Formation of two-dimensional crystals of icosahedral RNA viruses

Micron 39 (2008) 431–446 www.elsevier.com/locate/micron

Formation of two-dimensional crystals of icosahedral RNA viruses Bernard Lorber a,*, Marc Adrian b, Jean Witz c, Mathieu Erhardt d, J. Robin Harris e a

Architecture et Re´activite´ de l’ARN, Universite´ Louis Pasteur de Strasbourg, CNRS, IBMC, 15 rue Rene´ Descartes, 67084 Strasbourg, France b Laboratoire d’Analyse Ultrastructurale, Baˆtiment de Biologie, Universite´ de Lausanne, CH-1015 Lausanne, Switzerland c Formerly De´partement ‘‘Immunochimie des Peptides et des Virus’’, UPR 9021, Institut de Biologie Mole´culaire et Cellulaire, 15 rue Rene´ Descartes, F-67084 Strasbourg, France d Institut de Biologie Mole´culaire des Plantes, UPR 2357 du CNRS, 12 rue du Ge´ne´ral Zimmer, F-67000 Strasbourg, France e Institute of Zoology, University of Mainz, Saarstrasse 21, D-55099 Mainz, Germany Received 5 December 2006; received in revised form 8 February 2007; accepted 8 February 2007 This article is dedicated to the memory of Jean and Marie-The´re`se Witz.

Abstract The formation of 2D arrays of three small icosahedral RNA viruses with known 3D structures (tomato bushy stunt virus, turnip yellow mosaic virus and bromegrass mosaic virus) has been investigated to determine the role of each component of a negative staining solution containing ammonium molybdate and polyethylene glycol. Virion association was monitored by dynamic light scattering (DLS) and virus array formation was visualised by conventional transmission electron microscopy and cryo-electron microscopy after negative staining. The structural properties of viral arrays prepared in vitro were compared to those of microcrystals found in the leaves of infected plants. A novel form of macroscopic 3D crystals of turnip yellow mosaic virus has been grown in the negative staining solution. On the basis of the experimental results, the hypothesis is advanced that microscopic arrays might be planar crystallisation nuclei. The formation of 2D crystals and the enhancing effect of polyethylene glycol on the self-organisation of virions at the air/water interface are discussed. Synopsis: The formation of 2D arrays of icosahedral viruses was investigated by spectroscopic and transmission electron microscopic methods. # 2007 Elsevier Ltd. All rights reserved. Keywords: Icosahedral virus; RNA virus; Tomato bushy stunt virus (TBSV); Turnip yellow mosaic virus (TYMV); Brome mosaic virus (BMV); Transmission electron microscopy; Cryo-electron microscopy; Negative staining; Polyethylene glycol (PEG); Ammonium molybdate; Air/water interface; Adsorption; Nucleation; Crystallisation; Crystallogenesis

1. Introduction Under either normal physiological or pathological conditions, many bacterial, plant and animal cells contain crystalline protein or virus inclusions (for reviews, see e.g. Maramorosch (1977) and McPherson (1999), and references therein). Most of these natural 2 or 3D arrangements have too many imperfections to be suitable for in situ electron diffraction analyses. To overcome this limitation, methods have been implemented to produce in vitro arrays with a high degree of periodicity directly on the support grids used for

* Corresponding author at: Architecture et Re´activite´ de l’ARN, Universite´ Louis Pasteur de Strasbourg, CNRS, IBMC, 15 rue Rene´ Descartes, 67084 Strasbourg, France. Tel.: +33 3 8841 7008; fax: +33 3 8860 2218. E-mail address: [email protected] (B. Lorber). 0968-4328/$ – see front matter # 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.micron.2007.02.008

transmission electron microscopy. When the order of the crystalline lattice encompassing large numbers of identical particles is high enough, information on the packing contacts and the architecture of the individual unit particle can be gained (see e.g. Harris and Holzenburg, 1989; Vonck and van Bruggen, 1990; Brisson et al., 1994; Harris, 1997; Walz and Grigorieff, 1998). Horne and co-workers implemented a two-stage negative staining-carbon film technique with low concentrations of ammonium molybdate or uranyl acetate and noted some icosahedral viruses have a propensity to form either square or hexagonal arrays on the surface of mica sheets (Horne and Pasqualli-Ronchetti, 1974; Horne et al., 1975a,b). Subsequently, Wells et al. (1981) reported that the addition of polyethylene glycol (PEG) to the stain enhanced the formation of virus arrays. Mixtures of ammonium molybdate and PEG have also been utilised to produce 2D crystals from a number of

432

B. Lorber et al. / Micron 39 (2008) 431–446

proteins and enzymes (Harris, 1991, 2007; Harris et al., 1992; Harris and Holzenburg, 1995; Harris and Adrian, 1999). The combination of cryoelectron microscopy (Wilson and Glaeser, 1974) and negative staining has improved contrast and resolution of the images. It also provided the first indication that viruses can create 2D arrays within an unsupported thinlayer aqueous system (Adrian et al., 1998). Again, the addition of PEG was found to increase the tendency of viruses to form ordered 2D crystals prior to vitrification (Harris and Adrian, 1999; Harris, 2007). We have investigated how the two components of the negative-staining solution, the electron-dense ammonium molybdate solution and the PEG contribute to the formation of 2D arrays of three small icosahedral RNAviruses. The shells of these viral particles are highly symmetric: they possess 20 equivalent facets, 15 twofold, 10 threefold, and 6 fivefold axes. The bromovirus, bromegrass mosaic virus (BMV), the tombusvirus, tomato bushy stunt virus (TBSV), and the tymovirus, turnip yellow mosaic virus (TYMV) were chosen as models because their crystallographic 3D structures are known. The association of each type of virion in aqueous solutions containing the ingredients of the stain, either separately or in mixtures, has been monitored by non-invasive spectroscopy. The effect of these compounds on the formation of 2D virus array formation has been visualised by conventional negative-stain transmission electron microscopy (TEM) and by cryo-negative staining TEM. The structural properties of in vitro-prepared arrays of virions have been compared to those of microcrystals made of the same virions within the leaves of infected plants. Our experimental results have led us to raise the question whether the smallest virion arrays observed on the electron microscope grids might be planar crystal nuclei. Further, the mechanism by which PEG promotes the self-association of virions into 2D and 3D crystals is discussed.

2. Materials and methods 2.1. Chemicals Ammonium heptamolybdate tetrahydrate (Mr 1235.86, Cat. No. 1182 from Merck or A7302 from Aldrich) and uranyl acetate dihydrate (Mr 424.15, Cat. No. 8473, Merck) powders were used as received. Polyethylene glycol (PEG, Mr 1000) was from Hampton Research (50% m/v aqueous solution, Cat. No. HR2-523) or Sigma (solid, Cat. No. 81188). All solvents for light scattering analyses were prepared with sterile thricedistilled water. 2.2. Viruses Brome mosaic virus (BMV, common strain) was multiplied in barley Hordeum vulgare L., tomato bushy stunt virus (TBSV) in Datura stramonium, and turnip yellow mosaic virus (TYMV) in Chinese cabbage Brassica rapa c.v. Just Right. The three viruses were purified by isoelectric precipitation of the sap proteins at pH 4.8, followed by two or three cycles of low and high-speed centrifugation (Matthews, 1960). Concentrated virus samples were filtered through 0.2 mm pore diameter membranes (Millex, Millipore) and stored at 4 8C in 20 mM sodium acetate pH 5.8 with 0.1% (m/v) sodium azide. The quality of the virus preparations was estimated from the ratio of their absorbance at 260 nm over that at 280 nm (Table 1). Virion concentration was determined from absorbance at 260 nm (measured on 2 ml samples with a Nanodrop1 ND-100 spectrophotometer) and extinction coefficient (Table 1). 2.3. Crystallisation assays Virus preparations were tested for crystallisability in various solvents and at several virus concentrations using the vapour

Table 1 Structural and physical chemical properties of model viruses Virus

Virion with T = 3 geometrya

Genome

Capsid protein

Composition and Mr

Subunit Mr Copies Total (number of aa) Mrb

BMV Tripartite RNA, 1.0  10 6 20,385 (185) TBSV Monopartite RNA, 1.7  106 40,540 (387) TYMV Monopartite RNA, 1.9  106 20,088 (189)

