J. Cell Sci. 68, 271-284 (1984)
271
Printed in Great Britain © The Company of Biologists Limited 1984
THE THREE-DIMENSIONAL STRUCTURE OF THE
CELL WALL GLYCOPROTEIN OF CHLOROGONIUM
ELONGATUM
P. J. SHAW AND G. J. HILLS
John Innes Institute, Colney Lane, Norwich NR4 7UH, U.K.
SUMMARY
The green alga Chlorogonium elongatum, a member of the Volvocales, possesses a crystalline cell
wall composed of hydroxyproline-rich glycoprotein similar to the primary cell wall glycoproteins of
higher plants. Electron microscopy and computer image processing have been used to determine the
crystal structure of the Chlorogonium cell wall in three dimensions to a resolution of 2-0 nm. The
structure is composed of heterologous dimers. Each subunit of the dimer comprises a long, thin
spacer domain and a large globular domain, which is the site of the intra- and inter-dimer interactions. There are also sites of intersubunit interactions at the opposite ends of the rod domains. We
suggest that the rods are composed predominantly of glycosylated polyproline helix, as has been
suggested for higher plant cell wall glycoproteins and has been shown for the cell wall glycoprotein
of Chlamydomonas reinhardtii, which is closely related to Chlorogonium.
INTRODUCTION
Glycoproteins have been identified as important constituents of the extra-cellular
matrix of organisms ranging from the Archaebacteria to higher plants and animals.
Within the great range of glycoproteins that have been characterized, a distinction
may be made between those glycoproteins that contain hydroxyproline and those that
do not. It has been suggested (Lamport, 1977) that the acquisition of the enzymic
apparatus necessary to use molecular oxygen to hydroxylate proline residues
represents a distinct evolutionary step, and that this is reflected in the function and
distribution of the resulting glycoproteins. In the plant kingdom hydroxyproline
(Hyp)-containing glycoproteins have been shown to be important constituents of the
primary cell wall of higher plants, in /3-lectins (Allen, Desai, Neuberger & Geeth,
1978), and as the factors mediating sexual agglutination in certain algae
{Chlamydomonas reinhardtii (Cooper et al. 1983) and Chlamydomonas eugametos
(Pijst, Zilver, Musgrave & Van den Ende, 1983)).
The latter organisms are members of a fairly large group of motile eucaryotic algae,
the Volvocales, which possess cell walls" composed solely of glycoproteins. These
glycoproteins contain a high proportion of hydroxyproline and, at least in the case of
C. reinhardtii, there is good evidence that the major cell wall glycoprotein is organized
into Hyp-rich domains, showing a polyproline II helix structure, and domains containing little or no hydroxyproline (Homer & Roberts, 1979). It has been suggested
that the polyproline II helix, glycosylated by short oligosaccharides, is a common
structural feature of Hyp-containing plant glycoproteins, and evidence for the
272.
P. J. Shaw and G. J. Hills
presence of this structure has also been presented for potato lectin (Allen et al. 1978)
and the primary cell wall glycoprotein of higher plants (Lamport, 1977).
The cell wall glycoproteins of the Volvocalean algae are of particular interest to us
because, besides being Hyp-rich glycoproteins related to those found in higher plant
cell walls, they occur in vivo as crystalline lattices, which are thus amenable to detailed
structural study. Most of the glycoprotein of these cell walls forms the crystalline
outer layer and the remainder forms an amorphous and much more diffuse inner wall
layer. The inner layer is formed first during cell division and probably forms a template on which the outer layer crystallizes (Roberts, 1974).
Many species within the Volvocales have now been examined by high-resolution
electron microscopy, and all possess one of only four types of crystal structures in their
cell walls (Roberts, Hills & Shaw, 1982). The use of the cell wall structure as a
phylogenetic marker for the classification of the algae has been suggested (Roberts,
1974). We have previously determined the three-dimensional structure of a representative of one of these classes by electron microscopic analysis (Lobomonas piriformis;
Shaw & Hills, 1982) (class IV of Roberts et al. (1982)). We showed that the structure
is composed of two sets of dimers, each of which consists of two rod-shaped molecules
aligned in the plane of the crystal.
