.JOURNAL
OF STRUCTURAL
BIOLOGY
108,
62-68 (199%
Effects of Radiation MICHAEL W. M. Keck Center One Baylor Plaza,
for Computational Houston, Texas
Damage with 400-kV Electrons Hydrated Actin Bundles
F. SCHMID, JOANITA Biology, 77030; and Institute
JAKANA,
PAUL MATSUDAIRA,”
obtained from a few bundles that are tilted about an axis parallel to the bundle axis. Radiation damage limits the number of usable tilt images from each bundle. There have been theoretical suggestions that 400-kV electrons may provide an advantage over lOO-kV electrons from the point of view of radiation damage (Glaeser, 1985; Chiu et al., 1986). However, there has always been a counterargument that this would not be true because the ratio of elastic to inelastic (damage-producing) scattering is unaffected by the electron energies (Isaccson, 1977). Most of the radiation damage studies of crystals have been done by observing the fading of electron diffraction patterns as a function of cumulative electron dose (Glaeser and Taylor, 1978; Jeng and Chiu, 1984). These studies imply that the contrast of the images in a damage series decreases. Unwin (1974) has observed stain migration in images of stacked disk aggregates of tobacco mosaic virus coat protein as a function of dose. However, no quantitative information has yet been obtained about the effects of radiation on the phases, particularly in frozen, hydrated specimens. In the present study, we have undertaken a crystallographic analysis of structural changes in a characteristic image view of frozen, hydrated actin bundle taken at different accumulated doses with 400-kV electrons at a specimen temperature of - 168°C. We are taking the practical approach that the quality of the phases is the most important criterion in evaluating the reliability of image information. Data on the effects of damage allow us to set a limit on the number of usable images in a tilt series from a single bundle.
INTRODUCTION
The acrosomal process from Limulus polyphemus is a long cellular extension made up of actin and bundling proteins (Tilney, 1975). The native structure of the true discharge has recently been studied by electron cryomicroscopy and crystallographic analysis, which showed it to be crystalline with continuously varying views along the bundle axis (Schmid et aZ., 1991). One of the important biological questions about this intracellular structure is the mode of molecular interaction among the proteins which governs its formation, stability, and flexibility (Schmid et al., 1991). Because the thickness of the actin bundle is about 1000 A, 400 kV electron radiation may be useful for recording images for high-resolution three-dimensional structure determination. Although the apparent image contrast with 400 kV electrons is not as high as that at 100 kV, high-resolution features can be readily obtained (Schmid et al., 1991). In order to retrieve the three-dimensional structure of the acrosomal bundle with no a priori assumption of helical symmetry, we plan to combine images at various tilt angles. In the initial stage where we are merging low-resolution threedimensional data, our task is easier if the data are
MATERIALS Electron
AND METHODS
Cryomicroscopy
The acrosomal process was purified from the sperm of horseshoe crab, Limulus polyphemus, as described (Schmid et al., 1991). Frozen, hydrated actin bundles were prepared on freshly carbon-coated holey 400-mesh copper grids. The specimens were transferred to a Gatan 626 cryoholder and then into a JEOL 62
Inc. reserved.
Baylor College of Medicine, of Biology, Massachusetts
19, 1991, and in revised form December 9, 1991
Electron images can be used to provide amplitudes and phases for the structural determination of biological specimens. Radiation damage limits the amount of structural information retrievable by computer processing. A 400-kV electron microscope was used to investigate radiation damage effects on frozen, hydrated actin bundles kept at -168°C. The quality of phases within and among images in a damage series was evaluated quantitatively out to 16 A resolution. It was found that the phases of structure factors with good signal-to-noise ratio (IQ s 4) can be reliably retrieved from images taken at a cumulative dose of at least 25 electrons/A’. o lwz Academic PITSS, IX.