180 180 180

S20,wc (s)

4.7  106 87 9.0  106 132 5.5  106 116

dhe A2f E260 nmg A260 nm/ dn/dci (nm) (ml mol/g2) (mg/ml/cm) A280 nm h (ml/g)

D0d (cm2/s) 1.3  10 1.1  10 1.4  10

7 7 7

32 37 33

8  10 7  10 8  10

4 4 4

5.1 5.0 8.4

1.7 1.6 1.5

0.2 0.2 0.2

pI j 6.8 4.1 3.8

a The stability of the virions varies with the physical chemical properties of the solvent. BMV is stable from pH 3–6 and unstable at pH > 7. The removal of calcium ions bound to TBSV induces the reversible swelling. The capsid of TYMV swells when the pH exceeds 11. b Sum of the Mr of its protein and of its nucleic acid components. c Sedimentation velocity of infective virions in the ultracentrifuge. Data for water at 20 8C at pH < 7 from Description of Plant Viruses at URL http:// www.dpvweb.net/ (Adams and Antoniw, 2006). d Translational diffusion coefficient extrapolated at zero concentration derived from dynamic light scattering measurements performed on virus in 0.1 M sodium acetate solution at pH 4.5 and at 20 8C. e Hydrodynamic diameter deduced from translational diffusion coefficient assuming that particles are spheres. f Second virial coefficient given by the slope of the plot representing the scattered intensity (measured by static light scattering) vs. the virus concentration. Measurements were performed on virus in 0.1 M sodium acetate solution at pH 4.5 and at 20 8C. g Extinction coefficient at 260 nm. h Maximal quotient of absorbance at 260 nm on absorbance at 280 nm for pure virus. i Increment of refractive index determined by spectrometry and refractometry. j Isoelectric point taken from Description of Plant Viruses at URL http://www.dpvweb.net/ (Adams and Antoniw, 2006).

B. Lorber et al. / Micron 39 (2008) 431–446

diffusion technique (Ducruix and Giege´, 1999) in transparent Modular VDX crystallisation plates (Cat. No. HR3-198, Hampton Research). 5 ml droplets of virus suspension plus 5 ml crystallisation medium were deposited on 22 mm square transparent plastic coverslips (Cat. No. MD4-10, Molecular Dimensions, Ltd., U.K.) for equilibration at 20%C against a reservoir of 200 ml crystallisation medium. Assays were left undisturbed at 20 8C in a vibration-free incubator (Type IPP200, Memmert, Germany). 2.4. Light scattering and related methods Scattered light intensities (in static mode) and correlation times (in dynamic mode) (Schmitz, 1990) were measured at 20 8C with a Zetasizer1Nano S instrument (Malvern, UK). Samples were analysed in 45 ml quartz cuvettes with virus concentrations in the 0.01–0.1 mg/ml range. Toluene was the standard scatterer. All solvents were prepared with thricedistilled water and filtered through 0.2 mm pore diameter Millex membranes (Millipore) to eliminate dust particles. Experimental data were corrected for virus dn/dc, solvent viscosity and refractive index. Particle shape was assumed to be a sphere. Refractive indices were measured at 20 8C with an Abbe´ refractometer. The increment of refractive index with concentration (dn/dc) of the three viruses was found to be 0.2 ml/g. Solvent viscosities were measured at 20 8C with 3.5 ml micro-Ubbelohde capillary viscosimeter tubes (Schott, Germany). 2.5. Electron microscopy Integrity and association of virions was verified by negative staining with a 2% (m/v) (47 mM) aqueous uranyl acetate solution (solubility 7.6% m/v or 179 mM in water at 20 8C, pH 4.5). Nickel grids coated with a polyvinyl formal (Formvar1) film were covered with 2–20 ml virus suspension at 0.1–1 mg/ml. The excess was blotted after 2 min and 5 ml uranyl acetate solution was added for staining. After 10 s the excess stain was blotted by contact with a piece of filter paper and the grid was air-dried for 5 min at ambient temperature. Cryo-negative staining with ammonium molybdate was performed with an aqueous 16% (m/v) (130 mM) solution of ammonium heptamolybdate (solubility 40% m/v or 324 mM in water at 20 8C, initial pH 5.3) adjusted to pH 6 with concentrated ammonium hydroxide or sodium hydroxide. Virus 2D arrays were prepared by including the following modification of the original cryo-negative staining procedure (Adrian et al., 1998). A small volume of virus suspension (4 ml at 2–4 mg/ml – instead of 0.1–1 mg/ml – in distilled water containing 1–2% m/v PEG-1000) was applied during 30 s to a grid covered with a gold sputter-coated holey carbon film. This grid was then placed on the surface of a 100 ml drop of 16% (m/v) ammonium molybdate and 0.1–0.2% (m/v) PEG-1000 adjusted to pH 6 with NaOH. After 60 s the grid was removed and incubated 5 min at 20 8C to allow the formation of 2D viral arrays at the air–fluid interface. The excess of solution was blotted and the grid plunged in liquid ethane to vitrify the sample.

433

Negatively stained air-dried virus samples were examined with two transmission electron microscopes: a Hitachi H-600 operated at 75 kVequipped with a Hamamatsu CCD Advantage HR camera in Strasbourg and a Zeiss EM900 in Lausanne. Vitrified negatively stained samples were examined in Lausanne with a Philips CM12 transmission electron microscope equipped with a Gatan 626 cryo-holder and operated at 80 kV. Images were recorded on Kodak SO-163 film exposed at low dose with a 900 mm instrumental image defocus. Cryo-EM images were digitized and processed with the software Image J (Rasband, 1997). A fast Fourier transform (FFT) was generated to identify the geometric characteristics of the 2D arrays. 3. Results 3.1. Structural properties of model viruses The structural properties of the three small RNA viruses that served as models in this study are summarised in Table 1. The capsids of these viruses have diameters of 30 nm and possess icosahedral architectures with a T = 3 geometry. Each virus is assembled from 180 identical copies of a protein subunit and contains one or several ribonucleic acids. The nucleic acids of BMV and of TYMV begin with a tRNA-like fold that can be charged, respectively, by histidine or valine during the initiation of viral translation. A prerequisite of this study was that the virus preparations were of good quality. The high ratios of their absorbance at 260 nm and at 280 nm suggested that the virions had not been altered during purification or storage. All batches were mainly composed of individual and intact virions, as revealed by TEM, but pairs were also present (Fig. 1). When suspended in the solvents listed in Table 2, all viruses produced 3D crystals visible to the naked eye (Fig. 2). Altogether, these results confirmed the purity and homogeneity of the model viruses.

Fig. 1. Distribution of virions in BMV, TBSV and TYMV batches. The graphs represent the percentage of virions found as single particles, dimers, trimers or greater aggregates on TEM micrographs. Symbols are: (solid diamond) for BMV (n = 920), (open square) for TBSV (n = 1911) and (open triangle) for TYMV (n = 1443). Areas of the grid where dense clusters of particles had formed due to the unevenness of the carbon film were not taken into account.

434

B. Lorber et al. / Micron 39 (2008) 431–446

Table 2 Solvents for the production of 3D crystals of model viruses Virus

T (8C)

c (mg/ml)

Solvent compositiona,b

Crystal habitc

Refs.

