Threading structural model of the manganese-stabilizing protein PsbO reveals presence of two possible beta-sandwich domains.

Abstract

The manganese-stabilizing protein (PsbO) is an essential component of photosystem II (PSII) and is present in all oxyphotosynthetic organisms. PsbO allows correct water splitting and oxygen evolution by stabilizing the reactions driven by the manganese cluster. Despite its important role, its structure and detailed functional mechanism are still unknown. In this article we propose a structural model based on fold recognition and molecular modeling. This model has additional support from a study of the distribution of characteristics of the PsbO sequence family, such as the distribution of conserved, apolar, tree-determinants, and correlated positions. Our threading results consistently showed PsbO as an all-beta (beta) protein, with two homologous beta domains of approximately 120 amino acids linked by a flexible Proline-Glycine-Glycine (PGG) motif. These features are compatible with a general elongated and flexible architecture, in which the two domains form a sandwich-type structure with Greek key topology. The first domain is predicted to include 8 to 9 beta-strands, the second domain 6 to 7 beta-strands. An Ig-like beta-sandwich structure was selected as a template to build the 3-D model. The second domain has, between the strands, long-loops rich in Pro and Gly that are difficult to model. One of these long loops includes a highly conserved region (between P148 and P174) and a short alpha-helix (between E181 and N188)). These regions are characteristic parts of PsbO and show that the second domain is not so similar to the template. Overall, the model was able to account for much of the experimental data reported by several authors, and it would allow the detection of key residues and regions that are proposed in this article as essential for the structure and function of PsbO.

Full text

Page 1

Threading Structural Model of the Manganese-Stabilizing
Protein PsbO Reveals Presence of Two Possible ?-Sandwich
Domains
Florencio Pazos,1†Pedro Heredia,2†Alfonso Valencia,1and Javier de las Rivas2*
1Centro Nacional de Biotecnologı ´a, Consejo Superior de Investigaciones Cientı ´ficas, Madrid, Spain
2Instituto de Recursos Naturales y Agrobiologı ´a, Consejo Superior de Investigaciones Cientı ´ficas, Salamanca, Spain
ABSTRACT?
tein (PsbO) is an essential component of photosys-
tem II (PSII) and is present in all oxyphotosynthetic
organisms. PsbO allows correct water splitting and
oxygen evolution by stabilizing the reactions driven
by the manganese cluster. Despite its important
role, its structure and detailed functional mecha-
nism are still unknown. In this article we propose a
structural model based on fold recognition and
molecular modeling. This model has additional sup-
port from a study of the distribution of characteris-
tics of the PsbO sequence family, such as the distri-
bution of conserved, apolar, tree-determinants, and
correlated positions. Our threading results consis-
tently showed PsbO as an all-beta (?) protein, with
two homologous ? domains of approximately 120
amino acids linked by a flexible Proline-Glycine-
Glycine (PGG) motif. These features are compatible
with a general elongated and flexible architecture,
in which the two domains form a sandwich-type
structure with Greek key topology. The first domain
is predicted to include 8 to 9 ?-strands, the second
domain 6 to 7 ?-strands. An Ig-like ?-sandwich
structure was selected as a template to build the 3-D
