The crystal structure of annexin Gh1 from
Gossypium hirsutum
reveals an unusual S
3
cluster
Implications for cellulose synthase complex formation and oxidative stress response
Andreas Hofmann
1
, Deborah P. Delmer
2
and Alexander Wlodawer
3
1
Institute of Cell & Molecular Biology, The University of Edinburgh, Edinburgh, Scotland;
2
The Rockefeller Foundation,
New York, USA;
3
Macromolecular Crystallography Laboratory, NCI at Frederick, Frederick, Maryland, USA
The three-dimensional crystal structure of recombinant
annexin Gh1 from Gossypium hirsutum (cotton fibre) has
been determined and refined t o t he final R-factor of 0.219 at
the resolution of 2.1 A
˚
. This plant annexin consists of the
typical Ôannexin foldÕ and is similar to the previously solved
bell pepper annexin Anx24(Ca32), but significant differences
are seen when compared to the structure of nonplant
annexins. A comparison with the structure of the mamma-
lian annexin AnxA5 indicates that canonical calcium bind-
ing is geometrically possible within the membrane loops in

ÁÀ
2
), hydroxyl radical (OHÆ), and hydrogen peroxide
(H
2
O
2
). These s pecies are by-products of normal aerobic
metabolism and result from successive single electron
transfers from/to oxygen . Partially reduced oxygen species
are involved in DNA damage, lipid peroxidation, and
protein denaturation. Through apoptosis and necrosis,
these types of cellular damage can give rise to several
pathological symptoms observed in diseases such as cancer,
arthritis, and muscular dystrophy, as well as to genetic and
nervous disorders [1–4].
Mammalian annexins A1 [5], A5, and A 6 [6], as well as
plant annexins from Medicago sativa [7] and Arabidop-
sis thaliana [8,9], have been implicated in oxidative stress
response. In particular, it has been shown that an annexin-
like protein from Arabidopsis, Oxy5, is able to rescue
Escherichia coli DoxyR mutants from H
2
O
2
stress . Cot ton
fibre annexins have been shown to colocalize with cellulose
synthase and to have an inhibitory effect on glucan synthesis
[10]. In a recent study [11] a redox-dependent model for
cellulose synthase complex formation was proposed, which

homologues in Arabidopsis [15] raises the question whether
annexins in plants might also appear as a diverse multigene
family, in common with their mammalian relatives.
Calcium binding has been identified as a landmark
feature of animal, plant, and metazoan annexin proteins. As
structurally established for annexin A5 [16], the canonical
type II calcium binding sites are found within the AB loops
of each domain and are provided by the endonexin sequence
K-G-X-G-T-{38}-D/E [17]. Typically, the coordination
sphere around the cation is a pentagonal bipyramid with
a backbone carbonyl group and a water molecule in apical
positions. Another water molecu le, three backbone carbo-
nyl groups, and the acidic residue from the conserved motif
form the base of the bipyramid. Because only one side chain
is involved in creating this site, there is no stringent apriori
requirement that the side chains within the endonexin
sequence be conserved. A d ifferent amino acid sequence
with a suitable loop conformation might act as proper
calcium binding site as well.
Type III and AB¢ sites, in contrast, are constituted by
one or two backbone carbonyl groups and a neighbouring
bidentate acidic residue and coordinate the calcium ion
together with several water molecules. Type III binding sites
are the only ones observed with DE loops. It has been
concluded that calcium bound in the AB loops is responsible
for membrane a dsorption, while the calcium harboured in
DE sites increases the binding affin ity in general [18].
While the primary structure of plant annexins reflects the
characteristic fourfold repeat, there is variation in the loops
harbouring the endonexin sequence. The motif is conserved

recombinant Anx(Gh1) a nd determined its three-dimen-
sional crystal structure in the calcium-free form. A
comparison of the membrane binding loops with AnxA5
and Anx24(Ca32) reveals that canonical calcium binding
in the loops of domain I might be possible in Anx(Gh1),
in contrast to the bell pepper annexin. The protein
contains a highly unusual sulfur cluster formed by two
adjacent cysteine residues i n their reduced forms and a
methionine residue. The cluster is likely to be involved in
redox reactions and might constitute the m olecular basis
of oxidative stress response by annexins.
Materials and methods
Purification of recombinant protein
Cloning and construction of an N-terminal His
4
-fusion
protein has been described earlier [20]. T he recombinant
protein of Anx(Gh1) carried a hexapeptide extension
MAHHHH and was expressed in Escherichia coli
BL21(DE3) cells. A total of 8 L of LB medium (50 mgÆL
)1
ampicillin) were inoculated with an overnight culture of 1 L.
The cells were grown a t 37 °C until the absorbance at
600 nm exceeded 1.0. Induction was carried out with
0.5 m
M
isopropyl t hio-b-
D
-galactoside; at that t ime, the
concentration of ampicillin was increased twofold. Cell