BMV

15–25

10–50

0.02 M sodium acetate pH 4.5, 40% (m/v) PEG-8000 0.1 M magnesium acetate, 23% (v/v) PEG-550 MME pH 5.2 0.1 M magnesium acetate, 23% (v/v) PEG-550 MME pH 5.2 0.1 M sodium acetate pH 4.6, 30% (m/v) PEG-400 0.2 M sodium sulfate pH 7, 20% (m/v) PEG-3350 0.1 M sodium acetate pH 4.6, 30% (m/v) PEG-400

Facetted polyhedron Orthorhombic prism Rhombohedron Lens Rod and plate

Casselyn et al. (2001) Lucas et al. (2001) Lucas et al. (2002) This work This work

TBSV

4–25

20–100

0.5–1.0 M ammonium sulfate 0.05 M sodium acetate pH 4.5, 8–13% (m/v) PEG-8000

Dodecahedron Dodecahedron

Bawden and Pirie (1938) Ng et al. (1996)

TYMV

15–25

15–35

1.0–2.0 M ammonium phosphate, 0.1 M MES pH 3.9 0.1 M sodium acetate pH 4.5, 30% (m/v) PEG-400 0.2 M potassium sulfate pH 5.8, 20% (m/v) PEG-3350 0.2 M sodium phosphate pH 4.8, 20% PEG-3350 0.2 M potassium phosphate pH 5.8, 20% (m/v) PEG-3350 0.05 M ammonium molybdate pH 7.0, 2% (m/v) PEG-1000

Hexagonal dipyramide Dipyramide Hexagonal plate Dipyramide Plate Hexagonal plate

Canady et al. (1996) This work This work This work This work This work

a

Crystallization results obtained with the PEG screen from Nextal Biotech (Montre´al, Canada). Other solvents for BMV crystallization include: 0.1 M Na acetate pH 4.6 with 30% (v/v) PEG-400 and 0.1 M Na acetate pH 4.6 with 30% (m/v) PEG-550 MME. In both cases the habit is a rounded cube. c Several polymorphs may be present simultaneously. Abbreviations used: MES, morpholino ethane sulfonic acid; PEG, polyethylene glycol; MME, monomethylether. b

3.2. Viral association monitored by light scattering Dynamic light scattering (DLS) measurements performed on BMV, TBSVand TYMV in pure water revealed the presence of traces of viral aggregates with a mean diameter dh = 400 nm (Fig. 3a–c, upper left panels) that represented not more than 1% of the total number of virions (right panels in Fig. 3a–c). Such minute amounts were detected because the scattering intensity is proportional to the sixth power of the particle’s diameter. In water, viral association was also reflected by the higher second virial coefficients (i.e. A2 = 1.4  10 3 ml mol/g2 for BMV, 1.6  10 3 ml mol/g2 for TBSV, and 0.8  10 3 ml mol/g2 for TYMV) than in an aqueous solution containing a low salt concentration (i.e. 0.1 M sodium acetate, Table 1). 3.2.1. Effect of uranyl acetate In a 2% (m/v) uranyl acetate solution (47 mM, pH 4.5), single virions of BMV and TBSV were accompanied by traces of aggregates with a mean dh = 350 nm and 200 nm, respectively (Fig. 3a–c, second panels from top) and TYMV formed aggregates with dh 1500 nm. A2 values were comparable to those determined in sodium acetate solution. 3.2.2. Effect of ammonium molybdate The addition of 1% (m/v, or 166 mM, pH 6) ammonium molybdate to water dissolved the aggregates (Figs. 3a–c, central panels). Salting-in occurred at concentrations up to 5% (m/v) (result not shown). At salt concentrations between 5 and 16% (m/v), some aggregates with dh = 600 nm and others with dh = 6000 nm (upper detection limit of the instrument) appeared in BMV suspensions (Fig. 3a). Similarly, aggregates with dh = 300 nm were detected in TYMV solutions (Fig. 3c). All virions of TBSV were incorporated into aggregates with

small (dh = 80 nm) or large sizes (dh = 2000 nm)(Fig. 3b). A2 values of BMV and TYMV were greater (1.9  10 3 and 2.4  10 3, respectively) but that of TBSV was lower (0.5  10 3 ml mol/g2). 3.2.3. Effect of PEG-1000 At a concentration of 1% (m/v) in water this polymer led to the formation of a few aggregates (<1% in number of virions) in the samples of BMV (dh = 600 nm, Fig. 3a, second panels from bottom) and TBSV (dh = 350 nm, Fig. 3b). At variance, over 50% of the TYMV virions were present in aggregates 100 nm in diameter (Fig. 3c). Values of A2 for BMV were unchanged, while those for TBSV ( 0.1  10 3) and TYMV (0.3  10 3 ml mol/g2) were lower. 3.2.4. Combined effect of ammonium molybdate and PEG-1000 When the two compounds were present together, small amounts of large and very large aggregates formed in the samples of BMV (dh = 2000 nm, Fig. 3a, bottom panels) and TYMV (dh = 6000 nm, Fig. 3c). All TBSV virions were in medium (dh = 600 nm) and large aggregates (dh = 2000 nm, Fig. 3b). The values of A2 were not drastically altered: they were 0.6  10 3, 0.4  10 3 and 1.3  10 3 ml mol/g2 for BMV, TBSV and TYMV, respectively. 3.3. Electron microscopy of negatively stained virions Figs. 4–6 display the effects of a low concentration of PEG1000 on the formation of 2D arrays by the three model viruses as observed by cryo-electron microscopy. The randomly distributed virions of BMV seen in the presence of ammonium molybdate alone, organised themselves in a hexagonal layer after the addition of PEG (Fig. 5a and b). (For representations of

B. Lorber et al. / Micron 39 (2008) 431–446

435

Fig. 2. 3D crystals of model viruses. Crystals were grown in the presence of either a high PEG concentration (left-hand images), a high salt concentration (images in the center) or in the presence the negative staining solution made of ammonium molybdate and PEG-1000 (images on the right-hand side). Crystallisation solvents were: for BMV (a) 0.02 M sodium acetate pH 4.5 with 40% (m/v) PEG-8000; for TBSV(b) 0.1 M sodium acetate pH 4.5 with 10% PEG-8000, and (c) 0.7 M ammonium sulfate, and for TYMV (d) 0.2 M potassium sulfate pH 5.7 with 20% (m/v) PEG-3350, (e) 0.1 M morpholino ethanesulfonate pH 3.9 with 1.5 M ammonium phosphate or (f) negative staining solution containing 0.05 M ammonium molybdate pH 6.5 and 2% (m/v) PEG-1000. Note the growth of crystals with a tabular habit in the later condition. All images are at approximately the same scale and the largest crystals measured about 0.5 mm across. Virus concentrations and crystallisation temperatures are detailed in Table 2.

various arrangements of spheres, see e.g. Nicholas, 1965; Hammond, 1997.) The largest 2D arrays encompassed up to 150 particles. fast Fourier transforms of digitised images of the areas covering less than 10  10 virions exhibited reflections of the 5th order and above (Fig. 4c). When three virion layers were stacked (Fig. 4d), the interleaving one nestled in the interstices of the layers below and above and a hexagonal close-packed unit cell resulted.

At variance with BMV, TBSV formed 2D arrays in the absence of PEG (Fig. 5a and b). Virions organised themselves in either hexagonal (Fig. 5a) or square arrays (Fig. 5b) not exceeding 100 virions (Fig. 5b). When 1% (m/v) PEG-1000 was added to ammonium molybdate, the surface area of the resulting square arrays was enlarged by a factor 10–100 (Fig. 5c). Since no second virion layer was observed, it is not known if the unit cell is face-centred or body-centred cubic (in

436

B. Lorber et al. / Micron 39 (2008) 431–446

the first packing, the spherical particles are in closer contact than in the second one). The fast Fourier transforms of images of an area of less than 10  10 virions of the hexagonal array exhibited reflections of the 8th order and higher (Fig. 4c). When the same treatment was applied to a defect-free area of more than 20  20 virions, reflections of the 10th order were visible. In an exceptional case, a large and defect-free array gave a few reflections of the 15th order (result not shown). Indeed, the 2D arrays possessed defects resulting essentially from missing virions or from incorporation of damaged or supernumerary ones (Fig. 5c). As for TBSV, TYMV formed 2D arrays in the absence of PEG. The largest monolayers encompassed no more than 250 virions and had either a hexagonal or a square pattern. Fig. 4a displays a transition between a square lattice and two hexagonal lattices. The power spectra of the fast Fourier transforms of these small parts of the array exhibited reflections of at least the 4th (Fig. 6c) and 7th order (Fig. 6d), respectively. The addition of 1% (m/v) PEG-1000 doubled the surface area of the

hexagonal arrays. In the virus concentration range (2–4 mg/ml) used here, the array was always made of single and paired rows of less than 20 virions with many vacant positions (Fig. 6b). The disorder in these viral monolayers precluded further analysis by crystallographic image processing. 4. Discussion Interest in the formation of ordered monomolecular arrays and 2D micro- and macro-crystals extends across the disciplines of physics, chemistry (e.g. Wickman and Korley, 1998), polymer and colloid science (e.g. van Blaaderen et al., 1997), biochemistry, microbiology and structural biology and virology (e.g. Holzenburg, 1988; Harris, 1991; McPherson, 1999; Tihova et al., 2004; Tang et al., 2005), biotechnology (e.g. Pum et al., 1993), and medicine (e.g. Basu et al., 2004). In the present study, the crystallogenesis of three small quasi-spherical viruses was investigated by analysing the association behaviour of these highly symmetrical particles in

Fig. 3. Effect of solvent composition on BMV, TBSV and TYMV aggregation analysed by dynamic light scattering. The three panels on this page and two following correspond to (a) BMV, (b) TBSV and (c) TYMV. Histograms in the left-hand side panels represent the intensity particle size distribution and the ones in right-hand side panels the number particle size distribution vs. mean particle diameter. Data are mean values (with standard deviations) of measurements performed on five samples at different virus concentrations in the range 0.01–0.1 mg/ml. Experimental data take into account differences in solvent refractive index and viscosity. With BMV and TYMV the deconvolution algorithm generated a peak with dh = 1–2 nm that corresponded to ammonium molybdate or PEG (the protein subunit with Mr  20,000 has a dh  3–4 nm).

