towards high-resolution three-dimensional imaging of...

Journal of Structural Biology 153 (2006) 1–13

www.elsevier.com/locate/yjsbi

Towards high-resolution three-dimensional imaging of native mammalian tissue: Electron tomography

of frozen-hydrated rat liver sections

Chyong-Ere Hsieh a, ArDean Leith a, Carmen A. Mannella a, Joachim Frank a,b, Michael Marko a,¤

a Resource for Visualization of Biological Complexity, Wadsworth Center, New York State Department of Health, Empire State Plaza, Albany, NY 12201, USA

b Howard Hughes Medical Institute, Wadsworth Center, Empire State Plaza, Albany, NY 12201, USA

Received 26 April 2005; received in revised form 7 October 2005; accepted 8 October 2005Available online 28 November 2005

Abstract

Cryo-electron tomography of frozen-hydrated specimens holds considerable promise for high-resolution three-dimensional imagingof organelles and macromolecular complexes in their native cellular environment. While the technique has been successfully used withsmall, plunge-frozen cells and organelles, application to bulk mammalian tissue has proven to be diYcult. We report progress with cryo-electron tomography of frozen-hydrated sections of rat liver prepared by high-pressure freezing and cryo-ultramicrotomy. Improvementsinclude identiWcation of suitable grids for mounting sections for tomography, reduction of surface artifacts on the sections, improvedimage quality by the use of energy Wltering, and more rapid tissue excision using a biopsy needle. Tomographic reconstructions of frozen-hydrated liver sections reveal the native structure of such cellular components as mitochondria, endoplasmic reticulum, and ribosomes,without the selective attenuation or enhancement of ultrastructural details associated with the osmication and post-staining used withfreeze-substitution. 2005 Elsevier Inc. All rights reserved.

Keywords: Cryo-ultramicrotomy; Cryosections; High-pressure freezing; VitriWcation; Freeze-substitution; Rat liver; Mitochondria; Ribosomes; Endo-plasmic reticulum

1. Introduction and background

A major challenge for structural biology is the under-standing of macromolecular interactions in their cellularcontext. In order to achieve this, the cell must be imaged inthree dimensions in its native hydrated state and at suY-

cient resolution (3–10 nm) so that macromolecular com-plexes may be recognized. At present, cryo-electrontomography is the only available technique for this task.Considerable success with in situ identiWcation of macro-molecular complexes has been obtained with plunge-frozen

* Corresponding author. Fax: +1 518 474 7992.E-mail address: [email protected] (M. Marko).

1047-8477/$ - see front matter 2005 Elsevier Inc. All rights reserved.doi:10.1016/j.jsb.2005.10.004

specimens such as liposomes containing large protein com-plexes, isolated organelles, and small cells (Förster et al.,2005; Frangakis and Förster, 2004; Grünewald et al., 2002;Medalia et al., 2002; Rath et al., 2003; Wagenknecht et al.,2002). Larger cells and bulk tissue present a problem, sincespecimen thickness needs to be limited to 100–200 nm toachieve resolution suitable for macromolecular identiWca-tion by tomography. Thicker specimens are useful forunderstanding long-range 3-D cellular organization; how-ever, even using zero-loss energy Wltering, and high (300–400 kV) accelerating voltage, specimen thickness is limitedto about 1�m, because the electron dose needed to providea suYcient number of singly elastically scattered electronsfor phase-contrast imaging becomes excessive. As a result,

mailto: [email protected]

mailto: [email protected]

2 C.-E. Hsieh et al. / Journal of Structural Biology 153 (2006) 1–13

tomographic imaging of plunge-frozen large cells, such asthose of the slime mold Dictyostelium (Medalia et al., 2002),is usually limited to cellular margins or Xattened regions.

Recent progress in the Weld of cryo-ultramicrotomy hasmade it possible to prepare frozen-hydrated sections from avariety of biological specimens, thus overcoming the abovethickness limitation. Key technical contributions havecome from several laboratories, including those of Dubo-chet (Al-Amoudi et al., 2005; Chang et al., 1983), Mueller(Michel et al., 1991, 1992), Lepault (Erk et al., 1998), Som-lyo (Somlyo et al., 1977, 1985), and Leapman (Shi et al.,1996). This work is detailed in a comprehensive review bySitte (1996). Cryo-electron microscopy of frozen-hydratedsections has reached a stage of development that allows itsapplication to several types of biological specimens. Recentsuccessful applications of cryo-ultramicrotomy includeDNA and chromatin (Leforestier et al., 2001, 2005; Sartori-Blanc et al., 2001), rat liver (Hsieh et al., 2002), bacteria (Al-Amoudi et al., 2004; Matias et al., 2003; Zhang et al., 2004),yeast (Schwartz et al., 2003), human skin (Norlén et al.,2003; Norlén and Al-Amoudi, 2004), algae (Leis et al.,2005), and isolated chloroplasts (Nicastro et al., 2005).

We are pursuing electron tomography of frozen-hydratedmammalian tissue using two model systems of greatly diVer-ing metabolic activity and water content, mouse skin (Hsiehet al., 2004) and rat liver (Hsieh et al., 2002). The presentreport focuses on the latter tissue, which presents specialchallenges for cryo-preservation and cryo-ultramicrotomy.In a previous report on rat liver (Hsieh et al., 2002), wedescribed the Wrst successful application of electron tomogra-phy to frozen-hydrated mammalian tissue. Artifacts associ-ated with cryo-ultramicrotomy include knife marks,crevasses, chatter, and compression (Al-Amoudi et al., 2005;Chang et al., 1983; Frederik et al., 1982, 1984; Michel et al.,1991, 1992; Richter et al., 1991; Richter, 1994; Zierold, 1994).Our tomographic analysis revealed that the Wrst two types ofdefects were predominantly conWned to the section surface,and that there was considerable useful ultrastructural infor-mation in the interior of the sections (Hsieh et al., 2002).

In the present study, we report further progress in thedevelopment of techniques for tomographic imaging of fro-zen-hydrated mammalian tissue. (1) We demonstrate thatimaging frozen-hydrated sections using zero-loss energyWltering is eVective in increasing contrast without anincrease in electron dose. (2) We identify conditions underwhich crevasses on the surface of the section may bereduced or eliminated, even for sections thicker than150 nm. (3) We show that Quantifoil grids aVord severaladvantages for tomography of frozen-hydrated sections. (4)We employ the needle-biopsy technique for rapid excisionof tissue for high-pressure freezing.

In addition, we compare tomographic images of rat livertissue recorded from frozen-hydrated sections, with thoserecorded from plastic sections of freeze-substituted material.The structural preservation is equally good in both cases, butcomplications in structural interpretation caused by metalstaining are avoided with frozen-hydrated material.

2. Methods

2.1. High-pressure freezing

Tissue was prepared by high-pressure freezing (Moor,1987; Shimoni and Müller, 1998; Studer et al., 1989), themethod of choice for specimens in the 100- to 300-nmthickness range. Both EM-PACT (Leica, Vienna, Austria;Studer et al., 2001) and HPM 010 (Bal-Tec, Balzers, Liech-tenstein) high-pressure freezers were used, with equallygood results. The sample carrier used for the HPM 010 wasa 3-mm aluminum platelet with 2-mm cavity diameter, 0.3-mm depth, and 0.5-mm total thickness. The sample carrierfor the EM-PACT was a Xat platelet 3 mm in diameter and0.6 mm in thickness, with a central slot 0.3 mm wide and1.2 mm long.