1047-8477192 $3.00 Copyright 0 1992 by Academic Press, All rights of reproduction in any form
AND WAH CHIU
and Verna and Marrs McLean Department of Biochemistry, *Whitehead Institute for Biomedical Research and Department of Technology, Cambridge, Massachusetts 02142
Received September
on Frozen,
ELECTRON
DAMAGE
OF ACTIN
4000EX electron microscope. To ensure maximum stability of the cryoholder, 1.5 hr elapsed between transferring into the column and recording of the images. Untilted specimens were examined at 400 kV under low dose rate conditions at a specimen temperature of - 168°C (Brink and Chiu, 1991). The dose rate on each image was measured with a picoammeter recording the current on the fluorescent screen. The picoammeter was calibrated as described by Brink and Chiu (1991). Radiation
Damage
Series
Usually seven or more successive images from the same bundle area were recorded keeping all microscope settings constant throughout the radiation damage series. Each image was recorded at nominal x 20 000 magnification with a specimen dose of 3 electrons/A’. The cumulative dose was less than 25 electrons/ A” in the last image. The defocus was aimed at less than 1 pm under-focus. All images were recorded on Kodak SO-163 films developed for 12 min in 100% D19, and fixed 10 min in Kodak fixer at 20°C. The optical density of the image of the ice-embedded area is around 0.7. Images of actin bundles displaying the “h01” view were evaluated for their quality in an optical diffractometer. Image
Processing
images selected above were scanned on a PDS 1OlOM microdensitometer on a 16-pm raster with a scan interval equivalent to 8 A/pixel. Identical actin bundle areas in each image were boxed 612 pixels long by 204 pixels wide. A Fourier transform of the boxed area was calculated and the reciprocal lattice parameters were refined iteratively. A list of amplitudes and phases of the structure factors along with their IQ after proper phase shift was calculated from each image in the damage series (Robinson et al., 1988). IQ is a parameter which measures the signal-to-noise ratio of structure factors and ranges from 1 to 8, with 1 being the best (Henderson et al., 1986). Phase
Comparisons-Zntraimage
Symmetry
The presence and quality of various symmetry elements in a crystal image were evaluated for the calculated structure factors as previously described (Schmid et al., 1991). Except for the choice of whether to include a reflection or not based on its IQ, no other weighting scheme was used, and all reflections were considered out to the resolution limit as defined by the Sampling Theorem. Phase
Comparisons-Merging
the Damage
Series
Merging of the images in the damage series was carried out using the program ORIGTILT (Henderson et al., 1986). The images were merged using the symmetry of plane group pl. This ensures that the same reflections, not symmetry mates, are used to compare images in the damage series. As in the above phase calculations, after the choice of reflections based on IQ, no other weighting scheme was used. In both of these comparisons, the choice of reflections based on -heir IQ was done in two ways. In the first, the IQ of the reflection an the first image was used in the selection. Thus, the same reflections can be compared throughout the analysis, no matter what their IQ was after the first image. The other way of choosing reflections is to make the IQ selection on every image. This method is more like the one actually used in merging of images in a tilt series and, for instance, some of the reflections do not appear for every image. The results of the merging were used to shift the phase origins of the images until they were coincident, and two-dimensional reconstructions were carried out on several members of the damage series. RESULTS
Seven images of an actin bundle were obtained and processed. Figure 1 shows the first, third, fifth,
BUNDLE
AT
400 KV
63