The subject of the present investigation, Chlorvgonium elongatum, displays by far
the most common cell wall structure found in the Volvocales - denoted as type II by
Roberts. A preliminary investigation of the wall has been reported by Roberts & Hills
(1976). As the structure was also apparently the simplest among the wall types, it was
suggested that it was closest to that of a hypothetical ancestral type, from which the
other wall types had diverged. Table 1 shows all the species currently categorized as
type II.
Table 1. Species of algae currently known to have type II cell walls
Species
Brachiomonas submarina
Carteria crucifera
Carteria eugametos
Carteria incisa
Chlamydomonas applanata
Chlamydomonas chlamydogama
Chlamydomonas dorsoventralis
Chlamydomonas dysosmos
Chlamydomonas eugametos
Chlamydomonas
fimbriata
Chlamydomonas moetvusi
Chlamydomonas pulsatilla
Chlamydomonas reginae
Chlamydomonas rosae
Chlamydomonas sphaerella
Chlorvgonium elongatum
Chlorogonium euchlorum
Haematococcus capensis
Polytoma uvella
Collection no.
7/I a
8/7a
8/3
8/4
11/2
11 /48a
11/4
11/31
1 l/Sc
11 /69
1 l/16f
11 /44
11 /78
11 /66
11 /27
12/1
12/3
34/4b
62/2a
Three-dimensional structure of Chlorogonium cell wall
MATERIALS
AND
273
METHODS
Cultures of Ch. elongatum (Dangeard) (Strain 12/1) were obtained from the Culture Centre of
Algae and Protozoa, 36 Storey's Way, Cambridge, U.K. and were grown in sterile conditions in
liquid culture without aeration, using a Tris-buffered medium, supplemented with yeast peptone
(Catt, Hills & Roberts, 1978) under continuous illumination (80 ^Einstein mol~2 s in the waveband 400-700 nm). Cells were disrupted with glass ballotini in a Mickle shaker, layered onto a 5 %
to 120% (w/v) glucose gradient and centrifuged in a swinging-bucket rotor for 5 min at 1500 #.
The fastest sedimenting band consisted of almost pure cell walls.
Negatively stained specimens were made on 400 mesh copper grids coated with thin carbon films
made by evaporation onto a freshly cleaved mica surface and stripped off by flotation onto water.
The negative stains used were 2-5% (w/v) aqueous methylamine tungstate (MET) (Faberg6 &
Oliver, 1974) or methylamine tungstate/tannic acid made by the addition of 0-2mg/ml tannic acid
(Mallinckrodt, St Louis, U.S.A.) to 2-5% (w/v) methylamine tungstate solution (METT). The
latter solution must be made by the addition of solid tannic acid to the MET solution to avoid
precipitation, and the best results were obtained using the resulting solution within one day.
Specimens were examined on a JEOL JEM 1200 EX electron microscope fitted with a eucentric
single-axis tilt specimen holder or a tilt/rotation specimen holder. Micrographs were generally recorded at 80 kV, X 30 000, using Kodak SO 163 film. Since only a small proportion of specimen areas give
diffraction spots to high (2 nm) resolution, the following strategy was used for the collection of tilted
data. Micrographs were recorded from 10-20 areas of a suitable grid and the coordinates of each area
were noted from the microscope's digital display. The micrographs were developed and examined in
an optical diffractometer. The best areas, which were rare, contained spotsout to theA = 3row. Itwas
then possible to return to these areas and collect tilt series, generally about a single axis from +60 ° and
—60°, but occasionally about two perpendicular axes. Additional zero tilt images recorded after the
series showed that no significant degradation of the image occurred during this procedure.
Areas of the micrographs were masked off for scanning using optical diffractometry. Areas were
used only if the defocus level was such that the first zero of the contrast transfer function (Erickson,
1973) occurred outside the highest-resolution spots. If possible, exactly the same area of each
member of a tilt series was used.
Areas of the micrographs were digitized using a microdensitometer constructed in this laboratory
(Shaw, Garner & Parker, 1981) linked to a DEC PDP11/60 minicomputer, which was used for all
subsequent data processing. The scanning raster was set to 24 /im, which corresponds to 0-8 nm at
the specimen, and either 512x512, or more usually 384x384 points were scanned. (The adequacy
of the smaller scan size was checked by trimming down a 512x512 image and reprocessing the
smaller area. No significant difference was observed in the reconstructions.)