Specimens of liver tissue were obtained from white Spra-gue–Dawley rats shortly after sacriWcing. Tissue to befreeze-substituted was excised and diced in Ringer’s buVer,then placed in a specimen carrier for high-pressure freezingthat had been pre-Wlled with 1-hexadecene. The Wrst speci-men was frozen about 1 min after excision. Early specimensused for cryo-ultramicrotomy were prepared in the sameway, but the best results were obtained by needle biopsyusing kits from Leica (Vanhecke et al., 2003) and Bal-Tec(Hohenberg et al., 1996). When using needle biopsy, tissuecould be transferred to the specimen carrier and frozenwithin 40 s of cessation of blood Xow. For the recent workwith frozen-hydrated sections, the free space in the speci-men carrier was Wlled with 20% dextran (Sartori et al., 1993)in mammalian Ringer’s solution. We found that the dex-tran-containing buVer provided superior results with fro-zen-hydrated sections.

2.2. Freeze-substitution and preparation of plastic sections

For comparison with frozen-hydrated sections, weavoided specialized freeze-substitution procedures, andinstead employed a simple, straightforward procedure(Hohenberg et al., 1994; Van Harreveld and Crowell, 1964).Tissue was removed from the high-pressure freezing carri-ers under liquid nitrogen and was freeze-substituted in 2%OsO4 in acetone as follows: 8 h at ¡90 °C, 8 h at ¡60 °C, 8 hat ¡30 °C, and 1 h at 0 °C. This was followed by three 15-min rinses in acetone at 0 °C, followed by inWltration ofgraded Epon-Araldite at 30, 70, and 100%, at 4 °C, 12 h foreach step. Polymerization was at 60 °C. The procedure wascarried out with a Bal-Tec FSU 010 automated freeze-sub-stitution device.

Plastic sections were cut at 80- to 250-nm thickness withan Ultracut-S ultramicrotome (Reichert, Vienna, Austria)and a 45° diamond knife (Diatome, Biel, Switzerland). Thesections were picked up on 200-mesh copper grids coatedwith a 60-nm-thick Formvar Wlm and a 10-nm-thick carbonWlm, to which a suspension of 10-nm colloidal gold parti-cles was applied to serve as alignment Wducial markers fortomographic tilt series. Sections were stained with 2% aque-

C.-E. Hsieh et al. / Journal of Structural Biology 153 (2006) 1–13 3

ous uranyl acetate for 4–10 min and Reynolds lead citratefor 2–5 min, both at room temperature.

2.3. Cryo-ultramicrotomy

The method used for routine sectioning is very similarto that described earlier (Hsieh et al., 2002). Frozen-hydrated sections are cut with a UCT ultramicrotomewith an EM-FCS cryo kit (Leica, Vienna, Austria). Withinthe cryo-ultramicrotome chamber, the two halves of thehigh-pressure freezing specimen carrier are separated,exposing the tissue. The carrier is clamped, on edge,directly in a Xat microtome chuck. A 45° trimming dia-mond knife (Diatome) is used to cut away portions of thecarrier metal, and to trim the block face to approximately100 £ 100 �m for sectioning. If the block face does notappear perfectly smooth (indicating reasonably goodfreezing), the block is trimmed more deeply, or discarded.Sections are cut at ¡150 or ¡160 °C with a microtomefeed setting of 40–200 nm, with a 35° or 25° “dry” cryo-diamond knife (Diatome). The knife clearance angle is 6°,the cutting speed 0.2–0.6 mm/s, and the return cycle30 mm/s. An adjustable Static Line Ionizer II (Haug, Biel,Switzerland) is located 30 mm from the knife, and is usedonly during sectioning. We avoided high cutting speeds,recently recommended for reduction of chatter (Al-Amo-udi et al., 2005), because of rapid knife damage when cut-ting thick sections.

2.4. Mounting of frozen-hydrated sections for tomography

Two types of copper specimen grids are used. For long-term storage in liquid nitrogen, 200-mesh hexagonal fold-ing grids are used, as previously described (Hsieh et al.,2002). In our more current practice, when sections areexamined within a week of microtomy, sections are col-lected on 200-mesh R3.5/1 Quantifoil grids (QuantifoilMicrotools, Jena, Germany). The grids are coated with a10-nm-thick continuous carbon Wlm, and a mixture of 10-and 20-nm colloidal gold particles is then applied to thecarbon Wlm.

A short ribbon consisting of about eight sections is cut,and is left attached to the knife edge. The knife is thenmoved laterally, and another ribbon is cut. After a numberof sections or ribbons have accumulated on the knife edge,the static ionizer is turned oV, and the sections are trans-ferred to the grids using a white hair from a Dalmatian dog(Forbes, 1986). This type of hair resists breakage at lowtemperature.

To facilitate collection of the sections on grids, we fabri-cated a polished metal shelf that is installed just adjacent to,and at the same level as, the knife edge. Freshly glow-dis-charged grids, prepared as described above, are placed on afresh strip of 0.127-mm-thick indium foil (Alfa Inorganics,Beverly; Buchanan et al., 1993). Sections are placed as closeas possible to the center of the grid, and are also centeredon the grid squares, so as to maximize the tilt range during

tomography. About four ribbons of sections are collectedon each grid.

The grids are then pressed into the soft indium foil usingthe glass screw-press tool supplied with the microtome.During this process, the indium foil comes in contact withthe back side of the Quantifoil support Wlm, and supportsthe Wlm as the sections are pressed down. The screw-presstool is preferred over the polished metal rod press tool, alsosupplied with the microtome, because section damage andfrost accumulation is less. The grids are then placed in car-riers and transferred to the cryo-transfer specimen holderworkstation, or to storage under liquid nitrogen. If gridsare stored, rather than used on the same day as cut, thegrids are pressed again in the microtome cryo-chamberbefore transfer to the specimen holder workstation.

2.5. Electron tomography

Images were recorded at 400 kV acceleration voltageusing a JEM-4000FX electron microscope (JEOL, Tokyo,Japan) equipped with a GIF2002 imaging energy Wlter(Gatan, Pleasanton, CA). The energy Wlter is equipped witha 2048 £ 2048-pixel CCD camera that is Wber-optically cou-pled to a high-sensitivity phosphor scintillator. The CCDelement size is 30�m, and all images are binned to1024 £ 1024 pixels, thereby doubling the eVective CCDpixel size and compensating for the point-spread functionof the thick scintillator. The image pixel size relative to thespecimen was either 1.8 or 1.0 nm. All images were recordedusing zero-loss Wltering with a 15-eV slit. The objectiveaperture angle was 10 mrad.

All tilt series were recorded semi-automatically at auniform 1° or 2° interval from ¡60 to +60°, using soft-ware written in-house (M. Marko and A. Leith, unpub-lished). The software, which is controlled by a webpageuser interface, integrates an Emispec ESVision imagingsystem (FEI, Tempe, AZ) and the JEOL stage- and EM-control computers, as well as the Gatan Digital Micro-graph software. In addition to controlling the EM, thestage, and the CCD camera of the energy Wlter, the soft-ware automatically adjusts the energy oVset to keep thezero-loss peak centered in the slit during the tilt series.The acquisition software also integrates a FastScan 114frame-transfer CCD camera (TVIPS, Gauting, Germany).This camera is mounted next to the energy Wlter and isused for reduced-magniWcation re-focusing and trackingafter each tilt. These operations are done manually andon-axis, at an electron dose 1/100th of the exposure dose.The FastScan camera is also used for low-dose surveyingof the specimen and for recording of electron diVractionpatterns to verify the state of the ice.