and seventh images of the actin bundle in this damage series. The images show an area which includes the h01 view with a = b = 147 A, c = 762 A (Schmid et al., 1991). Slightly different orientations appear above and below the h01 zone. An area of each image was digitized and windowed according to features in the micrograph to ensure that exactly the same region of the bundle was processed for all members of the damage series. Almost 6 unit cells (4570 A) along the length of the bundle were included for each image. The actual magnification of the images was found to be x 15 750 by calibration with the unit cell parameters found for the actin bundle, leading to a scan size of about 7.5 A/pixel. The cumulative dose on the images in Figs. la-ld was 3, 10, 16, and 23 electrons/A’, respectively. The windowed area was transformed and the lattice parameters were independently refined for all seven images. The results of the lattice refinement are shown in Table I. Below each image is an IQ representation (Henderson et al., 1986) of the Fourier transforms of the bundle images after lattice refinement. IQ 1, 2, and 3 have the largest circles (S/n > 2.71, IQ 4 and 5 are represented by intermediate sized circles (2.7 > Sln > 1.61, and IQ 6 and 7 are smallest (1.6 > S/n > 1.0). No unbending or other corrections were applied to the data. Predictably, radiation decreases the amplitudes of the reflections (Jeng and Chiu, 1984). Figure 2 shows the fall-off of the amplitudes of the reflections in the Fourier transforms of the images. The average amplitude of all reflections (see below) in each resolution shell is plotted relative to its average amplitude in the first image, so amplitudes in all shells start at 100% in the first image. This averaging was done only for reflections whose IQ was less than 8 on the first image, because we wanted to start with a set of reflections which were actually measurable on the first film. The actin noncrystallographic symmetry creates absences for about half the reflections. Figures 3-6 show the results of phase residual calculations within and among the images. The notation “IQ 3,” “IQ 4,” etc. will hereafter refer to reflections with IQ 3 or better, 4 or better, etc. unless otherwise stated. Reflections are considered without regard to resolution to the limit of the data (16 A). Figure 3 shows the phase residual for 2, symmetry on each image using all reflections which were IQ 3, 4, etc. on the first image, whatever their S/n ratio on subsequent images. The same reflections are thus included for all comparisons in each IQ class. The same residual calculation is shown in Fig. 4 for reflections whose IQ on any image was as stated in the legend. In these curves, reflections of equal quality are being compared across the damage series. Due to the damage, fewer reflections are generally being compared at the end of the damage se-
SCHMID ET AL.
64
a
b
d
C
I .;.
;i
I
II
I:‘,
;
.
,,’ : ;*: :,i(.::*,’ . !!i t : ? . j :f$: .-If; .;,pf..
. : . 1 .
.
: ’
f
. .
.I:.i!!il.’ .!Ii:
1
.:;
.
ii!;!.
illi
h ..:I
:
: i i,.:,;I : 1 : . - . . . . . .
.
;:
e
: ‘h
;iiffiii: .I*,. 1.)
.!v.
. . . . : . .
.:;.
.
.’
h
g
(a-d) Electron images of frozen, hydrated actin bundles recorded successively with cumulative doses of 3, 10, 16, and 23 electrons/A2, respectively, with 400 kV electrons and - 168°C specimen temperature. (e-h) The IQ plots of each image in the damage series. IQ l-3 reflections are represented by the largest circles, 4 and 5 are smaller, and 6 and 7 are smallest. FIG.
1.
ries, especially for the IQ 3, 4, and 5 data. Figure 5 shows the results of merging the damage series in pl using reflections whose IQ on the first image was as indicated. Again, the same reflections are used on all images, as in Fig. 3. The merging statistics are cumulative with the program ORIGTILT, so for instance, the third image is merged with the vector sum of the phase shifted reflections from images 1 and 2. Figure 6 shows the merging of images as
above, this time using reflections whose IQs are as indicated for each image, as in Fig. 4. Using the results of the merging to determine the relative phase shifts of the images, projection reconstructions were calculated from the first, third, fifth,
TABLE I from Computed Fourier Transforms of the Images in a Damage Series Lattice
Parameters
(in A) Refined
Image no.
l/U*
c
1 2 3 4 5 6 7
125.8 125.8 125.2 125.4 124.9 124.7 124.7
772.4 773.3 774.4 772.9 771.2 771.5 772.8
5
10
15
20
25
Cumulative Dose (e-/A*) FIG. 2. The fall-off of amplitude with cumulative dose in various resolution ranges. Each resolution shell is scaled to its amplitude in the first image.