Hepes (pH 8.0) and applied to
a Q-Sepharose column. After a short washing, the protein
was eluted with a linear gradient 0–1
M
NaCl in 20 m
M
Hepes (pH 8.0). Anx(Gh1) eluted at 230–350 m
M
NaCl.
Concentration was carried out by ultracentrifugation using
Millipore Centricon devices.
Crystallization
Crystals of recombinant Anx(Gh1) were obtained using the
hanging-drop vapour-diffusion method. Droplets consisted
of 3 lLproteinand3lL reservoir solution equilibrated
against 300 lL reservoir solution at 285 K. The crystals
grew in about 8 weeks from 1.7
M
(NH
4
)
2
SO
4
,0.1
M
Hepes
(pH 7.0). Several crystals obtained from similar conditions
(pH 6.0–7.0) were soaked in mother liquor in the presence
of 2–15 m

were recorded, while in the second run the detector was
moved closer to the crystal to record reflections between 8
and 2.1 A
˚
.Dataprocessingwascarriedoutwiththe
programs
DENZO
and
SCALEPACK
[21] and the data collec-
tion statistics are summarized in Table 1.
The diffraction pattern indicated a trigonal space group
with approximate cell dimensions of a ¼ b ¼ 61 A
˚
and c ¼
215 A
˚
. Two-fold axes were detected parallel t o [210] and
[120] in the self-rotation function calculated with GLRF
[22]. This rendered P3
1
12 and P 3
2
12 as the possible space
groups.
The structure was solved by molecular replacement
with AMoRe [23] starting from a poly Ala model of
Anx24(Ca32) that excluded the IAB loop region. A unique
solution was found in the space group P3
1

simulated annealing, grouped and individual B-factor
refinement, and the final positional refinement. A flat
bulk-solvent model and overall anisotropic B-factor
correction were applied t hroughout the procedure. The
structure was refin ed to the final R-factor of 0.219 ( R
free
¼
0.280) with reasonable overall geometry, as monitored
with the program
PROCHECK
[28]. The refinement statistics
are summarized in Table 1. Coordinates and structure
factors have been deposited with the PDB under accession
number 1N00.
Figure preparation
Figures were prepared with
MOLSCRIPT
/
BOBSCRIPT
[29,30] using the
JAVA
application
BLUESCRIPT
for gener-
ating input scripts (A. Hofmann, unpublished data). The
objects created in such a manner were rendered with
POVRAY
[31].
Results and discussion
Crystallization

Number of independent reflections 27412
Completeness 100% (100%)
Multiplicity
a
7
R
merge
0.043 (0.433)
Refinement
No of reflections in working set/test set 24054 (3750)/
2656 (419)
Visible residues 4–321
Number of non-H atoms 2549
Solvent statistics: number of water
molecules/sulfate ions
181/3
R/R
free
b
0.219 (0.300)/
0.280 (0.377)
Average B-factor for all atoms (A
˚
2
) 51.9
Ramachandran plot: Residues
in most favoured/additionally
allowed/generously allowed region (%)
87.5/11.5/1.0
a