Images from frozen-hydrated sections were recorded at15�m underfocus, which corresponds to a spacing of 5 nmat the Wrst minimum of the contrast transfer function. Theaverage electron dose per image was scaled down from0.7 e¡/Å2 for a 2°-interval tilt series to 0.35 e¡/Å2 for a 1°interval, in order to keep the total dose to about 50 e¡/Å2.


Images from plastic sections were recorded at 1 �m under-focus and a dose of 10 e¡/Å2 per image. For a 1°-interval tiltseries, the total electron dose was 1300 e¡/Å2, whichincluded pre-irradiation to stabilize the section.

A cryo-transfer specimen holder (Model 626; Gatan,Warrington, PA) was used for both types of specimen, butthe plastic sections were imaged at room temperature, whilethe frozen-hydrated sections were imaged at ¡176 to¡179 °C. In both cases, dual liquid-nitrogen anticontami-nators were used.

Alignment of the tilt series images was done using Wdu-cial markers (Penczek et al., 1995), and the 3-D reconstruc-tions were made by weighted back-projection usingSPIDER (Frank et al., 1996). In single-tilt reconstructions,resolution varies along the three axes, as a function of theimage sampling size and microscope defocus (y-axis), thespecimen thickness and number of projections (x-axis;Crowther et al., 1970), and the maximum tilt angle (z-axis;Radermacher and Hoppe, 1980). In the current reconstruc-tions, the predicted resolution limits were, typically, 5 nmalong the tilt axis direction (y), 10–12 nm (depending onsection thickness) along the x-axis, and 15–18 nm along thez-axis.

The thickness (t) of frozen-hydrated sections was mea-sured both directly from the tomograms and by electronenergy-loss imaging, using the relationship t/� D ln (It/I0),where � is the mean-free path of 400 keV electrons in vitre-ous ice containing biological material, I0 was taken as themean gain-normalized CCD counts with the 15 eV slit in,and It the same with the slit out. The value for � was deter-mined to be 380 nm, based on comparison with thicknessdetermined from tomograms (results to be reported else-where).

3. Results

3.1. Contrast gain from zero-loss energy-Wltered imaging

The important advantage provided by energy Wltrationwhen imaging sections of unstained, frozen-hydrated tis-sue is demonstrated in Fig. 1 for a 220-nm-thick liver sec-tion. A zero-loss Wltered image and an unWltered imagefrom the same Weld, recorded using the same incident elec-tron dose, 1 e¡/Å2, are presented in Figs. 1A and C, respec-tively. Line scans in identical locations on both imagesindicate that the zero-loss image has about half the back-ground intensity as the unWltered image, while retainingsimilar signal intensity. For example, the mean signal-to-background ratio along the indicated line was 0.41 in theWltered image and 0.24 in the unWltered image, representingan increase by a factor of 1.7. Another measure of the con-trast gain can be made by comparing the relative intensi-ties of large, uniformly textured areas of diVering densityin the two images, using the relationship C D (I2 ¡ I1)/I1where C is contrast and I1 and I2 are the mean CCD countswithin boxes at the two areas. In the example, the contrastmeasured between the cytoplasm and the mitochondrialmatrix in the Wltered image is Cf D 0.091, and in the unWl-tered image, Cu D 0.030. The contrast increase is Cf/Cu D 3.Note that these methods of contrast and signal-to-back-ground estimation depend on the inherent contrast of thespecimen features, as well imaging conditions such as thedefocus setting. These results are consistent with animprovement in contrast, demonstrated using cross-corre-lation methods, using a variety of specimens (although notincluding frozen-hydrated specimens) imaged at 120 kV(Grimm et al., 1998).

Fig. 1. Contrast gain provided by zero-loss energy Wltering when imaging a 220-nm-thick frozen-hydrated section of rat liver. Comparison of Wltered (A)and unWltered (C) images taken at the same incident dose, 1 e¡/Å2. Mean CCD counts are 1127 for the Wltered image and 1966 for the unWltered image.Boxes show regions of the mitochondrial matrix (M) and cytoplasm used to measure contrast gain, as described in the text. Intensity proWles (B and D)across mitochondrial inner and outer membranes are along the 30-pixel ( D 30 nm) lines indicated in (A) and (C), respectively. Bar D 100 nm.


3.2. Correlation between crevasse formation and quality of high-pressure freezing

Images of typical thick sections of frozen-hydrated ratliver, recorded at 400 kV with zero-loss energy Wltering, arepresented in Figs. 2 and 3. Knife marks, which indicate thecutting direction, extend diagonally from lower left toupper right in Figs. 2 and 3C, and from upper left to lower

Fig. 2. Projection image of a 200-nm-thick frozen-hydrated section of ratliver tissue. Note the smooth membranes and the ribosomes studding theendoplasmic reticulum (arrowheads). Electron dense beads in this and thefollowing Wgures are colloidal gold particles used as Wducial markers foralignment of the tomographic tilt series. Bar D 200 nm.

right in Fig. 3A. The images show considerable variabilityin the extent of crevasse formation. Crevasses are seen asnarrow grooves, about 20 nm wide and 50–200 nm apart,running normal to the cutting direction. Crevasses, whenthey occur, may be reduced by cutting thinner sections, andthe depth of crevasses may be decreased by high (50–100 mm/sec) cutting speeds, although this increases thenumber of crevasses (Al-Amoudi et al., 2005). The sectionsin Figs. 2 and 3C are nearly free of crevasses, and the mito-chondria and endoplasmic reticulum (ER) are easily recog-nized. In contrast, the crevasses are so severe in Fig. 3A thatthose organelles are barely discernible. Using tomographicanalysis, we previously showed that crevasses typicallyextend only 20–50 nm into the section, and only from thesurface that was the block face before the section was cut(Hsieh et al., 2002). However, signiWcant crevasse formationcan impair the ability of the microscopist to select regionsof interest in sections for tomographic analysis.

We have found that the severity of crevasse formationtends to vary with the quality of tissue preservation. Whentissue is not optimally preserved, as characterized primarilyby scalloped membranes (Fig. 4), sections display moderateto severe crevasses and are more diYcult to cut in the cryo-ultramicrotome. Conversely, sections that show excellentultrastructural preservation are generally easier to cut, andtend to have very few crevasses (e.g., Figs. 2 and 3C), evenwhen sections 200 nm or thicker are cut. The tissue shownin Fig. 4 was prepared using our previous technique, minc-ing followed by freezing in 1-hexadecene, while all the otherexamples of frozen-hydrated sections are from materialprepared using needle biopsy followed by freezing in 20%dextran. Although crevasse-free sections are more oftenobtained using the latter technique, severe crevasses canstill occasionally be observed (Fig. 3A).

Fig. 3. Images and electron diVraction patterns of frozen-hydrated sections of rat liver. (A) 180-nm-thick section marked by numerous crevasses (examplesat arrows). (B) Electron diVraction pattern corresponding to (A). (C) Much smoother section, with the same thickness as in (A). (D) Electron diVraction
pattern from the specimen in (C). (E) Electron diVraction pattern from a 100-nm-thick Wlm of plunge-frozen pure water. DiVuse bands at spacings of 0.214and 0.370 nm correspond to vitreous ice. Bars D 200 nm.


