Content uploaded by James Barber
Author content
All content in this area was uploaded by James Barber on May 15, 2014
Content may be subject to copyright.
Architecture of the
Photosynthetic Oxygen-Evolving
Center
Kristina N. Ferreira,
1
* Tina M. Iverson,
2
* Karim Maghlaoui,
1
James Barber,
1
† So Iwata
1,2,3
†
Photosynthesis uses light energy to drive the oxidation of water at an oxygen-
evolving catalytic site within photosystem II (PSII). We report the structure of
PSII of the cyanobacterium Thermosynechococcus elongatus at 3.5 angstrom
resolution. We have assigned most of the amino acid residues of this 650-
kilodalton dimeric multisubunit complex and refined the structure to reveal its
molecular architecture. Consequently, we are able to describe details of the
binding sites for cofactors and propose a structure of the oxygen-evolving
center (OEC). The data strongly suggest that the OEC contains a cubane-like
Mn
3
CaO
4
cluster linked to a fourth Mn by a mono--oxo bridge. The details of
the surrounding coordination sphere of the metal cluster and the implications
for a possible oxygen-evolving mechanism are discussed.
Photosynthesis uses light energy to couple
the formation of molecular dioxygen to the
fixation of CO
2
. This process simultaneously
generates an aerobic atmosphere and produc-
es a readily usable carbon source, both of
which act to sustain almost all life on this
planet. Central to this process is photosystem
II (PSII), which catalyzes one of the most
thermodynamically demanding reactions in
biology: the photoinduced oxidation of water.
In this way it provides a primary source of
reducing equivalents (water-derived electrons
and protons), which, with an additional input
of energy from photosystem I (PSI), are used
to convert carbon dioxide to biomass, food,
and fuel. Molecular dioxygen is released as a
by-product. Given the essential role of PSII in
maintaining the biosphere, it is of consider-
able importance to elucidate its catalytic
mechanisms, particularly those involved in
the water oxidation process.
Much knowledge of the catalytic mecha-
nisms of PSII has been obtained from a wide
range of approaches (1–3), but the precise
molecular details of water oxidation cata-
lyzed by the oxygen-evolving center (OEC)
remain elusive. Two x-ray studies at 3.8 Å
(4) and 3.7 Å (5) have given the first direct
hints as to the structure of PSII, but these
models revealed neither the complete details
of the OEC nor the surrounding protein res-
idues, information vital for formulating a re-
action mechanism for water oxidation.
Here we report the crystal structure of a
cyanobacterial PSII complex at 3.5 Å res-
olution with more than 90% of its amino
acid residues traced. We have assigned all
the subunits of the PSII complex to specific
gene products and provide a description of
the protein environment of the various
redox-active cofactors and pigments within
this complex. We conclude that the OEC is
a cubane-like Mn
3
CaO
4
cluster with a
mono--oxo bridge to a fourth Mn ion. On
the basis of our proposed structure of the
metal center and its protein environment,
we discuss the possible mechanism of the
oxygen-evolving reaction.
Structure determination and overall
architecture. Dimeric PSII complexes were
isolated from Thermosynechococcus elongatus.
Details of sample preparation, crystallization,
and structure determination are provided in (6).
The molecular replacement trials using the co-
ordinates of earlier works [1FE1, 3.8 Å resolu-
tion, C
␣
atoms only, R factor 59% (4); and
1IZL, 3.7 Å resolution, mainly polyalanine, R
factor 53% (5)] produced density maps that
were highly biased to their models and there-
fore could not be used. For this reason, the
experimental phases for the structure presented
here were calculated by the method of multiple
isomorphus replacement with six heavy atom
derivatives (Table 1). After fitting to the exper-
imental map, the model was refined at 3.5 Å
resolution (Table 1 and fig. S1).
The crystallographic asymmetric unit con-
tains a dimer of PSII, where the two mono-
mers in the dimer are almost identical. In the
model, the monomer contains 19 protein sub-
units that have been clearly assigned to elec-
tron density, with the exception of the tenta-
tive assignment of PsbN, where the side
chains are not included in the model as a
result of disorder. Each monomer contains 36
chlorophyll a (Chl) and 7 all-trans carote-
noids assumed to be -carotene molecules.
However, two of the Chl molecules (Chl
35
and Chl
36
) are loosely attached to the com-
plex and disordered. Additionally, there is
unassigned electron density remaining that
could be carotenoid molecules or the fatty
acid tails of lipid molecules. Each monomer
model also includes one OEC, one heme b,
one heme c, two plastoquinones, two pheo-
phytins, one nonheme Fe, and two bicarbon-
ates (one is tentatively assigned as an un-
known nonprotein ligand at the OEC).
The PS II dimer (Fig. 1) has dimensions of
105 Å depth (45 Å in membrane), 205 Å
length, and 110 Å width. The overall structure
(Fig. 1) is similar to those reported previously
(4, 5); the root mean square (RMS) deviations
of the C
␣
backbone between the dimer structure
and 1FE1 (4) and 1IZL (5) are 2.1 Å (for 1359
C
␣
atoms) and 1.9 Å (for 3519 C
␣
atoms),
respectively. However, compared with previous
models (4, 5), the sequence assignment has
been substantially improved. We have newly
assigned or reassigned 3916 residues and built
the side chains. The improvement is evident in
the models for the extrinsic subunits, extrinsic
domains of D1, D2, CP43, and CP47, and the
small transmembrane subunits.
Within the dimer, the monomers are related
by a noncrystallographic twofold axis perpen-
dicular to the membrane plane. Our model of
PSII consists of 16 integral membrane subunits
composed of 35 transmembrane helices and
three lumenal subunits. The monomer is char-
acterized by pseudo-twofold symmetry, which
relates the D1, CP47, and PsbI subunits to the
D2, CP43, and PsbX subunits (Fig. 1B).
Protein subunits. The D1 and D2 sub-
units form the center of the PSII monomer
complex. Each of these subunits comprises
five transmembrane helices (A to E) orga-
nized in a manner almost identical to that of
the L and M subunits of the reaction center of
photosynthetic purple bacteria (bRC) (7, 8),
with RMS deviation between PSII and bRC
from Rhodopseudomonas viridis of 1.9 Å for
395 C
␣
atoms. However, the C-terminal do-
mains and the loops joining the transmem-
brane helices are more extended in the case of
the D1 and D2 subunits compared with bRC,
especially on the lumenal side close to the
OEC. Flanking the opposite sides of the D1/
D2 heterodimer are the CP43 (PsbC) and
CP47 (PsbB) subunits, each having six trans-
1
Department of Biological Sciences, Imperial College
London, London, SW7 2AZ, UK.
2
Division of Biomed-
ical Sciences, Imperial College London, London, SW7
2AZ, UK.
3
ATP System Project, ERATO, Japan Science
and Technology Corporation, 5800-3 Nagatsuta,
Midori-ku, Yokohama 226-0026, Japan.
