An alternative model for photosystem II/light harvesting complex II
in grana membranes based on cryo-electron microscopy studies
Robert C. Ford
1
, Svetla S. Stoylova
2
and Andreas Holzenburg
3
1
Department of Biomolecular Sciences, UMIST, Manchester, UK;
2
The Burnham Institute, La Joua, CA, USA;
3
Department of Biology and Department of Biochemistry and Biophysics, Microscopy and Imaging Center,
Texas A & M University, College Station, TX, USA
The photosynthetic protein complexes in plants are located
in the chloroplast thylakoid membranes. These membranes
have an ultrastructure that consists of tightly s tacked ÔgranaÕ
regions interconnected by unstacked membrane regions. The
structure of isolated grana membranes has been studied here
by cryo-electron microscopy. The data reveals an unusual
arrangement of the photosynthetic protein complexes,
staggered over two tightly stacked planes. Chaotrope treat-
ment of the paired grana membranes has allowed the sepa-
ration and isolation of two biochemically distinct membrane
fractions. These data have led us to an alternative model of
the u ltrastructure o f the gran a where segregation exists
within the grana itself. This a rrangement would change t he
existing view of plant photosynthesis, and suggests potential
links between c yanobacterial and plant photosystem II light
harvesting systems.

which in turn, pass it to the r eaction centre chlorophylls of
PSII. The l atter c onvert t he light energy into chemical
potential energy via e lectron transfer [7]. This chemical
potential is eventually used to carry out the universally
recognized functions of photosyn thesis, i.e. to ®x atmo-
spheric carbon dioxide for biomass, liberate oxygen from
water, and in general drive the energy requiring processes
in the plant.
The complete PSII/LHCII complex is thought to consist
of more than 20 different polypeptides and several hundred
(250±350) pigment molecules [8,9]. The total mass of PSII/
LHCII has been estimated at around 1 MDa. With such
complexity, it is understandable t hat knowledge o f the
structure of PSII/LHCII proteins has largely come from
studies of isolated components of the system [10] o r
subcomplexes that have been removed from the membrane
by detergent extraction [11,12]. However two-dimensional
crystalline arrays of PSII/LHCII have sometimes been
observed to form in situ in the grana membranes, and these
can be studied using electron crystallography techniques
[13±21]. Such native c rystals are inevitably smaller than
crystals of isolated proteins [10] or puri®ed complexes of
proteins [11]. However it has been shown using experimental
data [22] and by simulations [23] that averaging of cryo-
electron microscopy data for several small crystalline areas
is practical and results in structural data equivalent i n
quality to that obtained from much larger single crystalline
arrays.
In this article, we describe cryo-electron microscopy
studies of grana membranes, and show a projection

programs developed mainly at the Medical Research
Council Laboratory of Molecular Biology [26]. After
correction for lattice defects (lattice unbending) and for
the contrast transfer function (CTF), data was merged. The
program
PLOTALL
was used to calculate the phase errors for
the structure factors (kindly provided by W. Kuhlbrand,
MPI Biophysics, Frankfurt)
1
. Phase origins were determined
using the program
ORIGTILTD
with restriction to the lower
resolution (to 20 A
Ê
), high signal/noise (IQ3 or better)
re¯ections [26,46]. The IQ is a n integer value determined by
the peak-to-background ratio at a point in reciprocal space
determined by the reciprocal lattice, with a value of IQ1 for
a ÔgoodÕ ratio of 8 : 1 or more, with IQ values 2±8
identifying re¯ections with progre ssively worse peak to
background ratios. Finally IQ9 is assigned to re¯ections
with amplitudes below background. The phase origins were
then further re®ned using the lower resolution (to 20 A
Ê
)
averaged structure factors from the initial merging proce-
dure as the starting reference. Structure factors were
vectorially averaged using the program