It is known that smooth sections are not obtained fromfrozen-hydrated specimens containing crystalline ice (e.g.,Chang et al., 1983; Erk et al., 1998; Frederik et al., 1984).Specimens that cannot be completely vitriWed by high-pres-sure freezing contain ice microcrystals that range in sizefrom 2 to 30 nm (Echlin, 1992; Michel et al., 1991; Moor,1987; Sartori et al., 1993). Therefore, it is reasonable to con-sider that crevasse formation might be due to sub-optimalfreezing, which would also be reXected in poor ultrastruc-tural preservation. To test this hypothesis, we used electrondiVraction to characterize the state of the ice in frozen-hydrated liver sections that have drastically diVerentdegrees of crevasse formation, as shown in Figs. 3A and B.Both sections were found to contain predominantly vitre-ous ice, as indicated by the presence of two diVuse bands inthe patterns (Figs. 3B and D) at spacings characteristic ofvitreous ice (Dubochet et al., 1988). For comparison, thediVraction pattern of a 100-nm-thick layer of plunge-frozenpure water is shown in Fig. 3E. Thus, we were not able tocorrelate severe crevasses in the liver sections with a micro-crystalline ice content that is high enough to be detected assharp rings or spots in the diVraction patterns.

3.3. Attachment of frozen-hydrated sections to EM grids

Another important characteristic of frozen-hydratedsections, from the perspective of tomographic data collec-tion, is that they generally are not Xat (e.g., Richter et al.,1991). The sections may show a complicated curvature,often with several undulations roughly along the cuttingdirection. Cross-sections through three diVerent tomo-grams of frozen-hydrated rat liver, as shown in Fig. 5, illus-trate the problem caused by lack of section Xatness. Due tosection curvature, the areas of contact between the sectionand the support Wlm are few, and in many places there is agap between the two. This defect has two consequences:poor attachment of the section to the grid, causing instabil-

ity in the electron beam during tilt-series collection (seeFig. 5 in Frank et al., 2002), and reduced alignment accu-racy, due to the large distance between the section and thecarbon Wlm which carries the colloidal gold particles thatare used as Wducial markers. In the latter case, residualerrors in alignment of the tilt series, present in the plane of

Fig. 5. Tomographic cross-sections of frozen-hydrated rat liver sectionsshowing the separation between the support Wlm (white dotted lines) andthe section (black dotted lines). (A) 230-nm-thick section in good (uni-form) contact with the support Wlm. (B) 210-nm-thick section separatedfrom the support Wlm by 100 nm. (C) 190-nm-thick section that is curvedand separated from the support Wlm by 100–420 nm. Note that the sur-faces of these sections lack obvious grooves characteristic of crevasses, aswere seen in the cross-sections shown previously (Fig. 3E in Hsieh et al.,2002). Gold Wducial markers on the carbon Wlm are indicated by arrow-heads. Bars D 100 nm.

Fig. 4. Tomographic imaging of a 230-nm-thick section of rat liver tissue. (A) Untilted projection image, with several moderate crevasses evident (arrow-heads). (B) 1.8-nm-thick slice from tomogram showing clear deWnition of cellular features and absence of surface artifacts. Arrowhead indicates compres-sion of the space between the inner and outer mitochondrial membranes in the cutting direction. (C and D) Indications of mechanical disruption ofmembranes: (C) lateral dislocation (at arrow); (D) ruZing. Arrowheads in (A) indicate the sectioning direction. Bars D 200 nm; magniWcation doubled in(C) and (D).


the gold particles, are magniWed in the section when it issome distance from the plane of the markers. In an attemptto improve section Xatness and induce attachment, we rou-tinely press sections onto grids with a glass section press(see Section 2) or a metal rod (Frederik et al., 1982; McDo-wall et al., 1983). While these procedures may Xatten thesections to some degree, results are not consistent. Also,extensive pressing of the sections may cause damage, suchas deep vertical fractures sometimes observed in tomo-grams of thicker sections, usually along membrane surfaces.We Wnd that, in general, thin sections attach better to thegrid after pressing than do sections thicker than 150 nm.

We Wnd that Quantifoil grids oVer several advantagesfor electron tomography of frozen-hydrated sections, com-pared to the folding grids that we previously recommended(Hsieh et al., 2002). The local attachment of the section tothe support Wlm may be easily assessed at low magniWca-tion by observing the amount of movement of the specimenrelative to the array of holes as the stage is tilted. We rou-tinely record low-magniWcation on-axis images at 0° and30° tilt. If there is excessive displacement of the specimenrelative to the Quantifoil Wlm, a tilt series is not usuallyattempted. Since the thick holey carbon provides the mainspecimen support, only a thin (10 nm) carbon coat isneeded for carrying the colloidal gold particles over theholes. Indeed, for the highest-resolution work, tilt seriesmay be collected from sections over uncoated holes, relyingon gold particles on the edge of the hole, or a markerlessalignment method. The array of regular holes makes it easyto re-locate the specimen after in-plane rotation for collec-tion of double-tilt series. Since the grids are single-thick-ness, tilt series may be collected closer to the grid barswithout occlusion at high tilt. The disadvantage of singlegrids such as Quantifoil over double (folding) grids is thatthe sections tend to detach during long-term storage in liq-uid nitrogen, and they are more likely to accumulate frostduring handling since they cannot safely be wrapped inindium foil. However, the advantages far outweigh thepotential loss of sections for experiments that are not lim-ited by specimen availability.

3.4. Tomographic reconstruction of liver: morphology of mitochondrial cristae

A tomographic reconstruction of a small Weld from afrozen-hydrated rat liver section is presented in Fig. 4. Cre-vasses in this case were moderate (Fig. 4A) and structuraldetails in the Weld were easily recognized (unlike Fig. 3A). Aslice from the tomogram (Fig. 4B) reveals that the tissue inthe section has undergone noticeable compression in thecutting direction. Such compression is a common charac-teristic of frozen-hydrated sections prepared by cryo-ult-ramicrotomy. The compression of the specimen of Fig. 4 isreadily apparent in spacings between mitochondrial outerand inner membranes. The distance between these mem-branes is generally smaller when the membranes are alignednormal to the cutting direction (arrowhead in Fig. 4B; also

visible in Fig. 3C). In a few places, membranes display localstep-like dislocations (Figs. 4C and D; see also Matias et al.,2003). Fortunately, such severe membrane defects tend tobe localized, and occur at low frequency in well-frozentissue.

The overall structural preservation of the tissue in Fig. 4was good enough to warrant closer examination of the 3-Dmorphology of the membranes in the mitochondrion thatoccupies most of the Weld. The infoldings of the mitochon-drial inner membrane are called cristae, which until recentlywere represented as random baZe-like folds in the innermembrane. However, tomographic reconstructions of plas-tic-embedded specimens, Wrst of rat liver mitochondria(Mannella et al., 1994, 1997; Penczek et al., 1995), and laterof a wide variety of mitochondria (reviewed in Frey andMannella, 2000), have conclusively shown that the cristaeare normally connected by narrow openings to the periphe-ral region of the inner membrane. This structural principlewas later conWrmed for intact, plunge-frozen mitochondriaisolated from rat liver (Mannella et al., 2001), as well asfrom fungi (Nicastro et al., 2000). Close examination of sev-eral cristae in the tomographic reconstruction of Fig. 4, aswell as examples from a tomogram of another sample ofliver tissue, conWrms that the cristae in mitochondria in fro-zen-hydrated liver have the same circular or small slit-likeopenings (Fig. 6).

3.5. Comparison of frozen-hydrated and freeze-substituted rat liver tissue

It is informative to compare “typical best” images ofhigh-pressure frozen rat liver in frozen-hydrated sections,and in plastic sections following freeze-substitution. Thelatter is generally considered to be the best technique widelyavailable for preserving tissue in a close-to-native state forelectron microscopic imaging (Dahl and Staehelin, 1989;Giddings, 2003; McDonald, 1994, 1999; McDonald andMüller-Reichert, 2002; Studer et al., 1989).