*These authors contributed equally to this work.
†To whom correspondence should be addressed. E-
mail: s.iwata@imperial.ac.uk (S.I.); j.barber@imperial.
ac.uk (J.B.)
RESEARCH ARTICLES
www.sciencemag.org SCIENCE VOL 303 19 MARCH 2004 1831
membrane helices (I to VI) arranged in a
circular manner similar to the N-terminal do-
mains of the PsaA and PsaB proteins of PSI
(9). The transmembrane helices and chloro-
phyll molecules of each of these subunits
exhibit an internal pseudo-threefold symme-
try. Violating this internal symmetry in each
subunit is a large lumenal domain between
transmembrane helices V and VI. The large
domain of CP43 contains two long and three
short helices (fig. S2). In the case of CP47,
the lumenal domain contains two long and
four short helices and three  sheets (fig. S3).
Outside the pseudo-symmetric CP43/D1–
D2/CP47 core, there are 13 transmembrane he-
lices in the map, which are assigned to specific
low-molecular-weight subunits based on trac-
ing their amino acid sequences. (Fig. 1B). The
assignment of three of these helices, PsbE,
PsbF, and PsbK, where PsbE and PsbF are the
␣ and  subunits, respectively, of cytochrome
b559 (Cyt b559), agrees with earlier studies (4,
5, 10). The remaining 10 transmembrane heli-
ces are newly assigned to 9 low-molecular-
weight subunits, including PsbZ, which has two
transmembrane helices (I and II). One trans-
membrane helix reported previously (4, 5) ad-
jacent to Cyt b559, and designated PsbX (4),
was not present in our electron density map.
With the exception of Cyt b559, the functions of
most of these small subunits are unclear. Our
assignments, however, indicate that PsbL, PsbM,
and PsbT are involved in dimer formation. The
symmetrically related PsbI and PsbX proteins
stabilize the peripheral chlorophylls of the D1 and
D2 proteins (Chl
ZD1
and Chl
ZD2
). Four small
subunits (PsbJ, PsbK, putative PsbN, and PsbZ)
adjacent to CP43 may facilitate carotenoid bind-
ing, because four of the seven assigned -carotene
molecules are found in their vicinity (Fig. 1B).
Single copies of each of the three extrinsic
proteins, PsbO, PsbU, and PsbV, are located on
the lumenal surface (Fig. 1A and fig. S4). Tak-
en together, these extrinsic proteins and the C
terminus of the D2 protein form a “cap” over
the OEC. None are involved directly in the
ligation of the OEC. However, PsbO, which
forms an eight-stranded  barrel (fig. S4A),
stabilizes the backbone conformation of the AB
loop and C terminus of the D1 protein, which
provide the majority of the ligands to the OEC.
Additionally, our sequence assignment of the
PsbO subunit revealed that a large loop be-
tween strands 5 and 6 forms a part of a hydro-
philic pathway connecting the OEC with the
lumenal surface. Although the PsbO  barrel
resembles the  barrel of porins (5), it is not an
open tube but contains a number of bulky hy-
drophobic side chains.
Electron transfer. The reactions of PSII
are powered by light-driven primary and sec-
ondary electron transfer processes across the
reaction center (RC), composed of the D1 and
D2 subunits. Figure 2A shows the cofactors
involved in these electron transfer reactions.
Upon illumination, an electron is ejected
from the excited primary electron donor P680, a
Chl located toward the lumenal surface (P
D1
in
Fig. 2A is likely to be P680, as discussed
Table 1. Data collection, refinement, and phasing statistics for PSII structure determination. SLS, Swiss Light Source; ESRF, European Synchrotron Radiation
Facility; EMP, ethylmercury phosphate; PCMBS, p-chloromercuribenzoylsulfonate.
Data collection and phasing
Data set Native
Native
(Mn-edge)
Native
(for Ca detection)
CdCl
2
K
3
[Au(Cn)
6
]
Beam line SLS X06SA SLS X06SA SLS X06SA ESRF ID29 ESRF ID29
Wavelength (Å) 1.0076 1.89340 2.25430 1.0052 1.0052
Resolution (Å) 40.0–3.5 40.0–3.8 40.0–3.8 40.0–3.8 40.0–3.8
Total observation 298,731 168,066 212,461 245,580 221,188
Unique reflections 103,604 71,063 79,041 78,778 73,595
Completeness (%)* 87.3 (80.9) 74.6 (63.8) 83.6 (62.3) 82.0 (55.1†) 77.8 (61.7†)
Redundancy 2.88 2.37 2.68 3.11 3.00
R
sym
*‡ 0.08 (0.43) 0.11 (0.50) 9.0 (0.55) 0.08 (0.43) 0.09 (0.43)
I/(I)* 10.4 (2.0) 7.3 (2.2) 7.4 (2.1) 9.9 (1.6) 9.0 (2.0)
Phasing power§ N/A N/A N/A 0.65 0.53
Data set EMP1 EMP2 PCMBS TaBr
6
Beam line SLS X06SA SLS X06SA SLS X06SA SLS X06SA
Wavelength (Å) 0.96863 0.96863 0.96863 0.96863
Resolution (Å) 40.0–4.1 40.0–5.1 40.0–3.8 40.0–3.8
Total observation 215,331 68,313 179,833 172,328
Unique reflections 80,342 26,529 75,502 71,395
Completeness (%)* 73.0 (58.5†) 64.4 (48.9†) 73.5 (31.9†) 74.6 (53.8†)
Redundancy 2.68 2.57 2.38 2.41
R
sym
*‡ 0.10 (0.40) 0.08 (0.44) 0.12 (0.50) 0.08 (0.33)
I/(I)* 6.1 (1.3) 11.6 (1.8) 5.9 (2.1) 11.2 (2.6)
Phasing power§ 0.90 0.41 0.57 0.62
Refinement
Resolution (Å)* 20.0–3.5 (3.56–3.50)
Number of reflections 103,485 (85.7%)
R factor*㛳 0.303 (0.340)
R
free
*¶ 0.346 (0.384)
Average B values (Å
2
) 74.2
RMS deviations from ideal values
Bond length (Å) 0.014
Bond angles (
0
) 2.11
Dihedral angles (
0
) 19.4
Improper torsion angles (
0
) 2.04
*Values in parentheses are for the highest resolution shell. †The completeness of the highest shell is decreased from the overlap of the spots caused by the increased mosaicity
pronounced in heavy atom soaking experiments. ‡R
sym
⫽ ¥
h
¥
i
|I
i
(h) – ⬍I(h)⬎|/¥
h
¥
i
(h), where I
i
(h) is the ith measurement. §Phasing power is the RMS value of F
h
divided by
the RMS lack-of-closure error. 㛳R factor ⫽ ¥
h
||F(h)
obs
|–|F(h)
calc
||/¥
h
|F(h)|. ¶R
free
was calculated for 1% of reflections randomly excluded from the refinement.
