require alignment to a common frame of reference prior to reconstruction. As a
consequence, electron tomography reconstructions are not available in real-time.
Misalignment of projections is due to the fact that cross-correlation-based tracking
attempts to re-centre the region of interest at each tilt angle but in the case of
low-dose tomography, this function depends on the tracking region behaving in
exactly the same manner as the recording region. Even when microscope settings
are optimised, unreliable tracking or cumulative translational shifts (cf. drift/
blurring) might be due to a slight unevenness in the grid, for example.
Alignment of projections has traditionally relied on the presence of fiducial
markers to create a coordinate system for mutual alignment of the projections. In
this method, colloidal gold particles are added to the specimen before
plunge-freezing (cryo-tomography) or to one or both surfaces of sections (plastic or
cryo, see below). The spatial coordinates of each electron-dense marker are selected
in every projection to establish a common coordinate system. This requires at least
3 particles to be visible in all projections, and preferably at least 5 to account for
rotations that result from variations in lens current (a nominal magnification of
20,000X might actually be 20,002X in one projection followed by, say, 19,997X in
the next, an so on). Adsorbed gold colloids on plastic sections are typically
numerous and distributed randomly. This allows for considerable manipulation of
the aligned image stack prior to reconstruction, such that the reconstruction represents a globally corrected solution. For cryo- tomography, gold colloids are
typically titrated into a suspension sample prior to vitrification. Therefore, cryo
specimens typically have fewer markers per field of view, and the gold may bind
preferentially to certain components. The lack of shrinkage and mass loss experienced by these samples, however, means that this is usually sufficient. For cellular
samples grown on TEM sample grids, the markers can be deposited on the carbon
film support prior to incubation with the cells. Alternatively, colloidal gold can be
deposited onto vitreous specimens using an adaptation of the ingenious quantum
dot method demonstrated by Masich et al. [86]. Briefly, gold colloids are functionalised with triphenylphosphine to render them soluble in toluene. The stabilised
particles are dried in batches, and when required, the stabilised powder is resuspended in isopropane, which remains liquid at −150 °C. The vitrified sample grid is
dipped in the gold suspension, blotted with filter paper, and then transferred to a
cryo sample holder for tomography [64]. Gold colloids are preferred to quantum
dots for alignment because they offer constant contrast at all tilt angles. Quantum
dots are perfect crystals. During tilting, their contrast is minimal when the plane of
the crystal lattice coincides with the direction of the electron beam. They thus tend
to ‘blink’ on and off, and often blend in with (low contrast) electron-dense sample features when the markers’ contrast is not at its maximum. It may be preferable
to use mixtures of 10 nm and 20 nm gold colloids, in the event that the 10 nm
particles become difficult to track in highly tilted projections.
Marker-free alignment, usually of the feature-tracking variety, is essential when
fiducial markers are scarce or absent, when the markers are poorly distributed (e.g.
all near the tilt axis) or when the sample does not behave as a rigid body [87, 88],
such as a poorly attached cryosection. Of course, gold colloids can be selected as
14
A. Leis
consequence, electron tomography reconstructions are not available in real-time.
Misalignment of projections is due to the fact that cross-correlation-based tracking
attempts to re-centre the region of interest at each tilt angle but in the case of
low-dose tomography, this function depends on the tracking region behaving in
exactly the same manner as the recording region. Even when microscope settings
are optimised, unreliable tracking or cumulative translational shifts (cf. drift/
blurring) might be due to a slight unevenness in the grid, for example.
Alignment of projections has traditionally relied on the presence of fiducial
markers to create a coordinate system for mutual alignment of the projections. In
this method, colloidal gold particles are added to the specimen before
plunge-freezing (cryo-tomography) or to one or both surfaces of sections (plastic or
cryo, see below). The spatial coordinates of each electron-dense marker are selected
in every projection to establish a common coordinate system. This requires at least
3 particles to be visible in all projections, and preferably at least 5 to account for
rotations that result from variations in lens current (a nominal magnification of
20,000X might actually be 20,002X in one projection followed by, say, 19,997X in
the next, an so on). Adsorbed gold colloids on plastic sections are typically
numerous and distributed randomly. This allows for considerable manipulation of
the aligned image stack prior to reconstruction, such that the reconstruction represents a globally corrected solution. For cryo- tomography, gold colloids are
typically titrated into a suspension sample prior to vitrification. Therefore, cryo
specimens typically have fewer markers per field of view, and the gold may bind
preferentially to certain components. The lack of shrinkage and mass loss experienced by these samples, however, means that this is usually sufficient. For cellular
samples grown on TEM sample grids, the markers can be deposited on the carbon
film support prior to incubation with the cells. Alternatively, colloidal gold can be
deposited onto vitreous specimens using an adaptation of the ingenious quantum
dot method demonstrated by Masich et al. [86]. Briefly, gold colloids are functionalised with triphenylphosphine to render them soluble in toluene. The stabilised
particles are dried in batches, and when required, the stabilised powder is resuspended in isopropane, which remains liquid at −150 °C. The vitrified sample grid is
dipped in the gold suspension, blotted with filter paper, and then transferred to a
cryo sample holder for tomography [64]. Gold colloids are preferred to quantum
dots for alignment because they offer constant contrast at all tilt angles. Quantum
dots are perfect crystals. During tilting, their contrast is minimal when the plane of
the crystal lattice coincides with the direction of the electron beam. They thus tend
to ‘blink’ on and off, and often blend in with (low contrast) electron-dense sample features when the markers’ contrast is not at its maximum. It may be preferable
to use mixtures of 10 nm and 20 nm gold colloids, in the event that the 10 nm
particles become difficult to track in highly tilted projections.
Marker-free alignment, usually of the feature-tracking variety, is essential when
fiducial markers are scarce or absent, when the markers are poorly distributed (e.g.
all near the tilt axis) or when the sample does not behave as a rigid body [87, 88],
such as a poorly attached cryosection. Of course, gold colloids can be selected as
14
A. Leis