In general, images of frozen-hydrated and freeze-substi-tuted liver sections present comparable views of overall cel-lular architecture, in terms of the morphology anddistribution of components such as mitochondria, ER,ribosomes, and nuclear pores (Figs. 7 and 8). In well-frozenmaterial, the membranes of cells and organelles are smooth,not scalloped, in both frozen-hydrated and freeze-substi-tuted tissue. In the uncompressed direction, the center-to-center spacings between the peripheral (outer and inner)membranes and adjacent internal (cristae) membranes ofmitochondria in the frozen-hydrated tissue fall in the rangeof 7–12 nm. These membrane spacings are the same orslightly wider than those observed in freeze-substituted tis-sue. Note that the membranes in the latter specimens tendto be negatively stained, appearing white against a darkbackground. While negatively stained membranes are notalways found in high-pressure frozen, freeze-substitutedmaterial (e.g., Giddings, 2003), it is believed that thisappearance indicates optimal preservation. With optimal


freezing, the membrane proteins are intact, protecting the work (Matsko and Mueller, 2005) has shown that speci-
polar head-groups, and forming a barrier to osmium duringfreeze-substitution, such that subsequent post-stains arenot taken up. While with freeze-substitution using 2%osmium in acetone, this optimal preservation is only rarelyobserved (M. Mueller, personal communication), recent
mens stabilized by freeze-substitution in a mixture of epon/araldite in acetone reliably exhibit excellent preservation,with negative staining of membranes. However, this nega-tive staining of membranes in freeze-substituted specimenstends to make the membranes less distinct, especially

Fig. 6. Morphology of crista openings in mitochondrial inner membranes. (A, D, and G) 1.8-nm-thick slices from tomograms of frozen-hydrated sectionsof rat liver showing the attachment of cristae to the inner boundary membrane. (B, C, E, F, H, and I) Surface renderings of the cristae showing the innermembrane junctions. In each case, the opening is a small circle or slit. (A–C) and (D–F) are from the tomogram of Fig. 4; (G–I) is from a tomogram ofanother sample. Bar D 50 nm.

Fig. 7. Comparison of projection images of frozen-hydrated (A and C) and freeze-substituted (B and D) rat liver sections. Section thickness in both cases is100 nm. (A and B) Fields containing mitochondria and adjacent rough endoplasmic reticulum. Note in (B) that membranes are negatively stained, whilethe mitochondrial matrix and ribosomes are strongly positively stained. Arrowheads indicate individual examples of ribosomes lining the endoplasmicreticulum. (C and D) Fields that each contain the edge of a nucleus (lower right) and adjacent cytoplasm. Arrows indicate nuclear pores. In the nucleus in(C), note areas of heterochromatin outlined by dots, while in (D) heterochromatin (¤) is heavily stained. The white patches in the cytoplasm in (D) areunstained glycogen deposits. Bars (A and B) D 100 nm; (C and D) D 150 nm.


against a dense background. In this case, the crista mem- likely reXects the degree of anoxia to which the diVerent tis-
branes in images of frozen-hydrated liver sections are morevisible against the mitochondrial matrix than those in thefreeze-substituted sections.
The osmication and post-staining of proteins and nucleicacids in the freeze-substituted tissue strongly contrast themitochondrial matrix and the heterochromatin, while clus-ters of glycogen are visible as stain-excluding plaques in thecytosol. Glycogen granules cannot be unambiguouslyidentiWed in the frozen-hydrated tissue, suggesting the occur-rence of contrast matching with the cytosol. In the frozen-hydrated tissue, the density of the mitochondrial matrix issimilar to that in plunge-frozen isolated rat liver mitochon-dria (Mannella et al., 2001). Heterochromatin, obvious infreeze-substituted cells, is seen in frozen-hydrated cells asregions of comparable size and distribution that appear withconsiderably less contrast relative to the rest of the nucleo-plasm. Ribosomes are clearly visible in both cases.

It should be noted that there were diVerences in the Wrststeps of specimen preparation for the tissue specimens thatare compared here. These diVerences, in tissue excision andchoice of Wller for high-pressure freezing, are described inSection 2. They have no bearing on the issues discussedabove, which relate to contrast diVerences caused by osmi-cation and post-staining of the freeze-substituted speci-mens. There is another apparent diVerence between thefrozen-hydrated and freeze-substituted sections of rat liverin Fig. 7, namely a somewhat greater dilation of ER cister-nae in the frozen-hydrated tissue. This dilation wasobserved in both types of specimens to varying degrees, and

sue specimens were subjected prior to freezing. This physio-logical eVect may be preventable by keeping the animalsalive (under anesthesia) during tissue excision (e.g., by nee-dle biopsy), so that organs are perfused right up to excision.It is, however, worth noting that there were no indicationsof major cellular pathology, such as severe mitochondrialswelling, in these specimens.

4. Discussion

Considerable progress is being made in cryo-electrontomographic imaging of soft mammalian tissue such as liver.The Wrst critical step is routine preparation of suitable sec-tions. Success, as noted above, appears to partially depend onthe quality of freezing. For sections greater than 150nm inthickness, crevasses (when they occur) and knife marks areconWned to the surface and are not major problems for ana-lyzing the interior of the sections. The second critical step isattachment of the sections to the grid, and improvements inthis area are the focus of ongoing development eVorts.

4.1. Section attachment

Lack of section Xatness, resulting in poor attachment ofsections to the support Wlm, constitutes the single mostimportant obstacle to successful tomographic data collec-tion with these specimens. We have imaged hundreds offrozen-hydrated liver tissue sections, and have thus far col-lected 15 tomographic tilt series. In general, more than 50%

Fig. 8. Tomographic reconstructions of sections of 200-nm-thick frozen-hydrated (A–C) and freeze-substituted (D–F) rat liver tissue. 1.8-nm-thick slicesfrom tomograms. (B, C, E, and F) 1.5£ magniWcation of areas boxed in (A) and (D), respectively. As in Fig. 7, membranes are unstained in the freeze-substituted material, and ribosomes appear darker and than those in the frozen-hydrated section. The cisternae of the endoplasmic reticulum has a similarwidth in both preparations (B and E). The separation of mitochondrial cristae membranes (arrows) is the same in both (C and F). Bars D 200 nm.


of the sections may be successfully imaged in the electronmicroscope. However, fewer than 10% are suitable for elec-tron tomography. Most of the sections are unsuitablebecause of an unacceptable gap between the support Wlmand the section, or because of instability on electron irradi-ation. Use of Quantifoil grids aids in rapid screening forregions of good section attachment. Some sections areunusable for tomography for other reasons, such as poordistribution of gold particles, limited tilt range due tounsuitable location of the area of interest on the grid, orsub-optimal ultrastructural preservation.

Solutions to the attachment problem, such as surfacemodiWcation of supporting Wlms and the design of newtypes of grids, are being explored. Vonk (2000) reportedthat wrinkles in the carbon support Wlm at low tempera-tures can be avoided through the use of molybdenumrather than copper grids, and we have preliminary evidencethat this alternative helps with attachment of frozen-hydrated sections. In limited tests with molybdenum Quan-tifoil grids, using sections of cyanobacteria (Ting et al.,2005), we Wnd that there are more areas of good attachmentwhen the carbon Wlm is Xat, and that the increased stiVnessof molybdenum grids compared to copper grids helps pre-vent distortion of the support Wlm during handling andwhen pressing the sections onto the grid.

