Journal of Structural Biology 127, 169–176 (1999) Article ID jsbi.1999.4120, available online at http://www.idealibrary.com on

Tetrairidium, a Four-Atom , Is Readily Visible as a Density Label in Three-Dimensional Cryo-EM Maps of Proteins at 10–25 Å Resolution

N. Cheng,* J. F. Conway,* N. R. Watts,* J. F. Hainfeld,† V. Joshi,‡ R. D. Powell,‡ S. J. Stahl,§ P. E. Wingfield,§ and A. C. Steven*,1

*Laboratory of Structural Biology and §Protein Expression Laboratory, National Institute of Arthritis and Musculoskeletal and Skin Diseases, National Institutes of Health, Bethesda, Maryland 20892; †Department of Biology, Brookhaven National Laboratory, Upton, New York 11973; and ‡Nanoprobes Incorporated, Stony Brook, New York 11790

Received December 17, 1998; and in revised form April 2, 1999

interest. In addition to considerations of specificity Heavy clusters derivatized to bind to desig- and efficiency of labelling, the resolution achieved nated chemical groups on proteins have great poten- depends primarily on the size of the metal particle tial as density labels for cryo-electron microscopy. and the distance between its center-of-mass (the Smaller clusters offer higher resolution and pen- feature detected) and its point of attachment (the etrate more easily into sterically restricted sites, targeted site). For instance, in colloidal gold-based but are more difficult to detect. In this context, we immunolabelling (reviews, Griffiths, 1994; Ben- have explored the potential of tetrairidium (Ir4)asa density label by attaching it via maleimide linkage dayan, 1995), which is widely used to map the to the C-terminus of the hepatitis B virus (HBV) distributions of in cells, the gold particles capsid protein. Although the clusters are not visible are typically 50–200 Å in diameter, and there may be in unprocessed cryo-electron micrographs, they are up to 200–400 Å between their projected centers and distinctly visible in three-dimensional density maps the epitope marked, depending on whether interme- calculated from them, even at only partial occu- diaries such as protein A or secondary antibodies are pancy. The Ir4 label was clearly visualized in our used. maps at 11–14 Å resolution of both size variants of For structural analysis of isolated macromol- the HBV capsid, thus confirming our previous local- ecules, smaller labels and more precise localizations ization of this site with undecagold (Zlotnick, A., are needed. In this context, metal clusters deriva- Cheng, N., Stahl, S. J., Conway, J. F., Steven, A. C., tized to bind directly to groups such as sulfhydryls or and Wingfield, P. T., Proc. Natl. Acad. Sci. USA 94, amines are finding a growing number of applications 9556–9561, 1997). Ir penetrated to the interior of 4 (reviews, Hainfeld, 1996; Hainfeld and Powell, 1997; intact capsids to label this site on their inner sur- see other papers in this issue). Nanogold (Hainfeld face, unlike undecagold for which labelling was achieved only with dissociated dimers that were and Furuya, 1992) has a core of 55–75 or so gold ϳ then reassembled into capsids. The Ir cluster re- , 14 Å in diameter; in maleimide linkage to 4 ϳ mained visible as the resolution of the maps was cysteine, there is a spacing of 20 Å between its lowered progressively to ϳ25 Å. ௠ 1999 Academic Press center and the sulfur of the labelled amino acid. Key Words: metal cluster; cryo-electron micros- Undecagold (Bartlett et al., 1978; Hainfeld, 1989) copy; molecular marker; three-dimensional image has 11 atoms in a cluster ϳ8 Å across, and ϳ17 Å reconstruction; hepatitis B virus separate its center and the targeted residue. Nano- gold has sufficient scattering power to be directly visible in micrographs even against a background of INTRODUCTION negative stain (Wenzel and Baumeister, 1995; Grigori Heavy metal particles, whose high electron den- et al., 1997). Undecagold is smaller and only margin- sity renders them visible in electron micrographs, ally visible in unprocessed micrographs of ice- have been extensively used to label molecules of embedded specimens. However, it has been detected in averaged two-dimensional (Crum et al., 1994) and three-dimensional (Milligan et al., 1990; Steinmetz 1 To whom correspondence should be addressed at Building 6, et al., 1998) density maps at 15–30 Å resolution. Room B2-34, MSC 2717, National Institutes of Health, Bethesda, MD 20892-2717. Fax: (301) 480-7629; E-mail: Alasdair_Steven- Typically, the metal cluster of a labelling compound @nih.gov. is surrounded by an organic shell that contains the