Processing of the images was carried out essentially as described previously (Shaw & Hills, 1982;
Roberts, Shaw & Hills, 1981; Shaw, 1981). The reciprocal lattice vectors were initially determined
by an interactive program, then refined by a search procedure that examines every significant spot
in the transform and refines its position by a peak-profile analysis method. Amplitudes and phases
for each spot in the transform were then determined by the profile analysis procedure. Two further
quantities were output for each spot: the ratio of peak to background amplitudes, and the ratio of
the least-squares residual to average local background amplitude. This latter quantity is useful as
a measure of how well the spots agree with the theoretical sine profile (Shaw, 1981). For some
images, amplitude and phase data around each spot were printed out and the spots were examined
manually to check the validity of the automatic procedure.
Tilt angles and axes were determined by the calculation given by Shaw & Hills (1981). The
relative origins of the various data sets were determined using the phase search procedure described
by Unwin & Henderson (1975). The relative origins of four untilted images were determined to give
a zero tilt starting set to which the tilted data sets were correlated in order of increasing tilt angle.
Finally, each data set was refined in turn against all the others for three iterations of the entire
list, until no significant shifts in origins resulted. In refinement, only points on each z* line
closer than 0-025 nm" 1 were used. Amplitudes were scaled using an unweighted least-squares scale
factor:
k= _
Lrt'
summed over all points
274
P. J. Shaw and C. J. Hills
closer than 0-025 nm~'. Various weighting schemes were tried for this calculation, but in all cases
a slight drift of scale factors during iteration of the refinement was found. A similar problem has been
discussed by Fox & Holmes (1966).
Regularly spaced values along the 2* lines were interpolated by a least-squares fit using sine
functions (Crowther, de Rosier & Klug, 1970). This method worked well for all the phase curves,
but occasionally the procedure, an unconstrained least-squares fit, yielded what was judged to be
an unreasonably close fit to the data points for some of the weaker amplitude lines resulting in an
oscillating curve. These lines were smoothed manually. The interpolated data were used as input
to a three-dimensional Fourier transform progTam, and the reconstructed three-dimensional map
was contoured with equally spaced levels in sections of constant z.
RESULTS
Large, dislocation-free areas of the cell wall of Ch. elongatum can easily be obtained. This is a consequence of the shape of the cells and the way in which the twodimensional lattice is fitted around the closed surface of the cell. The cell wall is an
elongated ellipsoid with pointed ends, where large numbers of lattice faults occur.
The central portion of the ellipsoid is an approximate cylinder, however, and a nearly
Table 2. Statistics relating to the two-dimensional structure determination
A. Pairwise phase comparison of individual zero-tilt data sets
2012G
2032G
358G
317G
2012G
2032G
3S8G
317G
35-6
29-9
27-4
31-9
14-5
23-0'
25-4
26-4
9-9
10-7
34-3
26'1
12-7
10-7
11-0
33-7
In the lower left-hand triangle one film of each DElir has been rotated thro ueh 180°. 2012Gand
2032G stained with M E T T ; 358G and 317G stained with MET alone.
Phase residuals/?^ in degrees defines as: R$ = 1/2/V [£(4>u* —<p2**)]*, where <J>IA* are the phases for
the first image; 4>2A* are the phases for the second image with the origin shifted to give the best phase
correlation, summed over all spots whose peak-to-background ratio is greater than 2 0 .
• The factor of i is to give values comparable to those obtained with multiple film scaling in two
and three dimensions, where the discrepancy calculated is of individual films from an average or
interpolated value.
B. Overall statistics for two-dimensional scaling
4 data sets included
2-dimensional data included for 43 spots
fl<f>#= 14-3°
The summations are over all common points.