4.2. Section compression

Section compression lowers the Wdelity of representationof the native state of the tissue. The degree of compressionthat we observed in sections of frozen-hydrated rat liver tis-sue cut with a 35° knife falls in the range 30–50%, with thegreatest compression observed in sections thinner than100 nm. Reports in the literature for specimens such as livertissue, plant cells, yeast, bacteria, and skin (Al-Amoudiet al., 2005; Chang et al., 1983; Michel et al., 1992; Richter,1994; Shi et al., 1996) indicate a similar range of compres-sion for each of these specimens. Al-Amoudi and colleagues(2003) found that compression in frozen-hydrated sectionsmay be reduced by the use of an oscillating diamond knife,but only for a small fraction of sections. However, the the-ory behind the oscillating knife is sound, as proven by suc-cess with plastic sections (Studer and Gnägi, 2000). Forfrozen-hydrated sections, the eVect of the oscillating knifemay be specimen-dependent, and the beneWt may be greaterfor thicker sections. Thus, we feel that continued testing iswarranted.

Interestingly, we have observed that some structures infrozen-hydrated sections, such as desmosomes in skin, donot display the same degree of compression-related distor-tion as do neighboring structures, (Hsieh et al., 2004). Like-wise, ribosome proWles in rat liver are not obviouslyshortened in the compression direction; this was alsoobserved by Dubochet and Sartori Blanc (2001) in plant tis-sue. Other examples of macromolecular assemblies not dis-torted by section compression include nucleosome coreparticles (Leforestier et al., 2001), hexagonally packed

sperm chromatin (Sartori-Blanc et al., 2001), and clusters ofthe chemotaxis receptor Tsr in Escherichia coli (Zhanget al., 2004). Our conclusion is that the degree of compres-sion of various cellular structures within frozen-hydratedsections is nonuniform, with the relatively more-rigid struc-tures resisting large-scale distortion better than Xexiblestructures such as membranes. This structure-speciWc char-acteristic may make it practically impossible to computa-tionally correct tomograms for the eVects of compression.

4.3. Crevasse formation and freezing quality

The section in Fig. 2 is 200-nm thick (as measured byelectron energy-loss imaging), and the image was recordedafter a total electron dose of 5 e¡/Å2, well below the doseneeded to smooth out crevasses (Hsieh et al., 2002). Never-theless, crevasses are few. However, others (e.g., Al-Amoudiet al., 2005) report that crevasse-free sections are not to beexpected at thickness above 70–100 nm. We Wnd that excel-lent high-pressure freezing, as evidenced by especially goodpreservation of the biological structure, correlates withimproved sectioning quality. Unfortunately, high-pressurefreezing appears to give variable results. Even when the besttechnique is used, only about 40% of the blocks containoptimally frozen tissue, and this can only be identiWedthrough sectioning and electron microscopy.

It is known that the tendency to form ice microcrystalswithin high-pressure frozen material is related to the localwater content, with a greater tendency for microcrystalswhere the water content is higher (Michel et al., 1991;Moor, 1987; Sartori et al., 1993; Studer et al., 1995). In ourwork, we have observed that, with good high-pressurefreezing, specimens having a relatively low water content,such as yeast cells (except for their vacuoles) and mouseskin (Hsieh et al., 2004), are usually easier to section, andhave less severe crevasses than do liver tissue sections. Thisobservation is consistent with reports of high-quality fro-zen-hydrated sections obtained from specimens having rel-atively low water content, such as yeast (Al-Amoudi et al.,2003; Schwartz et al., 2003), bacteria (Matias et al., 2003;Zhang et al., 2004), and skin (Norlén et al., 2003; Norlénand Al-Amoudi, 2004). Nevertheless, with well-frozen livertissue, we do not see a diVerence in crevasse formationbetween the mitochondrial matrix and the surroundingcytoplasm, even though there is a considerable diVerence inwater content.

Using typical protocols, pre-Wxation treatment (e.g.,eVects due to anoxia) should not have a major eVect on theoverall water content of the tissue. Therefore, the consider-able variability we see in diVerent high-pressure frozenblocks from the same experiment must be due to factorssuch as the geometry of the individual specimen in the car-rier, the disposition of the Wller, and variations in the opera-tion of the machine from run to run. Echlin (1992) hassuggested that a mixture of vitreous and microcrystallineice may exist in aqueous specimens following high-pressurefreezing. Thus, it seems reasonable that the proportion of


vitreous to microcrystalline ice might vary depending onthe particular conditions during high-pressure freezing.Such a variation in the fraction of microcrystalline icecould account for observed diVerences in structural preser-vation and sectioning quality in the liver specimens. Thismay explain why crevasse-free thick sections are obtainedonly from optimally frozen tissue. While electron diVrac-tion failed to provide evidence for microcrystalline ice inhigh-pressure frozen liver sections (Fig. 3), reXections froma small fraction of crystalline ice might be missed due to thestrong, diVuse background scattering that results from athick layer of ice containing embedded biological material(McDowall et al., 1983; Sartori et al., 1993). Note that thetwo rings of vitreous ice, seen clearly at spacings corre-sponding to 0.370 and 0.214 nm from a thin layer of plunge-frozen pure water in Fig. 3E, are less distinct in the diVrac-tion patterns from the frozen-hydrated sections. Further,many of the strongest reXections from crystalline ice forms(Sartori et al., 1996) appear near the strong diVuse maximaof vitreous ice. Energy-Wltered electron diVraction experi-ments are planned, which might better detect smallamounts of crystalline ice in frozen-hydrated tissue byreducing background scattering in the diVraction patterns.

4.4. Comparison with freeze-substitution

It is clear that freeze-substitution of properly frozen tis-sue will continue to be a useful technique, even when sec-tioning of frozen-hydrated specimens becomes routine. Animportant advantage of freeze-substitution (in addition toease of sectioning and ease of tomographic data collection)is the enhanced contrast that is provided by section post-staining for some cellular components, such as heterochro-matin. Of course, this same characteristic also constitutesthe method’s major drawback, namely, that images ofstained specimens do not reXect true diVerences in massthickness between cellular components. There is a degree ofvariability in staining of certain cellular structures, such asmembranes, which sometimes stain positively, and some-times negatively. The heavy osmication that is associatedwith many freeze-substitution protocols can obscure fea-tures. Mitochondrial matrix granules and membranes ofthe cristae, when they stain positively, often cannot bedetected inside the densely stained matrix. With frozen-hydrated sections, Matias and colleagues (2003) were ableto detect a discrete peptidoglycan layer in the periplasm ofGram-negative bacteria that was obscured by stain in thefreeze-substituted material, while with freeze-substitution,they found that the outer membrane of Pseudomonas aeru-ginosa displayed a lipopolysaccharide layer, which was notseen in frozen-hydrated sections due to contrast matchingwith the dextran or sucrose media in which the cells hadbeen frozen. Similarly, a comparison of desmosomes in fro-zen-hydrated sections of skin (Al-Amoudi et al., 2004;Hsieh et al., 2004) with the same structure in freeze-substi-tuted material (He et al., 2003) reveals that intermediateWlaments attached to the desmosomal plaque are less

apparent in frozen-hydrated sections, while a low-densitygap between the cell membrane and the desmosomal plaqueis seen in frozen-hydrated, but not freeze-substituted sec-tions.