169 1047-8477/99 $30.00 Copyright ௠ 1999 by Academic Press All rights of reproduction in any form reserved. 170 CHENG ET AL. required for specific attachment to 500-mL column (LH-20, Pharmacia-Biotech, Piscataway, NJ) proteins and also governs its solubility. Nanogold eluted with methanol. Fractions with a yellow–orange color were and undecagold have triphenylphosphine shells that pooled and evaporated to dryness. Fractions with similar UV-Vis ϳ ϳ and IR spectra were combined. Pure product was characterized give them full diameters of 26 and 20 Å, respec- using [1H], [13C], and [31P]NMR spectral data. A total of 1.4 mg of tively. However, because this shell poses a steric this cluster and 11.4 mg (150-fold excess) of N-methoxycarbonyl- hindrance on a cluster’s ability to reach potentially maleimide were each dissolved in 0.5 mL of methanol, and the two labellable sites, it is of interest to have cluster labels were combined and left at 5°C for 30 min. The maleimido that are smaller than undecagold. cluster was isolated by gel filtration (LH-20 column, 80 mL, eluted with methanol at 2-mL/min flow rate). Fractions 8–11 (5-mL Two such candidates are the four- clusters, fractions) were pooled, and the solvent was quickly removed tetramercury (Hg4) and tetrairidium (Ir4). The mer- under reduced pressure. cury compound has been used to label the filamen- Preparation and labelling of capsids. Capsid protein con- tous bacteriophage fd (Lipka et al., 1979), but it was structs were expressed in Escherichia coli and isolated in a concluded that the Hg was too volatile under the procedure that involves dissociation to dimers, further purifica- tion, and reassembly (Wingfield et al., 1995; Zlotnick et al., 1997). electron beam to be generally suitable as an EM All labelling experiments were performed on preformed particles label, and little subsequent use has been made of prepared as follows: 2 mL of dimeric protein at 1 mg/mL in 100 this reagent. Ir4 was originally developed for X-ray mM NaHCO3,20mMDTT,pH9.6,wasmixedwith0.4mLof500 of large proteins whose native reflec- mM Hepes, 2 M NaCl, pH 5.0, and incubated for1hatroom tions are insufficiently perturbed by single heavy temperature. For labelling, the buffer was changed to 20 mM atom derivatives to provide adequate phases (Jahn, Hepes, 2 mM EDTA, pH 6.9, with a PD10 gel filtration column (Pharmacia-Biotech). The protein concentration was determined 1989; Weinstein et al., 1989). In EM, Ir has been 4 Ϫ1 Ϫ1 4 by absorbance, using ⑀280 ϭ 2.95 ϫ 10 M ·cm (Zlotnick et al., visualized in dark-field STEM micrographs (Furuya 1997). Fifty nanomoles of tetrairidium monomaleimide dissolved et al., 1988), but its scope as a protein label was in 200 µL of DMSO was immediately added to 50 nmol of protein limited by the state of preservation of freeze-dried in a final volume of 1200 µL. After an overnight incubation at specimens. Here, we report that Ir is a highly room temperature, MgCl2 was added to a final concentration of 4 100 mM to alleviate capsid aggregation. Unbound label and effective label for ice-embedded specimens, which unassembled protein were then removed by gel filtration on a retain their native conformations (Adrian et al., 12-mL, 15-cm-long column of CL-6B (Pharmacia-Biotech) equili- 1984). brated with 20 mM Hepes, 50 mM MgCl2, pH 8.0. Particles were Recently we localized the C-terminus of the hepati- concentrated to ϳ2 mg/mL by ultrafiltration (Ultrafree-15, Milli- tis B virus (HBV) capsid protein by appending a pore, Bedford, MA). Labelling was