model. The second domain has, between the strands,
long-loops rich in Pro and Gly that are difficult to
model. One of these long loops includes a highly
conserved region (between P148 and P174) and a
short ?-helix (between E181 and N188)). These re-
gions are characteristic parts of PsbO and show that
the second domain is not so similar to the template.
Overall, the model was able to account for much of
the experimental data reported by several authors,
and it would allow the detection of key residues and
regions that are proposed in this article as essential
for the structure and function of PsbO. Proteins
2001;45:372–381.
The manganese-stabilizing pro-
© 2001 Wiley-Liss, Inc.
Key words: fold recognition; manganese-stabilizing
protein;proteinstructure;PsbO;thread-
ing
INTRODUCTION
The higher plant photosyste m II (PSII) oxygen-evolving
complex has three extrinsic polypeptides associated with
the luminal side of the thylakoid membranes. These
polypeptides, with apparent molecular weights of 33, 23,
and 17 kDa, are important components of the oxygen-
evolving apparatus because they stabilize the correct
function of the manganese cluster, the site of water
splitting and oxygen evolution. These polypeptides are
encoded by three nuclear genes known as psbO, psbP, and
psbQ. They are all present in the photosyste m II of algae
and higher plants, but PsbO is theonly onealso present in
cyanobacteria,themostprimitiveoxyphotosyntheti corgan-
isms.
Experimental data on thespecific roleof thepolypeptide
PsbOassociated withthemanganeseclusterarecontrover-
sial because this protein is not directly involved in the
binding of the manganese or calcium required for oxygen
evolution. This lack of clear, functional data comes to-
gether with a limited knowledge of the PsbO structure.
Several groups have carried out analyses of the protein by
Fourier transform infrared (FTIR), circular dichroism
(CD), far-UV CD, and other biophysical techniques and
havereported diversedata concerning its secondarystruc-
ture.1–5TheFTIR and CD data seem to indicatethatPsbO
presents a major ?–structure component, but they differ
substantially about the explicit composition, with values
for the ?-helix content ranging from 8% to 27%. Hydrody-
namic studies haveindicated thattheshapeof themanga-
nese-stabilizing protein is elongated in solution, with
approximate dimensions of 12.6 nm ? 3.0 nm, yielding an
axial ratio of 4.2.6Folding studies on isolated PsbO have
suggested that the kinetics of unfolding and refolding of
this protein are similar to those of the immunoglobulin
lightchain.7Recentstudieshavepresente d anopendiscus-
sion about thestructureor conformation assigned to PsbO
in solution, with some authors suggesting that it has a
“natively unfolded” structure8,9and others proposing that
it attains a “molten globule” structure.10In a recent
publication the first X-ray structure of photosyste m II
from a thermophilic cyanobacteria (Synechococcus elonga-
tus) at 3.8 Å resolution presented.11This work provided
Grantsponsor: CICYT; Grantnumber: PB98-0480.
†Florencio Pazos and Pedros Heredia contributed equally to this
work.
*Correspondence to: Javier de las Rivas, Instituto de Recursos
Naturales yAgrobiologı ´a, Consejo Superior deInvestigaciones Cientı´-
ficas, c/ o Cordel de Merinas 40–52 (P.O. Box 257), 37071 Salamanca,
Spain. E-mail: jrivas@gugu.usal.es
Received 4 January2001; Accepted 6 August2001
Published online00 Month 2001; DOI 10.1002/prot.XXXX
PROTEINS: Structure, Function, and Genetics 45:372–381 (2001)
© 2001 WILEY-LISS, INC.