generally noisier for cryo-electron microscopy, whilst phases
are usually more reliable when compared to X-ray crystal-
lography. The approach developed by Perkins et al.
depends on oversampling (redundancy) followed by aver-
aging by vector summation. A measure of the redundancy
for each structu re factor is therefore an important indication
of the reliability of its vector sum phase and amplitude
values, with high redundancy correlating with better
accuracy. Within a data set derived from 21 crystal areas,
Fig. 1. Thylakoid membrane morphology.
(a) Transmission electron micrograph of an
ultrathin s ection of isolated barley chloroplast
membranes (thylakoids). Note the t ightly
stacked membranes (grana). Scale
bar  500 nm ( b) Zoomed re gio n of (a)
showing a single grana stack, sectioned s lightly
obliquely. (c) Explanation of the morphology
shownin(b)withunstackedregions(u)
and t ightly appressed membrane pairs forming
the stacked regions (s). A single membrane
pair (dotted line) is indicated. (d) Higher
magni®cation of two membrane pairs in a
stack in side view with the narrow partition
gap between the membranes highlighted
(white arrows). Scale bar  100 nm
(e) Micrograph of i solated membrane p airs in
face view, embedde d in negative s tain, and
displaying two-dimensional crystals of PSII/
LHCII. Scale bar  150 nm
Ó FEBS 2002 PSII/LHCII structure in situ (Eur. J. Biochem. 269) 327

individual (one-off) probability of observing data of IQ 7 o r
better by chan ce alone.
A redundancy of 5 .6 in this data set corresponds to
standard errors for t he mean (vector sum) phases of around
30° (see Tables 1±3)
2
. Standard error of the mean (vector
sum) phase appears to b e a more reasonable estimate for the
phase errors for this image processing procedure because
this measure i ncludes a weighting f or the number o f
observations, i.e. the redundancy of the data is taken into
account. In comparison, unweighted interimage phase
residuals do not take into account the redundancy of the
data and hence can g ive a misleading pessimistic impression
of the reliablility of oversampled data.
The three-dimensional data set was obtained using the
same approach as above and as described in Amos et al.
[26], but because of the very large body of data, we initially
restricted the analysis to the lower r esolution/higher ampli-
tude components. Table 1 lists the number of ®les employed
in the different tilt ranges, demonstrating that reciprocal
space is reasonably evenly sampled by the data. Neverthe-
less, the physical restriction imposed by the specimen holder
in the microscope means that there is a Ômissing coneÕ of data
corresponding to tilts beyond  60±70°. The effect th at this
missing data has on the three-dimensional reconstruction
has been discussed previously [26], with its main outcome
being some loss of resolution perpendicular to the crystal
plane.
A three-dimensional Coulomb d ensity map for the

No. of crystalline areas 21
No. of observations (to IQ8) 9824
(to IQ7) 4810
No. of structure factors 846
No. with FOM
2
> 0.8 734
No. used for map with FOM > 0.88 557
Mean redundancy (250±8 A
Ê
, to IQ7) 5.6
Table 2 .
8
Crystallographic image processing statistics for the 8 A
Ê
projection map over d ierent resolution ranges.
Resolution Rmerge
a
Mean FOM
b
SE (°)
c
Redundancy
d
% Complete
e
250±50 A
Ê
0.28 0.99 9.0 13.9 91%
50±30 A

error of the mean phase was calculated for each structure factor and then averaged over the given resolution range.
d
Average number
of observations of IQ7 or better, for structure factors within this resolution range.
e
Number of structure factors used (with FOM > 0.88)
for calculating the map vs. the number of structure factors actually expected in this resolution range.
Table 3. Crystallographic image processing statistics for the
9
30 A
Ê
three-
dimensional map.
Scan step at the specimen level 6.6 A
Ê
or 8.9 A
Ê
No. of crystalline areas 168
Maximum tilt angle  66°
No of ®les in tilt range 0±30° 69
30±40° 14
40±50° 28
50±60° 54
60±66° 2
No. of observations (to 30 A
Ê
) 5066
No. of structure factors used 470
Overall weighted phase residual to 30 A
Ê