ELECTRON Intra-image
symmetry
- First
DAMAGE
image’s
OF ACTIN
BUNDLE
AT
65
400 KV
IQ
Merging
residuals
- First
image’s
IQ
60 SO -$ 40 -0 ‘Z 2 30 2 2 a 20
--a-. . ..&.. --c - m-
OF STRUCTURAL
BIOLOGY
108,
62-68 (199%
Effects of Radiation MICHAEL W. M. Keck Center One Baylor Plaza,
for Computational Houston, Texas
Damage with 400-kV Electrons Hydrated Actin Bundles
F. SCHMID, JOANITA Biology, 77030; and Institute
JAKANA,
PAUL MATSUDAIRA,”
obtained from a few bundles that are tilted about an axis parallel to the bundle axis. Radiation damage limits the number of usable tilt images from each bundle. There have been theoretical suggestions that 400-kV electrons may provide an advantage over lOO-kV electrons from the point of view of radiation damage (Glaeser, 1985; Chiu et al., 1986). However, there has always been a counterargument that this would not be true because the ratio of elastic to inelastic (damage-producing) scattering is unaffected by the electron energies (Isaccson, 1977). Most of the radiation damage studies of crystals have been done by observing the fading of electron diffraction patterns as a function of cumulative electron dose (Glaeser and Taylor, 1978; Jeng and Chiu, 1984). These studies imply that the contrast of the images in a damage series decreases. Unwin (1974) has observed stain migration in images of stacked disk aggregates of tobacco mosaic virus coat protein as a function of dose. However, no quantitative information has yet been obtained about the effects of radiation on the phases, particularly in frozen, hydrated specimens. In the present study, we have undertaken a crystallographic analysis of structural changes in a characteristic image view of frozen, hydrated actin bundle taken at different accumulated doses with 400-kV electrons at a specimen temperature of - 168°C. We are taking the practical approach that the quality of the phases is the most important criterion in evaluating the reliability of image information. Data on the effects of damage allow us to set a limit on the number of usable images in a tilt series from a single bundle.
INTRODUCTION
The acrosomal process from Limulus polyphemus is a long cellular extension made up of actin and bundling proteins (Tilney, 1975). The native structure of the true discharge has recently been studied by electron cryomicroscopy and crystallographic analysis, which showed it to be crystalline with continuously varying views along the bundle axis (Schmid et aZ., 1991). One of the important biological questions about this intracellular structure is the mode of molecular interaction among the proteins which governs its formation, stability, and flexibility (Schmid et al., 1991). Because the thickness of the actin bundle is about 1000 A, 400 kV electron radiation may be useful for recording images for high-resolution three-dimensional structure determination. Although the apparent image contrast with 400 kV electrons is not as high as that at 100 kV, high-resolution features can be readily obtained (Schmid et al., 1991). In order to retrieve the three-dimensional structure of the acrosomal bundle with no a priori assumption of helical symmetry, we plan to combine images at various tilt angles. In the initial stage where we are merging low-resolution threedimensional data, our task is easier if the data are
MATERIALS Electron
AND METHODS
Cryomicroscopy
The acrosomal process was purified from the sperm of horseshoe crab, Limulus polyphemus, as described (Schmid et al., 1991). Frozen, hydrated actin bundles were prepared on freshly carbon-coated holey 400-mesh copper grids. The specimens were transferred to a Gatan 626 cryoholder and then into a JEOL 62
Inc. reserved.
Baylor College of Medicine, of Biology, Massachusetts
19, 1991, and in revised form December 9, 1991
Electron images can be used to provide amplitudes and phases for the structural determination of biological specimens. Radiation damage limits the amount of structural information retrievable by computer processing. A 400-kV electron microscope was used to investigate radiation damage effects on frozen, hydrated actin bundles kept at -168°C. The quality of phases within and among images in a damage series was evaluated quantitatively out to 16 A resolution. It was found that the phases of structure factors with good signal-to-noise ratio (IQ s 4) can be reliably retrieved from images taken at a cumulative dose of at least 25 electrons/A’. o lwz Academic PITSS, IX.