R ESEARCH A RTICLES
19 MARCH 2004 VOL 303 SCIENCE www.sciencemag.org1832
below). The electron is quickly transferred to-
ward the stromal surface to the final electron
acceptor, plastoquinone Q
B,
through Chl
D1
,
pheophytin (Pheo
D1
), and plastoquinone Q
A
.
After accepting two electrons and undergoing
protonation, Q
B
is released from PSII into the
membrane matrix. The cationic radical P680
•⫹
is reduced by a redox-active tyrosine, known as
Tyr
Z
(Tyr
161
of D1 subunit), to generate a
neutral tyrosine radical Tyr
Z
•
, which acts as an
oxidant for the water oxidation process at the
OEC. The oxidation of two water molecules to
produce a dioxygen proceeds in a stepwise
manner as described in the S-state cycle (11).
With the exception of Q
A
, all the redox-active
cofactors involved in the electron transfer pro-
cesses are located on the D1 side of the RC.
The radical cation P680
•ⴙ
has a very high
oxidizing potential, recently estimated to be 1.3
to 1.4 V (12), which is required for the water-
splitting reaction. This contrasts with 0.4 V
produced by the bacterial equivalent consis-
ting of a “special pair” of bacterio-
chlorophylls (BChl), which is reduced by a
cytochrome after photooxidation. The two chlo-
rophyll molecules in PSII equivalent to the
special pair of the bRC are denoted P
D1
and P
D2
(4). Although the overall organization of P
D1
and P
D2
is very similar to the special pair, the
relative orientation angle between their tetra-
pyrrole heads differs slightly from their bacte-
rial equivalents. The head groups of P
D1
and
P
D2
are in direct van der Waals contact, with a
Mg-Mg distance of 8.2 Å (Fig. 2B). This is
slightly longer than the 7.5 Å observed for the
special pair of bRC, which could, together with
the difference in the head-group orientation,
explain why P680 shows more monomeric
character and weaker electronic coupling than
its bacterial counterpart (13).
P
D1
and P
D2
are located close to Chl
D1
and
Chl
D2
(Fig. 2B), which are equivalent to the
accessory BChls of the bRC. The P680 excited
state is delocalized over the four chlorophylls
(14, 15), and Chl
D1
, which is the chlorophyll
closest to the active Pheo
D1
, is involved in the
initial primary charge separation (16, 17). Un-
like accessory BChls in the bRC, which are
ligated by histidines, there is no protein ligand
for Chl
D1
and Chl
D2
; the closest residue to
Chl
D1
is D1 Thr
179
and to Chl
D2
is D2 Ile
178
.
The electrons are transferred from Pheo
D1
to Q
A
, which is a firmly bound plastoquinone.
Even though PSII and bRCs use different pri-
mary quinone acceptors, the Q
A
binding pock-
ets are structurally similar. In PSII, the Q
A
is
hydrogen bonded to the main-chain amide
group of D2 Phe
261
and D2 His
214
, where the
latter also serves as a ligand to the nonheme Fe
(Fig. 3, A and B). The quinone ring is accom-
modated within a hydrophobic cavity com-
posed of residues, including D2 Ile
213
,D2
Thr
217
,D2Met
246
,D2Ala
249
,D2Trp
253
,D2
Ala
260
, and D2 Leu
267
.
The nonheme Fe, which mediates electron
transfer from Q
A
to Q
B
, is positioned on a
pseudo-twofold axis of the D1/D2 heterodimer.
Like the bRC, the nonheme Fe has four histi-
dine ligands: D1 His
215
,D1His
272
,D2His
214
,
and D2 His
268
(Fig. 3A). In bRC, a glutamate
serves as a fifth ligand, whereas in PSII this
ligand is not from the protein. It has been
suggested that bicarbonate may function as the
fifth ligand to the nonheme Fe in PSII (18) and
that bicarbonate has a regulatory function in-
volving electron flow from Q
A
to Q
B
as well as
facilitating the protonation of Q
B
(2, 19). In our
PSII structure, nonheme Fe is associated with
an electron density that is sufficient to accom-
Fig. 1. Overall structure of PSII. (A) View of the PSII dimer perpendicular to the membrane
normal. Helices are represented as cylinders with D1 in yellow; D2 in orange; CP47 in red; CP43
in green; cyt b559 in wine red; PsbL, PsbM, and PsbT in medium blue; and PsbH, PsbI, PsbJ, PsbK,
PsbX, PsbZ, and the putative PsbN in gray. The extrinsic proteins are PsbO in blue, PsbU in
magenta, and PsbV in cyan. Chlorophylls of the D1/D2 reaction center are light green,
pheophytins are blue, chlorophylls of the antenna complexes are dark green, -carotenes are
in orange, hemes are in red, nonheme Fe is red, Q
A
and Q
B
are purple. The oxygen-evolving
center (OEC) is shown as the red (oxygen atoms), magenta (Mn ions), and cyan (Ca
2⫹
) balls.
(B) View of the PSII monomer along the membrane normal from the lumenal side. A part of
the other monomer in the dimmer is shown to emphasize the region of monomer/monomer
interaction along the dotted line. The pseudo-twofold axis perpendicular to the membrane
plane passing through the nonheme Fe relates the transmembrane helices of the D1/D2
heterodimer, the low molecular subunits, PsbI and PsbX, and CP43 and CP47 as emphasized by
the black lines encircling these subunits. Coloring is the same as in (A).
R ESEARCH A RTICLES
www.sciencemag.org SCIENCE VOL 303 19 MARCH 2004 1833
modate this anion. Close to this nonprotein den-
sity are D1 Tyr
246
and D2 Lys
264
, positioned and
oriented such that they could stabilize the bicar-
bonate by hydrogen bonding (Fig. 3A).
There is less conservation between the Q
B
sites of PSII and bRC than for the Q
A
sites.
There are similarities in that the PSII Q
B
is
hydrogen bonded to D1 Ser
264
and D1 His
215
,
which is also a ligand for the nonheme Fe, and
possibly to the main-chain amide group of D1
Phe
265
. However, the Q
B
sites in PSII and the
bRC show a 3.5 Å displacement when super-
imposed using C
␣
atoms of D1/D2 and the L/M
subunits. This is because there is a conforma-
tional difference in the loops containing D1
Ser
264
and the equivalent in bRC, which is
caused by the insertion of one residue. As a
consequence, the volume of the Q
B
binding
pocket, composed of the residues including D1
Met
214
,D1Leu
218
,D1Ala
251
,D1Phe
255
, and
D1 Leu
271
, is slightly larger in PSII than in the
bRC. This could explain the difference in her-
bicide specificity between PSII and bRC (20).