linear sucrose gradient composed of 0±2
M
sucrose i n 0.75
M
Tris-base, 3
M
urea, pH 8.8. After centrifugation for 2 h at
110 000 g in a Beckman SW41 rotor, green bands corre-
sponding to different membrane fractions were harvested.
Membranes were diluted 1 : 1 with distilled water and then
centrifuged at 110 000 g for2htoobtainpellets.After
resuspension in buffer A, the membranes w ere analysed by
absorbance spectroscopy, SDS/PAGE and electron micros-
copy. Absorbance spectra were recorded with a Kontron
spectrophotometer (model Uvikon 943) with 1-cm path-
length cuvettes. SDS/PAGE was carried o ut as described
previously [19,21].
RESULTS
PSII/LHCII structure
Figures 1a,b shows the morphology of the thylakoid
membranes we employed, with the characteristic stacked
membranes o f t he grana. Isolation of tightly stacked
membrane pairs (Fig. 1c,d) is readily achieved, and two-
dimensionally ordered arrays of PSII/LHCII present in
these membranes (Fig. 1e) can be observed [13±21]. C ryo-
electron microscopy of such two-dimensional arrays gener-
ated projection and three-dimensional structures of the
PSII/LHCII complex. Figure 2a shows the signal-to-noise
ratios of reciprocal lattice points after averaging 21 untilted
crystalline arrays. The data is relatively complete and the

ÔcoreÕ domain in the 8-A
Ê
map is a
distinctive S-shaped region formed by several strong density
peaks. The S-shaped region could be the location of the
reaction centre of PSII, which, on the basis of its predicted
similarity to the bacterial reaction centre [24], has been
observed in three-dimensional density maps obtained for
PSII core complexes [11,12] (Fig. 4B). The overall dimen-
sions of the core domain (140 ´ 100 A
Ê
) match very closely
to the dimensions of one monomer of the cyanobacterial
PSII core complex (130 ´ 100 A
Ê
)determinedbyX-ray
crystallography [12], as shown in Fig. 3B. This supports the
conclusion that the higher plant PSII complex i s monomeric
in vivo, a s suggested previously [14,15,19±21,23]. Clearly,
caution must be exercised in a more detailed comparison of
the two projection maps especially regarding the identi®ca-
tion of transmembrane helices, because in the native PSII
structure, additional extrinsic proteins and loops will be
superimposed (compare to Figure 4), w hereas in the c urrent
deposition of the cyanobacterial structure, only the trans-
membrane regions and t wo of the extrinsic subunits are
de®ned [12]. Similarly, the 8-A
Ê
resolution projection map of
the PSII core complex derived by R hee & coworkers [25],

ology [11,26,27]. Details regarding the image processing
statistics are given in Table 3. The three-dimensional
structure has been calculated to a resolution o f  30 A
Ê
.
This cu t-off is suitable for comparison with earlier studies of
negatively stained PSII/LHCII, which have a similar
resolution. Three-dimensional d ata beyond 30 A
Ê
have been
collected and processed, but further crystals need to be
included in the analysis to adequately oversample three-
dimensional reciprocal space to higher (8 A
Ê
) resolution.
Figure 4 s hows different views of t he PSII/LHCII complex,
with a surface generated at a suitable threshold for
discrimination of protein density. The main features of the
140 ´ 100 A
Ê
core domain c orrespond closely to those
described earlier for negatively stained spec imens [14,15].
The distinctive cavity on the lumenal side of the complex is
apparent, surrounded by four prominent lumenal domains,
some of which were previously assigned to extrinsic PSII
proteins that enhance oxygen e volution. Sequential removal
of these extrinsic proteins, followed by structural analysis
has identi®ed domains I, II and III as the approximate
locations of oxygen evolution enhancing (OEE)
3

The third (12 kDa) extrinsic polypeptide of the c yanobac-
terial complex was not resolved in the publishe d structure
[12], but is likely to appear in later density maps (P. Orth,
FU, Berlin, personal communication)
4
.Theoveralldimen-
sions and shape of the cyanobacterial PSII core complex and
the higher plant PSII core region are very similar at 30 A
Ê
resolution, again supporting the idea that the higher plant
PSII complex is monomeric in situ.
The location of the connecting densities that bridge
between the core domains was unexpected. It is clear from
Fig. 4 that the connecting domains lie in a separate p lane to
the main core region. All these small domains align almost
exactly along a single plane, which immediately suggests
that they are not due to random noise or poor sampling of
three-dimensional space. The most likely explanation for
this observation, given the double-layered nature of the
crystals, i s t hat the connecting domains occupy a membrane
that is separate to the one housing the core domain.
A n arrow but distinct gap between th e two planes of density
is  0.5±1 nm across, which would c orrespond closely t o the
width o f the partition region that can be identi®ed between
pairs of closely appressed g rana membranes in ultrathin
sections (Fig. 1d). The overall size (4 nm height ´ 3nm
Fig. 3.
6
Projection maps of the entire PSII/
LHCII complex. (A) Maps are calculated to