motifs seen in mammalian annexins (Table 2). The
intermodular salt bridge Glu113-Arg271 (IIB-IVB) is
conserved in both p lant annexins, as are the intramodular
salt bridges Asp93-Arg118 (IIB-IIC) and 276–280 (IVB-
IVC). Additionally, an interaction not seen in AnxA5 is
hydrogen bonding between CO117 and Arg276, thereby
tying together domains IIB, I VB and IVC.
The (artificially elongated) N-terminal tail consisting of
17 amino acids is visible in the current structure, apart from
the first three residues. The tail runs smoothly along the
concave surface of the protein and is anchored there by van
der W aals contacts (Leu9-Trp85) and by several hydrogen
bonds (CO10-His45, CO227-NH8, CO315-Thr8). In par-
ticular, the co ntact between CO10 and the iminium nitrogen
of His45, already identified in the structure of Anx24(Ca32),
seems to play an important role for the interaction between
core and N-terminal domain of plant annexins, since His45
is strictly conserved in t he plant subfamily.
As observed before with Anx24(Ca32), the globular
structure of Anx(Gh1), unlike that of mammalian annexins,
clearly shows separation of t he two modules ( I/IV and
II/III) leading to greater accessibility of the intermodular
space than in the case of AnxA5 and to formation of a
groove on the c onvex side (cf. Figure 1B). Located at the
entrance of the groove between domains III and IV is a
U-shaped, positively charged patch. The patch is formed
by five l ysine and three arginine residues ( Lys223, L ys226,
Arg238, Lys242, Lys249, Lys253, Arg256, and Arg291) and,
in the crystal structure, binds two sulfate ion s to compensate
for the excessive positive charge. The surface location in a

Table 2. Conservation of salt bridges.
Anx(Gh1) Anx24(Ca32) AnxA5
IE-IIA Arg80-Glu99 – –
IIB-IVB Glu113-Arg271 Glu116-Arg272 Glu112-Arg271
IIB-IVB CO117-Arg276 CO120-Arg277 –
IIA-IIB Asp39-Arg118 Asp96-Arg121 Asp92-Arg117
IVB-IVC Arg276-Asp280 Arg277–281 Arg276-Asp280
2560 A. Hofmann et al. (Eur. J. Biochem. 270) Ó FEBS 2003
Membrane binding loops
For reasons of homology, it is likely that the AB and DE
loops on the convex surface will serve as membrane binding
loops in the plant annexins, as was previously observed for
their mammalian relatives. In addition, the conservation of
aromatic and positively charged residues sticking out of the
convex surface (see Table 3) of plant annexins emphasizes
a possible functional role for membrane adsorption. Apart
from the loops IIIDE and the IVDE, all other membrane-
binding loops c arry conserved residues, which might either
interact with th e phospholipid headgroup or the glycerol
backbone region.
With respect to possible calcium binding in the membrane
loops, the recent crystal structure of Anx24(Ca32) raised the
question of how this might be accomplished by plant
annexins. As mentioned earlier, the endonexin sequence as a
constituent of canonical calcium binding sites in annexins is
conserved in domain I only and is present in a modified
form in the fourth domain. Despite extensive efforts, we
have not yet been successful in obtaining a calcium-bound
structure of Anx(Gh1) by soaking or cocrystallization
methods (data not shown). Analysis of possible molecular

respectively. While positioned adjacent to each other, both
side chains exist in the reduced (thiol) form, although
formation of a disulfide bridge is sterically possible (Fig. 3).
This is even more remarkable since the protein was never
kept under reducing conditions. S imilar s ituations have
been observed i n other proteins, such as the fatty acid
binding protein [32] and cyclophilins [ 33]. The electron
density in this region c learly shows no additional p eaks,
which would indicate a dithioether linkage between both
side chains. The torsion angles N–C
a
–C
b
–S of the two
residues are )62° for Cys116 and )75° for Cys243, and the
sulfur atoms are separated by 5.5 A
˚
. As verified by
molecular modelling, a simple rotation a round the C
a
–C
b
cysteine side chain bonds would enable formation of a
dithioether linkage (N–C
a
–C
b
–S ¼ 65° for Cys116 and 102°
Table 3. Conservation of surface-exposed residues. Residue s in italic
indicate lack of conservation.

. In their protonated forms, both
sulfhydryl groups interact with the methionine-S via
hydrogen bonding, establishing a 3S-2H topology with
almost tetragonal coordination on the methionine-S. This
S
3
cluster is located in the lower part of the annexin core in
module II/III and is acc essible only from the hydrophilic
cleft between modules I/IV and II/III, where Tyr250
provides shielding against direct interaction with solvent
molecules. Its plane is almost perpendicular to the S
3
plane
and the distance between the sulfur of Cys243 and the
tyrosine ring is 3.6 A
˚
.
While no experimentally proven chemical function of this
newly discovered cluster has been postulated so far, one can
easily imagine its involvement in the electron transfer
reactions. Oxidation of both cysteine residues to yield a
dithioether bond sets fr ee two electrons, w hich might be
donated to an oxidizing reagent, putatively a partly reduced
oxygen species. Hydrogen bonding of both sulfhydryl g roups
to the methionine certainly shifts the thiol-thiolate equili-
brium to the deprotonated side and therefore increases the
redox potential of the Cys2 system to more negative values.
Thus, Met112 acts as a factor to increase the red ox reactivity
of Cys116-Cys243. Tyr250 might be involved in these puta-
tive reactions by shuffling electrons from/to the S