In addition, the proton pathway connecting the
Q
B
site to the stromal space is considerably
different in PSII compared with bRC. The PSII
Q
B
site is much closer to the surface than in
bRC because of the absence of an equivalent to
the bRC H subunit. Because D1 His
252
is with-
in hydrogen-bonding distance of D1 Ser
264
, this
residue could aid Q
B
protonation (Fig. 3).
The antenna system and carotenoids. We
have identified 14 and 16 Chls bound to the
transmembrane regions of CP43 and CP47, re-
spectively (Fig. 4). Although most Chls are
arranged in two layers on opposite sides of the
membrane, two (Chl
9
of CP43 and Chl
20
of
CP47) are positioned at almost an equal dis-
tance from both surfaces. As a result of this,
they form a stack with adjacent Chls positioned
on the stromal and lumenal sides (Fig. 4A). The
function of these stacked Chls is not clear, but
they may facilitate the fast energy transfer
known to occur within CP43 and CP47 and
could be the origin of the long-wavelength–
absorbing Chls of these proteins (21).
Many Chls in CP43 and CP47 are approx-
imately related by an internal pseudo-twofold
axis of the PSII monomeric complex and, with-
in each protein, Chls are arranged around the
internal pseudo-threefold axis. Ten Chls of
CP43 and 13 Chls of CP47 are ligated by
histidine, and one in CP43 has an asparagine
ligand. Both proteins have two Chls in equiva-
lent positions associated with either methionine
or serine side chains. These side chains are not
close enough to form direct ligands but could
facilitate such an interaction by means of bound
water molecules. The Mg
2⫹
of two loosely
attached Chls (Chl
35
and Chl
36
) are close to the
main-chain carbonyl oxygen and cysteine side
chain, respectively, but it is not clear whether
these groups are ligands because of disorder.
The overall organization of Chls in PSII
differs considerably from that in PSI. In PSI,
the N-terminal domains of PsaA and PsaB are
equivalent to CP43 and CP47, and the C-
terminal domains of PsaA and PsaB are equiv-
alent to D1 and D2 subunits (22). The distribu-
tion of the peripheral antenna Chls of PSI
bound to the N-terminal domains of PsaA and
PsaB is similar to that for CP43 and CP47.
However, the C-terminal domains of PsaA and
PsaB, together with other subunits, coordinate
43 Chls, which form a central antenna sur-
rounding the electron transfer system of the PSI
RC. This arrangement is notably different from
PSII, which only coordinates two peripheral RC
Chls (Chl
ZD1
and Chl
ZD2
). Figure 4B clearly
shows that the locations of these Chls are not
optimized to mediate energy transfer from
CP43 and CP47 to the RC as is the case for the
central antenna system of PSI. This difference
probably explains the well-known slow trap-
ping of excitation energy in PSII compared
with PSI (21). However, Chl
ZD1
and Chl
ZD2
may have other functions, including protection
of PSII against photoinduced damage (23).
Seven -carotenes are assigned to the
density, although it is possible that there are
more as indicated by biochemical analyses
(24); potential -carotene densities are ob-
served at the D1/CP43 and the dimer inter-
faces. -carotene can be photooxidized when
water splitting is inhibited with a msec time
constant (25, 26), which suggests that at least
one carotenoid must be positioned about 20 Å
from P680. Moreover, there is spectroscopic
evidence indicating that -carotene facilitates
long-distance electron flow from Cyt b559 to
P680
•⫹
(27) and from Chl
ZD2
or Chl
ZD1
(28,
29). According to our assignment, the head
group of one all trans–-carotene is in direct
contact with Chl
ZD2
and is located between
Cyt b559 and the RC Chls (Figs. 2A and 4B).
It is therefore likely that this -carotene is
involved in electron transfer from Cyt b559
and Chl
ZD2
to P680, a conclusion also made
by Kamiya and Shen, who placed one all
trans– and one cis–-carotene in a similar
position in their structure (5).
Of the remaining assigned carotenoids,
four are positioned between Cyt b559 and
CP43, possibly stabilized by the cluster of
small transmembrane subunits in that region,
and two are located on the CP47 side. The
electron density map also indicates that CP47
may contain 2 to 3 -carotenes close to the
monomer/monomer interface in the dimer.
Six of the seven assigned -carotene mole-
Fig. 2. Cofactors involved in electron transfer. (A) Electron transfer
cofactors shown perpendicular to the internal pseudo-twofold. Coloring
scheme is the same as in Fig. 1. The phytol tails of the chlorophylls and
pheophytins have been removed for clarity. The side chains of Tyr
Z
(D1
Tyr
161
) and D1 His
190
are shown in yellow, and Tyr
D
(D2 Tyr
160
) and D2
His
189
are in orange. The four chlorophylls comprising P680 are in direct
van der Waals contact, and other electron transfer distances are given in
Å. (B) The P680 dimer of chlorophylls (P
D1
and P
D2
) and accessory Chls
(Chl
D1
and Chl
D2
). Coloring scheme is the same as in Fig. 1, except that
the protein main chain is depicted in light gray, whereas the side-chain
bonds and carbon atoms follow the coloring of the protein subunits (D1,
yellow; D2, orange). The histidine ligands D1 His
198
and D2 His
197
are
shown, as well as the redox-active Tyr
Z
–D1 His
190
and Tyr
D
–D2 His
189
pairs. The view is down the pseudo-twofold axis from the stromal side.
R ESEARCH A RTICLES
19 MARCH 2004 VOL 303 SCIENCE www.sciencemag.org1834
cules are in close contact with Chl head
groups, as required for facilitating energy
transfer from -carotene to Chl and for
quenching Chl triplets. Some of these carote-
noids, and perhaps others not yet assigned,
may serve to quench singlet oxygen known to
be produced by P680 triplets formed by the
recombination of P680
•⫹
Pheo
•⫺
(30).
The oxygen-evolving center. The den-
sity assigned to the OEC, consisting of a
large globular domain connected to an ex-
tended region (Fig. 5A), is located close to
the surface of the CD lumenal helix of the
D1 subunit. The location and shape of the
density are similar to that reported in pre-
vious x-ray studies and interpreted to have
three Mn ions in the larger domain and one
in the extended region (4, 5). However, in
our electron density map, it is clear that
four metal ions the size of Mn can be
accommodated at the corners of a tetrahe-
dron in the large globular density, whereas
one metal ion is located in the center of the
extended region. This arrangement is
strongly supported by the difference densi-
ty map (simulated-annealing omit map),
which clearly shows the peak for the re-
spective metal when each metal atom of the
cluster is omitted from the model (fig.