Page 2

some new information about PsbO, such as the assignment
of an electron density region of the extrinsic luminal side of
photosystem II crystals to half the molecular mass of
PsbO. This region showed a cylindrical structure 35 Å in
length and corresponded to a group of ?-strands of uncer-
tain connectivity.11Because these X-ray data were low
resolution, they did not provide clear knowledge about the
PsbO fold. The model published (PDB ? 1FE1) only gave
coordinates for alpha-carbons without assignment of resi-
due types.11For this reason it was not possible to know
which part of the PsbO sequence corresponded to the
electron density observed. This experimental work,11pub-
lished after the first submission of our threading model,
showed good agreement with our model, which in turn
may help to further the assignment of residues in the PSII
X-ray density map.
Protein structure prediction using computational meth-
ods is an expanding field that can provide interesting
insight into protein function and biochemistry. In the
current work, in which we used fold recognition methods,
wemadesignificantprogressonpreviousstructuralpredic-
tions about PsbO.12“Threading,” “fold recognition,” and
“remote homology modeling” are names given to a set of
techniques to predict a protein’s three-dimensional struc-
ture in the absence of obvious sequence similarity with
proteins of known structure. Threading methods attempt
to accommodate the problem sequence into the 3-D coordi-
nates of several proteins of known structure and, with the
aid of an appropriate scoring system, to select the best
sequence–structure fit. Many scoring schemas have been
proposed, ranging from simple concordance of secondary
structure elements13to potentials of mean force,14statisti-
cal potentials,15fitness of sequence profiles,16and many
others. More sophisticated methods, combining different
potentials using neural networks technology, followed the
initial approaches.17According to the results reported at
several CASP (Critical Assessment of Techniques for
Protein Structure Prediction) meetings, threading meth-
ods are increasingly reliable and a genuine alternative
method for predicting protein structures.18According to
reports at recent CASP and CAFASP events it has been
also shown that using a combination of several thread-
ing methods for fold recognition considerably increases
the correctness of predictions (http://www.cs.bgu.ac.il/
?dfischer/CAFASP2).
There were three steps in our process of making a
prediction about the PsbO structure: (1) obtaining possible
models by using a series of complementary threading
methods—TOPITS, GONPM, H3P2, 3-PSSM, HFR,
THREADER2, and HMM; (2) comparing the results to
select the most consistent predictions; and (3) assessing
the possible models using information from the study of
the PsbO protein structure family, including the distribu-
tion of conserved, apolar, correlated, and subfamily-
specific residues. This approach is largely based on our
previous systematic assessment of threading models with
sequence-derived properties.19Finally, we compared the
best model with the available experimental information
and made suggestions to further experimental work. Do-
ingstructuralanalysesonPsbOincreasedourunderstand-
ing of the relations between the structure and function of
this protein.
METHODS
Threading Methods
Following are descriptions of the basic principles of the
seven threading methods used in this work, including
explanations of their scoring systems and associated reli-
ability, that is, their chances of being correct in predicting,
or percentage of certainty, along with the addresses of
their Web sites. Four methods use Z-score schemas:
z-score ? (raw ? mean) / s, where raw is the query-
template alignment score, mean is the mean score of the
query with all the possible folds in the fold library, and s is
the standard deviation of the score distribution:
● TOPITS13(http://dodo.cpmc.columbia.edu/predictpro-
tein/) is based on the matching between the predicted
secondary structure and solvent accessibility of the
problem sequence and the known secondary structure
and accessibility of the templates. The results are
ranked by Z-score. Hits with Z-scores of more than 3.0
are expected to correspond to correct predictions in more
than 60% of the cases.
● GONPM20(http://fold.doe-mbi.ucla.edu/)usessequence–
sequencereplacementtablesandsequence-derivedprop-
erties of the query protein, including the predicted
secondary structure. The results are also arranged by a
Z-score value, and when it is greater than 4.0, it
indicates reliable predictions. The authors do not pro-
vide a quantitative value of certainty for the scores.
● H3P221(http://fold.doe-mbi.ucla.edu/) is similar to the
previous ones and uses a five-dimensional sequence-
structuresubstitutionmatrixderivedfromknownstruc-
tures and predicted secondary structure. Z-scores are
also used to classify the results. The authors presented
an evaluation of the reliability associated with the
Z-scores performed by counting the number of false
positives in 243 examples using a fold library of 811
proteins. The probability of a false positive for Z-scores
of 2.5 was 1.2 ? 10?3, decreasing to 1.9 ? 10?4for
Z-scores of more than 3.0.
● 3D-PSSM22(http://www.bmm.icnet.uk/?3dpssm/) com-
bines multiple-sequence profiles with structure-based
profiles (which include solvation potentials derived from
known structures and predicted secondary structure).
The results are sorted by e-values. The authors indicate
that an e-value of 0.97 corresponds to a certainty of
correct prediction of 71.7% and an e-value of 1.11 to a
certainty of 67.2%. Therefore, e-values ? 1.0 are taken
as significant.
● HFR23(http://www.cs.bgu.ac.il/?bioinbgu/) is a hybrid
method that collects results from five threading pro-
grams, combining them in a search for the most consis-
tent fold prediction among them. It also takes into
consideration evolutionary information. The ranking of
the hits is based on a combined score, which for values of
more than 9.0 corresponds to correct predictions. A
STRUCTURAL MODEL OF PSBO PROTEIN
373