quantitated as follows: the concentration of Ir4 was calculated from the absorbance at 320 nm cysteine residue at this site and labelling it with (corrected for the contribution by light scattering alone, estimated undecagold (Zlotnick et al., 1997). HBV capsids are from the UV spectrum of the reduced unlabelled particles) using 4 Ϫ1 Ϫ1 icosahedral shells of two sizes—280 Å in diameter ⑀320 ϭ 2.55 ϫ 10 M ·cm . Protein concentrations were calcu- (90 dimers of 17-kDa subunits) and 320 Å (120 lated from the absorbance at 280 nm, corrected for the contribu- ϭ dimers), respectively. Unusually for viral capsids, tion by Ir4 at this wavelength, using A280/A320 2.08. these shells are perforated with sizable holes whose Cryo-electron microscopy. Capsid samples were frozen in a ϳ thin film of vitrified ice suspended over holey carbon films, as diameters, at 20 Å, are similar to that of undeca- described by Zlotnick et al. (1996). Micrographs were recorded at a gold. Our attempts to label intact capsids were magnification of ϫ38 000 on a CM200-FEG microscope (Philips, unsuccessful, but dissociated dimers were readily Eindhoven, Netherlands), fitted with a model 626 cryoholder labelled and could be reassembled into capsids. In (Gatan, Pleasanton, CA) and operating at 120 keV. Pairs of the density map, undecagold was conspicuously pre- micrographs at different defocus values were recorded using low-dose techniques. In the focal pairs analyzed in calculating the sent at a specific site on the inner surface of the reconstructions, the first exposure had a defocus value in the capsid; on this basis, we attributed the nonlabelling range 0.8–1.1 µm, so that the first zeros of the contrast transfer of intact capsids to failure of the cluster to diffuse function (CTF) were at spatial frequencies of (17 Å)Ϫ1 Ϫ (20 Å)Ϫ1, through the holes in the capsid wall. In the experi- which was increased by 0.4 µm for the second exposure. ments reported here, we were able to label intact Calculation of density maps. Micrographs were digitized on an capsids with Ir , demonstrating the greater penetrat- SCAI scanner (Zeiss, Englewood, CO) at 7 µm per pixel, corre- 4 sponding to 1.8 Å per pixel at the specimen. Particle images were ing power of this smaller cluster. extracted and processed using a semiautomated procedure (Con- way et al., 1993). Their orientations were determined by means of MATERIALS AND METHODS the PFT algorithm (Baker and Cheng, 1996), using as starting models preexisting maps of Cp147 capsids: T ϭ 4 (Conway et al., ϭ Preparation of the tetrairidium cluster. To prepare Ir4(CO)85P( p- 1997); T 3 (Conway et al., 1998). The CTF was corrected and the C6H4C(O) NHMe)363(P( p-C6H4C(O)NH Me)2 ( p-C6H4C(O)NH(CH2) focal pairs were combined as described by Conway et al. (1997). 3NH2)6, and its maleimido derivative. One-hundred milligrams of The maps were calculated using methodology reviewed by Fuller Ir4(CO)12, 69 mg of Tris( p-N-methylcarboxamidophenyl)phos- et al. (1996), including all particles with correlation coefficients of phine, P( p-C6H4C(O)NHMe)3, and 62 mg of the amino-substituted 0.5 or higher for each of the three statistics calculated by the PFT phosphine, P(p-C6H4C(O)NHMe)2 (p-C6H4C(O)NH(CH2)3NH2), were program (Baker & Cheng, 1996). The number of particles included refluxed for4hin50mLofanhydrous toluene under N2, cooled to in each reconstruction and the resulting resolution are listed in room temperature, and then separated by gel filtration on a Table I. TETRAIRIDIUM AS A CRYO-EM DENSITY LABEL 171