and D2 is highlighted in the cyanobacterial
map (dashed ellipse). This region is tentatively
assigned in the higher plant map (ellipse), and
is centred on a rough twofold symmetry axis.
The transmembrane helices of the a ccessory
polypeptides CP47 and CP43 can not be
readily identi®ed in the higher plant map,
however, as < 5 0% of the ma ss of thes e
subunits is contained in the tran smembrane
helices, then t heir identi®cation in a projection
map is unlikely because of convolution with
overlying densities.
Ó FEBS 2002 PSII/LHCII structure in situ (Eur. J. Biochem. 269) 331
width ´ 4 nm length) and number (4±5) of the connecting
domains immediately suggested that they could be periph-
eral LHCII proteins [10], although the resolution was
insuf®cient for unambiguous id enti®cation, and one cannot
exclude the possibility that these densities may be due to
ordered peripheral proteins. If the assignment to LHCII is
correct, then the observation of only 4±5 densities rather
than 8±12 implies that only a subset of the LHCII
population is involved in the contacts between core
complexes.
Biochemical evidence for two grana membrane fractions
Biochemical evidence for the presence of two different
membrane types in grana thylako id membrane fractions is
scant. A s earch f or conditions that would allow t he
disruption of t he paired membranes w ithout membrane
solubilization was carried out. Se veral procedures employ-
ing chaotropes and/or proteases were found to give some

but it contained less material when c ompared to the control
(band b ). Two further distinct chlorophyll-containing bands
were observed for the Tris/urea-treated material: mem-
branes separating as a broad band located at  1.4
M
sucrose (band c) showed a radically changed absorption
spectrum, being depleted in the Chl b absorption bands at
650 and 480 nm, consistent with a lack of light-harvesting
Chl a/b proteins (LHCII). Membranes located slightly
above the main band at  1.0
M
sucrose (band a) had a
similar absorption spectrum to the main band, but with a
slightly increased Chl b absorption.
SDS/PAGE of the Tris-treated membranes is shown in
Fig. 5C. The fraction isolated from around 1.4
M
sucrose
(band c in panel A) was signi®cantly depleted in LHCII
polypeptides, but was enriched in the core polypeptides
D1,D2, CP43 and CP47 (right track). No bands due to
extrinsic polypeptides of PSII (33, 23, 17, 10 kDa) could be
observed, but these polypeptides will be removed by the
chaotrope treatment. The f raction isolated at 1.0
M
sucrose
(band a in panel A ) is signi®cantly depleted in the D1,D2,
CP43 and CP47 polypeptides (left track) whilst retaining
intensely staining LHCII polypeptides. These data therefore
suggest that separation of grana membranes into denser

membranes as determined by SDS/PAGE and
Coomassie staining. The left lane shows band
a and the right lane band c. Molecular mass
markers are ind icated o n t he left of the panel.
Ó FEBS 2002 PSII/LHCII structure in situ (Eur. J. Biochem. 269) 333
membrane patches that contained large  14 nm diameter)
particles (Fig. 6a, insert) consistent with core PSII. The
packing of these complexes is very tight (2300 parti-
clesálm
)2
), considerably higher than that observed for
untreated grana membranes (1300±1500 particlesálm
)2
).
The LHCII-enriched density gradient fraction contained
rolled-up membrane ÔtubesÕ withsmallinternalfeatures.
(Fig. 6b,c).
DISCUSSION
Interpretation of the three-dimensional data
The data presented here provide i nformation for the
complete PSII/LHCII complex observed under co nditions
that preserve its native s tate [27]. In earlier structural studies,
negative stain was employed where d ehydration and
shrinkage are known t o be problems [14,21] as well as
differential staining of upper and lower surfaces of t he
specimen [28]. These combined factors may explain why
previous studies did not readily identify two planes of
density. Negatively stained PSII/LHCII c rystals in spinach
grana [14] do d isplay some small domains that lie in a lower
plane than the main core of the complex [14], but there was