Three-dimensional structure of Chlorogonium cell wall
275
perfect crystal lattice can be laid down in this region. The purification procedure
produces mainly entire or nearly entire envelopes, and back-to-back double layers are
thus usually observed in the electron microscope. However, with careful searching
areas of single layers can be found. Unfortunately these often show poor crystal order,
which we believe is due to inadequate preservation by the negative stain. With simple
methylamine tungstate as a stain we have only ever obtained three or four areas with
diffraction spots extending to 2-0nm (this included one tilt series). The addition of
Fig. 1. Electron micrograph of portions of the cell wall of Chlorogonium stained with
METT. Bar, 100 nm.
Fig. 2. Optical diffraction pattern of negatively stained Chlorogonium cell wall showing
spots in addition to the basic lattice spots.
Fig. 3. Optical diffraction pattern of an area of the Chlorogonium wall that does not display
any extra spots.
276
P. J. Shaw and G. J. Hills
a small amount of tannic acid to the methylamine tungstate markedly improved the
specimen preservation, however, and in our later work about one area in every ten
surveyed was found to be suitable for analysis. The pictures produced by staining with
METT were identical to those using simple MET, at least to the 2-0 nm resolution
that was available. This was confirmed quantitatively by pairwise comparison of two
METT and two MET data sets, all of which had good diffraction to 2-0 nm (see Table
2). MET and METT images agree with each other as well as they agree among
themselves.
Tannic acid has been widely used as a fixative in sectioning (Mitzuhira &
Futaesaku, 1972). Akey & Edelstein (1983) have used tannic acid fixation followed by
negative staining with uranyl acetate, and have shown that a pattern of stain
distribution identical to that obtained with simple negative staining is produced.
When tannic acid and uranyl acetate are mixed in solution, however, a dense
precipitate results. Providing the solid tannic acid is added to a solution of MET, no
precipitate is produced and a light straw-coloured solution is obtained, which is stable
for some time; presumably, a soluble complex of tungstate and tannic acid is formed.
The resulting solution may be used in the same way as conventional negative stains.
A micrograph of the cell wall negatively stained with METT is shown in Fig. 1,
together with the optical diffraction pattern from this area (Fig. 2) and from another
typical image area (Fig. 3). As can be seen in Fig. 2, spots in addition to the main
reciprocal lattice spots are often observed. The intensity of the extra spots is very
variable and they are sometimes absent. We believe they arise from a super-lattice in
Fig. 4. Two-dimensional reconstruction of the Chlorogonium cell wall obtained by averaging the data from four images. The unit cell chosen has been outlined.
277
Three-dimensional structure of Chlorogonium cell wall
500
(-5,0)
*
/+
+/
180-
i
j,
1
1
4
.hi
r *^
-1801
-0-4
-0-2
1
—1
nm-i
1——1 1 —
0-2
0-4
0-4
-0-2
-i
0-2
-0-4
3000-
-0-4
-0-4
0-4
-0-2
Fig. 5. Selection of lattice lines obtained during the three-dimensional reconstruction.
the wall structure (see below), but since their total intensity is small we ignored them
in our initial structural analysis. Although the indexing system we have used is not the
most obvious one, it is the only one that gives integral h indices for these extra spots.
The unit cell dimensions for this lattice are:
a = 28-6nm
= 7-2nm
y = 128-5°.
P. J. Shaw and G. J. Hills
278
Table 3. Statistics relating to the three-dimensional structure determination
4 untilted images were included
37 tilted images were included
30 2* line were finally included (only those lines with more than 20 data points)
summed over all N data points.
-0-5
-0-6
Fig. 6. Reciprocal space resolution plot of the 2* lines used in the three-dimensional
reconstruction.
Three-dimensional structure of Chlorogonium cell wall
279
Table 2 shows statistics on the agreement of four individual two-dimensional images.
A two-dimensional reconstruction produced by averaging data from the four images
is shown in Fig. 4. The basic structure is of rows of subunits packed closely together
along the b axis, but rather widely separated by less-dense fibrillar interconnections
in the other dimension. Although the structure appears to have approximate twofold
symmetry, there are very distinct and quite consistent departures from this symmetry,
Table 2 shows this in a quantitative fashion, by pairwise comparison of images with
and without relative twofold rotations. The structure must therefore be assigned to
the two-sided plane group PI (Holser, 1958).