Page 3

quantitative value of certainty associated with the score
was not provided.
● THREADER215,24(http://insulin.brunel.ac.uk) is quite
different from previous methods; it combines sequence
information with pseudo-energies obtained from solva-
tion and contact potentials previously derived from
known protein structures. This method provides a
Z-score to calibrate the significance of a prediction. The
authors indicate that for a Z-score of greater than 3.5
the result is very significant, with a more than 60%
chance of being correct.
● HMM16
(http://www.cse.ucsc.edu/research/compbio/
HMM-apps/) is perhaps the more different approach
because it is based on a direct sequence comparison
potential provided by a Hidden Markov Model engine.
In this approach the query sequences are compared
with collections of profiles derived for each one of the
known protein structures (fold library). The score pro-
vided for each fold measures the probability that a
query sequence belongs to such class or group of se-
quences with the same fold. A high score (such as ?9.0)
indicates that the sequence of interest is probably a
member of the selected class. Explicit certainty values
associated with the scores are not provided.
Evaluating Threading Models by Sequence
Features
Threading models were evaluated according to the fit-
ting of sequence-extracted features derived from the study
of the PsbO protein family. The clustering of apolar
residues, of conserved residues, of correlated residues,19
and of tree-determinant amino acids (Pazos & Valencia,
unpublished) were the four properties studied. The idea
was to discard those models in which, for example, apolar
residues do not form a cluster, as statistically expected
from the analysis of known globular proteins.
The tendency of sets of residues to cluster together was
calculated as the difference between the distribution of
distances among all pairs of residues and the distance
distribution of the residues on which we are focusing:
apolar, conserved, correlated, and tree-determinants. The
explicit calculation was carried out with the previously
derived formula25:
Xd ??
i ? 1
n
Pic? Pia
di? n
where n is the number of distance bins (there are 15
equally distributed bins from 4 to 60 Å); diis the upper
limit for each bin (e.g., 8 for the 4–8 bin, normalized to 60);
Picis the percentage of apolar, conserved, correlated, or
tree-determinant pairs with distance between diand di?1;
and Piais the same percentage for all pairs in the model.
Defined in this way, Xd ? 0 indicates no separation
between the two distance populations, Xd ? 0 indicates
positive cases where the population of apolar, conserved,
correlated, or tree-determinant pairs is shifted to smaller
distances with respect to the population of all pairs
(clustering).
RESULTS AND DISCUSSION
Threading Consistently Assigns an All-? Structure
to PsbO
We explored the possible threading models for PsbO by
applying seven programs to three divergent PsbO se-
quences: PsbO-spiol, a sequence from the plant Spinacea
oleracea;PsbO-chlre,fromthealgaeChlamydomonasrein-
hardtii; and PsbO-synec, of the cyanobacteria Synechocys-
tis sp. PCC6803. These three sequences cover the main
range of known PsbO sequences down to 43% sequence
identity.
The results obtained (Table I) are given in terms of the
possible framework protein in PDB,26followed by the score
proposed for each query-template alignment obtained with
each threading program (e.g., 1fnf 2.32) and the structural
characteristics that define each template according to
SCOP,27that is, structural class (e.g., all-?), type of fold
(e.g., sandwich), number of ?-strands of a domain (e.g.,
seven strands), topology of the domain (e.g., Greek key),
number of domains, and SCOP code (e.g., 1.002.001.002.
001.003 for 1fnf, included for comparison purposes and not
as an absolute reference because it cannot be expected to
be stable over time).
All the PDB proteins selected by the threading methods
as candidates for modeling PsbO are all-? proteins. Seven
of the 11 proteins with significant scores (bold in Table I)
are all-? sandwich proteins. Despite the consistent detec-
tion of an all-?-type structure, two methods (TOPITS and
H3P2) did not predict any protein with a significant score
and two other methods (3D-PSSM and THREADER2)
produced only one best prediction above their confidence
threshold, as shown in Table I. This could indicate that
PsbO is a difficult protein to predict and that the more
simple automatic programs cannot find any reasonable
model for it. The best relative scores were produced by
HFR and THREADER2 for models based on 1prr and 1hnf.
The relative score 4.39/3.5, given by THREADER2, corre-
sponds to a certainty greater than 80%.
According to the meaning of the scores (see the Methods
section) and considering the agreement observed in Table
I, we can estimate about a 90% level of confidence in the
prediction of PsbO as an all-? protein, and about an 80%
level of confidence in the prediction as an all-? sandwich
protein. This estimation is prudent but has to be taken
with caution because the accuracy of different threading
methods has not been compared quantitatively in any
published study, making it difficult to make a precise
estimation of certainty.
PsbO Seems to Match Ig-Like ? Sandwich
Fold Type
Even if all the proteins presented in Table I belong to the
very general all-? structural class, it would be necessary to
do additional investigation into the possibility of predict-
ing a more specific fold. For this reason the relative scores
for different templates of the most significant methods
(3D-PSSM, HFR, and THREADER2) are presented in
Figure 1. The results consistently show the abundance of
all-? sandwich and Ig-like ? sandwich among the top five
374
F. PAZOS ET AL.