B. Lorber et al. / Micron 39 (2008) 431–446

437

Fig. 3. (Continued )

solution and by comparing the properties of the 2D arrays formed on the grids of the electron microscope with the periodic arrangements found in infected plants. 4.1. Virus aggregation in solution 4.1.1. Salt solutions Supersaturation is the driving force that governs the transition of molecules from the soluble to crystalline phase. It must be sufficient to overcome the electrostatic potential barrier and to trigger the formation of a critical nucleus that will ultimately become a crystal (Chernov, 1984). The degree of virus supersaturation is influenced by numerous variables, including the nature and concentration of the crystallising agent. Salt ions playing this role, have been ranked in the Hofmeister series, according to their ability to salt-out (i.e. insolubilise) or precipitate proteins (for an English translation of Hofmeister’s original papers, see Kunz et al., 2004; for a review on the Hofmeister series, see Cacace et al., 1997). The so-called counter-ions shield the charges on the protein surface and force the protein molecules to associate through electrostatic, ionic and/or hydrophobic interactions. The balance between protein hydration and salt binding is subtle (see Arakawa and Timasheff, 1984). Divalent and higher valency

ions may establish bridges between macromolecules that may aid in establishing inter-particle contacts. As shown in Fig. 2, ammonium molybdate at moderate concentration plays the role of a crystallising agent at an adequate virus concentration. The difference between the solution pH and the isoelectric point (pI) of the particle is another variable that critically influences the solubility of a virus. The pI is the pH at which the net charge of the particle, and consequently its mobility in an electric field, is zero (Bostro¨m et al., 2005). Our models differ in their isoelectric points: the pI of BMV (6.8) is close to neutrality and those of TBSV (4.1) and TYMV (3.8) are more acidic (Table 1). Thus, in an ammonium molybdate solution at pH 6, the charge of BMV is expected to be close to zero and that of TBSV and TYMV to be negative. Our dynamic light scattering results do not support the prediction that BMV should be less soluble. On the contrary, it appears to be the most soluble of the three viruses in all solvents (Fig. 3a–c). 4.1.2. Effect of PEG At constant temperature, the presence of a co-solute like PEG is another crystallisation variable. The three viruses crystallise in a wide range of conditions, in the presence of PEGs with various chain lengths (Table 2). The fact that BMV requires PEG to arrange as stable arrays highlights the

438

B. Lorber et al. / Micron 39 (2008) 431–446

Fig. 3. (Continued ).

importance of this additive to enhance associative interactions between virions (Fig. 4b). PEG alone also induced TBSV to form large ordered aggregates (Fig. 5c) and the mixture of salt and PEG strongly aggregates these virions in solution (Fig. 3b). Polyethylene glycols with the general formula –[CH2–CH2– O]n–CH2–CH2–OH are non-ionic and hydrophilic chains possessing a flexible random coil structure. Upon addition to biological particles they form coacervates. This partioning effect is widely applied during the purification of proteins (Hebert, 1963; Leberman, 1966). In solution, PEG molecules interact with water as a dynamic network (Vergara et al., 1999). Preferential solvent interactions between proteins and PEG exclude the polymer from the protein domain (Atha and Ingham, 1981; Lee and Lee, 1981; Arakawa and Timasheff, 1985), but depletion of bound water on the protein surface may occur. Alone or together with salts, PEGs are efficient crystallising agents for biomacromolecules and viruses (McPherson, 1976; Arakali et al., 1995). Much larger crystals of muscle phosphoglucomutase could be grown when low concentrations of short-chain PEG molecules were mixed with high salt concentrations, because most disordered aggregates appearing during protein precipitation were solubilised and were no longer potential nucleation sites (Ray, 1992 and references therein).

In this context, in X-ray scattering a decrease of the second virial coefficient in protein and virus solutions has been interpreted as a consequence of the strengthening of intermolecular interactions (Bonnete´ and Vivare`s, 2002). Particle crowding produced by PEGs and hydration depletion interactions of polymer segments in the vicinity of macromolecules can enhance protein self-association, precipitation or crystallisation (Vergara et al., 2002; Vivare`s and Bonnete´, 2002). Apparently, the larger the protein particles, the stronger are the attractive interactions (Tardieu et al., 2002). In our study, BMV formed arrays only when 1–2% (m/v) PEG-1000 was present in the 16% (m/v) ammonium molybdate solution although light scattering measurements did not indicate any synergic effect of both compounds (i.e. aggregates were not larger in the presence of both chemicals than in the presence of one alone, as shown in Fig. 3a). A small angle X-ray scattering study of BMV demonstrated that the combination of PEG and salt creates a synergetic increase of attractive interactions for small proteins (Casselyn et al., 2001 and references therein). With TBSV we have observed a similar strong synergy, since all virions were aggregated in the mixture of ammonium molybdate and PEG and the mean diameter of the aggregates was greater than in pure solutions (Fig. 3b). This was correlated with the enlargement of the 2D arrays of TBSV

B. Lorber et al. / Micron 39 (2008) 431–446

439

Fig. 4. Effect of PEG on the formation of 2D arrays of virions of BMV visualised by cryo-negative staining with ammonium molybdate. (a) Close-up view of the random distribution of virions in the absence of PEG; (b) close-up view of hexagonal 2D arrays after addition of 2% (v/v) PEG-1000 at pH 6. The capsids seem to be covered with tiny spikes. (la) indicates a second layer of virions laying on top of the first one; (c) fast Fourier transform of the marked 6  8 virions area in the upper left corner of panel b. The arrow points toward a reflection of the 5th order with the corresponding resolution; (d) cryo-electron micrograph showing a superposition of up to three layers of BMV particles. Numbers next to circled areas indicate the number of virions layers that are superimposed at that place.

(Fig. 5). As to TYMV, PEG did not contribute to increase the size of its aggregates in DLS but the 2D arrays observed in electron microscopy had larger dimensions. Moreover, macroscopic 3D crystals were reproducibly prepared in the molybdate/PEG staining solution (Table 2, Fig. 2). 4.2. Properties of virus arrays formed on electron microscopy grids The pioneers of viral electron microscopy introduced tungsten, cadmium, molybdenum, paladium, osmium and uranium salts to embed viruses as electron-dense positively stained particles and crystals, enhance differences with the background and reveal details of the structure of well-preserved assemblies (e.g. Brenner and Horne, 1959). Flash-freezing and analysis under cryogenic conditions further improved contrast and resolution. By varying the composition of the negative stain for use in cryo-electron microscopy, it was found that a 16% (m/v) ammonium molydbate solution adjusted to pH 6 can yield high resolution protein and viral images (Adrian et al., 1998). Here the addition of a low concentration (1–2% m/v) of PEG-1000 favoured the self-organisation of virions during cryo-negative staining. This 2D crystallisation phenomenon is

essentially similar to the one observed for a number of proteins, including human erythrocyte catalase (Harris and Holzenburg, 1989; Harris et al., 1993; Harris and Holzenburg, 1995), using the negative staining-carbon film technique. In the case of BMV, randomly distributed virions or groups of virions were replaced by well-ordered hexagonal arrays (Fig. 4a and b). The superimposition of further monolayers of virions indicated a tendency to create thin 3D crystals (Fig. 3d). The effect was even stronger with TBSV; these viral particles self-organised as 10 to 100-fold larger 2D arrays in the presence of the PEG (Fig. 4a–c). TYMV also formed larger arrays but they had many vacant positions (Fig. 5a and b). An explanation for this might simply be that the space between double rows and single rows remained due to the lack of available virions from the bulk solution to complete the monolayer lattice. Another possibility might be that virion movement from the fluid to the surface monolayer was hindered while the complex 2D pattern was established. It is noteworthy that TBSV was mainly arranged in paired rows of orientated virions in the absence of PEG (Fig. 5b) (see Adrian et al., 1998, Fig. 7) and that TYMV associated in long adjacent rows only in the presence of the polymer (Fig. 6b). As shown in Fig. 1, a good proportion of virions were engaged in pairs after staining with uranyl acetate. This preferential viral