Some statistics of the three-dimensional structure determination are shown in
Fig. 7. Stereo pair of the stack of contoured sections of the three-dimensional reconstruction. A and B, large and small subunits.
Fig. 8. Wooden model of the three-dimensional reconstruction showing the envelope
included within the lowest positive contour in Fig. 7.
Fig. 9. The wooden model taken apart to show the subunit structure.
280
P. J. Shaw and G. J. Hills
to
s
o
u
$
z(nm)
Fig. 10. Graph of the maximum contrast (i.e. the difference between the maximum and
minimum density) plotted against z.
Table 3, a selection of z* lines in Fig. 5, and a projection plot showing the distribution of
data measured in reciprocal space in Fig. 6. The lack of symmetry in most of the#* lines,
particularly in amplitude, confirms that the symmetry group is only PI. A stereo-pair
picture of the stacked contour plot is shown in Fig. 7, and a wooden model of the envelope enclosed by the lowest positive contour in Figs 8 and 9. The crystalline layer has
a total thickness of 6—7 nm; Fig. 10 shows a graph of image contrast as a function of z.
Delineation of the subunits from which the structure is composed is quite straightforward and there is little or no ambiguity. There are two types of subunit, which,
however, are very similar. The subunits comprise two domains: a large globular
domain of approximate dimensions 4nmX6nmX6nm and a very thin long, rod-like
domain about 7—8nm in length, which bends and thickens at the end. The chief
difference between the two types of subunit lies in the size of the large globular
domains; the A subunits in Fig. 7 containing significantly more stain-excluding
material. The subunits are associated as heterologous dimers with the dimer interface
being formed by the large domains, and the dimers in turn packing to form lines of
the large domains. The rod-like arms project out of this line on either side in the plane
of the crystal and interlock at their ends with those from the neighbouring rows in a
second, much less dense line of structure. The effect is to give on one side of the crystal
rows of raised units that project about 3-0—4-0 nm from the surrounding structure,
and on the other side a rather flat network. The two types of pore through the
structure, although crystallographically different, are nevertheless almost identical in
size and shape, being approximately circular and 4-0—5-0 nm in diameter. The structure is illustrated diagrammatically in Fig. 11.
Three-dimensional structure of Chlorogonium cell wall
281
Fig. 11. Diagrammatic representation of the subunit structure of the Chlorogonium cell
wall.
For the three-dimensional analysis we ignored the extra spots that are frequently
observed. They are very variable in intensity and often entirely absent from the
transform, and at their strongest are of rather small intensity compared to the simple
lattice spots. With the definition of b* we have used, these spots may be indexed by
simple fractional k indices. We have observed image transforms with extra spots
corresponding to k = 2n/5, k — 2w/6 and k = 2n/7, as well as transforms where the
spots are 'smeared' in the direction parallel to b*. The results from analysis of an
image with particularly strong and coherent super-lattice spots are shown in Figs
12—14. The two-dimensional reconstruction using only the basic lattice spots is
shown in Fig. 12. A reconstruction including the extra spots is shown in Fig. 13. The
changes resulting from the extra spots are quite small, and are centred mainly around
the subsidiary lines of structure, where the subunit 'tails' interdigitate. Fig. 14 shows
reconstruction using data only from the extra peaks, i.e. it represents the difference
between Figs 12 and 13. This shows very clearly rows of associated positive and
negative peaks. We interpret this as being due to a periodic displacement of the ends
of the tails, which repeat in this case every three, unit cells (or every six tails). We
presume that the spots that we have observed on other images at k = 2n/5, k = 2n/l
arise from a similar replacement repeating every five and every seven tails, respectively. Two explanations of this phenomenon have occurred to us. The first is that
this is a genuine super-lattice structure arising from longer-range interactions than
nearest neighbour, and is displayed by the wall structure in vivo in the cell. The
second is that it is produced by flattening an originally curved surface onto the
specimen grid. This would introduce strains into the sheet, which for certain
geometries might be relieved by periodic distortions along what is likely to be the
weakest part of the structure. We favour the first explanation, chiefly because there
does not appear to be any correlation between the presence of the super-lattice and
position on the wall; an area near the 'equator' where curvature is minimal is just
as likely to display the super-lattice as an area near the end where curvature is
greater.