S5A). One of the five metal ions is likely to
be Ca
2⫹
, which is associated with the OEC
(2). We have identified the metals using
x-ray anomalous difference maps at the Mn
absorption edge (1.89 Å) and at 2.25 Å
wavelength, where Ca
2⫹
has an anomalous
difference (f ⬙ ) 3.9 times as strong as Mn
(Fig. 5A and fig. S5B). The Mn anomalous
map only correlates with one metal in the
small domain and three in the large globu-
lar domain, whereas the 2.25 Å wavelength
map covers the remaining one metal ion in
the large domain. On the basis of the ap-
proximate positions between the metal ions
and their coordination properties, we sug-
gest that the OEC is a cubane-like
Mn
3
CaO
4
cluster with each metal ion in
this cluster having three--oxo bridges (the
large domain) connected to another Mn ion
by a mono--oxo bridge in the extended
region (Fig. 5). In our model of the OEC,
the metal-to-metal distances within the cu-
bane-like cluster are ⬃2.7 Å for Mn-Mn
Fig. 3. Electron acceptor side of
PSII. (A) Overall view of the
nonheme iron, Q
A
and Q
B
. Col-
oring scheme is as in Fig. 1, with
protein main chains depicted in
gray and with side-chain bonds
and carbon atoms following the
coloring of the protein subunit
as used in Fig. 1. The bicarbon-
ate completing the coordination
sphere of the nonheme Fe is
shown with magenta bonds and
is probably hydrogen bonded to
D2 Lys
264
and D1 Tyr
246
.(B) The
Q
A
binding pocket. The hydro-
phobic residues forming this
pocket are shown. The O
1
of the plastoquinone head group is likely to be
hydrogen bonded to the nonheme Fe ligating D2 His
214
by its ␦-nitrogen,
whereas the O
4
atom may hydrogen bond to the backbone amide nitrogen
of D2 Phe
261
.(C) The Q
B
binding pocket. Q
B
binds deep into a cavity lined
with the hydrophobic residues. O
1
is likely to be hydrogen bonded to the
␦-nitrogen of D1 His
215
, which also forms a ligand to the nonheme Fe,
whereas O
4
may form hydrogen bonds with the amide nitrogen of D1
Phe
265
and the side chain ␥-oxygen of D1 Ser
264
.D1Ser
264
appears to make
further hydrogen-bonding contact with D1 His
252
. Probable hydrogen bonds
are shown as dotted lines; solid lines represent ligands.
Fig. 4. Pigment organization in the PSII monomer complex. (A) Energy transfer in the PSII monomer complex.
View is perpendicular to the membrane normal. Chlorophylls (green from RC and dark green from antenna)
have the phytol tail omitted for clarity. Chlorophylls that are stacked with the planes of the porphyrin head
group parallel and within van der Waals contact are highlighted in red. -carotenes (orange) are shown as
ball-and-stick models and tend to have one head group in direct van der Waals contact with a chlorophyll.
(B) View of the pigments along the pseudo-twofold axis, perpendicular to the view in Fig. 2A.
R ESEARCH A RTICLES
www.sciencemag.org SCIENCE VOL 303 19 MARCH 2004 1835
and ⬃3.4 Å for Mn-Ca
2⫹
, which are
typical for the oxo bridges proposed and
compatible with distances derived from
extended x-ray absorption fine structure
(EXAFS) studies (31). However, higher
resolution data will be required to deter-
mine the precise distances and to investi-
gate whether some of the bridging oxygen
atoms are protonated. The distance between
the Mn ion (Mn
4
) in the small domain and
the two Mn ions (Mn
2
and Mn
3
)inthe
cubane-like cluster is ⬃3.3 Å, which is a
typical distance for a mono--oxo bridge
between Mn ions and also is in line with
EXAFS measurements (31).
Although the arrangement of Mn ions in
our model fits one of the possible models
derived from EXAFS (31), a cubane-like
Mn
3
CaO
4
cluster linked to a fourth Mn by
a mono--oxo bridge as proposed here has
not been specifically suggested. However,
a regular cubane structure for the Mn
4
clus-
ter was proposed some time ago (32), and
recently Dismukes and colleagues have
given evidence for a distorted cubane core
for higher S states (33), in line with an
earlier paper (34). The 3 ⫹ 1 arrangement
for the Mn ions has been proposed by
Peloquin and Britt (35).
The electron density shows that the
Mn
3
CaO
4
cluster has four side chains as
ligands: D1 Asp
342
for Mn
1
,D1Glu
189
and
D1 His
332
for Mn
2
, and CP43 Glu
354
for
Mn
3
(fig. S5, C and D, and Fig. 5, A and
B). Additionally, the carboxyl group of C-
terminal D1 Ala
344
is located close to
Ca
2⫹
, although in the current density map
these are not connected (fig. S1C). This
could, however, be a possible ligand of
Ca
2⫹
at some stage during the S-state cy-
cle. D1 His
337
seems to interact with the
cluster by a hydrogen bond to the O2 oxy-
gen of the cubane-like structure (Fig. 5B).
Mn
4
, located in the small domain, has two
direct protein ligands, D1 Asp
170
and D1
Glu
333
, with a third residue D1 Asp
61
pos-
sibly interacting through a water molecule.
Furthermore, between this metal and Ca
2⫹
,
a nonprotein density indicative of an anion-
ic ligand is observed (Fig. 5A and fig.
S5D). According to the coordination geom-
etry of Ca
2⫹
and Mn and difference Fourier
studies, we tentatively fit a bicarbonate
molecule to this density, where it seems to
be acting as a tridentate bridge between
Mn
4
and Ca
2⫹
; bicarbonate has been sug-
gested to play a role in the assembly of the
OEC (36). It is very likely that this is the
water oxidation site and, in the active state,
this nonprotein ligand is replaced by water
molecules, X
1
for Mn
4
, and X
21
and X
22
for
Ca
2⫹
(Fig. 5B), although one of the Ca
2⫹
ligands could be a Cl
–
, which is a cofactor
for the OEC (1–3). The coordination num-
ber is usually six or seven for Ca
2⫹
and six
for Mn. It is possible to fulfill these coor-
dination patterns assuming water molecules
or hydroxide are associating with the Mn
and Ca
2⫹
in addition to the ligands men-
tioned above. Although water or hydroxide
cannot be observed at the current resolu-
tion, the proposed OEC model can accom-
modate these molecules without conflicting
with other atoms in the cluster. The hydrox-
ide ions could play a role in maintaining the
electroneutrality of the cluster.
The assignment of the metal ligands is
consistent with a wide range of mutational
studies (37, 38). The involvement of the
C-terminal domain of the D1 protein in the
assembly and stabilization of the OEC was
first emphasized by Diner and colleagues
(39), followed by several studies suggest-
Fig. 5. The oxygen-evolving center (OEC). (A) Stereo view of the OEC with side-chain ligands
and possible catalytically important side-chain residues. Mn ions, Ca
2⫹
, and oxygen atoms are
shown in magenta, cyan, and red, respectively. One unidentified nonprotein ligand to the OEC
is colored in green. The protein main chain is depicted in light gray; the side-chain bonds and
carbon atoms follow the coloring of the protein subunits (D1, yellow; CP43, green).