440

B. Lorber et al. / Micron 39 (2008) 431–446

Fig. 5. Effect of PEG on the formation of 2D arrays of TBSV. Close-up views of (a) hexagonal and (b) square arrays visualised by cryo-negative staining with 16% (m/v) ammonium molybdate alone at pH 6. In panel b, viral pairing leads to the formation of paired rows and a disruption of the order within the array. Arrows point toward grain boundaries (g). (c) The addition of 1% (m/v) PEG-1000 is accompanied by an enlargement of the dimensions of the square 2D array. Only about one third of the surface area of the actual crystal is shown. The surface of the capsids appears to be rough. For a 3D reconstruction of the virus particle, see Fig. 7 in (Adrian et al., 1998). In panel c, some lattice defects are marked; they include linear defects such as dislocations (d), point vacancies (v), incorporation of damaged virions (appearing darker because the stain has filled the empty protein capsid) (p) and of supernumerary particles (s). The lattice type is indicated by a white hexagon or square. (d,e) fast Fourier transform of the areas of the hexagonal (9  8 virion) and square (24  22 virion) arrays delineated in panels a and c, respectively. Arrows point toward a reflection plane for which the corresponding resolution is given. A few faint reflections of the 10th order (resolution 3 nm) were visible on the original image.

association had the disadvantage that it could be the source of the misalignments in the final arrays. Fast Fourier transforms produced from the above 2D arrays present on the electron microscope grids enables one to distinguish 2D close-packing in monolayer arrays from true 2D crystals. The FFTs of the former arrays should be composed of only first-order reflections and those of the latter should also display second-order and higher reflections. As shown here, the arrays of BMV (Fig. 4c), TBSV (Fig. 5d and e) and TYMV (Fig. 6c and d) share the common property to yield FFTs with

reflections of at least the 4th to the 10th order (depending on the dimensions of the area analysed). Experimentation with virus samples at 2–4mg/ml (i.e. from 2 to 40 times more concentrated than those used by Adrian et al., 1998), indicated that, despite the high particle density, the size and the crystalline order of the arrays were always limited. The best result was obtained with TBSV: the FFT of its largest 2D crystals exceptionally contained reflections up to the 15th order. For BMV and TYMV the arrays produced in the presence or absence of PEG were always of smaller size, similar to those of TBSV in the

B. Lorber et al. / Micron 39 (2008) 431–446

441

Fig. 6. Effect of PEG-1000 on the formation of 2D arrays of virions of TYMV during negative staining with ammonium molybdate. (a) Close-up view of small 2D arrays in the absence of PEG. Close-up view of the transition between a square lattice and two hexagonal lattices; (b) loose network with a hexagonal lattice in the presence of 1% PEG-1000 at pH 6. The meshing is essentially made of single and paired rows of virions. Empty capsids have a darker center and capsomers can be seen on the surface of intact virions; (c,d) fast Fourier transforms of the (c) square (4  3 virion) and (d) hexagonal (10  6 virion) areas delineated in panel a. A few weak reflections of the 5th and of the 8th order, respectively, were visible on the original images. Symbols are as in Figs. 4 and 5.

absence of polymer. In view of the extensive progress made by virus crystallography and cryo-electron microscopy single particle analysis (for reviews, see e.g. Fuller et al., 1996 and Lee and Johnson, 2003), no further 2D crystallographic analysis has currently been pursued in the context of this study of virus crystal nucleation and growth. 4.3. In vitro-produced versus naturally occurring virus crystals TBSV was the first virus for which regular arrangements assimilated to crystals have been found in an infected plant (Smith, 1956, Astier et al., 2001). Subsequently, carbon replicas of natural TYMV microcrystals were obtained (Cornuet, 1959) and within a decade osmiophilic arrays of TYMV were

observed in the leaves of Chinese cabbage (Rubio-Huertos et al., 1967). Negatively stained ultra-thin sections of leaves of inoculated oat or barley also revealed crystalline arrangements of BMV (Paliwal, 1970). In various host plants, TBSV was seldom found in the crystalline state (Russo and Martelli, 1972). Table 3 compares the dimensions and the lattice properties of microcrystals existing in vivo and those crystalline virus arrays grown in vitro. The thin sections of plant tissues show the lattices of 3D crystals perpendicularly to a multitude of sectioning planes. Natural and in vitro-grown viral crystals reach comparable dimensions, usually not exceeding 2 microns. In both environments square (in 2D, or cubic in 3D) and hexagonal lattices are predominant (for models of the various arrangements of spheres, see e.g. Nicholas, 1995 and

442

B. Lorber et al. / Micron 39 (2008) 431–446

Fig. 7. Amorphous aggregates vs. organised 2D arrangements of BMV, TBSV and TYMV virions. (Top row) Random virus aggregation visualised by transmission electron microscopy after negative staining with uranyl acetate. Besides single virions, dimers, trimers, larger oligomers and aggregates without any ordered pattern can be recognised. (Bottom row) Cryo-electron micrographs of the smallest ordered arrays visualised after negative staining with ammonium molybdate in the absence or presence of PEG-1000. As soon as a few viral particles come together they arrange as planar arrays with geometrical characteristics.

Table 3 Dimensions and properties of crystalline virus arrays in vitro and in vivo Virus

Dimensionsa, number of virions in crystals and per mm2 and lattice typeb of 2D arrays in vitro

Dimensionsc, number of virionsc in crystals and per m2, and lattice type of arraysb with location in infected plant leavesd with references

BMV

No arrays in the absence of PEG. In the presence of 1–2% (m/v) PEG-1000, numerous small arrays of 0.3  0.4 mm2, 150 virions (1200 v/mm2) with a hexagonal lattice and a thickness of 1–3 layers.

Arrays of 2  0.8 mm2, 1900 virions (1200 v./mm2). Several layers of virions stacked in hexagonal or deformed hexagonal lattice in the cytoplasm of epidermal, xylem and mesophyll cells in oat (Paliwal, 1970).

TBSV

In the absence of PEG, Arrays of 0.4  0.2 mm2, 100 virions (1100 v./mm2), with either square or hexagonal lattice. In the presence of 1–2% (m/v) PEG-1000, Arrays of 2.5  3.5 mm2, 9000 virions (1100 v./mm2) with a square or a hexagonal lattice.

Small clusters of small arrays with no recognizable pattern besides regularly spaced alignments (Smith, 1956). Arrays of 0.6 mm  0.4 mm, 200 virions (900 v./mm2). Loose hexagonal lattice in the cytoplasm of palisade cells, vacuoles and nuclei of parenchyma cells in Datura (Russo and Martelli, 1972).

TYMV

In the absence of PEG, Arrays of 0.3 mm  0.6 mm, 250 virions (1300 v./mm2) Hexagonal or square lattice. In the presence of 1–2% (m/v) PEG-1000, Less ordered arrays with incomplete hexagonal lattice made of single and paired rows of max. 20 virions.