RESULTS

Labelling HBV Capsids with Ir4. The capsids studied represent variants of the HBV assembly domain expressed in E. coli. The complete assembly domain is 149 residues long. In the native capsid protein, it is coupled to a 34-residue, C-terminal domain that is highly basic and binds RNA in nucleocapsid assembly. The wildtype assembly do- main has cysteines at positions 48, 61, and 107 that are replaced by alanines in the 149-residue con- struct, Cp*149. Cp*150 has an additional residue, a cysteine, appended at the C-terminus. Labelling was performed by adding Ir4 (Fig. 1) dissolved in DMSO to preassembled, freshly reduced capsids at a 1:1 molar ratio of Ir4 clusters:protein monomers and incubating for 12 h at room tempera- ture. At lower temperatures or at Ir4 concentrations much above 50 nmol/mL, the reagent began to precipitate. In early experiments, gave erratic estimates of labelling levels, which were explained by the presence of microprecipitates of Ir4 seen by cryo-EM (e.g., Fig. 2d). Subsequently, precipi- tation of the reagent was largely alleviated by includ- ing up to 20% DMSO or 2-propanol, additives that are well tolerated by most (but not all) proteins. No adverse effects of the DMSO on HBV capsid struc- ture were observed, at least to ϳ11 Å resolution (see below). After the incubation, unbound label was removed