Page 4

proteins obtained with these methods. In some cases the
prediction was all-? barrel: for example 1avgI, predicted
by 3D-PSSM for Chlamydomonas PsbO. This kind of
structure is less represented than the all-? sandwich.
Indeed, the models of PsbO based on all-? barrel struc-
tures are usually partial, leaving important parts of PsbO
out of the model (data not shown). The comparison of three
threading methods in Figure 1 reflects that 3D-PSSM
gives less confident predictions than do HFR and
THREADER2, indicated by the lower values of the relative
scores. As can be inferred from the discussion of the
scoring schemas presented in Methods, a score/threshold
parameter of 1 corresponds to a certainty level of around
60% to 70%.
PsbO Seems to Include Two Similar ? Domains
A common and interesting feature of the threading
models based on all-? sandwich proteins is that most of
them correspond to two-domain proteins, with a few others
TABLEI
Method
TOPITSRost(1995)
GONPMFischer&
Eisenberg(1996)
H3P2Rice&
Eisenberg(1997)
3D-PSSMSternberg
(1999)
HFRFischer(2000)
THREADER2Jones
(1999)
HMMKarplus(1998)
METHODBASEDON
binarystructureandaccessiblityorSOLVATION
OTHERMETHODS
Scores(confidence
threshold)
?3.0
?4.0
?2.5
?1.0
?9.0
?3.5
??9.0
PsbO-spiol
248aa
plant
sequenceidentity
100%
1fnf(2.32)
All-?sandwich
7strands
greek-key
4?domains(7?4)
1.002.001.002.001.003
fibronectin
celladhesionprotein
1mfn4.2
All-?sandwich
7strands
greek-key
2?domains(7?8)
1.002.001.002.001.004
fibronectin
celladhesionprotein
2ayh(2.2)
All-?sandwich
12–14strands
complex
1.002.026.001.002.002betaglucanasa
?1.0
1prr9.90
All-?sandwich
8strands
greek-key
2?-domains(8?8)
1.002.010.001.002.001
sporecoatproteins
bindingprotein
8fabA(3.03)
All-?sandwich
7strands
greek-key
2?-domains(9?7)
1.002.001.001.001.0121.002.001.001.002.022
immunoglobulin
Fabfragment
1cne?9.01
All-?barrel
6strands
greek-key
1.002.038.001.001.007
nitratereductase
coredomain
PsbO-chlre
239aaalgae
sequenceidentity63%
?3.0
1mfn4.3
All-?sandwich
7strands
greek-key
2?domains(7?8)
1.002.001.002.001.004
fibronectin
celladhesionprotein
?2.5
1avgI0.70
All-?barrel
8strands
meander
1.002.053.001.003.001
thrombininhibitor
1prr14.2
All-?sandwich
8strands
greek-key
2?-domains(8?8)
1.002.010.001.002.001
sporecoatproteins
bindingprotein
1hnf4.39
All-?sandwich
7strands
greek-key
2?domains(9?7)
1.002.001.001.001.007 1.002.001.001.003.006
T-lymphocyte
adhesionglycoprotein
1cne?9.92
All-?barrel
6strands
greek-key
1.002.038.001.001.007
nitratereductase
coredomain
PsbO-synec
246aa
cyanobacteria
sequenceidentity46%
?3.0
1mfn4.2
All-?sandwich
7strands
greek-key
2?domains(7?8)
1.002.001.002.001.004
fibronectin
celladhesionprotein
?2.5
1fnhA(2.70)
All-?sandwich
7strands
greek-key
3?dom.(*7?7?6)
1.002.001.002.001.003
fibronectin
integrinbinding
1fnhA9.10
All-?sandwich
7strands
greek-key
3?dom.(7?7?6)
1.002.001.002.001.003
fibronectin
integrinbinding
?3.5
2bb2?9.03
All-?sandwich
8strands
greek-key
2?domains(8?8)
1.002.010.001.001.005
eyelensprotein
Fig. 1.
proteins selected by three threading methods (3D-PSSM, HFR, and
THREADER2)forPsbOsequenceofspinach(PsbO-spiol),Chlamydomo-
nas (PsbO-chlre), and Synechocystis (PsbO-synec). The bar size corre-
sponds to the values of the score given for each protein over the
confidence threshold of the corresponding threading method. The gray
bars correspond to all-? proteins, the bars in black correspond to Ig-like ?
sandwich protein, and the bars in white correspond to other folds.
Graphical representation of the threading results. The first five
STRUCTURAL MODEL OF PSBO PROTEIN
375