Arrays of 1.4 mm  0.9 mm, 1450 virions (1200 v./mm2) with a hexagonal close-packed lattice (Cornuet, 1959). Loose arrays of 1  0.5 mm2, 400 virions (800 v./mm2) with a hexagonal pattern in some regions found in chloroplast, cytoplasm and X-bodies in Chinese cabbage (Rubio-Huertos et al., 1967).

a

Maximal size measured on cryo-electron microscope grids prepared as described under Section 2. The symmetry depends on the plane along which the lattice is cleaved. c Greatest dimensions of 2D arrays and approximative numbers of virions estimated on micrographs of thin sections displayed in the respective references. No attempt was made to estimate the volume of 3D crystals. d No crystalline array of any of these viruses has been reported to exist in healthy plants. b

B. Lorber et al. / Micron 39 (2008) 431–446

Hammond, 1997). These forms of viral packing achieve highest densities and are close to one another in energy, as illustrated by the transition displayed in Fig. 6a. The in vitro-produced crystals are characterised by the presence of various lattice imperfections; some examples are displayed in Fig. 5c. Although no such imperfections could be identified in crystals that have grown in vivo, processing of digitised images of virus crystals published in literature (see Table 3 for references) was not successful because of the irregularities generated in the virion layers by thin sectioning of the plant tissue. Recently, atomic force microscopy (AFM) analyses have revealed that, as a consequence of their larger size, virus crystals can accommodate a wider range of lattice defects (such as defective virions like smaller or empty capsids and satellite viruses with greater dimensions) than do proteins (Kuznetsov et al., 2005). 4.4. Crystal nuclei trapped in thin films From the crystallogenesis point of view, small ordered virus arrays similar to those displayed in the panel of the bottom row of Fig. 7, might be interpreted as association states corresponding to the start of the crystallisation/nucleation process. Similar planar arrangements (generally termed nuclei) of 20–50 molecules in one or two monomolecular layers were found in supersaturated solutions of the quasi-spherical irontransporter protein ferritin using atomic force microscopy (Yau and Vekilov, 2000). Closer inspection of our cryo-electron micrographs revealed several features common to the three viruses. Firstly, is the presence of several small-sized planar arrangements with rectangular, square or trapezoidal geometries that contain only 6  4 = 24 BMV virions, 5  4 = 20 TBSV virions, or 5  3 = 15 TYMV virions (lower row of panels in Fig. 7). A second feature is that these structures can be clearly distinguished from amorphous aggregates (upper row of panels in Fig. 7). Thirdly, another common point is the quasiabsence individual virions and any smaller virion association states, such as dimers, trimers, tetramers, etc. For the growth of ferritin crystals two pathways have been proposed (Yau and Vekilov, 2000). In the first, single protein molecules progressively bind to the planar nucleus and in the second a building block that is a symmetrical arrangement of a few molecules is required for subsequent crystal growth. Our negatively stained images suggest that the 2D nucleation of quasi-spherical viruses on the holey carbon film is more likely to proceed via the first pathway. No ordered virion aggregates that could play the role of a nucleation site have been identified. The addition of monomers (or perhaps dimers or very small 2D planar groups of virions) onto linear or 2D arrays seems more probable than the incorporation of 3D aggregates possessing a well-defined size and geometry. This view is reinforced by the fact that almost all virions are engaged in at least one interaction and that almost no free virions are present with the 2D arrays. In our study, BMV had a tendency to stack, forming up to three layers (Fig. 4d) and TYMV formed small square 2D networks as soon as relatively few virions were grouped together, even at low virus concentration (Fig. 6).

443

4.5. Insights into virus and protein crystal nucleation The examination of cryo-EM images has leads us to suggest a scenario for the formation of 2D virus crystals during the negative staining-carbon film procedure combining air-dry negative staining across holes and cryo-negative staining. In this process, the rate of water evaporation and the strength of capillary, electrostatic and surface tension forces existing at the air–water interface of the ultra-thin fluid film containing the biological material appear to be crucial for the formation of the 2D crystals. The rapid evaporation of a microvolume of solvent is accompanied by a shrinking of the thickness of the suspended film down to the diameter of a virion (for a schematic representation of the phenomenon on a solid surface, see Denkov et al., 1993). Hence, the concentration of the biological particles dramatically increases. To minimise the number of charged groups exposed to the air, the hydrophobic regions of proteins and viral particles will preferentially locate at the interface between the air and the solution. Hence, their local concentration becomes so high that they are supersaturated and driven to interact. At this stage, close packing may precede nucleation, and a transition from a less ordered packing to a more ordered packing may occur through mobility and a rearrangement of particles. This is suggested by the coexistence of regions with a clear periodicity next to disordered ones within large monolayers (see e.g. micrographs of TBSV and TYMV in Figs. 5 and 6a, respectively, and Fig. 8). However, the small square arrays obtained with TYMV in the absence of PEG (see right-hand panel on the bottom of Fig. 7) indicate that nuclei can form readily. Obviously, the moment at which flashfreezing occurs by plunging the grid in liquid ethane is the critical step since it vitrifies the thin film of highly concentrated and viscous solution and immobilizes the structures contained in it. Three examples are given in Fig. 8. Thus, the plunge freezing procedure is an excellent way to trap transient intermediates of the crystallisation process. Harris and coworkers, who studied the multiple 2D crystal forms of human erythrocyte catalase obtained by the negative staining-carbon film technique on mica sheets (Harris and Holzenburg, 1989, 1995; Harris et al., 1993) came to a similar conclusion from their air-dried samples following analysis by transmission electron microscopy, namely that the 2D crystallization occurred at the fluid–air interface rather than on the mica surface (see also Harris, 1997). They suggested that early protein-protein associations involved in 2D crystal nucleation could be captured on the electron micrographs. They also deduced from their observations that uncontrolled and unavoidable minor fluctuations happening during the preparation of the grids enabled the crystallisation event to be seen under slightly different conditions on different micrographs. These authors have postulated that the slow drying of a protein solution containing ammonium molybdate and PEG on the surface of the mica substrate first generates close-packed randomly associated molecules (or imperfectly organised precrystalline/paracrystalline arrays) and that the latter then transform in ordered arrays with crystallographic properties. They also noted that under appropriate conditions, few single

444

B. Lorber et al. / Micron 39 (2008) 431–446

Fig. 8. Cryo-negative staining images of 2D virus crystallisation in the presence of 1% (m/v) PEG-1000, in thin films spread across the holes of a perforated carbon film. (Left) View of 2D arrays of BMV in a hole of the perforated carbon film. The arrow points toward a second layer of virions (la) on top of the first one despite sufficient free space around the crystals. (Center) View of four 2D arrays of TBSV (marked A1 to A4) that had nucleated in close proximity and are on the point of merging together. In addition to the imperfections described in Fig. 5, the arrays A1 and A2 are parallel, A2 and A3 converge at right angle and the misalignment of array A4 (at an angle of 308 with respect to A2) is obvious. Important rearrangements would have been necessary to align perfectly the four lattices. Frequently a small number of nuclei resulted in large arrays when enough free space was available around the growing crystal. (Right) A 2D meshing made of single and double rows of TYMV virions in a window of the carbon film. Nucleation has occurred spontaneously at several (almost equidistant) places and it was arrested at a time where not all available positions were occupied. The growth rate in 1D (i.e. along the rows) was superior to that in 2D (i.e. in the plane) and in 3D (i.e. in thickness) and to the incorporation rate of virions in vacant sites. For all three panels, experimental conditions were as described in Figs. 4b, 5c and 6b.