TABLE I Summary of Three-Dimensional Reconstructions FIG. 1. Schematic diagram of the tetrairidium cluster com- pound used. The diameter of the cluster core is 5 Å and its full Resolution diameter, including the organic shell, is ϳ17 Å (Weinstein et al., Total (Å)b 1989). The electron density in the metal core is 4.7 electrons/Å3. number of The average Ir–Ir bond distance is expected to be 2.736 Å. Sample particlesa FSC FRC Although the Ir atom at the apex of the cluster is directly bonded to six atoms or , it satisfies the 18-electron rule (as applied 13.2 13.7 (2) 1041 149 ء T ϭ 3: Cp ϩ to coordinate-covalent bonds in organometallic clusters), as do the ء ϭ T 3: Cp 150 Ir4: Expt 1 1096 (4) 14.8 14.5 other three iridium atoms (Ros et al., 1986). T ϭ 4: Cp149 1296 (2) 11.3 10.9 ϩ Ir4: Expt 1 1460 (4) 12.6 11.5 150 ء T ϭ 4: Cp ϩ Ir : Expt 2 600 (4) 14.7 10.5 150 ء T ϭ 4: Cp 4 by gel filtration and capsid labelling was assessed by a The total number of particles used in each reconstruction is measuring absorbance by Ir4 at 320 nm and by given. The number of focal pairs combined to yield these data is protein at 280 nm. These data reproducibly indi- given in parentheses. All particles present were analyzed; of cated a labelling efficiency of ϳ11% for Cp*150 under these, all that had all three PFT correlation coefficients (PFT, PRJ, and CMP—see Baker and Cheng, 1996) of 0.5 or higher were the conditions used. Nonspecific binding, estimated included in the reconstruction. This fraction varied from ϳ40 to by treating Cp*149 capsids—which have no cys- ϳ80% in different focal pairs. teines—in the same way, was found to be less than b Resolution was calculated by two criteria—the Fourier shell 0.5%. correlation (FSC, van Heel, 1987) and the Fourier ring correlation coefficient (FRC, Saxton and Baumeister, 1982; Conway et al., Cryo-EM of Ir4-labelled capsids. Micrographs of 1993). In both cases, the criterion used to define the resolution labelled and unlabelled capsids are compared in Fig. limit was the frequency at which the correlation dropped below 2a and Figs. 2b and 2c, respectively. There is no 0.5 for the last time. The relatively large discrepancy between the clear-cut difference between them, and we conclude two measures for the last reconstruction (T ϭ 4, Expt 2) simply reflects that the correlation hovers about the 0.5 value over a that individual Ir4 clusters are not directly visible in considerable range, i.e., between approx (14 Å)Ϫ1 and (10 Å)Ϫ1,in cryo-micrographs. The images shown were recorded this analysis. quite close to focus (ϳ0.8 µm), conditions that we 172 CHENG ET AL.

considered to be conducive to visualizing the small (5 Å) metal clusters while still conveying the capsids, albeit with relatively low contrast. The clusters were no more visible at other defocus values used, up to ϳ2 µm (data not shown).