Acknowledgments

We thank Drs. Kent McDonald, Martin Mueller, Hel-mut Gnägi, and Jan Hoek for helpful discussions andadvice, and Drs. Kathleen Kinnally, Christian Renken, andAbigail Synder-Keller for assistance with obtaining rat livertissue and useful discussions. We are grateful to Ms. AnnKorsen and Leica USA for loan of an EMPACT high-pres-sure freezer. We thank the Wadsworth Center’s ElectronMicroscopy Core for general support with the Bal-Techigh-pressure freezer. This work was supported by NIH/NCRR Biomedical Research Technology Program GrantRR01219 (P.I. J. Frank), which funds the Wadsworth Cen-ter’s Resource for Visualization of Biological Complexity.

References

Al-Amoudi, A., Dubochet, J., Gnägi, H., Lüthi, W., Studer, D., 2003. Anoscillating cryo-knife reduces cutting-induced deformation of vitreousultrathin sections. J. Microsc. 212, 26–33.

Al-Amoudi, A., Norlén, L.P.O., Dubochet, J., 2004. Cryo-electron micros-copy of vitreous sections of native biological cells and tissues. J. Struct.Biol. 148, 131–135.

Al-Amoudi, A., Studer, D., Dubochet, J., 2005. Cutting artifacts and cut-ting process in vitreous sections for cryo-electron microscopy. J. Struct.Biol. 150, 109–121.

Buchanan, R.A., Leapman, R.D., O’Connell, M.F., Reese, T.S., Andrews,S.B., 1993. Quantitative scanning transmission electron microscopy ofultrathin cryosections: subcellular organelles in rapidly frozen liver andcerebellar cortex. J. Struct. Biol. 110, 244–255.

Chang, J.-J., McDowall, A.W., Lepault, J., Freeman, R., Walter, C.A., Dubo-chet, J., 1983. Freezing, sectioning and observation artifacts of frozenhydrated sections for electron microscopy. J. Microsc. 132, 109–123.

Crowther, R.A., DeRosier, D.J., Klug, A., 1970. The reconstruction of athree-dimensional structure from its projections and its application toelectron microscopy. Proc. R. Soc. Lond. (A) 317, 319–340.

Dahl, R., Staehelin, L.A., 1989. High-pressure freezing for the preservationof biological structure: theory and practice. J. Electron Microsc. Tech.13, 165–174.

Dubochet, J., Adrian, M., Chang, J.J., Homo, J.-C., Lepault, J., McDowall,A.W., Schultz, P., 1988. Cryo-electron microscopy of vitriWed speci-mens. Q. Rev. Biophys. 21, 129–228.

Dubochet, J., Sartori Blanc, N., 2001. The cell in absence of aggregationartifacts. Micron 32, 91–99.

Echlin, P., 1992. Low-temperature Microscopy and Microanalysis. PlenumPress, New York.

Erk, I., Nicolas, G., CarloV, A., Lepault, J., 1998. Electron microscopy offrozen biological objects: a study using cryosectioning and cryosubsti-tution. J. Microsc. 189, 236–248.

Forbes, M.S., 1986. Dog hairs as section manipulators. EMSA Bull. 16, 67.Förster, F., Medalia, O., Zauberman, N., Baumeister, W., Fass, D., 2005.

Retrovirus envelope protein complex structure in situ studied by cryo-electron tomography. Proc. Natl. Acad. Sci. USA 102, 4729–4734.

Frangakis, A.S., Förster, F., 2004. Computational exploration of structuralinformation from cryo-electron tomograms. Curr. Opin. Struct. Biol.14, 325–331.

Frank, J., Radermacher, M., Penczek, P., Zhu, J., Li, Y., Ladjadj, M., Leith,A., 1996. SPIDER and WEB: processing and visualization of images in3D electron microscopy and related Welds. J. Struct. Biol. 116, 190–199.


Frank, J., Wagenknecht, T., McEwen, B.F., Marko, M., Hsieh, C.-E., Man-nella, C.A., 2002. Three-dimensional imaging of biological complexity.J. Struct. Biol. 138, 85–91.

Frederik, P.M., Busing, W.M., Persson, A., 1982. Concerning the nature ofthe cryosectioning process. J. Microsc. 125, 167–175.

Frederik, P.M., Busing, W.M., Persson, A., 1984. Surface defects on thincryosections. Scan. Electron Microsc. 1, 433–443.

Frey, T.G., Mannella, C.A., 2000. The internal structure of mitochondria.Trends Biochem. Sci. 25, 319–324.

Giddings, T.H., 2003. Freeze-substitution protocols for improved visuali-zation of membranes in high-pressure frozen samples. J. Microsc. 212,53–61.

Grimm, R., Typke, D., Baumeister, W., 1998. Improving image quality byzero-loss energy Wltering: quantitative assessment by means of imagecross-correlation. J. Microsc. 190, 339–349.

Grünewald, K., Medallia, O., Gross, A., Steven, A., Baumeister, W., 2002.Prospects of electron cryotomography to visualize macromolecularcomplexes inside cellular compartments: implications of crowding.Biophys. Chem. 100, 577–591.

He, W., Cowin, P., Stokes, D., 2003. Untangling desmosomal knots withelectron tomography. Science 320, 109–113.

Hohenberg, H., Mannweiler, K., Müller, M., 1994. High pressure freezingof cell suspensions in cellulose capillary tubes. J. Microsc. 175, 34–43.

Hohenberg, H., Tobler, M., Müller, M., 1996. High-pressure freezing of tis-sue obtained by Wne-needle biopsy. J. Microsc. 183, 1–7.

Hsieh, C.-E., Marko, M., Frank, J., Mannella, C.A., 2002. Electron tomo-graphic analysis of frozen-hydrated tissue sections. J. Struct. Biol. 138,63–73.

Hsieh, C., He, W., Marko, M., Stokes, D.L., 2004. 3D Tomographic map ofdesmosome from frozen-hydrated skin sections. Microsc. Microanal.10 (Suppl. 2), 1188CD.

Leforestier, A., Dubochet, J., Livolant, F., 2001. Bilayers of nucleosomecore particles. Biophys. J. 81, 2414–2421.

Leforestier, A., Mangenot, S., Livolant, F., 2005. Combination of chirality andelectrostatic eVects in the supramolecular organization of nucleosomecore particles assemblies. Microsc. Microanal. 11 (Suppl. 2), 136–137.

Leis, A., Andrees, L., Gruska, M., Al-Amoudi, A., Sartori, A., Dubochet, J.,Baumeister, W., 2005. Cryo-electron tomography and Xuorescencemicroscopy of unicellular algae in vitreous cryosections. Microsc.Microanal. 11 (Suppl. 2), 330CD.

Mannella, C.A., Marko, M., Penczek, P., Barnard, D., Frank, J., 1994. Theinternal compartmentation of rat-liver mitochondria: tomographicstudy using the high-voltage transmission electron microscope.Microsc. Res. Tech. 27, 278–283.

Mannella, C.A., Buttle, K., Marko, M., 1997. Reconsidering mitochondrialstructure: new views of an old organelle. Trends Biochem. Sci. 22, 7–38.

Mannella, C.A., PfeiVer, D.R., Bradshaw, P.C., Moraru, L.I., Slepchenko,L.B., Loew, L.M., Hsieh, C.-E., Buttle, K., Marko, M., 2001. Topologyof the mitochondrial inner membrane: dynamics and bioenergeticimplications. IUBMB Life 52, 93–100.

Matsko, N., Mueller, M., 2005. Epoxy resin as a Wxative during freeze-sub-stitution. J. Struct. Biol. 152, 92–103.

Matias, V.R.F., Al-Amoudi, A., Dubochet, J., Beveridge, T.J., 2003. Cryo-transmission electron microscopy of frozen-hydrated sections of gram-negative bacteria. J. Bacteriol. 185, 6112–6118.

