!

4!

thought to be pleomorphic. Similar analysis could also be used to study objects that possess some local symmetry. 1.3 Tilt Series Acquisition In order to produce a cryo-electron tomogram, sequential images of the specimen embedded in vitreous ice are obtained by rotation about a single axis of the specimen stage in the electron microscope. A number of parameters need to be carefully considered before acquiring tilt series data, including magnification, defocus, angular increment and electron dose. A magnification should be chosen such that the area of the specimen viewed on the detector is maximized, while ensuring that the pixel size at the specimen level does not limit the resolution. The pixel size at the specimen level (ps) can be calculated by:

ps = pa / M

(1.1)

where pa is the actual pixel size of the detector and M is the magnification. For a 2,048 x 2,048 pixel detector with! actual pixel size of 15 m, a magnification of 20,000 would an yield images with a pixel size of 0.75 nm and an area of 1.5 x 1.5 m at the specimen level. Because of the technical limitations in electron tomography (discussed in Section 1.5), a reasonable expectation for the resolution of a tomogram with virions embedded in vitreous ice might be 5.0 nm. For images taken under the conditions described above, the pixel size would not be the limiting factor, since the Nyquist limit would be 1.5 nm. It is useful set the microscope defocus such that the first zero in the contrast transfer function would occur near the expected resolution limit for the reconstructed tomogram. The contrast transfer function, CTF(k), is given by:

!
&# ) CTF(k) = "sin( Cs$3 k 4 + #%f$k 2 + '2 *

5!

(1.2)

where Cs is the spherical aberration constant for the microscope, ! is the voltage
! dependent wavelength of the electrons, k is the spatial frequency and !f is the defocus!

(Frank, 2006a). A typical tomographic reconstruction may typically be limited to a resolution of 5.0 nm. For a 120 kV instrument with a spherical aberration constant of 2.0, a defocus 6.0 m would put the first node at a spatial frequency of ~1/4.5 nm-1. If an appropriate defocus is chosen, the contrast can be maximized over the useful resolution range (Fig. 1.1a). Obtaining images too close to focus would limit the contrast (Fig. 1.1b), while operating too far from focus would place a node within the resolution range of interest (Fig. 1.1c). The angular increment should be chosen so that it will not be the limiting factor in the resolution of the tomogram. The theoretical resolution limit, d, based upon the angular increment is given by the Crowther criterion: (1.3) where D is the diameter of the object being reconstructed and N is the number of equally spaced projection images (Crowther et al., 1970b). For a tilt series about a single axis, this can be rewritten as:

d=

"D#$ 180

(1.4)

where !" is the angular increment in degrees. Assuming that the macromolecule under

! study has a 50 nm diameter and that our maximum expected resolution is 5.0 nm, an
angular increment of 5 would yield a theoretical resolution limit of ~4.4 nm, which

6!

would be a sufficient angular increment for this specimen. For Newcastle disease virus, virions can approach 200 nm in diameter and an angular increment of 1.5 would be required to maintain a theoretical resolution limit near ~5.0 nm. It should be noted that equation 1.4 only holds true for spherical objects or cylindrical objects rotated about their long axes. In reality, a specimen consisting of iceembedded virions has a slab-like geometry. Equation 1.4 also implies that the data will cover the complete angular range (+/- 90). For electron tomography, the Crowther criterion serves only as an approximation of the theoretical resolution limit, since it does not take into account the increase in apparent thickness of a slab-like specimen upon specimen tilting, nor the missing wedge of data due to the limited tilt range of the microscope (see Section 1.5). The Crowther criterion, however, is a useful rule of thumb for determining the angular increment to be used in a tomographic examination of spherical viruses. Choosing the total dose for a tilt series is not straight-forward. As the total dose increases, gains in contrast will be offset by an increase in irradiation damage. However, the maximum total dose that a particular specimen can tolerate is variable (Tocheva et al., 2010). Generally, it is useful to collect several tilt series with varying total electron dose. In order to maintain image quality throughout a tilt series, it may be helpful to adjust the electron dose in proportion to the tilt angle. This topic is discussed in detail in Section 1.5. If the specimen is at the eucentric height for the microscope, the rotation of the stage from one tilt series image to the next should induce only a minimal movement of the specimen relative to the detector. If no mechanical instabilities were present, the

