HIGH-RESOLUTION ELECTRON MICROSCOPY OF ...

jf. Cell Sci. 51, 295-321 (1981) Printed in Great Britain © Company of Biologists Limited 1981

295

HIGH-RESOLUTION ELECTRON MICROSCOPY OF GLYCOPROTEINS: THE CRYSTALLINE CELL WALL OF LOBOMONAS K. ROBERTS, P. J. SHAW AND G. J. HILLS John limes Institute, Colney Lane, Norwich NR4 yUH, U.K.

SUMMARY Lobomonas piriformis is a member of an order of green algae (Volvocales) that have crystalline glycoprotein cell walls. As part of a program of investigation of these glycoproteins and their architecture we have studied the cell wall of Lobomonas by a variety of chemical, electronmicroscopical and image-analysis techniques. Lobomonas and Vitreochlamys incisa show a very similar structure in their cell walls and represent 1 of the 4 classes into which all the structures of the walls of these algae that we have so far examined fall. The 2 classes that we have previously studied in detail, represented by Chlamydomonas reinhardii and Chlorogonium elongatum, have a crystalline component of the wall that is a more or less smooth continuous surface overlying an amorphous inner wall layer. Although Lobomonas also has this 2-layer structure, the crystalline layer consists of distinct plate3, each of which is built around a single, very coherent crystal lattice. The polar nature of the architecture of the cell wall is shown by sectioning and by examination of the cell-wall surface by metal-shadowing of carbon replicas, both of intact cells and of isolated cell-wall plates. There are great similarities in chemical composition between the glycoproteins of the cell wall of C. reinhardii and those of Lobomonas. Both have a large content of hydroxyproline in their amino acid composition and a sugar/hydroxyproline ratio of about 60, and both contain sugar sulphates. Lobomonas however has a large gluco3e content, whereas Chlamydomonas has almost none. Electron micrographs of walls stained with methylamine tungstate and shadowed specimens show that the Lobomonas crystal structure is entirely different from that of C. reinhardii, and that there is a distinctly different structure in the centre of the plates from that at their edges, although the transition between the 2 areas occurs with no distortion of the crystal lattice. Computer image analysis has been used to calculate reconstructed images of the 2 areas, and by using minimal-dose techniques has yielded 2-dimensional maps of the negatively stained structure at a resolution of 1 -8 nm. The 2-sided plane group of both areas of the crystal is P2, and the centre area contains 2 distinct structural units, both centred on dyad axes, together with other more complex features. In the edge structure, one of the structural units appears unchanged, but the other unit has a considerably different appearance. The most likely interpretation of this is as a conformational or positional change in one of the subunits. However, because the underlying lattice is so accurately maintained across this transition, it seems probable that the basic structural arrangement that defines the lattice is common to the 2 areas. Some of the computational and mathematical techniques used in the image analysis have not been previously published and are described in detail and compared with published techniques in an Appendix.

INTRODUCTION

The green algae of the order Volvocales have a cell wall characterized by a crystalline glycoprotein lattice. The glycoproteins involved show remarkable similarity to the hydroxyproline-containing glycoproteins characteristic of higher plant cell walls (Lamport, 1980) and provide a unique model system for studying their structure in

296

K. Roberts, P. J. Shaw and G. J. Hills

detail. The cell-wall architecture is species-dependent and at present 4 distinct types of cell wall have been described on the basis of their crystal structures (Roberts, 1974). (In fact, 5 cell-wall groups were originally described but one of these, Chlamydomonas angulosa, has now been re-examined and found to be the same as class I, i.e. Chlamydomonas reinhardii. This means that there are now only 4 classes represented by the following species: class I, C. reinhardii; class II, Chlorogonium elongatum; class III, Chlamydomonas asymmetrica and class IV, Lobomonas piriformis.) It is proposed that these 4 basic cell-wall types have not only taxonomic but also evolutionary significance. For example, all the multicellular species examined (Volvox, Pandorina, Eudorina) have a cell-wall type (class I) identical to the single-celled C. reinhardii (Roberts, 1974; Roberts, Gurney-Smith & Hills, 1972; Hills, Gurney-Smith & Roberts, 1973), suggesting a common ancestral origin for these species. Most algae examined have a class II cell-wall structure, which is postulated as being the original ancestral type of wall from which the others have evolved (Roberts & Hills, 1976). The other 2 groups of algae (III and IV), which contain few representatives, have so far received only a brief descriptive treatment (Roberts, 1974). This paper provides a more rigorous structural and biochemical analysis of one of these remaining cell-wall classes (IV), using as the type organism L. piriformis. A similar cell wall is found in Vitreochlamys incisa (formerly Chlamydomonas incisa) and a preliminary description of both has been published (Roberts, 1974). The Lobomonas cell wall consists of relatively rigid crystalline plates, with a high degree of long-range order, covering the outer surface of a thin amorphous inner cell-wall layer, and we concentrate here on a description of the glycoprotein arrangement within a 2-dimensional projection of the crystalline layer to a resolution of i-8 nm. This has been determined by a combination of low-dose electron microscopy and digital image processing. The results are discussed in terms of the suggestion that the glycoproteins found in these algal cell walls are the ancient ancestors of the hydroxyproline-containing cell-wall glycoproteins and arabinogalactan proteins of higher plants (Lamport, 1980).

