Hawkes P.W., Spence J.C.H. (Eds.) Science of Microscopy. V.1 and 2

Подождите немного. Документ загружается.

Chapter 7 Cryoelectron Tomography (CET) 545

and Hegerl, 2001; Jiang et al., 2003b). The main purpose of ﬁ ltering/

“de noi si ng” is to improve the visualization of the raw data and to

facilitate interpretation primarily at the level of the ultrastructure.

However, interpretation at the molecular level requires more quantita-

tive methods.

Figure 7–5. First electron tomographic investigation of a eukaryotic cell; the slime mould Dictyostel-

ium discoideum embedded in vitriﬁ ed ice. a) Phase contrast image and corresponding ﬂ uorescence

image (inset) of cells on TEM grids. b) Cryo-electron micrograph at 0° tilt (conventional 2D

projection) of a ∼200 nm thin peripheral region of the cell. c) Tomographic reconstruction from a

complete tilt-series (120 images) and d) visualization by segmentation. Large macromolecular com-

plexes, e.g. Ribosomes are shown in a green color, the actin ﬁ lament network in orange-red and the

cells’ membrane in blue. Cryo-tomograms of Dictyostelium discoideum cells grown directly on carbon

support ﬁ lms have provided unprecedented insights into the organization of actin ﬁ laments in an

unperturbed cellular environment. The tomograms show, on the level of individual ﬁ laments, their

modes of interaction (isotropic networks, bundles, etc.), they allow us to determine the branching

angles precisely (in 3-D), and they reveal the structure of membrane attachment sites (Medalia

et al., 2002). (See color plate.)

546 J.M. Plitzko and W. Baumeister

A prerequisite for understanding the interaction of proteins is the

localization of macromolecules in the cell. Practical ways to achieve

this goal are labeling of macromolecules by speciﬁ c and electron-dense

markers (Koster and Klumperman, 2003) or recognition of individual

particles by their known structural signature. The former is very hard

to realize under in vivo conditions, whereas the latter requires sufﬁ -

ciently high resolution. Moreover, labeling techniques can be used for

only a few macromolecules simultaneously since the different labels

need to be distinguishable and speciﬁ c. In the context of structural

proteomics this makes recognition by means of the structural signa-

ture favorable since simultaneous recognition of molecules under at

least close-to-native conditions is mandatory.

It has been shown previously that the currently achieved resolution

of cryo-ET should sufﬁ ce to identify large protein complexes in the

range of 1 MDa (Boehm et al., 2000). Moreover, detection of protea-

somes (Lowe et al., 1995; Zwickl, 2000) and thermosomes (Klumpp

et al. 1997; Klumpp and Baumeister, 1998) in phantom cells, which

mimic real cells in size and composition with the advantage of being

biochemically well deﬁ ned, showed that this approach is also feasible

experimentally (Frangakis et al., 2002). The approach used in these

works requires structural knowledge from other sources; matched

ﬁ ltering of the tomograms with a structural template yields the loca-

tions and orientations of these complexes in the cell (see Section 3.5.4).

Therefore, cryo-ET requires structural data from other techniques

such as X-ray crystallography or single-particle analysis to obtain

molecular information. The principal dose limitation of cryo-ET does

not permit high-resolution structure determination on the basis of the

tomogram alone. The dose fractionation theorem (Hegerl and Hoppe,

1976; McEwen et al., 1995) states that tomography images each volume

element with the same statistical signiﬁ cance as a 2D projection

acquired with the same overall dose. This makes tomographic data

highly advantageous for location of features since the information is

distributed over three dimensions. However, the statistical signiﬁ -

cance of each volume element is generally not sufﬁ cient to resolve

macromolecules at the level of tertiary or quaternary structural ele-

ments initially. However, the combination of cryo-ET and single-par-

ticle averaging can offer a remedy for this; by averaging identical

copies of complexes, they can potentially be resolved to at least a few

nanometers. First realizations of the combination of cryo-ET and

single-particle averaging have been reported for isolated proteins and

yielded structures of 800-kDa protein complexes to ∼2 nm resolution

(Nitsch et al., 1998); averaging of complexes directly from tomograms

of organelles or viruses is a novel approach (Figure 7–6). First 2D

results on the envelope complexes of dried and stained retroviruses

indicated the potential of this approach (Zhu et al., 2003). Recent 3D

results on the structure of the nuclear pore complex, a membrane

complex that is difﬁ cult to assess by other techniques, taken from

tomograms of entire nuclei, showed that this approach can yield

structural characterization of unprecedented resolution (Beck et al.,

2004, 2007).