Density maps of Ir4-labelled capsids. In an experi- ment in which 11% labelling of Cp*150 was obtained, density maps were calculated for both the T ϭ 4 and the T ϭ 3 capsids to 12 and 14 Å resolution, respectively (Table I, Fig. 3). In both cases, conspicu- ous focal densities are present beneath the fivefold and quasi-sixfold lattice sites. These densities are absent from Cp*149 control capsids that were sub- jected to the same labelling reaction (cf., Fig. 3). This localization of the Ir4 clusters and, consequently, of the C-termini of the capsid protein confirms the result previously obtained by visualizing undecagold bound to Cp*150 at ϳ20 Å resolution (Zlotnick et al., 1997). Consistent results were obtained for a T ϭ 4 capsid reconstruction in an earlier labelling experi- ment in which the DMSO was omitted (Table I, cf., Figs. 4a and 4b). Although the labelling efficiency of the latter experiment could not be determined reli- ably by spectroscopy (see above), it must have been very similar to that of the experiment illustrated in Fig. 3, to judge by the intensity of the cluster- associated densities in the respective maps. Dependence of cluster visibility on resolution. To estimate the minimum resolution required to visual- ize Ir4 labels by cryo-EM, we progressively lowered the resolution of these density maps. As shown in Fig. 4, the cluster-associated density becomes pro- gressively more diffuse as the resolution falls, but it remains significantly above the level of residual background noise to at least 25 Å resolution.

DISCUSSION Visibility of tetrairidium as a density label. Hith- erto, undecagold is the smallest metal cluster to have been used as a density label in cryo-EM. Ir4 is significantly smaller (2 vs 6.5 kDa, 4 vs 11 heavy atoms). Although it was essentially invisible in micro- graphs covering a considerable range of defocus values (Fig. 2), Ir4 was clearly visualized in three- dimensional density maps, even at incomplete occu- pancy and moderate resolution (Figs. 3 and 4), FIG. 2. Cryo-micrographs, recorded relatively close to focus reflecting the large improvement in signal-to-noise (⌬f ϳ0.9 µm), of (a) Ir4-labelled Cp*150 capsids and (b, c) unlabelled capsids. The two controls are (b) Cp*149 capsids that ratio in such a density map compared to the original were subjected, with negative results, to the labelling protocol and projection images. (c) Cp147 capsids that were not so treated. Also shown (d), We reproducibly obtained a labelling stoichiom- attached to a carbon film, are microprecipitates of Ir4 clusters ϳ formed when the labelling reaction was attempted in the absence etry of 11% for Ir4 relative to the capsid protein of DMSO, which are large enough to be directly visible. Bar ϭ under our conditions of incubation, corresponding to 500 Å. ϳ26 clusters per T ϭ 4 capsid and ϳ20 clusters per T ϭ 3 capsid. However—as with undecagold (Zlot- TETRAIRIDIUM AS A CRYO-EM DENSITY LABEL 173

FIG. 4. Effect of progressively lowering the resolution on the visibility of the Ir4 cluster label on the T ϭ 4 Cp*150 capsid. (a, b) Reconstructions from independent labelling experiments at a resolution of ϳ11 Å (see Table I). (d, e, f ) The resolution of reconstruction a was lowered to (d) 15 Å, (e) 20 Å, and (f ) 25 Å, respectively. (c) A 25-Å resolution reconstruction based on 100 particles (cf., the 1460 particles used for the reconstructions in a, d, e, and f ), whose statistical properties (residual noise level) is more typical of a density map at this resolution. Bar ϭ 50 Å. nick et al., 1997)—the bound clusters are located on of the cluster density as depicted in the reconstruc- the fivefold and quasi-sixfold axes, where occupancy tion. by one cluster blocks the binding of clusters to the Initially, we were surprised that a label as small as other four (or five) subunits surrounding the lattice Ir4 should be so conspicuous under these conditions point. Thus, the maximum binding to be expected is of observation (Fig. 3). However, the difference in ϭ ϭ 42 (T 4) and 32 (T 3) clusters per capsid, so that specific density between the cluster and the adjacent the binding obtained represents about 60% of maxi- protein (⌬␳ ϳ21.5 ϭ 22.8 vs 1.3) is so much higher mum. than that between protein and vitreous ice (⌬␳ As visualized (Figs. 3, 4a, and 4b), the 5 Å- ϳ0.35 ϭ 1.3 vs 0.95) that it remained distinctly diameter clusters appear considerably larger than visible in the map, despite attenuation by partial expected; in fact, their full width at half-height is occupancy and delocalization arising from averaging ϳ20 Å. Only part of this broadening may be attrib- (see above). We anticipate that Hg , which is similar uted to the limited resolution (11 Å), because a step 4 to Ir in electron scattering power, should also be function 5 Å across has a full width at a half-height 4 of ϳ9 Å when band-limited to this resolution. The detectable under similar conditions, i.e., in 3D den- remainder of the broadening must arise from other sity maps of ice-embedded specimens imaged at low effects, such as local disorder of the C-termini, or electron doses. slight displacement of each cluster from the adjacent Resolution of metal cluster labels. The resolution symmetry axis in a direction that depends on which attainable with these markers depends more on the of the surrounding subunits it is bound to. The latter spacing between the center of the cluster-associated effect would give rise to averaging-related smearing density and the labelled residue than it does on the 174 CHENG ET AL.