A
weighted 2兩F
o
兩 – 兩F
c
兩 density is shown as a light-blue wire mesh contoured at 1.5. Anomalous
difference Fourier maps at 1.89340 Å (Mn edge, contoured at 10) and 2.25430 Å (highlights
Ca
2⫹
, contoured at 7) wavelengths are shown in magenta and blue-green, respectively. (B)
The same as (A) but with a rotation around the y axis of 40° and without electron density and
anomalous difference maps. (C) Schematic view of the OEC. Residues in D1, D2, and CP43
subunits are shown in yellow, orange, and green, respectively. X
11
,X
21
, and X
22
are possible
substrate water binding positions to Mn
4
(X
11
) and to Ca
2⫹
(X
21
and X
22
), identified from the
position of the nonprotein ligand and coordination pattern of Mn and Ca
2⫹
ions. Possible
water molecules, which are not visible at the current resolution, are indicated as W. Possible
hydrogen bonds are shown as light-blue dotted lines. (D) Hydrophilic pathway between the
active site and lumen (blue arrow). The residue coloring is the same as in (B).
R ESEARCH A RTICLES
19 MARCH 2004 VOL 303 SCIENCE www.sciencemag.org1836
ing that D1 Asp
342
and the D1 C-terminal
carboxy group of D1 Ala
344
could be pos-
sible metal ligands (37 ). All mutations of
D1 His
332
prevent photoautotrophic growth
and, again, this residue has been suggested
to be a Mn ligand (40). Site-directed mu-
tagenesis of D1 Glu
189
,D1His
332
,D1
Glu
333
, and D1 His
337
have all indicated
that these residues are involved directly or
indirectly in stabilizing the OEC (40).
There have been many mutations of D1
Asp
170
, indicating its involvement in the
assembly of the Mn
4
cluster by providing a
ligand to the high-affinity site that binds
the first Mn ion during the assembly (39).
To our knowledge, there has been no sug-
gestion that CP43 Glu
354
could be a poten-
tial ligand for the Mn
4
cluster, but site-
directed mutagenesis of the corresponding
residue in the cyanobacterium Synechocys-
tis significantly reduces PSII activity (41).
This glutamate is a part of a 3
10
helix
contained within the motif Gly-Gly-Glu-
Thr-Met-Arg-Phe-Trp-Asp, which is con-
served in all known CP43 sequences. This
motif is located in the large lumenal do-
main joining helices V and VI of CP43 (fig.
S2) and seems to form a “lid” over the
OEC. This finding emphasizes the impor-
tant role played by the large extrinsic do-
main of this chlorophyll-binding protein in
water oxidation as predicted in (41).
Tyr
Z
(D1 Tyr
161
) and Glu
189
(both on the
X
21
side), and CP43 Arg
357
and D1 Gln
165
(both on the X
22
side) are closely associated
with the nonprotein ligand binding site and
thus possibly form a hydrogen bond to sub-
strate water molecules during the reaction
cycle. Tyr
Z
is oriented to form a hydrogen
bond to D1 His
190
(Fig. 5A), in contrast to the
conclusions of Fromme et al.(42). This hy-
drogen bond has been suggested to be impor-
tant in stabilizing the tyrosine radical Tyr
Z
•
generated by the reduction of P680
•⫹
(43).
The site is deep in the cavity holding the OEC
and does not have an obvious connection to
the lumenal space.
In the D2 protein, there is an equivalent to
Tyr
Z
, which is denoted Tyr
D
(D2 Tyr
160
).
This tyrosine is also oxidized by P680
•⫹
to
generate a long-lived tyrosine radical (Tyr
D
•
),
which is not involved directly in water oxi-
dation but may help bias the electron transfer
reactions to the D1 side of the RC (44). It is
oriented in a similar way as Tyr
Z
and posi-
tioned such that it is likely to form a hydro-
gen bond with its neighboring histidine
residue (D2 His
189
), which is assumed to
stabilize Tyr
D
•
. The environment of Tyr
D
is
very hydrophobic compared with that of
Tyr
Z
. In the D2 protein, the site equivalent
to the OEC in the D1 protein is occupied by
the bulky side chains of D2 Phe
169
,D2
Phe
184
,D2Phe
185
,D2Phe
188
, CP47
Phe
362
, and CP47 Phe
363
(fig. S3). The two
CP47 residues are contained in the highly
conserved motif Phe-Phe-Glu-Thr-Phe-Ser-
Val-Leu-Val of the extrinsic domain link-
ing helices V and VI.
Mechanism of water oxidation. A
tightly bound Ca
2⫹
located close to a Mn ion
is a prerequisite for the mechanism of water
oxidation suggested by Siegbahn and col-
leagues (45, 46) and Brudvig and colleagues
(47). They propose that during the S-state
cycle only one of the Mn ions binds a water
substrate molecule and, before dioxygen for-
mation, produces a highly reactive electro-
philic intermediate, either a Mn(IV) oxyl
radical (46) or a Mn(V)oxo (47). The in-
volvement of Mn(V) in water oxidation has
also been advocated by Pecoraro et al.(48).
Our structure for the OEC strongly suggests
that Mn
4
is this reactive Mn ion. Indeed, the
X
1
site of Mn
4
detected in our electron den-
sity map could be occupied by a water mol-
ecule or an oxygen intermediate during the
water oxidation reaction.
According to the mechanisms cited
above, the O⫽O bond formation is pro-
posed to occur by a nucleophilic attack
from a second-substrate water molecule li-
gated to Ca
2⫹
, where this metal ion may
also act in part as a weak Lewis acid,
possibly aided by the Mn
3
cluster and by a
Cl
–
ligand (47). In our structure, the X
21
or
X
22
ligands of Ca
2⫹
, which are close to the
X
1
site, seem an ideal binding niche for the
second-substrate water molecule. The near-
by residues, D1 Gln
165
and CP43 Arg
357
,
may provide hydrogen bonds for stabilizing
intermediates prior to O⫽O bond forma-
tion. Indeed, mutation of CP43 Arg
357
to a
serine totally inhibited the evolution of ox-
ygen and prevented photoautotrophic
growth (49). Thus, the reaction schemes
proposed by Siegbahn, Brudvig, Pecoraro,
and colleagues are broadly consistent with
our proposed structure of the OEC. How-
ever, their reaction mechanisms also incor-
porate the proposal of Hoganson and
Babcock (50) that the oxidation of water
involves proton-coupled electron transfer
to the Tyr
Z
radical, with the hydrogen bond
between Tyr
Z
and D1 His
190
being impor-
tant not only for stabilizing Tyr
Z
•
but also
in providing the exit pathway for protons
derived from water oxidation. Although
Tyr
Z
is close to the water oxidation site and
could also provide a hydrogen bond to
reaction intermediates, our structure sug-
gests that protons are more likely to exit by
another route, as suggested by Junge and
colleague (51). Water molecules and some
residues, including Glu
189
, could link the
proposed water oxidation site with a proton
channel formed by polar residues connect-
ing D1 Asp
61
(which is closely associated
with Mn
4
) to the lumenal surface (Fig. 5, B
and C) without the involvement of Tyr
Z
and
D1 His190, which are positioned on the
other side of the OEC.