CEL68
282
P.J. Shaw and G.J. Hills
Figs 12—14. Two-dimensional reconstructions of an area of the Chlorogomum cell wall
showing strong super-lattice spots.
Fig. 12. Without the additional spots.
Fig. 13. With the additional spots.
Fig. 14. With only the additional spots included. The contour level is approximately
half of that in Figs 12 and 13.
Three-dimensional structure of Chlorogonium cell wall
283
DISCUSSION
It is a somewhat unexpected feature of our structural results that the dimer should
be heterologous. However, this finding is consistent with results obtained by SDS
(sodium dodecyl sulphate)/polyacrylamide gel electrophoresis of the cell wall
glycoprotein (Roberts, 1974). For Chlorogonium two bands were obtained, both with
approximate relative molecular mass of 150-200( X103). This figure is consistent with
the volume enclosed by the lowest positive contour in our reconstruction. Further
support for this correlation of the SDS/polyacrylamide gel patterns with the structure
comes from another species showing the same class of cell wall structure,
Chlamydomonas dysosmos. This wall gives only one band on SDS/polyacrylamide
gel electrophoresis, with a mobility close to that of the larger of the two Chlorogonium
bands (Roberts, 1974). Our preliminary analysis of micrographs of this cell wall has
shown that although the overall structure is very similar to that of Chlorogonium, the
C. dysosmos structure shows good twofold symmetry. We therefore surmise that this
structure is composed of symmetric dimers. We suggest that a species such as C.
dysosmos appears to have a single cell wall glycoprotein subunit that is closest to a
hypothetical ancestral type. One might then propose a simple evolutionary step from
a glycoprotein subunit that dimerizes (i.e. self—self recognition) to two closely similar
glycoproteins, which recognize each other and thus form heterologous dimers.
The rod-like arms of the subunits are comparable in width to the resolution of the
reconstruction (2 nm) and would be entirely consistent with the model for Hyp-rich
glycoproteins proposed by Lamport (1980). In this model regions very high in
hydroxyproline residues form a polyproline II helix, and the oligosaccharides
attached to the hydroxyl groups of the proline residues pack around the backbone
helix to give stable and strong rods, which would have a diameter of approximately
2 nm. There is strong evidence for the presence of a large domain of this structure in
the cell wall glycoprotein of the closely related species C. reinhardtii, both from
circular dichroism measurements of the intact glycoprotein and large proteolytic
fragments (Homer & Roberts, 1979), and from chemical studies of amino acid composition and patterns of glycosylation (Roberts, 1979). The model we suggest for each
of the subunits of the Chlorogonium cell wall glycoprotein, therefore, is of a long rod
composed of a glycosylated polyhydroxyproline helix, which acts as a structural
spacer, together with a large globular domain, which forms strong dimer associations
and inter-dimer contacts. There is then a region at the opposite end of the rods that
forms a second pair of intersubunit associations. The Chlorogonium wall glycoprotein
appears to display in a particularly simple manner structural features that may well
be shared by other hydroxyproline-containing plant glycoproteins; namely, long rodlike spaced domains together with globular domains that form interaction and association sites. We might expect a similar general structure for such glycoproteins as the
flagellar sexual agglutination factors in Chlamydomonas (Cooper et al. 1983; Pijst et
al. 1983), and the class of plant cell wall glycoproteins that has been termed extensin
(Lamport, 1977). There is already some evidence for this type of architecture in
potato lectin (Allen et al. 1978).
284
P. J. Shaw and G. J. Hills
REFERENCES
AKEY, C. W. & EDELSTEIN, S. J. (1983). Equivalence of the projected structure of thin catalase
crystals preserved for electron microscopy by negative stain, glucose embedding or embedding
in the presence of tannic acid. J . molec. Biol. 163, 575-612.
ALLEN, A. K., DESAI, N. N., NEUBERGEB, A. & GEETH, J. M. (1978). Properties of potato lectin
and the nature of its glycoprotein linkages. Biochem.J. 171, 665-674.
CATT, J. W., HILLS, G. J. & ROBERTS, K. (1978). Cell wall glycoproteins from Chlamydomonas
reinhardii, and their self-assembly. Planta 138, 91-98.