Synthesis of b-1,4-glucan chains (cellulose) in plants requires
a chain elongation step during glucan polymerization,
which most likely is catalysed by cellulose synthase (CesA)
proteins. These proteins are components of plasma m em-
brane-bound CesA complexes w ith sixfold symmetry and
usually referred to as r osettes. Current models assume that
the active site of plant CesA proteins is formed on the
cytoplasmic face of t he plasma membrane by three Asp
Fig. 3. The sulfur cluster. Spatial arrangement of the S
3
cluster formed
by Met112, Cys116, and Cys243. The electron density shown was
calculated as omit map and is contoured at 1.5 r. Helices IIB and IIIE
are shown as Ca traces. Inset: T he distances between the individual
sulfur atoms are given in A
˚
.
Fig. 4. Amino acid sequence alignment. Aminoacidsequencesofdif-
ferent plant and mammalian annexins are aligned to show conserva-
tion of residues Met112, Cys116, Cys243, a nd Tyr250 of Anx(Gh1).
The sulfur-containing residues are marked red and the aromatic resi-
due (T yr or Phe) i s marked in cyan . All mammalian s equenc es shown
refer to the human proteins.
2562 A. Hofmann et al. (Eur. J. Biochem. 270) Ó FEBS 2003
residues together with a Q-X-X-R-W motif, both of which
are conserved. Eight transmembrane helices create a
channel through which the synthesized glucan chain is
secreted. The cytoplasmic N-terminal domain of CesA
proteins contains two zinc finger motifs, which recently have
been shown to bind zinc in a redo x-dependent manner (cf

crystal structure of Anx(Gh1) from cotton emphasizes the
high conservation of the unique annexin fold even among
the members of the plant subfamily of annexin proteins. The
fold is comprised of the arrangement o f four a-helical
domains into two modules, which are held together by polar
interactions. Despite this overall conservation, the fold
allows for subtle differences, s uch as t he generation of a
groove on the convex side of the plant proteins, w hich is
not observed with non-plant an nexins s ince the modules
are packed much tighter.
A comparison of the current structure of Anx(Gh1) with
the structures o f Anx24(Ca32) and AnxA5 reveals that the
cotton annexin, in contrast to the bell pepper protein,
provides canonical calcium binding sites in the first domain.
The observed conformation of the other domains does not
allow binding of divalent c ations. The molecular mechanism
of calcium binding of plant annexins requires further studies
and work aimed at investigation of this matter is currently in
progress. The crystallization behaviour of Anx(Gh1) and
the results obtained in this study are certainly promising for
succeeding in determination of a calcium-bound structure of
a plant annexin.
A feature of particular interest in Anx(Gh1) is the
occurrence of two adjacent cysteine r esidues in helices IIB
and IIIE, which are observed in the present structure in
their reduced states, although formation of a dithioether
bond is possible by simple rotation a round the C
a
–C
b