MATERIALS AND METHODS Cultures of L. piriformis (Piingsheim) (strain 45/i) and V. incisa (Pringsheim) (strain 11/10, were obtained from the Culture Centre of Algae and Protozoa, 36 Storeys Way, Cambridge) U.K. Cells were grown in sterile conditions in liquid culture without aeration, using a Trisbuffered, yeast and peptone-supplemented medium (Catt, Hills & Roberts, 1978) under continuous illumination (80 /iEinsteins mol"1 s"1 in the waveband of 400-700 nm). Cell walls were harvested from the medium (as cast-off mother cell walls) by differential centrifugation as described by Hills, Phillips, Gay & Roberts (1975). Negatively stained specimens were made on 400-mesh copper grids coated with thin evaporated carbon films stripped from freshly cleaved mica. Stains used were 5 % ammonium molybdate, 5 % sodium tungstate (pH 6-8), or 2-5 % methylamine tungstate (Faberg6 & Oliver, 1974). Rotary-shadowed specimens were made using platinum/palladium at an angle of 50. For thin sections, walls and cells were fixed either by the method of Franke, Krien & Malcolm-Brown (1969) or using a modified tannic acid procedure (Catt et al. 1978). Subsequent processing is described elsewhere (Roberts et al. 1972). Specimens were examined in either an AEI EM6B, a JEOL JEM 100B or a Siemens 102 electron microscope, all calibrated using catalase. Optical diffraction was carried out as described previously (Roberts et al. 1972). For routine examination normal doses of 5000-10000 electrons per square nm (e/nm1) were used. For high-resolution

