEF2-sor,
respectively. The excellent agreement in the volumes for all three model and the fact that
the buried interface area in eEF2-cryo (1761 Å2) is comparable to that in apo-eEF2 (1534
Å2) validate the fitted eEF2. There are also no distances shorter than 2.1 Å across the
interface between domains I/G’/II and domains III-V. Furthermore, during fitting the
bond was broken between Val480 and Met481 in the domain II-III linker. After fitting,
the distance between the C-atoms of Val480 and Val489 (the first residue firmly
anchored in domain III) became 20.4 Å, compared to 22.5 and 24.0 Å in eEF2-sor and
apo-eEF2, respectively. Although the linker region cannot be fitted into the density map
at our resolution, this strongly indicates that residues 481-488 can still connect domains II
and III in eEF2-cryo, suggesting that the relative orientation of domains II and III found
by fitting is indeed possible.
Supplemental Table: Summary of the image processing parameters.
Electron microscope Number of Number of Number of Defocus range
micrographs particles defocus groups (μm)
Philips 420 38 22,166 14 1.2 – 3.2
F20 36 28,997 21 2.4 – 4.6
F30 12 12,735 8 1.0 – 3.1
Legends for Supplemental Figures:
Fig. S1: Resolution curve
Fourier shell correlation (FSC) curve for the cryo-EM map of the yeast
80SeEF2sordarin complex. FSC = 0.5 indicates 11.7 Å resolution. Some of the
structural information extends to 7.8 Å according to the more lenient 3 criterion.
Fig. S2: Accuracy of the docking of atomic coordinates from eEF2
The atomic structure of eEF2 in complex with sordarin (PDB code: 1N0U) was manually
fitted, allowing relative movements between domains, into the isolated density for eEF2
(Figure 2A, B). Once the fitting was completed, its accuracy was judged by shifting the
atomic coordinates of the altered EF-G in an arbitrary direction (“shift” along the X-axis),
and computing the correlation coefficient (“cc” on the Y-axis) between the displaced
coordinates and the cryo-EM density (Valle et al., 2003b). The result of the analysis,
shown in the plot, indicates a positional accuracy better than ± 2 Å.
Fig. S3: Ratchet-like intersubunit rearrangement in the yeast 80S ribosome
In order to represent the RSR that takes place between the 80SeEF2sordarin complex
and the POST 80S ribosome (Spahn et al., 2001a), the corresponding models for the
evolutionary conserved part of the 18S rRNAs are superposed. The 18S rRNA of the
80SeEF2sordarin complex is shown as green ribbon, the 18S rRNA of the POST 80S
ribosome (Spahn et al., 2001a) as yellow ribbon. The models are shown in stereo display
from the 60S side (a, upper panel) and from the solvent side (b, lower panel).
Fig. S4: Movement of the stalk region
Close-up on the stalk region. Elements of 25S rRNA and 60S ribosomal proteins fitted
into the 80SeEF2sordarin complex are shown as blue and orange ribbons, respectively.
The cryo-EM density of the 80SeEF2sordarin complex is shown as blue wire-mesh.
The density of the stalk region of the POST 80S ribosome (Spahn et al., 2001a) is
superposed in as yellow wire-mesh.