this mechanism experimentally will be required.
Acknowledgements
We thank Zbigniew Dauter (NCI and NSLS, Brookhaven National
Laboratory) for help with data collection on beamline X9B and Robert
O. Gould and Malcolm Walkinshaw for helpful discussions.
References
1. Behl, C., Davis, J., Lesley, R. & Schubert, D. (1994) Hydrogen
peroxide mediates amyloid beta protein toxicity. Cell 77, 817–827.
2. Edgington, S.M. (1994) As we live and breathe: free radicals and
aging. Correlative eviden ce from a number of fields su ggests they
may be key. Biotechnology 12, 37–40.
3. Chiueh, C. (2001) Iron overload, oxidative stress, a nd axonal
dystrophy in brain disorders. Pediatr. Neurol. 25, 138–147.
4. Gharavi, A.E., Pierangeli, S.S., Levy, R.A. & Harris, E.N. (2001)
Mechanisms of pregnancy l oss in antiphospholipid syndrome.
Clin. Obstet. Gynecol. 44, 11–19.
5. Rhee, H.J., Kim, G.Y., Huh, J.W., Kim, S.W. & Na, D.S. (2000)
Annexin I is a stre ss p rotein induced by h eat, o xidative stress and a
sulfhydryl-reactive agent. Eur. J. Biochem. 267, 3220–3225.
6. Sacre, S.M. & M oss, S.E. (2002) Intracellular localization of
endothelial cell annexins is differentially regulated by oxidative
stress. Exp. Cell Res. 274, 254–263.
7. Kovacs, I., Ayaydin, F., Oberschall, A., Ipacs, I., Bottka, S.,
Pongor,S.,Dudits,D.&Toth,E.(1998)Immunolocalizationofa
novel annexin-like protein encoded by a stress and abscisic acid
responsive gene in alfalfa. Plant J. 15, 185–197.
8. Gidrol, X., Sabelli, P., Fern, Y. & Kush, A. (1996) Annexin-like
protein from Arabidopsis thaliana rescues d elta oxyR mutant of
Escherichia coli from H
2

¨
misch, J. & Paques, E.
(1990) The calcium-binding sites in human annexin V by crystal
structure analysis at 2.0 A
˚
resolution. FEBS Lett. 275, 15–21.
17. Geisow, M., Fritsche, U., Hexham, J., Dash, B. & Johnson, T.
(1986) A con sensus amino acid s equence repeat in Torpedo and
mammalian calcium-dependent membrane binding proteins.
Nature 320, 636–638.
18. Jost, M., Weber, K. & Gerke, V. (1994) Annexin II contains two
types of Ca
2+
-binding sites. Biochem. J. 298, 553–559.
19. Hofmann,A.,Proust,J.,Dorowski,A.,Schantz,R.&Huber,R.
(2000) Annexin 24 from Capsicum annuum. X-ray structure and
biochemical characterization. J. Biol. Chem. 275, 8072–8082.
20. Hofmann,A.,Ruvinov,S.,Hess,S.,Schantz,R.,Delmer,D.P.&
Wlodawer, A. (2002) Plant annexins form calcium-independent
oligomers in solution. Protein Sci. 11, 2033–2040.
21. Otwinowski, Z. & Minor, W. (1997) Processing of X-ray diffrac-
tion data collected in oscillation mode. Methods Enzymol. 276,
307–326.
22. Tong, L. & Rossmann, M. (1990) The locked rotation function.
Acta Cryst. A46, 783–792.
23. Navaza, J. (1994) AMoRe: an automated package for molecular
replacement. Acta Crystallogr. A50, 157–163.
24. Matthews, B. (1986) Solvent content of protein crystals. J. Mol.
Biol. 33, 491–497.
25. Chen, Z., Baruch, P., Mathews, F., Matsushita, K., Yamashita,

32. Hohoff, C., Bo
¨
rchers, T., Ru
¨
stow,B.,Spener,F.&vanTilbeurgh,
H. (1999) Expression, purification, and crystal structure determi-
nation of recombinant human epidermal-type fatty acid binding
protein. Biochemistry 38, 12229–12239.
33. Dornan, J., Page, A., Taylor, P., Wu, S., Winter, A., Husi, H. &
Walkinshaw, M. (1999) Biochemical and structural characteriza-
tion of a divergent loop cyclophilin from Caenorhabditis elegans.
J. Biol. Chem. 274, 34877–34883.
34. Burger, A., Berendes, R., Liemann, S., Benz, J., Hofmann, A.,
Go
¨
ttig,P.,Huber,R.,Gerke,V.,Thiel,C.,Ro
¨
misch, J. & Weber,
K. (1996) The crystal structure and ion channel activity of human
annexin II, a peripheral mem brane protein. J. Mol. Biol. 257,
839–847.
35. Bru
¨
nger, A. (1992) Free R value: a novel statistical quantity for
assessing the accuracy of crystal structures. Nature 355, 472–475.
36. Nicholls, A., Bharadwaj, R. & Honig, B. (1993) GRASP: Gra-
phical representation and analysis of surface properties. Biophys.
J. 64, A166.
2564 A. Hofmann et al. (Eur. J. Biochem. 270) Ó FEBS 2003