References and Notes
1. J. Barber, Q. Rev. Biophys. 36, 71 (2003).
2. B. A. Diner, G. T. Babcock, in Advances in Photosyn-
thesis: The Light Reactions. D. R. Ort, C. F. Yocum, Eds.
(Kluwer Academic Publ., Dordrecht, The Netherlands,
1996), vol. 4, pp. 213–247.
3. B. A. Diner, F. Rappaport, Annu. Rev. Plant Biol. 53,
551 (2002).
4. A. Zouni et al., Nature 409, 739 (2001).
5. N. Kamiya, J. R. Shen, Proc. Natl. Acad. Sci. U.S.A.
100, 98 (2003).
6. Materials and methods are available as supporting
material on Science Online.
7. J. Deisenhofer et al., Nature 318, 618 (1985).
8. J. P. Allen et al., Proc. Natl. Acad. Sci. U.S.A. 85, 8487
(1988).
9. P. Jordan et al., Nature 411, 909 (2001).
10. B. Hankamer et al., J. Struct. Biol. 135, 262 (2001).
11. B. Kok, B. Forbush, M. McGloin, Photochem. Photo-
biol. 11, 457 (1970).
12. F. Rappaport et al., Biochemistry 41, 8518 (2002).
13. L. M. C. Barter, J. R. Durrant, D. R. Klug, Proc. Natl.
Acad. Sci. U.S.A. 100, 946 (2003).
14. J. R. Durrant et al., Proc. Natl. Acad. Sci. U.S.A. 92,
4798 (1995).
15. J. P. Dekker, R. van Grondelle, Photosyn. Res. 63, 195
(2000).
16. J. Barber, M. D. Archer, Photochem. Photobiol. A-
Chem. 142, 97 (2001).
17. B. A. Diner et al., Biochemistry 40, 9265 (2001).
18. R. Hienerwadel, C. Berthomieu, Biochemistry 34,
16288 (1995).
19. Govindjee, J. J. S. van Rensen, in The Photosynthetic
Reaction Center, vol.1 (Academic Press, San Diego,
CA, 1993), pp. 357–389.
20. W. Oettmeier, in The Photosystems: Structure, Function,
and Molecular Biology, J. Barber, Ed. (Elsevier, Amsterdam,
The Netherlands, 1992), vol. 11, pp. 349–408.
21. F. L. de Weerd et al., Biophys. J. 82, 1586 (2002).
22. W. D. Schubert et al., J. Mol. Biol. 280, 297 (1998).
23. D. H. Stewart, G. W. Brudvig, Biochim. Biophys. Acta
1367, 63 (1998).
24. C. A. Tracewell et al., Arch. Biochem. Biophys. 385,61
(2001).
25. A. Telfer, J. De Las Rivas, J. Barber, Biochim. Biophys.
Acta 1060, 106 (1991).
26. J. Hanley et al., Biochemistry 38, 8189 (1999).
27. P. Faller, A. Pascal, A. W. Rutherford, Biochemistry 40,
6431 (2001).
28. C. A. Tracewell et al., Biochemistry 40, 193 (2001).
29. C. A. Tracewell, G. W. Brudvig, Biochemistry 42, 9127
(2003).
30. A. Telfer et al., Biochemistry 33, 14469 (1994).
31. J. H. Robblee, R. M. Cince, V. K. Yachandra, Biochim.
Biophys. Acta 1503, 7 (2001).
32. J. C. de Paula et al., J. Am. Chem. Soc. 108, 4002
(1986).
33. T. G. Carrell, A. Tyryshkin, G. C. Dismukes, J. Biol.
Inorg. Chem. 7, 2 (2002).
34. J. B. Vincent, G. Christou, Inorg. Chim. Acta 136, L41
(1987).
35. J. M. Peloquin, R. D. Britt, Biochim. Biophys. Acta
1503, 96 (2001).
36. S. V. Baranov et al., Biochemistry 39, 6060 (2000).
37. R. J. Debus, Biochim. Biophys. Acta 1503, 164 (2001).
38. B. A. Diner, Biochim. Biophys. Acta 1503, 147 (2001).
39. B. A. Diner, P. J. Nixon, J. W. Farchaus, Curr. Opin.
Struct. Biol. 1, 546 (1991).
40. H.-A. Chu, P. Nguyen, R. J. Debus, Biochemistry 34,
5859 (1995).
41. C. Rosenberg et al., Biochemistry 38, 15994 (1999).
42. P. Fromme et al., Philos. Trans. R. Soc. London Ser. B
357, 1337 (2002).
43. C. Tommos, G. T. Babcock, Biochim. Biophys. Acta
1458, 199 (2000).
44. P. Faller et al., Proc. Natl. Acad. Sci. U.S.A. 98, 14368
(2001).
45. P. E. Siegbahn, R. H. Crabtree, J. Am. Chem. Soc. 121,
117 (1999).
46. P. E. Siegbahn, Curr. Opin. Chem. Biol. 6, 227
(2002).
R ESEARCH A RTICLES
www.sciencemag.org SCIENCE VOL 303 19 MARCH 2004 1837
47. J. S. Vrettos, J. Limburg, G. W. Brudvig, Biochim.
Biophys. Acta 1503, 229 (2001).
48. V. L. Pecoraro et al., Pure Appl. Chem. 70, 925
(1998).
49. N. Knoepfle, T. M. Bricker, C. Putnam-Evans, Biochem-
istry 38, 1582 (1999).
50. C. W. Hoganson, G. T. Babcock, Science 277, 1953
(1997).
51. M. Haumann, W. Junge, Biochim. Biophys. Acta 1411,
86 (1999).
52. The authors dedicate this paper to G. T. Babcock and
M. P. Klein. J.B. and S.I. acknowledge support from the
Biotechnology and Biological Research Science Coun-
cil. We thank the Centre for Structural Biology and
Bioinformatics Facility at Imperial College London for
technical support; and C. Schulze-Briese and T. Tomi-
zaki at PX06SA/SLS, Paul Scherrer Institute, Villigen,
Switzerland; B. Shepard at ID29; and M. Iwata of the
ERATO ATP System Project for help with data collec-
tions. T.M.I. has been a Life Sciences Research Foun-
dation Fellow of the Howard Hughes Medical Insti-
tute, a European Molecular Biology Organization
Long-Term Fellow, and a Ruth L. Kirschstein National
Research Science Award Fellow and acknowledges
support from D. C. Rees during the initial stages of
the study. We wish to thank P. Siegbahn, M. Lund-
berg, P. Nixon, C. Dismukes, and A. Telfer for com-
ments and helpful discussions. The coordinates, to-
gether with the structure factors native, Mn-edge and
long wavelength (2.25 Å, for Ca detection), and the
experimental multi-wavelength anomalous disper-
sion (MAD) phases have been deposited in the Pro-
tein Data Bank (entry 1S5L) and will be available
upon publication.