Chapter 7 Cryoelectron Tomography (CET) 547

3 Major Difﬁ culties in Cryoelectron Tomography

Despite the fact that the acquisition process for ET sounds very simple

and straightforward, the acquisition of a tomographic tilt series, espe-

cially from biological samples is, even today, a major technical chal-

lenge. The explanation will be given in the following sections addressing

the ﬁ ve major obstacles (“core problems”) of ET: (1) sample preparation

and preservation, (2) radiation sensitivity, (3) tilting geometry, (4)

instrumentation (EM, energy ﬁ lters, CCD detectors, etc.) and automa-

tion, and (5) alignment, reconstruction, and visualization.

3.1 Sample Preparation

3.1.1 Basics

Biological objects such as cells consist mainly of water (more than 70%).

Built up from carbon-based molecules, they form the essential frame-

work of life but for a closer structural investigation with the EM they

have to withstand a constant “bombardment” of fast electrons under

ultrahigh vacuum conditions: clearly not the best starting points

Figure 7–6. CET in combination with the single particle approach of transport-competent Dictyostel-

ium discoideum nuclei. a) Three-dimensional reconstruction of the peripheral rim of an intact nucleus.

X-y slice of 10 nm thickness along the z axis through a typical tomogram. Side views of nuclear pore

complexes (NPCs) are indicated by arrows. Ribosomes connected to the outer nuclear membrane are

visible (arrowheads). Inset displays a phase-contrast image and the corresponding ﬂ uorescence image.

b) Surface rendered representation of a segment of nuclear envelope (NPCs in blue, membranes in

yellow). c) Structure of the Dictyostelium NPC after classiﬁ cation and averaging of subtomograms.

Cytoplasmic face of the NPC (upper left); the cytoplasmic ﬁ laments are arranged around the central

channel. Nuclear face of the NPC (upper right); the distal ring of the basket is connected to the nuclear

ring by the nuclear ﬁ laments. Cross sectional view of the NPC (bottom). The dimensions of the main

features are indicated. All views are surface-rendered (nuclear basket in brown; Beck et al. 2004). (See

color plate.)

548 J.M. Plitzko and W. Baumeister

and clearly conditions that are far from natural and extremely

inhospitable.

Indeed, the central goal in any EM study is the observation of struc-

tures down to the atomic level. Unlike material science studies, where

the samples remain steady, biological specimens tend to be more

“resistant” to an investigation with the EM. First they are exposed to

ionizing radiation, namely electrons with speeds usually above

200,000 km/s, according to the acceleration voltage used. Second, the

ultrahigh vacuum generates an ambient pressure far below the atmo-

spheric pressure, and thus literally “sucks” out every shred of liquid,

subsequently resulting in the implosion of the biological structure.

Third, they are build up from carbon-based compounds plus elements

with very low atomic numbers like hydrogen, oxygen, and nitrogen,

and minute amounts of other low Z-elements, all very weak scatterers,

in regard to their interaction with electrons, resulting in low-contrast

images. Fourth, if we regard cellular structures of higher organisms,

such as mammalian cells, they can easily reach sizes of tenth of microm-

eters in all three dimensions, and are therefore almost or effectively

impenetrable to the electron beam. Fifth, the secrets of these structures

are hidden in a highly crowded environment, the cytoplasm, where

functional units like proteins, protein complexes, or even molecular

machines are “densely packed,” literally touching each other. This

“molecular crowding” (Ellis, 2001; Ellis and Minton, 2003) makes it

difﬁ cult to separate them for structural characterization. Sixth and last,

every cell is different, just as every human has a different face. Thus,

single 2D projections will not provide a complete 3D characterization,

and averaging techniques as in the so-called single-particle approach

are therefore impractical. All six of these facts have to be addressed in

any life science study and especially when investigating biological

structures in a close-to-life state, quasi in vivo.