7 object of interest would remain centered on the detector throughout the acquisition of the series of tilted images without the need for re-centering. However, this is usually not the case. During acquisition of the tilt series data, the object of interest will appear to move relative to the detector if no corrections are made. A process known as tracking must be performed in order to keep the object of interest roughly centered on the detector. Once an area of interest is chosen and the imaging parameters are decided upon, the procedure for obtaining tilt series images is a three-step automated process that consists of tracking, focusing, and finally exposure of the area of interest (obtaining the actual data). These three steps are repeated for each angular step in the tilt series. Tracking and focusing are usually performed along the tilt axis, but the electron beam is shifted several microns away from the area of interest, so that no unnecessary electron dose is imparted into the area of interest. The tracking process determines the distance the object of interest has moved between successive angular steps due to mechanical instabilities of the specimen holder. A computational search determines the translational elements that bring the tracking images of two successive angular steps into register. This search is scored by crosscorrelation coefficient, r:

$[" (x, y) #
1

p1 ][ " 2 (x, y) # p2 ]
(1.5)

r=

x,y

) %1/ 2 ' ' 2 2 *$[ "1 (x, y) # p1 ] $[ " 2 (x, y) # p2 ] & ' ' x,y ( x,y +

where !1(x,y) and !2(x,y) are the density height for a particular pixel in two successive

! tracking images. The average density heights over all pixels in the first and second image
are denoted as <!1> and <!2>, respectively. Upon finding the relative translation of the

8!

two tracking images that generates the highest coefficient, the current in the image shift coils can then be adjusted by the appropriate amount to produce an image shift that would return the object of interest to the center of the detector. Since the tracking images are taken prior to the exposure image for each angular step, the calculated shifts can be applied during the exposure step, and the object of interest should not appear to move from frame to frame in the tilt series. It is also necessary to keep the defocus constant throughout the tilt series. Like the tracking process, auto-focusing is also performed at a location on the specimen a few microns away from the object of interest, usually in the opposite direction along the tilt axis from the tracking area. Automated focusing takes advantage of the fact that tilting the electron beam induces a shift in the image that depends upon the defocus, according to:

s = M(Cs" 2 # $f )"

(1.6)

where s is the displacement in the direction of the beam tilt angle !, M is the

! magnification, Cs is the spherical aberration constant for the microscope, and !f is the
defocus (Koster et al., 1992b). The displacement, s, between two images with opposite electron beam tilts can be determined from a cross-correlation search, and the defocus can then be calculated by solving equation 1.6 for !f. The defocus can then be adjusted until the displacement determined via the cross-correlation search matches the expected value of s for the desired defocus. As stated previously, focusing, tracking, and exposure of the object of interest are repeated, in that order, to obtain an entire tilt series. It is usually most simple to begin with the specimen un-tilted and obtain consecutive tilts in one direction (positive or

9!

negative tilting). The stage can then be returned to the 0 position and the remainder of the images can be obtained by tilting in the opposite direction. Furthermore, it may be possible to skip the tracking and focusing and focusing steps for some angular steps in order to speed-up the data collection process. 1.4 Tilt Series Alignment and Tomographic Reconstruction Once a tilt series is acquired, it is necessary to align the image data prior to the three-dimensional reconstruction process, as the tracking procedure only roughly keeps the object or objects of interest centered on the detector. A search can be carried out to determine the translational and in-plane rotational elements that maximize the crosscorrelation coefficient between sets of successive images. As the tilt angles are roughly known from the microscope goniometer, a rough three-dimensional map can be calculated directly from the cross-correlation aligned images. However, the relationships among the tilt series images are usually refined "#$%& fiducial gold markers, which are added during specimen preparation (Kremer et al., 1996a; Luther et al., 1988). The positions of the fiducial markers are manually recorded for each tilt series image and a three-dimensional model of the fiducial marker positions is generated and refined. The x and y positions of gold markers in the initial three-dimensional model are assumed from the x, y positions of corresponding gold markers in the un-tilted projection image. Also, it is initially assumed that all gold markers are in the same plane, where z=0. In order to accurately determine the rotational and translational relationships among the tilted images, the minimum value of a function ( f ) is sought. This function measures how well the manually determined x, y marker positions in the micrographs match the x, y