McDonald, K.L., 1994. Electron microscopy and EM immunochemistry.Methods Cell Biol. 44, 411–444.

McDonald, K.L., 1999. High pressure freezing for preservation of high res-olution Wne structure and antigenicity for immunolabelling. MethodsMol. Biol. 117, 77–97.

McDonald, K., Müller-Reichert, T., 2002. Cryomethods for thin sectionelectron microscopy. Methods Enzymol. 351, 96–123.

McDowall, A.W., Chang, J.J., Freeman, R., Lepault, J., Walter, C.A., Dub-ochet, J., 1983. Electron microscopy of frozen hydrated sections of vit-reous ice and vitriWed biological samples. J. Microsc. 131, 1–9.

Medalia, O., Weber, I., Frangakis, A.S., Nicastro, D., Gerisch, G., Baumei-ster, W., 2002. Macromolecular architecture in eukaryotic cells visual-ized by cryoelectron tomography. Science 289, 1209–1213.

Michel, M., Hillman, T., Müller, M., 1991. Cryosectioning of plant mate-rial frozen at high pressure. J. Microsc. 163, 3–18.

Michel, M., Gnägi, H., Müller, M., 1992. Diamonds are a cryosectioner’sbest friend. J. Microsc. 166, 43–56.

Moor, H., 1987. Theory and practice of high pressure freezing. In: Steinbr-echt, R.A., Zierold, K. (Eds.), Cryotechniques in Biological ElectronMicroscopy. Springer, Berlin, pp. 175–191.

Nicastro, D., Frangakis, A.S., Typke, D., Baumeister, W., 2000. Cryo-elec-tron tomography of Neurospora mitochondria. J. Struct. Biol. 129, 48–56.

Nicastro, D., Austin, J., Pierson, J., Gaudette, R., Schwartz, C., Ladinsky,M., Staehelin, L.A., McIntosh, J.R., 2005. Visualizing the macromolec-ular organization of chloroplast membranes using cryo-electrontomography. Microsc. Microanal. 11 (Suppl. 2), 150–151.

Norlén, L., Al-Amoudi, A., Dubochet, J., 2003. A cryotransmission elec-tron microscopy study of skin barrier formation. J. Invest. Dermatol.120, 555–560.

Norlén, L., Al-Amoudi, A., 2004. Stratum corneum keratin structure, func-tion and formation—the cubic rod-packing and membrane templatingmodel. J. Invest. Dermatol. 123, 715–732.

Penczek, P., Marko, M., Buttle, K., Frank, J., 1995. Double-tilt electrontomography. Ultramicroscopy 60, 393–410.

Radermacher, M., Hoppe, W., 1980. Properties of three-dimensionallyreconstructed objects from projections by conical tilting compared tosingle axis tilting. In: Proceedings of the 7th European Congress onElectron Microscopy, Den Haag, vol. 1, pp. 132–133.

Rath, B.K., Hegerl, R., Leith, A., Shaikh, T.R., Wagenknecht, T., Frank, J.,2003. Fast 3D motif search of EM density maps using a locally normal-ized cross-correlation function. J. Struct. Biol. 144, 95–103.

Richter, K., Gnägi, H., Dubochet, J., 1991. A model for cryosectioningbased on the morphology of vitriWed ultrathin sections. J. Microsc. 163,19–28.

Richter, K., 1994. Cutting artefacts on ultrathin cryosections of biologicalbulk specimens. Micron 25, 297–308.

Sartori, N., Richter, K., Dubochet, J., 1993. VitriWcation depth can beincreased more than 10-fold by high-pressure freezing. J. Microsc. 172,55–61.

Sartori, N., Bednar, J., Dubochet, J., 1996. Electron-beam-induced amor-phization of ice III or IX obtained by high-pressure freezing. J.Microsc. 182 (3), 163–168.

Sartori-Blanc, N., Senn, A., Leforestier, A., Livolant, F., Dubochet, J.,2001. DNA in human and stallion spermatozoa forms local hexagonalpacking with twist and many defects. J. Struct. Biol. 134, 76–81.

Schwartz, C., Nicastro, D., Ladinsky, M.S., Mastronarde, D., O’Toole, E.,McIntosh, J.R., 2003. Cryo-electron tomography of frozen-hydratedsections of eukaryotic cells. Microsc. Microanal. 9 (Suppl. 2), 1166–1167.

Shi, S., Sun, S.Q., Andrews, B., Leapman, R.D., 1996. Thickness measure-ment of hydrated and dehydrated cryosections by EELS. Microsc. Res.Tech. 33, 241–250.

Shimoni, E., Müller, M., 1998. On optimizing high-pressure freezing: fromheat transfer theory to a new microbiopsy device. J. Microsc. 192, 236–247.

Sitte, H., 1996. Advanced instrumentation and methodology related tocryoultramicrotomy: a review. Scanning Microsc.(Suppl. 10), 387–466.

Somlyo, A.V., Shuman, H., Somlyo, A.P., 1977. Elemental distribution ofstriated muscle and the eVects of hypertonicity. J. Cell Biol. 74, 824–857.

Somlyo, A.P., Bond, M., Somlyo, A.V., 1985. Calcium content of mito-chondria and endoplasmic reticulum in liver frozen rapidly in vivo.Nature 314, 622–625.

Studer, D., Michel, M., Müller, M., 1989. High pressure freezing comes ofage. Scanning Microsc.(Suppl. 3), 253–269.

Studer, D., Michel, M., Wohlwend, M., Hunziker, E.B., Buschmann, M.D.,1995. VitriWcation of articular cartilage by high-pressure freezing. J.Microsc. 179 (3), 321–332.

Studer, D., Graber, W., Al-Amoudi, A., Eggli, P., 2001. A new approachfor cryoWxation by high-pressure freezing. J. Microsc. 203, 285–294.


Studer, D., Gnägi, H., 2000. Minimal compression of ultrathin sectionswith use of an oscillating diamond knife. J. Microsc. 197, 94–100.

Ting, C.S., Hsieh, C., Sundararaman, S., Mannella, C., Marko, M., 2005.Comparative three-dimensional imaging of environmentally criticalcyanobacteria through cryo-electron tomography. Microsc. Microanal.11 (Suppl. 2), 332CD.

Van Harreveld, A., Crowell, J., 1964. Electron microscopy after rapid freezingon a metal surface and substitution Wxation. Anat. Rec. 149, 381–386.

Vanhecke, D., Graber, W., Herrmann, G., Al-Amoudi, A., Eggli, P., Studer,D, 2003. A rapid microbiopsy system to improve the preservation of bio-logical samples prior to high-pressure freezing. J. Microsc. 212, 3–12.

Vonk, J., 2000. Parameters aVecting specimen Xatness of two-dimensionalcrystals for electron crystallography. Ultramicroscopy 5, 123–129.

Wagenknecht, T., Hsieh, C.-E., Rath, B., Fleischer, S., Marko, M., 2002.Electron tomography of frozen-hydrated isolated triad junctions. Bio-phys. J. 83, 2491–2501.

Zhang, P., Bos, E., Heyman, J., Gnägi, H., Kessel, M., Petere, P.J., Subr-amaniam, S., 2004. Direct visualization of receptor arrays in frozen-hydrated sections and plunge-frozen specimens of E. coli engineered tooverproduce the chemotaxis receptor Tsr. J. Microsc. 216, 76–83.

Zierold, K., 1994. The morphology of ultrathin cryosections. Ultramicros-copy 14, 201–210.

towards high-resolution three-dimensional imaging of...

Documents