1047-8477192 $3.00 Copyright 0 1992 by Academic Press, All rights of reproduction in any form
AND WAH CHIU
and Verna and Marrs McLean Department of Biochemistry, *Whitehead Institute for Biomedical Research and Department of Technology, Cambridge, Massachusetts 02142
Received September
on Frozen,
ELECTRON
DAMAGE
OF ACTIN
4000EX electron microscope. To ensure maximum stability of the cryoholder, 1.5 hr elapsed between transferring into the column and recording of the images. Untilted specimens were examined at 400 kV under low dose rate conditions at a specimen temperature of - 168°C (Brink and Chiu, 1991). The dose rate on each image was measured with a picoammeter recording the current on the fluorescent screen. The picoammeter was calibrated as described by Brink and Chiu (1991). Radiation
Damage
Series
Usually seven or more successive images from the same bundle area were recorded keeping all microscope settings constant throughout the radiation damage series. Each image was recorded at nominal x 20 000 magnification with a specimen dose of 3 electrons/A’. The cumulative dose was less than 25 electrons/ A” in the last image. The defocus was aimed at less than 1 pm under-focus. All images were recorded on Kodak SO-163 films developed for 12 min in 100% D19, and fixed 10 min in Kodak fixer at 20°C. The optical density of the image of the ice-embedded area is around 0.7. Images of actin bundles displaying the “h01” view were evaluated for their quality in an optical diffractometer. Image
Processing
images selected above were scanned on a PDS 1OlOM microdensitometer on a 16-pm raster with a scan interval equivalent to 8 A/pixel. Identical actin bundle areas in each image were boxed 612 pixels long by 204 pixels wide. A Fourier transform of the boxed area was calculated and the reciprocal lattice parameters were refined iteratively. A list of amplitudes and phases of the structure factors along with their IQ after proper phase shift was calculated from each image in the damage series (Robinson et al., 1988). IQ is a parameter which measures the signal-to-noise ratio of structure factors and ranges from 1 to 8, with 1 being the best (Henderson et al., 1986). Phase
Comparisons-Zntraimage
Symmetry
The presence and quality of various symmetry elements in a crystal image were evaluated for the calculated structure factors as previously described (Schmid et al., 1991). Except for the choice of whether to include a reflection or not based on its IQ, no other weighting scheme was used, and all reflections were considered out to the resolution limit as defined by the Sampling Theorem. Phase
Comparisons-Merging
the Damage
Series
Merging of the images in the damage series was carried out using the program ORIGTILT (Henderson et al., 1986). The images were merged using the symmetry of plane group pl. This ensures that the same reflections, not symmetry mates, are used to compare images in the damage series. As in the above phase calculations, after the choice of reflections based on IQ, no other weighting scheme was used. In both of these comparisons, the choice of reflections based on -heir IQ was done in two ways. In the first, the IQ of the reflection an the first image was used in the selection. Thus, the same reflections can be compared throughout the analysis, no matter what their IQ was after the first image. The other way of choosing reflections is to make the IQ selection on every image. This method is more like the one actually used in merging of images in a tilt series and, for instance, some of the reflections do not appear for every image. The results of the merging were used to shift the phase origins of the images until they were coincident, and two-dimensional reconstructions were carried out on several members of the damage series. RESULTS
Seven images of an actin bundle were obtained and processed. Figure 1 shows the first, third, fifth,
BUNDLE
AT
400 KV
63