COOPER, J. B., ADAIR, W. S., MECHAM, R. P., HEUSER, J. E. & GOODENOUGH, U. W. (1983).
Chlamydomonas agglutinin is a hydroxyproline-rich protein. Proc. natn. Acad. Set. U.SA. 80,
5898-5901.
CROWTHER, R. A., D E ROSIER, D. J. & KLUG, A. (1970). The reconstruction of a 3-dimensional
structure from projections and its application to electron microscopy. Proc. R. Soc. Land. A 317,
319-340.
ERICKSON, H. P. (1973). The Fourier transform of an electron micrograph - first order and second
order theory of image formation. Adv. opt. Elect. Microsc. 5, 163-199.
FABERGE, C. & OLIVER, R. M. (1974). Methylamine tung9tate, a new negative stain. J . Microsc.
20, 241-246.
Fox, G. C. & HOLMES, K. C. (1966). An alternative method of solving the layer scaling equations
of Hamilton, Rollett & Sparks. Ada crystallogr. 20, 886-891.
HOLSER, W. T . (1958). Point groups and plane groups in a two-sided plane and their subgroups.
Z. Kristallogr. Kristallgeom. 110, 266-281.
HOMER, R. B. & ROBERTS, K. (1979). Glycoprotein conformation in plant cell walls. Planta 146,
217-222.
LAMPORT, D. T. A. (1977). Structure, biosynthesis and significance of cell wall glycoproteins. In
Recent Advances in Phytochemistry, vol. II (ed. Loewus & Runeckles). New York: Plenum.
LAMPORT, D. T . A. (1980). Structure and function of plant glycoproteins. In The Biochemistry of
Plants, vol. 3 (ed. P. K. Stumpf & E. E. Conn), pp. 501-541. New York, London: Academic
Press.
MITZUHIRA, V. & FUTAESAKU, Y. (1972). New fixation for biological membranes using tannic acid.
Acta Histochem. Cytochem. 5, 233-236.
PIJST, H. L. A., ZILVER, R. J., MUSGRAVE, A. & VAN DEN ENDE, H. (1983). Agglutination factor
in the cell body of Chlamydomonas eugametos. Planta 158, 403—409.
ROBERTS, K. (1974). Crystalline glycoprotein cell walls of algae: their structure composition and
assembly. Phil. Trans. R. Soc. Land. B 268, 129-146.
ROBERTS, K. (1979). Hydroxyproline: its asymmetric distribution in a cell wall glycoprotein.
Planta 146, 275-279.
ROBERTS, K. & HILLS, G. J. (1976). The crystalline glycoprotein cell wall of the green alga
Chlorogonium elongatum: a structural study. J. Cell Sci. 21, 59-71.
ROBERTS, K., HILLS, G. J. & SHAW, P. J. (1982). The structure of algal cell walls. In Electron
Microscopy of Proteins, vol. 3 (ed. Harris). New York, London: Academic Press.
ROBERTS, K., SHAW, P. J. & HILLS, G. J. (1981). High-resolution electron microscopy of
glycoproteins: the crystalline cell wall of Lobomonas. J. Cell Sci. 51, 295-313.
SHAW, P. J. (1981). A peak profile analysis procedure for extracting unit cell transform data from
the Fourier transforms of periodic arrays. J. Cell Sci. 51, 314—321.
SHAW, P. J., GARNER, R. T . & PARKER, E. A. (1981). A flat bed scanning microdensitometer for
computer processing of electron micrographs. Micron 12, 103-114.
SHAW, P. J. & HILLS, G. J. (1981). Tilted specimens in the electron microscope: a simple specimen
holder and the calculation of tilt angles for crystalline specimens. Micron 12, 279—282.
SHAW, P. J. & HILLS, G. J. (1982). Three-dimensional structure of a plant cell wall glycoprotein.
J. molec. Biol. 162, 459-471.
UNWIN, P. N. T. & HENDERSON, R. (1975). Molecular structure determination by electron microscopy of unstained crystalline specimens..?, molec. Biol. 94, 425—440.
{Received 16 January 1984-Accepted 2 February 1984)