The electron beam represents a form of ionizing radiation that is

harmful to the health of any living organism. Clearly, we cannot

shield the biological substance but we can try to extend its “life-span”

during investigation before structural damage occurs. Researchers in

the past have been ingenious in inventing preparation techniques for

EM studies of biological samples. In the late 1950s staining techniques

were introduced that addressed the ﬁ rst three issues at once (Brenner

and Horne, 1959). Staining with salts from heavy metals (usually

osmium or uranium salts) envelopes the structure of interest and after

insertion into the EM the negative imprint of the structure can be

observed. Moreover, it increases the contrast dramatically, according

to the high atomic numbers of the metals used. However, the biologi-

cal structure desiccates in the vacuum environment of the microscope

and illumination by the electron beam literally incinerates whatever

is left of the biological substance, leaving behind the imprint of the

structure outlined by the staining substance. Staining and dehumidi-

ﬁ cation alter the structure and, because of the nanocrystalline nature

of the staining salts, limit the resolution, in the best cases, to approxi-

mately 2 nm. The major advantage of this technique is its simplicity

(no special instrumentation is required), its speed, and the fact that

Chapter 7 Cryoelectron Tomography (CET) 549

the negative “shadow” of the sample is almost unaffected by radiation

damage. The obvious disadvantage is the highly “unnatural” state in

which the investigation is performed, leading to artifacts and possible

misinterpretations.

To study larger objects such as cells or organelles staining has been

combined with embedding in epoxy resins and subsequent sectioning

with the help of ultramicrotomes (Porter and Blum, 1953). Despite the

fact that almost all our ultrastructural knowledge of the cell interior is

based on this technique, there are some obvious limitations. First the

substitution material, e.g., epoxy resin, shrinks during hardening, thus

deforming the original structure. Second the stain applied after sec-

tioning reduces the resolution (as described above) and third a second

shrinkage step of the section is observed after initial illumination in

the EM. This effect can be as large as 30–40% in the direction of the

electron beam and 5–10% in the plane perpendicular to it. Shrinkage

of the initial section dimensions is caused by outgassing of the solid

polymer in the ultrahigh vacuum and to a larger extent by the loss of

mass due to radiation damage (Luther, 1992). However, shrinkage

occurs early in the illumination process, so that preexposing the sample

until the contraction is complete can help to stabilize the section. The

total electron dose is therefore increased dramatically and can lead to

dose-induced stain aggregation. After preexposure the sample remains

steady for further investigation of the modiﬁ ed structures and can be

reused almost indeﬁ nitely.

3.1.2 Cryopreparation

In the 1970s researchers experimented on 2D protein crystals with the

possibility of preserving specimen hydration in the EM (Taylor and

Glaeser, 1973, 1974). However, In 1981 Dubochet and McDowall pub-

lished their work on the vitriﬁ cation of pure water for EM and in 1984

Adrian et al. reported the ﬁ rst successful investigation of a shock-

frozen biological object, namely viruses, embedded in vitriﬁ ed ice, now

known as the cryopreparation technique. There is no doubt that this

invention is one of the important milestones in life science EM studies,

if not the most important. With this technique, scientists were, for the

ﬁ rst time, able to investigate biological objects in a close-to-life state,

quasi in vivo.