fraction, and LHCII complexes can be observed in another
membrane fraction. A survey of previous structural studies
of thylakoid membranes [13,16,17,21, 29±32] suggests that
they may be newly interpreted in terms of the alternative
model of thylakoid structure. A review of these studies is
beyond the scope of this paper and will be presented
elsewhere.
The alternative model, if correc t, has several implica-
tions for understanding PSII function ranging from light
harvesting control [33±38] to the optimization of diffusion
of PSII and of components around PSII [39±44]. A
discussion of these implications is again beyond the scope
of this paper, and will be addressed in a separate review.
However we note that migration of light energy to the PSII
core in a direction perpendicular to the membrane plane
would not be unique to plants. The more ancient
cyanobacterial PSII does not have LHCII proteins, but
rather it depends on water-soluble light harvesting proteins
that are attached as a ÔphycobilisomeÕ to the stromal
surface of the PSII core [36]. Other photosynthetic
bacteria, such as the green sulphur bacteria, also move
light excitation energy from chlorosomes to the membrane
in which the reaction centre is found [37].
Testing the model
This paper h ighlights a discord between the structural data
and the existing model of PSII/LHCII and grana archi-
tecture, and this should now open a debate on the merits
of the alternative models. We note that Ômacro-domainsÕ of
LHCII in plants have already been proposed to explain
data derived from several biophysical techniques [45], and

PSII core whilst the remaining occupy a separate
membrane.
Progress is slowly being made towards processing a
higher resolution three-dimensional data set for the PSII/
LHCII crystals. When this is complete, the data should
reveal much more concerning the nature of the contacts in
the crystals and offer further insight into the interplay
between PSII s tructure and function in the thylakoid
membrane.
ACKNOWLEDGEMENTS
We would like to thank Dr M. F. Rosenberg for his assistance with
software and Dr S. Prince, Dr S. V. Rue and Prof. G. Garab for
useful suggestions and debate. T. D. Flint is thanked f or plant g rowth
and specimen p reparation as well as L. Child and P. McPhie for expert
technical assistance. The data collection phase of this work was
supported by the UK Biotec hnology and Biological Science s Research
Council.
REFERENCES
1. Andersson, B. & Anderson, J.M. (1980) Lateral heterogeneity in
the distribution of chlorophyll±protein complex es of the thylakoid
membranes of spinach chloroplasts. Biochim. Biophys. Acta 593,
427±440.
2. Anderson, J.M. & Andersson, B. (1982) The architecture of the
photosynthetic membrane: lateral and transverse organisation.
Trends Biochem. Sci. 7, 288±292.
3. Barber, J. (1980) An explanation for the relationship between salt-
induced t hylakoid stack ing and the chloro phyll ¯u orescenc e
changes associated in spillover of energy from photosystem II to
photosystem I. FEBS Lett. 118, 1±10.
4. Green, B.R. & Dunford, D.G. (1996) The chlorophyll-carotenoid

Resolution. Nature 409 , 739±
743.
13. Staehelin, L.A. (1975) Chloroplast membrane structure. Biochim.
Biophys. Acta 408, 1±11.
14. Holze nburg, A., Bewley, M.C., Wilson, F.H., Nicholson, W.V. &
Ford, R.C. (1993) Three-dimensional s tructure of photosystem II.
Nature 363, 470±472.
15.Ford,R.C.,Rosenberg,M.F.,Shepherd,F.H.,McPhie,P.&
Holzenburg, A. (1995) Photosystem II 3D structure and the role of
the extrinsic subunits in photosynthetic oxygen evolution. Micron
26, 133±140.
16. Tsvetkova, N.M., Apostolova, E.L., Brain, A.P.R., Williams,
W.P. & Quinn, P.J. ( 1995) Facto rs in¯uencing PSII p article array
formation in Arabidopsis thaliana chloroplasts and the relationship
of such arrays to the thermostability o f PSII. Biochim. Biophys.
Acta 1228, 201±210.
17. Semenova, G. (1995) Particle regularity on thylakoid fracture
faces is in¯uence d by sto rage cond itions. Can. J. Bot. 73, 1676±
1682.
18. Marr, K.M., M cFeeters, R.L. & Lyon, M.K. (1996) Isolation and
structural analysis of two-dimensional crystals of photosystem II
from Hordeum vulgare viridis zb63. J. Struct. Biol. 117, 86±98.
19. Stoylova, S., Flint, T.D., Ford, R.C. & Holzenburg, A. (1997)
Projection structure of photosystem II in vivo studied by cryo-
electron microscopy. Micron 28, 439±446.
20. Stoylova, S., Flint, T.D., Ford, R.C. & Holzenburg, A. (1998)
Comparison of photosystem II 3D structure as determined by
electron crystallography of frozen-h ydrated and n egatively stained
specimens. Micron 29, 341±348.
21. Stoylova, S., Flint, T.D., Ford, R.C. & Holzenburg, A. (2000)