Supporting Online Material
www.sciencemag.org/cgi/content/full/1093087/DC1
Materials and Methods
Figs. S1 to S5
References and Notes
29 October 2003; accepted 22 January 2004
Published online 5 February 2004;
10.1126/science.1093087
Include this information when citing this paper.
The Structure and Receptor
Binding Properties of the 1918
Influenza Hemagglutinin
S. J. Gamblin,
1
* L. F. Haire,
1
* R. J. Russell,
1
* D. J. Stevens,
1
B. Xiao,
1
Y. Ha,
2
† N. Vasisht,
1
D. A. Steinhauer,
1
‡ R. S. Daniels,
1
A. Elliot,
1
D. C. Wiley,
2
J. J. Skehel
1
§
The 1918 influenza pandemic resulted in about 20 million deaths. This enormous
impact, coupled with renewed interest in emerging infections, makes characterization
of the virus involved a priority. Receptor binding, the initial event in virus infection, is
a major determinant of virus transmissibility that, for influenza viruses, is mediated by
the hemagglutinin (HA) membrane glycoprotein. We have determined the crystal
structures of the HA from the 1918 virus and two closely related HAs in complex with
receptor analogs. They explain how the 1918 HA, while retaining receptor binding site
amino acids characteristic of an avian precursor HA, is able to bind human receptors
and how, as a consequence, the virus was able to spread in the human population.
The HAs of influenza viruses mediate receptor
binding and membrane fusion, the first stages
of virus infection (1). The receptors that they
recognize are sialic acids of cell-surface glyco-
proteins and glycolipids, and the nature of the
interactions involved in determining binding
specificity has been described in biochemical,
genetic, and structural studies (2–8). Sialic ac-
ids are usually found in either ␣2,3 or ␣2,6
linkages to galactose, the predominant penulti-
mate sugar of N-linked carbohydrate side
chains. The binding preference of a given HA
for one or other of these linkage types correlates
with the species specificity for infection. Thus,
the HAs of all 15 antigenic subtypes found in
avian influenza viruses bind preferentially to
sialic acid in ␣2,3 linkage (9), and it is this form
of the sialosaccharide that predominates in avi-
an enteric tracts where these viruses replicate
(10). Swine influenza viruses are reported to
bind sialic acid in ␣2,6, and sometimes also
␣2,3, linkages (11), and sialic acid in both
linkages is detected in porcine tracheae (10).
Human viruses of the H1, H2, and H3 subtypes
that are known to have caused pandemics in 1918,
1957, and 1968, respectively, recognize ␣2,6-
linked sialic acid (5), the major form found on
cells of the human respiratory tract (12, 13).
Because an avian origin is proposed for the
HAs of swine and human viruses (14 ), changes in
binding specificity are required for cross-species
transfer. The mechanism that human viruses have
used to achieve these changes appears to be dif-
ferent for different subtypes. For the HAs of the
H2 and H3 human viruses, a minimum of two
changes in receptor binding site amino acids,
Gln
226
to Leu
226
and Gly
228
to Ser
228
, correlates
with the shift from avian to human receptor bind-
ing (15, 16). By contrast, HAs of human H1
viruses acquire the ability to bind to human recep-
tors while retaining Gln
226
and Gly
228
(11). To
understand how they do this, we determined the
structures of HAs from the 1918 pandemic virus
(1918-human) with the use of HA expressed from
DNA of the sequence recovered from tissues in-
fected with virus in 1918 (17) and from the pro-
totype human (1934-human) and swine (1930-
swine) H1 influenza viruses, A/Puerto Rico/8/34
and A/swine/Iowa/30, respectively (Fig. 1).
A/Puerto Rico/8/34 was one of the first human
influenza viruses isolated in the Americas (18)
and has been widely used in laboratory investiga-
tions of influenza. A/swine/Iowa/30 was the first
influenza virus isolated from mammals in 1930
(19), 3 years before the first isolate was recovered
from humans (20).
Overall structure. The structures were
solved by molecular replacement, and crystallo-
graphic statistics are given in Table 1 and table S1.
The overall trimeric structures of the three H1
HAs are similar (Fig. 2), but they show notable
differences to HAs of other subtypes with respect
to the arrangements of the receptor binding, ves-
tigial esterase, and membrane fusion subdomains,
both within the HA trimer and also within indi-
vidual monomers (21) (table S2). As predicted
from their placement in the same phylogenetic
and structure-based clade, the H1 HAs are most
similar to those of the H5 subtype (21). We have
examined the structures in detail in relation to their
receptor binding and membrane fusion activities
and to the antigenic variation that occurred in both
periods of human H1 virus prevalence, 1918 to
1957 and 1977 to date (Fig. 2), and we will
present a detailed description of this analysis else-
where (22). We have, however, concluded that the
receptor binding properties are the most distinc-
tive features of the 1918 virus HA, and we focus
on these here.
The receptor binding subdomain. The
receptor binding sites are located at the mem-
brane-distal tip of each subunit of the HA trimer
(Fig. 2). Three secondary structure elements—the
190 helix (residues 190 to 198), the 130 loop
(residues 135 to 138), and the 220 loop (residues
221 to 228)—form the sides of each site, with the
base made up of the conserved residues Tyr
98
,
Trp
153
, His
183
, and Tyr
195
(1) (Fig. 2B). The
conformations adopted by the 130 and 220 loops
of the three H1 HAs are similar, but they are
significantly different from those of the equivalent
loops in the HAs of other influenza subtypes (2, 3)
(Figs. 2B, 3, and 4 and table S2). To understand
the structural basis of the receptor specificity of
H1 HAs, we determined the structures of the
1934-human and the 1930-swine HAs in com-
plex, with ␣2,3- and ␣2,6-linked sialopentasac-
charides as analogs of avian and human receptors,
1
Medical Research Council (MRC) National Institute for
Medical Research, The Ridgeway, Mill Hill, London NW7
1AA, UK.
2
Department of Molecular and Cellular Biolo-
gy, Howard Hughes Medical Institute, Harvard Univer-
sity, 7 Divinity Avenue, Cambridge, MA 02138, USA.
*These authors contributed equally to this work.
†Present address: Department of Pharmacology, Yale
University School of Medicine, 333 Cedar Street, New
Haven, CT 06520, USA.
‡Present address: Department of Microbiology and
Immunology, Emory University School of Medicine,
1510 Clifton Road, Atlanta, GA 30322, USA.
§To whom correspondence should be addressed. E-
mail: mbrenna@nimr.mrc.ac.uk
R ESEARCH A RTICLES
19 MARCH 2004 VOL 303 SCIENCE www.sciencemag.org1838