The object-holders are commercially available copper grids with a

diameter of 3 mm (Figure 7–7A). They can be purchased in different

geometries, mesh formats, and with different coatings (Plano GmbH,

Wetzlar, Germany; Ted Pella, Inc., Redding, CA; Quantifoil Micro Tools

GmbH, Jena, Germany; Ermantraut et al., 1998). Typically these grids

are coated with a very thin support foil, usually amorphous carbon

with thicknesses ranging between 4 and 20 nm. Shock-freezing has to

be done very fast and therefore a special apparatus is necessary. These

“g uillotine”-like instruments, so called plungers, permit freezing of the

sample in milliseconds to a temperature of liquid nitrogen (∼90 K). To

ensure a direct phase transformation from liquid to an amorphous

(vitriﬁ ed) solid, cryogens with very high cooling rates, like ethane or

less frequently propane, are used (Figure 7–7B and C).

550 J.M. Plitzko and W. Baumeister

The grid is clamped in an inverse tweezer and vertically mounted

into the plunging instrument. A droplet of a couple of microliters of

the sample solution, either isolated protein complexes or whole cells in

their buffer solution, is applied to the surface of the copper grid. Since

only thin objects in the range of a couple of hundred nanometers (t ≤

500 nm) are suitable for investigation with the EM, the sample volume,

typically in the range of 1–4 µl, has to be reduced. This is done with a

ﬁ lter paper carefully brought either to one side or to both sides of the

grid to remove excess liquid. This process is called blotting. To prevent

any mechanically induced disruption of, e.g., cellular structures, blot-

ting can be done from the backside of the grid (Figure 7–8). Depending

on the duration of the blotting process, the viscosity of the sample

solution, and the size of the object, the thickness of the resulting ice

layer can be adjusted. It guarantees a complete vitriﬁ cation, without

the formation of unwanted and harmful ice crystals, which can already

be observed with ﬂ uid layers exceeding a couple of micrometers.

Figure 7–7. Plunge freezing instrumentation. a) Typical arrangement of cells cultured on TEM grid just

a home-build plunge freezing apparatus (Images courtesy of R. Gatz, MPI of Biochemistry, Martinsried

(near Munich), Germany). The small reservoir (yellow) in the middle of the dewar (green) contains the

liquid ethane, which is chilled by a surrounding bath of liquid nitrogen. The forceps, which hold the

grid are attached to a weighted arm. When the arm is released by means of a foot-trigger, the grid is

gravity-plunged into the ethane. The cross-sectional scetch (left part of image B) illustrates the ‘guillo-

tine-like’ arrangement. c) The Vitrobot (FEI company, Eindhoven, The Netherlands), a ‘robot’ for vitri-

ﬁ cation, allows the control of environmental and all necessary processing parameters. (See color

plate.)

prior to plunging. The schematic indicates the dimensions. b) CAD (computer aided design) image of

Chapter 7 Cryoelectron Tomography (CET) 551

Existing cryoplungers are mainly home-built and vary in design.

Quite recently commercially available instruments have been intro-

duced with the additional ability to adjust and control some of the

preparation parameters, such as humidity, temperature, and air pres-

sure, in specially designed environmental chambers. Moreover the

whole process, except for the application of the sample, which has to

be done manually, is computer controlled and automated, to facilitate

the sample preparation process and expand its reproducibility (Vitro-

bot, FEI company, Eindhoven, Netherlands).

After vitriﬁ cation, the sample is transferred in a cryoholder, specially

designed to ensure that the sample remains in the frozen state at

approximately liquid nitrogen temperature. If, by any chance, the tem-

perature drops below −160°C the possibility that the vitriﬁ ed ice will

start to crystallize increases. Therefore during transfer and during the

subsequent investigation, frozen-hydrated samples have to be kept at

all times at liquid nitrogen temperature and below −160°C, which

resulted in the name cryoelectron microscopy.

Compared to the commonly used techniques, cryopreparation allows

us to preserve the native structure of the cytoplasm and the whole

arrangement within the cell and even the physiological state at the

moment of plunging. We thus circumvent alterations or modiﬁ cations

as seen with staining or epoxy resin embedding. The technique is

Figure 7–8. Steps to be taken in the plunge freezing process. I Apply the solu-

tion containing protein complexes or cells (TEM grid illustrated in a cross

sectional way). To remove the excess solution and thus adjust the ice layer

thickness, II blotting is done with a ﬁ lter paper, carefully brought to the front-

or backside of the TEM grid. After blotting the remaining solution on the grid

is rapidly plunged III into liquid ethane, and afterwards kept at all times at

liquid nitrogen temperature to prevent contamination, crystallization or smelt-

ing of the amorphous (vitriﬁ ed) ice layer. (See color plate.)