bene®ts. Micron 25, 5±13.
29. Simpson, D.J. (1979) Freeze fracture studies on barley plastid
membranes III. Carlsberg Res. Commun. 44 , 305±336.
Ó FEBS 2002 PSII/LHCII structure in situ (Eur. J. Biochem. 269) 335
30. Simpson, D.J., Vallon, O. & Von Wettstein, D. (1989) Freeze
fracture studies on barley plastid membranes VIII. Biochim. Bio-
phys. Acta 975, 164±174.
31. Olive, J., Recouvrer, M., Girard-Bascou, J. & Wollman, F.A.
(1992) Further identi®cation of the exoplasmic face p articles on
the freeze-fractured thylako id membranes. Eur. J. Cell Biol. 59,
176±186.
32. Rosenb erg, M.F., Holzenburg, A., Shepherd, F.H., Nicholson,
W.V., Flint, D. & Ford, R.C. (1997) Rebinding of the extrinsic
proteins of photosystem II studied by e lectro n m icroscopy and
single particle alignment. Biochim. Biophys. Acta 1319, 119±132.
33. Allen, J.F. (1992) Protein phosphorylation in the regulation of
photosynthesis. Biochim. Biophys. Acta 1098, 275±335.
34. Horton, P. (1999) Are grana necessary for regulation of light
harvesting? Aus. J. Plant. Physiol. 26, 6 59±669.
35. Campbell, D.A. & Hayden, D.B. (1992) Cross-linking of photo-
system-II light-harvesting complexes between appressed maize
thylakoids Plant Physiol. Biochem. 30, 723±732.
36. Delorimer, R.M., Smith, R.L. & Stevens, S.E. (1992) R egulation
of phycobilisome structure and gene expression by light int ensity.
Plant. Physiol. 98, 1003±1010.
37. Olson, J.M. (1998) Chlorophyll organization and function in green
photosynthetic bacteria. Photoche m. Pho tob iol. 67 , 61±75.
38. Kyle, D.J., Haworth, P. & Arntzen, C.J. (1982) Thylakoid mem-
brane phosphorylation leads to a decrease in connectivity between
photosystem II reaction centres. Biochim. Biophys. A cta 680,336±

1476.
46. Glaeser, R.M. & Downing, K.H. (1992) Assessment of resolution
in biological electron crystallography. Ultramicroscopy 47,256±
265.
47. Brillinger, D.R., Downing, K.H. & Glaeser, R.M. (1990) Some
statistical aspects of low-dose electron imagin g of crystals. J. Stat.
Plan. Inf. 25, 535.
48. Boekema, E.J., van Breemen, J.F.L., van Roon, H. & Dekker, J.P.
(2000) Arrangement of photosystem II supercomplexes in crys-
talline macrodomains within the thylakoid membrane of green
plant chloroplasts. J. Mol. Biol. 301 , 1123±1133.
49. Boekema, E.J., Hankamer, B., Bald, D., Kruip, J., Nield, J.,
Boonstra, A.F., Barber, J. & Roegner, M. (1995) Supramolecular
structure of the phot osystem II complex from green plan ts and
cyanobacteria. Proc. Natl. Acad. Sci. USA 92, 175±179.
50. Hankamer, B., Nield, J., Zheleva, D., Boekema, E.J., Jansson, S.
& Barber, J. (1997) Isolation and biochemical characterisation of
monomeric a nd dimeric photosystem II complexes from spinach
and their relevance to the organisation of photosystem II in vivo.
Eur. J. Biochem. 243, 422±429.
51. Boekema,E.J.,vanRoon,H.,Calkoen,F.,Bassi,R.&Dekker,
J.P. (1999 ) Multiple types of association of photosystem II and its
light-harvesting antenna in partially solubilized photosystem II
membranes. Biochemistry 38, 2233±2239.
336 R. C. Ford et al. (Eur. J. Biochem. 269) Ó FEBS 2002