protein molecules accompanied the 2D crystals. Another case of conversion from a paracrystal to an ordered crystal by lateral ordering of chains of molecules was reported for the protein annexin p68 crystallising in PEG-4000 (Gabriel et al., 1991). Similar to the negative staining-carbon film technique used by Harris and Holzenburg (1995), array formation in our virus suspensions that are thinly spread over the holes of a carbon film occurs in at the fluid–air interface (see Fig. 8). Thus, the general mechanism that we propose for the formation of viral 2D arrays involves trapping at the air–water interface during the evaporation of the solvent on the non-blotted side of the holes of the perforated carbon film. This is essentially the same as the selective location of viruses at the air–fluid interface, as demonstrated by Johnson and Gregory, 1993, Cyrklaff et al., 1994). It is noteworthy that high virus concentrations (e.g. 2– 4 mg/ml) were needed to produce larger ordered arrays in the presence of PEG (e.g. 0.1–1 mg/ml, cf Adrian et al. (1998). This indicates that the higher viscosity and the water-retention property of PEG probably slowed the interface evaporation and contributed to the formation of superior 2D viral crystals. When the cryo-negative staining ammonium molybdate/ PEG system applied here, was transferred to air-dry negative staining across holes in the presence of ammonium molybdate and trehalose, Harris and Scheffler (2002) found that for both proteins and viruses, inclusion of PEG could also create 2D crystals. In addition, under these conditions the creation of higher-order molecular assemblies can occur (Meissner et al., 2007). Air-drying of specimens prepared across holey carbon films with 5% (m/v) ammonium molybdate and 1% (m/v) trehalose as the negative stain also generates 2D crystals in the presence of PEG, but the achievable electron image resolution

is inferior to that obtainable from frozen-hydrated negatively stained specimens (Harris, 1997, 2007; Adrian et al., 1998; Harris and Adrian, 1999; Harris and Scheffler, 2002). 5. Conclusion Three small quasi-spherical RNA viruses have served as models to investigate the formation of 2D arrays. The 60 symmetry elements of the viral capsids should facilitate selforganisation as 2D crystals, by providing more possibilities to establish packing contacts than might be expected from asymmetric particles. Aggregation phenomena, monitored by spectroscopy correlate with the observation that PEG favours the formation of larger virion arrays. Small and geometrical ordered arrays of virions observed on electron micrographs share some similarity in size and structure with the planar structures identified by atomic force microscopy as crystal nuclei, in the case of the quasi-spherical protein apoferritin (Yau and Vekilov, 2000). A succession of association steps that take into account the stages observed on negatively stained cryo-electron micrographs is proposed for the process of 2D virus crystal formation from nuclei. This is discussed above within the broad context of literature data dealing with protein crystallisation. Finally, it must be stressed no simple negative staining procedure can be provided that is immediately applicable for the production of 2D crystals from all biological particles. As with 3D crystallisation, the experimenter must define for each virus or protein sample the optimal conditions that yield the best 2D crystallisation, i.e. with respect to virus or protein concentration, purity and stability, ammonium molybdate

B. Lorber et al. / Micron 39 (2008) 431–446

concentration, PEG mass, grade and concentration, temperature, solution pH, environmental humidity, and the time and rate of evaporation from the fluid–air interface. Acknowledgements We are grateful to Prof. Jacques Dubochet (Laboratoire d’Analyse Ultrastructurale, Universite´ de Lausanne, Switzerland) for the use of the cryo-electron microscopy facility and for his in-depth comments on the manuscript. The expert advice of J. Mutterer with ImageJ was greatly appreciated. We have also appreciated discussions with Dr. Pierre Pfeiffer (IBMP, Strasbourg) and Dr. Herve´ Lecoq (INRA, Avignon, France). We acknowledge a grant from Universite´ Louis Pasteur, Strasbourg and support from the French Ministry for Research through grants ACI-BCMS 042358 and PIR-MME. References Adams, M.J., Antoniw, J.F., 2006. DPVweb: a comprehensive database of plant and fungal virus genes and genomes. Nucleic Acids Res. 34, D382–D385 (Database issue). Adrian, M., Dubochet, J., Fuller, S., Harris, J.R., 1998. Cryo-negative staining. Micron 29, 145–160. Arakali, S.V., Luft, J.R., DeTitta, G.T., 1995. Non-ideality of aqueous solutions of polyethylene glycol: consequences for its use as a macromolecular crystallizing agent in vapor-diffusion experiments. Acta Crystallogr. D 51, 772–779. Arakawa, T., Timasheff, S.N., 1984. Mechanism of protein salting-in and salting out by divalent cation salts: balance between hydration and salt binding. Biochemistry 23, 5912–5923. Arakawa, T., Timasheff, S.N., 1985. Mechanism of poly(ethylene glycol) interaction with proteins. Biochemistry 24, 6756–6762. Astier, S., Albouy, J., Maury, Y., Lecoq, H., 2001. Principes de Virologie Ve´ge´tale. INRA Editions, Paris, p. 444. Atha, D.H., Ingham, K.C., 1981. Mechanism of precipitation of proteins by polyethylene glycols. J. Biol. Chem. 256, 12108–12117. Basu, S.K., Govardhan, C.P., Jung, C.W., Margolin, A.L., 2004. Protein crystals for the delivery of biopharmaceuticals. Expert Opin. Biol. Ther. 4, 301–317. Bawden, F.C., Pirie, N.W., 1938. Crystalline preparations of tomato bushy stunt virus. Br. J. Exp. Path. 19, 251–263. Bonnete´, F., Vivare`s, D., 2002. Interest of the normalized second virial coefficient and interaction potentials for crystallizing large macromolecules. Acta Crystallogr. D58, 1571–1575. Bostro¨m, M., Tavares, F.W., Finet, S., Skouri-Panet, F., Tardieu, A., Ninham, B.W., 2005. Why forces between proteins follow different Hofmeister series for pH above and below pI. Biophys. Chem. 117, 115–122. Brenner, S., Horne, R.W., 1959. A negative staining method for high resolution electron microscopy of viruses. Biochim. Biophys. Acta 34, 103–110. Brisson, A., Olofsson, A., Ringler, P., Schmutz, M., Stoylova, S., 1994. Twodimensional crystallization of proteins on planar lipid films and structure determination by electron crystallography. Cacace, M.G., Landau, E.M., Ramsden, J.J., 1997. The Hofmeister series: salt and solvetn effects on interfacial phenomena. Q. Rev. Biophys. 30, 241–277. Canady, M.A., Larson, S.B., Day, J., McPherson, A., 1996. Crystal structure of turnip yellow mosaic virus. Nat. Struct. Biol. 3, 771–781. Casselyn, M., Perez, J., Tardieu, A., Vachette, P., Witz, J., Delacroix, H., 2001. Spherical plant viruses: interactions in solution, phase diagrams and crystallization of brome mosaic virus. Acta Crystallogr. D57, 1799–1812. Chernov, A.A., 1984. Modern Crystallography. III. Crystal Growth. Springer Verlag. Cornuet, P., 1959. Plate #1 in Maladies a` Virus des Plantes Cultive´es et Me´thodes de Lutte. INRA, Paris.