552 J.M. Plitzko and W. Baumeister

much more sophisticated and the result of the preparation is not known

until a look is taken through the EM. The task of determining suitable

processing parameters for the material at hand, to get, in the end, the

best preparation result for an in-depth investigation of the biological

structure, can therefore be very tedious.

3.1.3 Cryosectioning

While plunge-freezing is now almost routinely used for the prepara-

tion of puriﬁ ed and isolated macromolecular complexes as well as for

prokaryotic and very small eukaryotic cells, larger bulky structures,

such as big mammalian cells or larger organelles, cannot be addressed

as a whole for EM investigations. However, the technique of sectioning

using ultramicrotomes, likewise known from epoxy resin-embedded

samples, has been adapted for the use in combination with high-

pressure frozen samples. To obtain a vitriﬁ ed sample of a specimen

tenths of micrometers in dimension, plunge-freezing as described

above is not sufﬁ cient, because of its limited “depth of vitriﬁ cation,”

explained by the freezing rate, which is, for example 10

°C/s for pure

water. If the sample dimensions exceed approximately 10 µm, vitriﬁ ca-

tion is incomplete, which will result in the formation of cubic ice

crystals. Therefore, high-pressure freezing is the method of choice,

because the increase in pressure will lower the freezing rate (Moore,

1987). The lowest values are observed around 2000 bar, where the

depth of vitriﬁ cation increases up to 10 times as compared to freezing

at ambient pressures (Sartori et al., 1993). The sample material, e.g., a

cell suspension, is forced into a 15-mm-long copper tube with an inner

diameter of approximately 0.3 mm. Afterward a pressure of 2000 bar

builds up in a few milliseconds and, as soon as the pressure is estab-

lished, cooling of the specimen, e.g., with a jet of pressurized liquid

nitrogen, takes place. Before cutting the sections the copper jacket has

to be removed using the cryo-ultramicrotome, designed to enable

cutting at liquid nitrogen temperature. This is normally done with a

diamond trimming device to form a square face of about 150–200 µm.

Subsequent sectioning can be carried out using a diamond or a good

glass knife (Figure 7–9). Cutting conditions such as knife-angle, cutting

speed and stroke time, and temperature of the specimen, of the knife,

and of the surrounding liquid nitrogen gas can be adjusted according

to the specimen properties and the desired section thickness. However,

the optimum sectioning conditions have to be found empirically in

each experiment.

In contrast to conventional room temperature sectioning, where the

sections are ﬂ oated on water, cryosections have to be transferred man-

ually to an EM grid with an eyelash, a tedious procedure that requires

patience and great skill. The sections tend to attract contaminating ice

crystals in the preparation chamber of the cryoultramicrotome and

during transfer to the EM. Furthermore they are generally not ﬂ at and

exhibit cutting artifacts, e.g., deformation, crevasses, chatter, and knife

marks (for a complete description of cryosectioning and related cutting

artifacts see Dubochet et al., 1988; Frederik et al., 1991; Michel et al.,

1991; Sitte, 1996; Al-Amoudi et al., 2005).