Page 5

(1fnhA, 1fnf) corresponding to proteins with three or four
domains. Moreover, the best threading scores for the PsbO
sequence were obtained with two-domain proteins. We
interpret these results as a first indication of the presence
of two structural domains in PsbO.
To use some independent or different information to test
further the idea of a two-domain protein structure, we
searched for possible domain boundaries in a multiple
sequence alignment containing 15 PsbO sequences.12A
clue came from the presence of a conserved PGG motif,
from position 118 to position 120 in the pea sequence
(117–119 in the spinach sequence). The flexible structure
of this motif and its likely role as a domain linker or hinge
can indicate the presence of two parts or regions in a
protein with 118 and 127 residues, respectively.
A multiple alignment of four PsbO plant sequences (pea,
potato, tomato, and rice) was generated to check the
internal identity of the proposed domains. The result,
shown in Figure 2, yielded a remarkable homology (20
totally conserved residues, with 16.9% identity and 47.4%
similarity). It is surprising that such a symmetry had not
been detected previously. Perhaps this is because the
similarity between the two domains in the algae and
cyanobacteria sequences is much lower, in turn making
the similarity less obvious when all the known PsbO
sequences are compared. The discovery of a symmetry in
the protein, with two regions of clear sequence similarity,
strongly supports the two-domain structure of PsbO ini-
tially suspected from analyzing the threading models. This
observed two-domain structure cannot be detected using
the standard domain databases ProDom28and Pfam.29
Separated Threading of Two PsbO Regions Also
Selects Ig-Like ? Sandwich Proteins
Because different methods—structure- and sequence-
based—had indicated a two-domain structure for PsbO,
independent threading runs with each of the two PsbO
regions were carried out: first, with the N-terminal region,
from residue 1 to residue 120, and second, with the
C-terminal region, from residue 117 to residue 248. The
threading results with each separated region pointed
again to all-? sandwich proteins as the most favorable
models, with a similar distribution to the one presented in
Table I, in which the threading experiments had been
carried out with the full sequence. Some templates pro-
posed were: 1ksr ? actin binding protein 120, for the
N-terminal region, and 1srdA ? Cu,Zn-superoxide dis-
mutase, for the C-terminal second region, both of which
Fig. 2.
regions from pea, potato, tomato, and rice. The first region includes 118 residues and the PGG final motif,
marked within a gray box. The second region includes 127 residues. In each region the conserved Pro and Gly
are marked with boxes. There are 20 positions fully conserved in the alignment, six of which are Pro.
Domain distribution shown of the multiple-sequence alignment of selected PsbO sequences: two
376
F. PAZOS ET AL.