Crystalline cell wall of Lobomonas

297 1

negative-stained pictures, minimal-dose conditions (500-1000 e/nm ) were used (see Hills, 1981). Images were selected for further processing by examining their optical diffraction patterns. The selected images had sharp unsplit spots extending to a resolution of about z nm (up to 9 orders in b*), and focussing was such that the first zero of the contrast transfer function (Erickson, 1973) occurred at a higher resolution than the highest resolution spots. The effect of the contrast transfer function could therefore be neglected. Computer processing was carried out on a DEC PDP 11/60 mini-computer using one 28 Mbyte disc (RK07), running the RSX 11M operating system. Programs were either written specifically for this machine or modified to run on it. Most programs were written to take their input parameters and file specifications from the command line used to involve them. This was so as to take advantage of the indirect command-file feature of the operating system; files containing sequences of image-processing operations could be set up and run as a single job or, alternatively, the programs could be invoked interactively as image-processing 'commands'. This approach offers a convenient and easily implemented compromise between imageprocessing systems consisting of' stand-alone' programs (Unwin & Henderson, 1975; Crowther & Klug, 1975) and elaborate fully integrated systems with specialized console monitor programs such as SPIDER (Frank & Shimkin, 1978) and SEMPER (Saxton, Pitt & Homer, 1979). All images (scanned, intermediate or reconstructed) were stored in a common format, and programs were written so as to be able to operate on any image as far as possible. Some programs, such as scanning, trimming and interpolation, operate only on real images; others, such as spot lattice refinement, and amplitude and phase calculation, only on transform images. Other programs were written so as to operate on either type of image: grey-scale output; contour output; fast Fourier transformation (FFT). A final set of programs, for example, origin refinement, symmetry averaging and film-to-film scaling, operates on files comprising lists of indices, amplitudes and phases. Images were scanned and stored on magnetic tape using, initially, a Scandig rotating-drum densitometer (courtesy of W. O. Saxton) or, later, directly read onto disk using a flat-bed computer-controlled densitometer constructed in this laboratory (Shaw, Garner & Parker, 1981). The scanning sample interval was matched as closely as possible to one third of the minimum spatial period required (i.e. about 0 7 nm, since the diffraction maxima extended to 2 0 nm). Generally 512 by 512 points were scanned, although this was occasionally less when only a smaller area of the image was suitable for analysis. A mixed radix F F T program was used: an implementation for this computer of the routines written by Ten Eyck (1973). Floating of the images (De Rosier & Moore, 1970) was not generally necessary since the image filled the whole of the Fourier transform interval and so there were no large edge effects. Occasionally there were slight optical density gradients across the image, which gave rise to small origin streaks in the transform. In these cases a floating procedure was used in which a plane was fitted to the data by a least-squares calculation and then subtracted so as to give a mean optical density of zero across the image. In practice this made very little difference to the final results. The reciprocal lattices were determined and refined, and values for the amplitudes and phases of the peaks were generally determined by a profile-fitting procedure. This differs from other published methods and is compared with them in the Appendix. The best data sets were averaged together for the final maps of the negative-stained structure, and the 2-fold symmetry was imposed exactly by setting the phases to the nearest theoretical phase for a 2-fold symmetric structure (o° or 180°). For scanning electron microscopy, cells were fixed, dehydrated, carbon dioxide critical-point dried from amyl acetate (Anderson, 1951), coated with carbon and gold, and viewed in a Phillips type 501 scanning electron microscope. Light microscopy was carried out on a Zeiss Photomicroscope II. Polyacrylamide gel electrophoresis on 5 % gels has been described (Catt, Hills & Roberts, 1976) and carbohydrate assays were by the phenol/sulphuric acid method (Dubois et al. 1956). Hydroxyproline and other assays were as previously described (Roberts, 1979). Sugars were estimated by gas chromatography of their trimethylsilylated derivatives (Bhatti, Chambers & Clamp, 197°) on a column of 3 % SE30 on GasChromQ. Amino acids were analysed as their isobutyl iV(0)-heptafluorobutyl esters on the same column (MacKenzie & Tenaschuk, 1975). Sulphate content was determined using the barium chloride/gelatin technique (Dodgson & Price, 1962).

298

K. Roberts, P. J. Shaw and G. J. Hills

r-r-1; , ^ ; ,-

5

Crystalline cell wall of Lobomonas

299

RESULTS AND DISCUSSION General morphology

L. piriformis is an approximately spherical biflagellate green alga of the family Chalmydomonacae and under the culture conditions used ranges in size from 10-20 fim in diameter (Fig. 1). When the cells divide, 4 daughter protoplasts are usually formed within the mother cell wall, which splits and releases the progeny cells. The mother cell walls remain free in the growth medium and under the light microscope reveal a faceted surface (Fig. 2). Both in scanning electron micrographs (Fig. 3) and in thin sections it can be seen that these facets reflect the fact that the wall is made up from a series of abutting plates (Fig. 6). By differential centrifugation the cast-off cell walls can be separated from the cells and provide the highly pure cell-wall preparations (Fig. 4) that were used for all the work described here. Thin sections of cell walls fixed in the presence of tannic acid reveal further details of cell-wall construction (Figs. 8-11). Tannic acid was used as a cofixative to enhance the strong subsequent positive staining of the negatively charged carbohydrate-rich wall. The plates, of which there may by 50-400 to a cell, vary in size between 2 and 5 /o after aveiaging (°)

Weighted

Unweighted

13-6 0-115 90 io-o

25-6 0133 13 2 18-0

(58 independent spots were included in the final reconstruction) >

r>

/

• u^

j\

• Rj (weighted)

=

R^g (unweighted) = ^ where ^»i is the measured phase of spot (h, k) shifted to the best phase origin, and 2-fold symmetric phase (0° or 1800) closest to this value. fRP

(weighted)

0OJU

is the

=

RP (unweighted) = where F u is the average amplitude for spot {h, k) taken over all the images, and F(At is the amplitude for spot (h, k) in the image i. X R4 (weighted)

= XF>t|^-^m|

R4 (unweighted) = E — 10At - m\, where (pu, is the average phase for spot (h, k) taken over all the images, and 0