Chapter 7 Cryoelectron Tomography (CET) 553

While the idea of high-pressure freezing and cryosectioning is rela-

tively old (Moore and Riehle, 1968; Bernhard, 1965) and commercially

available instruments have been available for almost 20 years, its impact

on biological and biomedical research has been comparatively small,

owing to the fact that its use in combination with cryoultramicrotomy

is still technically difﬁ cult, the yield of good micrographs is low, and

the cutting artifacts in most of the sections complicate subsequent

image analysis. However, in recent years cryosectioning was somehow

rediscovered, and improvements have been made to facilitate the freez-

ing and cutting procedure and to minimize the inﬂ uence of cutting

artifacts (Studer and Gnaegi, 2000; Al-Amoudi et al., 2003). Cryosec-

Figure 7–9. Cryo-sectioning. a) Cross-sectional schematic of the cryo-sectioning process with a

magniﬁ cation of the trimmed copper tube and the ‘ribbon’ formed after several cutting cycles. c)

Low-magniﬁ cation image of a cryo-section placed on a TEM grid. d) Cryo-electron micrograph at

tesy of A. Leis and L. Andrees, MPI of Biochemistry, Martinsried, Germany). The cutting artefacts are

clearly visible.

cryo-ultramicrotome. b) View into the cryo-ultramicrotome during sectioning. The inset shows a

higher magniﬁ cation of the sectioned cells (unicellular red algae Cyanidioschyzon merolae; images cour-

554 J.M. Plitzko and W. Baumeister

tioning to fabricate thicker sections of 150–200 nm for cryo-ET studies

or even serial sectioning, likewise used in room temperature experi-

ments (Soto et al., 1994), in combination with cryo-ET would deﬁ nitely

help in the 3D investigation of larger structures. The production of

undistorted vitriﬁ ed sections remains a specialized task and it is by

no means a routinely used technique. Nevertheless the prospects for

the future of cryosectioning in combination with cryo-ET are promis-

ing (Hsieh et al., 2002; Leis et al., 2005).

3.2 Radiation Sensitivity—Beam Damage

When biological specimens are irradiated by the electron beam in the

EM, the specimen structure is damaged as a result of ionization and

subsequent chemical reactions. Ionization occurs if energy is deposited

in the specimen as a result of inelastic scattering events. This can pri-

marily induce the heating of the sample in the irradiated area, radio-

chemical decomposition of water (radiolysis), as well as secondary

chemical reactions (breaking of chemical bonds, formation of new

molecules or even radicals). Unlike materials science, where beam

damage is almost negligible, radiation damage in life sciences is the

fundamental limitation in any cryo-EM investigation.

The damage that occurs depends on the number of electrons trans-

mitted through the sample. Therefore the current density per unit area

j (A/cm

) multiplied by the exposure time t (s) has been chosen as an

appropriate measure for the electron dose (C/cm

, or e

−

/Å

, e

−

= elec-

trons). This is not the same as the deﬁ nition in radiochemistry, where

the dose is deﬁ ned in units of gray; the energy adsorbed per unit weight

(Gy = J/kg). According to the last deﬁ nition, a carbon sample exposed

to 50 e

−

/Å

generated by a 300 kV source would correspond to a dose of

1.6 × 10

Gy, which, e.g., can be observed in the vicinity of a nuclear

reaction!

Since the amount of damage is proportional to the applied dose, and

the amount of exposure received by the specimen escalates with the

square of the magniﬁ cation, it is clear that this is the fundamental

limitation in HREM of biological materials. However, several measures

can be taken to partially evade this dilemma. Although the ionization

processes are not temperature dependent, the resulting beam damage,

e.g., bubbling, mass loss, diffusion of free radicals, is. Therefore, by

lowering the temperature to, for example, liquid nitrogen temperature

(∼190°C), radiation damage effects can be drastically reduced. Cooling

the sample from room temperature to 90 K, the “lifetime” (the time the

sample can be investigated before structural damage is observed) can

be increased by a factor of approximately nine, the so called cryoprotec-

tion factor C

(Knapek, 1982; Conway et al., 1993; Stark et al., 1996). This

cryoprotection factor is given by the ratio of the critical dose N

at dif-

ferent temperatures (N

is deﬁ ned as the dose at which the diffraction

intensity has fallen to 1/e (∼37%) of its original value (Hayward and

Glaeser, 1979). Further cooling to liquid helium temperature (∼4 K) has

been proven to be advantageous for very thin specimens as used for

electron crystallography or even in selected cases for single particle