FIG. 3. Density maps of Ir4-labelled and unlabelled HBV capsids. (Top) T ϭ 4 capsids of the labelled Cp*150 construct and the control Cp*149 construct. The small black arrowhead (middle panel, top row) points to the organic linker connecting the Ir4 cluster to the capsid protein. (Bottom) A similar comparison for the corresponding T ϭ 3 capsids. For each density map, isodensity renderings of the outer and inner surfaces are shown, together with a central section depicting local variations in density. The capsids are viewed along a twofold symmetry axis and are contoured to enclose 100% of expected mass. The specifics of the reconstructions are given in Table I. For both capsids, labels are visible at corresponding sites on their inner surfaces. Bar ϭ 50 Å. TETRAIRIDIUM AS A CRYO-EM DENSITY LABEL 175

that is inaccessible to undecagold. An alternative approach to this problem of steric impedance may be to engineer variants of undecagold or Ir4 with longer linkers between the relatively bulky label and the functional group that may be better able to reach into restricted sites, albeit at possible cost in terms of resolution (see above). Preferential labelling of B-subunits in capsids assembled from labelled dimers; implications for regulation of capsid assembly. In the T ϭ 4 capsid, pairs of three nonequivalent subunits surround the quasi-sixfold axis. They are called the B-, C-, and D-subunits, respectively (Fig. 5). For capsids as- sembled from dimers labelled in with undeca- gold, it was observed that the linker connecting the cluster to the B-subunit was markedly more visible than that for the C- and D-subunits, implying that the single cluster per quasi-sixfold site was preferen- tially associated with a B-subunit (Zlotnick et al., 1997). Although this linker is also discernible in our FIG. 5. Distribution of protein subunits over the surface map of preformed capsids labelled with Ir4 (Figs. 3 lattice of the HBV T ϭ 4 capsid, adapted from Zlotnick et al. and 4), the preference for B-subunits is much less (1997). The symmetry axes are marked and the four quasiequiva- pronounced. This distinction may reflect a feature of lent subunits (A, B, C, D) are distinguished. The building blocks the assembly pathway; for instance, binding of the are A–B and C–D dimers. Because six quasiequivalent subunits are clustered around the twofold axis, it is referred to (Discussion) cluster to the C-termini of unassembled dimers as the quasi-sixfold axis. might switch them into a state of enhanced receptiv- ity for binding other dimers, leading to a cooperative aggregation around a local symmetry axis, and in so size of the cluster. With undecagold, this spacing is doing, may mimic an effect of RNA binding to the C-terminal domain in nucleocapsid assembly. If the ϳ17 Å, and with Ir4, it is slightly less, at ϳ15 Å. However, it is possible to achieve more precise symmetry axis in question were fivefold, this would localizations than these figures imply because, un- explain the preferential labelling of B-subunits, with der favorable circumstances, the organic linker con- additional dimers subsequently adding on to connect necting the cluster to the labelled protein may be the resulting pentamers, thereby completing the ϭ ϭ visualized (arrowhead in Fig. 3, middle panel, top formation of either a T 4oraT 3 capsid. row; see also Fig. 4 of Zlotnick et al., 1997). Extrapo- Prospects for mapping the 3D structures of protein lating an appropriate distance along this vector from complexes by cysteine scanning and cluster labelling. the cluster center should give a precise localization Although ample scope remains for further develop- of the residue of interest, i.e., to within Ϯ5Åorso. ment of clusters with other functionalities, the reac- While clusters as small as Ir4 are readily visible in tion of maleimide with the sulfhydryl of cysteine has density maps at 25–30 Å resolution (cf., Fig. 4), one been found to work well with undecagold (Hainfeld, advantage of extending their resolution to 10–15 Å is 1988), Nanogold (Hainfeld and Furuya, 1992), and enhanced prospects of visualizing this linker and, Ir4 (Jahn, 1989). In addition to naturally occurring consequently, of effecting a more precise localization cysteines, current techniques of the targeted residue. readily allow the insertion or substitution of this Although shifting to progressively smaller clusters amino acid at virtually any site of interest. In does not bring a major gain in resolution, these principle, therefore, the possibility already exists to reagents have the advantage of being less suscep- use cluster labelling to delineate the overall path of a tible to steric constraints that may limit their reactiv- polypeptide chain in any protein complex amenable ity. The relatively bulky organic shell surrounding to analysis by cryo-EM. Given enough constraints of the cluster prevents it from accessing confined spaces this kind, it should then be possible to fold in in much the same way that an atomic force micro- information on conformational preferences of spe- scope is limited in the vertical dimension by the size cific peptides, secondary structure predictions, and of its tip. Thus, smaller clusters have superior other data such as intramolecular cross-links to penetrating power, as illustrated by the observed build detailed molecular models, starting from positive Ir4 labelling of a residue inside HBV capsids cryo-EM density maps in the 10 Å-resolution range 176 CHENG ET AL.