and seventh images of the actin bundle in this damage series. The images show an area which includes the h01 view with a = b = 147 A, c = 762 A (Schmid et al., 1991). Slightly different orientations appear above and below the h01 zone. An area of each image was digitized and windowed according to features in the micrograph to ensure that exactly the same region of the bundle was processed for all members of the damage series. Almost 6 unit cells (4570 A) along the length of the bundle were included for each image. The actual magnification of the images was found to be x 15 750 by calibration with the unit cell parameters found for the actin bundle, leading to a scan size of about 7.5 A/pixel. The cumulative dose on the images in Figs. la-ld was 3, 10, 16, and 23 electrons/A’, respectively. The windowed area was transformed and the lattice parameters were independently refined for all seven images. The results of the lattice refinement are shown in Table I. Below each image is an IQ representation (Henderson et al., 1986) of the Fourier transforms of the bundle images after lattice refinement. IQ 1, 2, and 3 have the largest circles (S/n > 2.71, IQ 4 and 5 are represented by intermediate sized circles (2.7 > Sln > 1.61, and IQ 6 and 7 are smallest (1.6 > S/n > 1.0). No unbending or other corrections were applied to the data. Predictably, radiation decreases the amplitudes of the reflections (Jeng and Chiu, 1984). Figure 2 shows the fall-off of the amplitudes of the reflections in the Fourier transforms of the images. The average amplitude of all reflections (see below) in each resolution shell is plotted relative to its average amplitude in the first image, so amplitudes in all shells start at 100% in the first image. This averaging was done only for reflections whose IQ was less than 8 on the first image, because we wanted to start with a set of reflections which were actually measurable on the first film. The actin noncrystallographic symmetry creates absences for about half the reflections. Figures 3-6 show the results of phase residual calculations within and among the images. The notation “IQ 3,” “IQ 4,” etc. will hereafter refer to reflections with IQ 3 or better, 4 or better, etc. unless otherwise stated. Reflections are considered without regard to resolution to the limit of the data (16 A). Figure 3 shows the phase residual for 2, symmetry on each image using all reflections which were IQ 3, 4, etc. on the first image, whatever their S/n ratio on subsequent images. The same reflections are thus included for all comparisons in each IQ class. The same residual calculation is shown in Fig. 4 for reflections whose IQ on any image was as stated in the legend. In these curves, reflections of equal quality are being compared across the damage series. Due to the damage, fewer reflections are generally being compared at the end of the damage se-
SCHMID ET AL.
64
a
b
d
C
I .;.
;i
I
II
I:‘,
;
.
,,’ : ;*: :,i(.::*,’ . !!i t : ? . j :f$: .-If; .;,pf..
. : . 1 .
.
: ’
f
. .
.I:.i!!il.’ .!Ii:
1
.:;
.
ii!;!.
illi
h ..:I
:
: i i,.:,;I : 1 : . - . . . . . .
.
;:
e
: ‘h
;iiffiii: .I*,. 1.)
.!v.
. . . . : . .
.:;.
.
.’
h
g
(a-d) Electron images of frozen, hydrated actin bundles recorded successively with cumulative doses of 3, 10, 16, and 23 electrons/A2, respectively, with 400 kV electrons and - 168°C specimen temperature. (e-h) The IQ plots of each image in the damage series. IQ l-3 reflections are represented by the largest circles, 4 and 5 are smaller, and 6 and 7 are smallest. FIG.
1.
ries, especially for the IQ 3, 4, and 5 data. Figure 5 shows the results of merging the damage series in pl using reflections whose IQ on the first image was as indicated. Again, the same reflections are used on all images, as in Fig. 3. The merging statistics are cumulative with the program ORIGTILT, so for instance, the third image is merged with the vector sum of the phase shifted reflections from images 1 and 2. Figure 6 shows the merging of images as
above, this time using reflections whose IQs are as indicated for each image, as in Fig. 4. Using the results of the merging to determine the relative phase shifts of the images, projection reconstructions were calculated from the first, third, fifth,
TABLE I from Computed Fourier Transforms of the Images in a Damage Series Lattice
Parameters
(in A) Refined
Image no.
l/U*
c
1 2 3 4 5 6 7
125.8 125.8 125.2 125.4 124.9 124.7 124.7
772.4 773.3 774.4 772.9 771.2 771.5 772.8
5
10
15
20
25
Cumulative Dose (e-/A*) FIG. 2. The fall-off of amplitude with cumulative dose in various resolution ranges. Each resolution shell is scaled to its amplitude in the first image.
ELECTRON Intra-image
symmetry
- First
DAMAGE
image’s
OF ACTIN
BUNDLE
AT
65
400 KV
IQ
Merging
residuals
- First
image’s
IQ
60 SO -$ 40 -0 ‘Z 2 30 2 2 a 20
--a-. . ..&.. --c - m-