445

Cyrklaff, M., Roos, N., Gross, H., Dubochet, J., 1994. Particle-surface interaction in thin vitrified films for cryo-electron microscopy. J. Microsc. 175, 135–142. Denkov, N.D., Velev, O.D., Kralchevsky, P.A., Ivanov, I.B., Yoshimura, H., Nagayama, K., 1993. Two-dimensional crystallization. Nature 361, 26. Ducruix, A., Giege´, R., 1999. Crystallisation of Nucleic Acids and Proteins. IRL Press, Oxford. Fuller, S.D., Butcher, S.J., Cheng, R.H., Baker, T.S., 1996. Three-dimensional reconstruction of icosahedral particles—the uncommon line. J. Struct. Biol. 116, 48–55. Gabriel, B.L., Taylor, K., Creutz, C.E., Kretsinger, R.H., 1991. Microcrystals of the annexin p68: paracrystal to crystal transition and molecular packing as determined by electron microscopy and image reconstruction. J. Struct. Biol. 107, 29–37. Hammond, C., 1997. The Basics of Crystallography and Diffraction; IUCr Texts on Crystallography. 3. Int. Union of Crystallography. Oxford Science Publications, Oxford University Press. Harris, J.R., 1991. The negative staining-carbon film procedure: technical considerations and survey of macromolecular applications. Micron Microsc. Acta 22, 341–359. Harris, J.R., 1997. Negative Staining and Cryoelectron Microscopy: The Thin Film Techniques, Royal Microscopical Society Microscopy Handbook No. 35. BIOS Scientifc Publishers, Oxford, UK. Harris, J.R., 2007. Negative staining of thinly spread biological samples. In: Kuo, J. (Ed.), Methods in Molecular Biology, Electron Microscopy: Methods and Protocols, second ed., 369. Humana Press Inc., pp. 107–142. Harris, J.R., Adrian, M., 1999. Preparation of thin-film frozen-hydrated/vitrified biological specimens for cryoelectron microscopy. Methods Mol. Biol. 117, 31–48. Harris, J.R., Holzenburg, A., 1989. Transmission electron microscopic studies on the quaternary structure of human erythrocyte catalase. Micron and Microsc. Acta 20, 223–238. Harris, J.R., Holzenburg, A., 1995. Human erythrocyte catalase: 2-D crystal nucleation and production of multiple forms. J. Struct. Biol. 115, 102–112. Harris, J.R., Scheffler, D., 2002. Routine preparation of air-dried negatively stained and unstained specimens on holey carbon support films: a review of applications. Micron 33, 461–480. Harris, J.R., Pfeifer, G., Pu¨hler, G., Baumeister, W., 1992. Production of 3-D Crystals from Thermoplasma acidophilum Multicatalytic Proteinase/Proteasome by the Negative Staining Carbon Film Technique. Electron Microscopy, vol. 1. EUREM 92, Granada, Spain, pp. 383–384. Harris, J.R., Engelhardt, H., Volker, S., Holzenburg, A., 1993. Electron microscopy of human erythrocyte catalase: new two-dimensional crystal forms. J. Struct. Biol. 111, 22–33. Hebert, T.T., 1963. Precipitation of plant viruses by polyethylene glycol. Phytopathology 53, 362. Holzenburg, A., 1988. Preparation of two-dimensional arrays of soluble proteins as demonstrated by bacterial D-ribulose-1,5-bisphosphate carboxylase/oxygenase. Methods Microbiol. 20, 342–356. Horne, R.W., Pasqualli-Ronchetti, I., 1974. A negative staining-carbon film technique for studying viruses in the electron microscope. I. Preparative procedures for examining icosahedral and filamentous viruses. J. Ultrastruct. Res. 47, 361–383. Horne, R.W., Pasqualli-Ronchetti, I., Hobart, J.M., 1975a. A negative stainingcarbon film technique for studying viruses in the electron microscope. II. Application to adenovirus Type 5. J. Ultrastruct. Res. 51, 233–252. Horne, R.W., Hobart, J.M., Pasqualli-Ronchetti, I., 1975b. A negative staining-carbon film technique for studying viruses in the elctron microscope. III. The formation of two-dimensional and three-dimensional crystalline arrays of cowpea chlorotic mottle virus. J. Ultrastruct. Res. 53, 319– 330. Johnson, R.P.C., Gregory, D.W., 1993. Viruses accumulate spontaneously near droplet surface. A method to concentrate viruses for electron microscopy. J. Microscopy 171, 125–136. Kunz, W., Henle, J., Ninham, B.W., 2004. Zur Lehre von der Wirkung der Salze (about the science of the effect of salts): Franz Hofmeister’s historical papers. Curr. Opin. Coll. Interf. Sci. 9, 19–37.

446

B. Lorber et al. / Micron 39 (2008) 431–446

Kuznetsov, Y.G., Makino, D.L., Malkin, A., McPherson, A., 2005. The incorporation of large impurities into virus crystals. Acta Crystallogr. D61, 720– 723. Leberman, R., 1966. Virology 30, 341–347 The isolation of plant viruses by means of ‘‘simple’’ coarcervates. Lee, K.K., Johnson, J.E., 2003. Complementary approaches to structure determination of icosahedral viruses. Curr. Opin. Struct. Biol. 13, 558–569. Lee, J.C., Lee, L.L.Y., 1981. Preferential solvent interaction between proteins and polyethylene glycol. J. Biol. Chem. 256, 625–631. Lucas, R.W., Kuznetsov, Y.G., Larson, S.B., McPherson, A., 2001. Crystallization of brome mosaic virus and T = 1 brome mosaic virus particles following a structural transition. Virology 286, 290–303. Lucas, R.W., Larson, S.B., McPherson, A., 2002. The crystallographic structure of brome mosaic virus. J. Mol. Biol. 317, 95–108. Maramorosch, K., 1977. The Atlas of Insect and Plant Viruses, Ultrastructure in Biological Sciences Series, vol. 8. Academic Press, New York. Matthews, R.E.F., 1960. Properties of nucleoprotein fractions isolated from turnip yellow mosaic virus preparations. Virology 12, 521–539. McPherson, A., 1976. Crystallization of proteins from polyethylene glycol. J. Biol. Chem. 251, 6300–6303. McPherson, A., 1999. Crystallization of Biological Macromolecules. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY. Meissner, U., Schro¨der, E., Scheffler, D., Martin, A.G., Harris, J.R., 2007. Formation, TEM study and 3D reconstruction of the human erythrocyte peroxidredoxin-2 dodecahedral higher-order assembly. Micron 38, 29–39. Ng, J.D., Lorber, B., Witz, J., The´obald-Dietrich, A., Kern, D., Giege´, R., 1996. The crystallization of biological macromolecules from precipitates: evidence for Ostwald ripening. J. Crystal Growth 168, 50–62. Nicholas, J.F., 1965. An Atlas of Models of Crystal Surfaces. Gordon and Breach, New York. Paliwal, Y.C., 1970. Electron microscopy of bromegrass mosaic virus in infected leaves. J. Ultrastruct. Res. 30, 491–502. Pum, D., Weinhandl, M., Ho¨do¨l, C., Sleytr, U.B., 1993. Large-scale recrystallization of the S-layer of Bacillus coagulans E38-66 at the air/water interface and on lipid film. J. Bacteriol. 175, 2762–2766. Rasband, W.S., 1997–2006. Image J. U.S. National Institutes of Health, Bethesda, Maryland. , http://rsb.info.nih.gov/ij. Ray Jr., W.J., 1992. Effect of polyethylene glycol-400 at low concentration on long-term growth of muscle phosphoglucomutase crsytals from concentrated salt solutions. Proteins: Struct., Funct. Genet. 14, 300–308.

Rubio-Huertos, M., Vela, A., Lopez-Abella, D., 1967. Crystalline arrays of spherical particles in turnip yellow mosaic virus-infected cells. Virology 32, 438–444. Russo, M., Martelli, G.P., 1972. Ultrastructural observations on tomato bushy stunt virus in plant cells. Virology 49, 122–129. Smith, K.M., 1956. Virology 2, 706-709. Apparent crystallization of tomato bushy stunt virus in cells of infected plants. Schmitz, K.S., 1990. An Introduction to Dynamic Light Scattering by Macromolecules. Academic Press, Boston. Tardieu, A., Bonnete´, F., Finet, S., Vivare`s, D., 2002. Understanding salt or PEG induced attractive interactions to crystallize biological macromolecules. Acta Cryst. D58, 1549–1553. Tihova, M., Dryden, K.A., Le, T.L., Harvey, S.C., Johnson, J.E., Yeager, M., Schneeman, A., 2004. Nodavirus coat protein imposes dodecahedral RNA structure independent of nucleotide sequence and length. J. Virol. 78, 2897– 2905. van Blaaderen, A., Rue, R., Wiltzius, P., 1997. Template-directed colloidal crystallization. Nature 385, 321–323. Vergara, A., Paduano, L., D’Errico, G., Sartorio, R., 1999. Network formation in polyethyleneglycol solutions. An intradiffusion study. Phys. Chem. Chem. Phys. 1, 4875–4879. Vergara, A., Paduano, L., Sartorio, R., 2002. Mechanism of protein–poly(ethylene glycol) interaction from a diffusive point of view. Macromolecules 35, 1389–1398. Vivare`s, D., Bonnete´, F., 2002. X-ray scattering studies of Aspergillus flavus urate oxidase: towards a better understanding of PEG effects on the crystallization of large proteins. Acta Cryst. D58, 472–479. Vonck, J., van Bruggen, E.F.J., 1990. Electron microscopy analysis of twodimensional crystals and single molecules of alcohol oxidase from Hansenula polymorpha. Walz, T., Grigorieff, N., 1998. Electron microscopy of two-dimensional crystals of membrane proteins. J. Struct. Biol. 121, 142–161. Wells, B., Horne, R.W., Shaw, P.J., 1981. The formation of two-dimensional arrays of plant viruses in the presence of polyethylene glycol. Micron 12, 37–45. Wickman, H.H., Korley, J.N., 1998. Colloid crystal self-organization and dynamics at the air/water interface. Nature 393, 445–447. Wilson, K.A., Glaeser, R.M., 1974. Electron diffraction of frozen, hydrated protein crystals. Science 186, 1036–1037. Yau, S.T., Vekilov, P.G., 2000. Quasi-planar nucleus structure in apoferritin crystallization. Nature 406, 494–497.