(Bo¨ttcher et al., 1997; Conway et al., 1997; Trus et al., covalently attached to antibodies improves immunolabeling, J. 1997). Histochem. Cytochem. 40, 177–184. Hainfeld, J. F. (1996) Labeling with Nanogold and undecagold: We thank Drs. A. Zlotnick, D. Belnap, and B. Trus for helpful Techniques and results, Scanning Microsc. Suppl. (Proc. 14th discussions and Dr. T. Baker (Purdue) for software. Pfefferkorn Conf.) 10, 309–322. [Malecki, M., and Roomans, G. M. (Eds.), Scanning Microscopy Int’l, Chicago]. REFERENCES Hainfeld, J. F., and Powell, R. D. (1997) Nanogold technology: New frontiers in gold labeling, Cell Vision 4, 408–432. Adrian, M., Dubochet, J., Lepault, J., and McDowall, A. W. (1984) Jahn, W. (1989) Synthesis of water soluble tetrairidium clusters Cryo-electron microscopy of viruses, Nature 308, 32–36. suitable for heavy atom labelling of proteins, Z. Naturforsch. B Baker, T. S., and Cheng, R. H. (1996) A model-based approach for 44, 79–82. determining orientations of biological macromolecules imaged by cryoelectron microscopy, J. Struct. Biol. 116, 120–130. Lipka, J. J., Lippard, S. J., and Wall, J. S. (1979) Visualization of polymercuri-methane-labeled fd bacteriophage in the scanning Bartlett, P. A., Bauer, B., and Singer, S. J. (1978) Synthesis of transmission electron microscope, Science 206, 1419–1421. water-soluble undecagold cluster compounds of potential impor- tance in electron microscopic and other studies of biological Milligan, R. A., Whittaker, M., and Safer, D. (1990) Molecular systems, J. Am. Chem. Soc. 100, 5085–5092. structure of F-actin and location of surface binding sites, Nature 348, 217–221. Bendayan, M. (1995) Colloidal gold post-embedding immunocyto- , Prog. Histochem. Cytochem. 9, 1–25. Ros, R., Scrivanti, A., Albano, V. G., Braga, D., and Garlaschelli, L. (1986) Chemistry of tetrairidium carbonyl clusters. Part 1. Bo¨ttcher, B., Wynne, S. A., and Crowther, R. A. (1997) Determina- Synthesis, chemical characterization, and nuclear magnetic tion of the fold of the core protein of hepatitis B virus by electron resonance study of mono- and di-substituted phosphine deriva- cryomicroscopy, Nature 386, 88–91. tives. X-Ray crystal structure determination of the diaxial Conway, J. F., Trus, B. L., Booy, F. P., Newcomb, W. W., Brown, isomer of [Ir4(CO)7(µ-CO)3(Me2PCH 2CH2PMe2)], J. Chem. Soc. J. C., and Steven, A. C. (1993) The effects of radiation damage Dalton Trans. 2411–2421. on the structure of frozen hydrated HSV-1 capsids, J. Struct. Saxton, W. O., and Baumeister, W. (1982) The correlation averag- Biol. 111, 222–233. ing of a regularly arranged bacterial cell envelope protein, J. Conway, J. F., Cheng, N., Zlotnick, A., Wingfield, P. T., Stahl, S. J., Microsc. (Oxford) 127, 127–138. and Steven, A. C. (1997) Visualization of a 4-helix bundle in the Steinmetz, M. O., Stoffler, D., Muller, S. A., Jahn, W., Wolpens- hepatitis B virus capsid by cryo-electron microscopy, Nature inger, B., Goldie, K. N., Engel, A., Faulstich, H., and Aebi, U. 386, 91–94. (1998) Evaluating atomic models of F-actin with an undecagold- Conway, J. F., Cheng, N., Zlotnick, A., Stahl, S. J., Wingfield, P. T., tagged phalloidin derivative, J. Mol. Biol. 276, 1–6. and Steven, A. C. (1998) Localization of the N-terminus of Trus, B. L., Roden, R. B., Greenstone, H. L., Vrhel, M., Schiller, J. hepatitis B virus capsid protein by peptide-based difference T., and Booy, F. P. (1997) Novel structural features of bovine mapping from cryoelectron microscopy, Proc. Natl. Acad. Sci. papillomavirus capsid revealed by a three-dimensional recon- USA 95, 14622–14627. struction to 9 Å resolution, Nat. Struct. Biol. 4, 413–420. Crum, J., Gruys, K. J., and Frey, T. G. (1994) Electron microscopy Van Heel, M. (1987) Similarity between images, Ultramicroscopy of cytochrome c oxidase crystals: Labeling of subunit III with a 21, 95–100. monomaleimide undecagold cluster compound, 33, 13719–13726. Weinstein, S., Jahn, W., Hansen, H., Wittmann, H. G., and Yonath, A. (1989) Novel procedures for derivatization of ribo- Fuller, S. D., Butcher, S. J., Cheng, R. H., and Baker, T. S. (1996) somes for crystallographic studies, J. Biol. Chem. 264, 19138– Three-dimensional reconstruction of icosahedral particles: ‘‘The 19142. uncommon line,’’ J. Struct. Biol. 116, 48–55. Wenzel, T., and Baumeister, W. (1995) Conformational constraints Furuya, F. R., Miller, L. L., Hainfeld, J. F., Christopfel, W. C., and in protein degradation by the 20S proteasome, Nat. Struct. Biol. Kenny, P. W. (1988) Use of Ir 4(CO)11 to measure the lengths of 2, 199–204. organic molecules with a scanning transmission electron micro- scope, J. Am. Chem. Soc. 110, 641–643. Wingfield, P. T., Stahl, S. J., Williams, R. W., and Steven, A. C. (1995) Hepatitis core antigen produced in Escherichia coli: Griffiths, G. (1994) Fine Structure Immunocytochemistry, Subunit composition, conformational analysis, and in vitro Springer-Verlag, Berlin. capsid assembly, Biochemistry 34, 4919–4932. Gregori, L., Hainfeld, J. F., Simon, M. N., and Goldgaber, D. (1997) Zlotnick, A., Cheng, N., Conway, J. F., Booy, F. P., Steven, A. C., Amyloid beta-protein inhibits ubiquitin-dependent protein deg- Stahl, S. J., and Wingfield, P. T. (1996) Dimorphism of hepatitis radation in vitro, J. Biol. Chem. 272, 58–62. B virus capsids is strongly influenced by the C-terminus of the Hainfeld, J. F. (1989) Undecagold-antibody method, in Hayat, capsid protein, Biochemistry 35, 7412–7421. M. A. (Ed.), Colloidal Gold: Principles, Methods, and Applica- Zlotnick, A., Cheng, N., Stahl, S. J., Conway, J. F., Steven, A. C., tions., Vol. 2, pp. 413–429, Academic Press, San Diego. and Wingfield, P. T. (1997) Localization of the C terminus of the Hainfeld, J. F. (1992) Site specific cluster labels, Ultramicroscopy assembly domain of hepatitis B virus capsid protein: Implica- 46, 135–144. tions for morphogenesis and organization of encapsidated RNA, Hainfeld, J. F., and Furuya, F. R. (1992) A 1.4-nm gold cluster Proc. Natl. Acad. Sci. USA 94, 9556–9561.