Page 6

corresponded to the Ig-like ? sandwich fold. These results
can be viewed as confirming the proposed domain struc-
ture and fold. For the second domain the 1srdA template
provided a better alignment than did the templates ob-
tained for the whole protein. This protein (1srdA, superox-
ide dismutase) has longer loops between the ?-strands
that fit better with the PsbO loops in this region.12As a
whole, the threading runs with the individual domains did
not lead to higher scores, and they produced models that
only covered half the protein.
Selection of the Best Frameworks Among the all-?
Sandwich Candidates
A search of the FSSP database to compare protein
structures30(http://www2.embl-ebi.ac.uk/dali/fssp/fssp.
html), using the seven all-? sandwich protein candidates
(Table I), led us to distinguish two groups of all-? sand-
wiches. In the first group were five proteins—1hnf, 8fabA,
1fnf, 1fnhA, and 1mfn; in the second group two, 1prr and
2bb2. These groups correspond to two different folds in the
SCOP classification, the first to Ig-like ? sandwiches and
the second to crystallines. To select the optimum possible
template to model PsbO, each of these ? proteins was
assessed using two criteria: threading alignment coverage
and conservation of PsbO sequence properties.
The threading alignments of PsbO with each template
protein showed that the coverage of the aligned sequence
(248 aa) was quite complete for the first group of proteins
(Ig-like ? sandwich fold) but only partial for the second
group (crystalline fold). Such partial alignment is reflected
in the number of residues not aligned in the N-terminus or
C-terminus of the models:
PDBCode
1fnf
1fnhA
1hnf
8fabA
1mfn
2bb2
1prr
NotAlignedResidues
3NtorCt
3NtorCt
19Ct
9Ntand15Ct
39Nt
55Ct
85Nt
1st—
2nd—
These differences in model completeness may indicate
that PsbO fits better to the number of secondary structure
elements and topology characteristic of an Ig-like ? sand-
wich fold.
Additional information was required to make the selec-
tion of a specific template from among different possible
Ig-like ? sandwich frameworks for modeling PsbO. Our
strategy was to evaluate the distribution of a number of
relevant PsbO sequence characteristics in each model. We
evaluated in detail the distribution of the conserved,
apolar, tree-determinant, and correlated residues that
were calculated from the sequences of the PsbO protein
family. The visualization of the sequence characteristics in
the various models and the numerical evaluation of the
significance of their distribution (see Methods section) was
carried out with the program Threadlize.31
The distribution of the various residues (Table II) shows
1mfn and 1hnf as the best candidate models. The table also
shows that 1prr and 2bb2 have negative scores, indicating
they are not appropriate models for PsbO, which confirmed
our previous deduction. In general, the lowest values were
obtained for non-Ig-like folds. A combined score adding the
four sequence features (Table II, last column) also selects
some of the folds as better frameworks for holding the
PsbO sequence attributes in a consistent spatial proxim-
ity. Such an addition of the fitting values produced the best
numbers for 1mfn (5.351) and 1hnf (4.449). Several rea-
sons formed the basis for the final selection of the best
template from between these two proteins: (1) as indicated
above, the coverage of the alignment of PsbO with 1mfn
leaves out 39 N-terminal residues but only 19 C-terminal
residues with 1hnf; (2) the 1mfn PDB corresponds to 20
structures determined by NMR that are more difficult to
use for modeling than is the 1hnf X-ray structure; and (3)
the relative threading score (Table II, second column) was
best for 1hnf.
FSSP Structural Family Assigned to PsbO and
Some Functional Clues
Because we had inferred there was a two-domain struc-
ture for PsbO, the structural families that best fit each
domain had to be considered in the model building. Both
are homologous Ig-like ? sandwich domains, but they have
some topological differences in their ?-strands. The first
domain of 1hnf belongs to the FSSP structural family led
by1qa9A(seetheFSSPdatabase30athttp://jura.ebi.ac.uk:
8765/holm/qz?filename?/data/research/fssp//1qa9A.fssp),
which includes 8fabA, 1fnf, 1fnhA, 1cdy, and 1dr9A. These
proteins are mentioned because several were detected as
possible templates for PsbO by the threading methods (see
8fabA, 1fnf, and 1fnhA in Table I) or because they are
structurally and functionally very closely homologous to
1hnf (1cdy and 1dr9A). The second domain fits better in
TABLEII
PDBId
8fabA
1fnf
1fnhA
1hnf*
1mfn
2bb2
1prr
Threading
score/threshold
0.866
0.773
1.011
1.254*
1.050
1.003
1.100
FittingvaluesofPsbOpropertiesineachtemplate
Conserved
residuesresidues
0.6851.008
0.3851.093
0.3760.305
0.387 1.067
0.2581.193
?1.1290.377
1.306
?0.053
¥
properties
2.630
?0.648
?1.079
4.449*
5.351
?1.412
?1.935
Apolar Tree-
determinants
0.467
0.063
?1.110
0.553
0.832
?0.246
?0.667
Correlated
mutations
0.470
?2.189
?0.650
2.442
3.068
?0.414
?2.521
STRUCTURAL MODEL OF PSBO PROTEIN
377

Page 7

the FSSP structural family led by 1mfmA (see the FSSP
database30athttp://jura.ebi.ac.uk:8765/holm/qz?filename?/
data/research/fssp//1mfmA.fssp), which also includes pro-
teins 8fabA, 1fnhA, 1fnf, 1cdy and 1dr9A.
From a study of these two FSSP families we noted some
important features about them. First, both families in-
clude proteins with very different functions. For instance,
T-lymphocyte glycoprotein fragment, T-cell receptor frag-
ment, fibronectin, superoxide reductase, and cytochrome f
(cited in order of decreasing Z-value in the FSSP) are all
part of the 1qa9A family; and superoxide dismutase,
fibronectin, Fab fragment from human Ig, beta-amylase,
and T-cell surface glycoprotein fragment are members of
the 1mfmA family. Second, many of these functions are
related to cell adhesion or macromolecule interaction. And,
third, we observed that the folds of the redox-related
enzymes superoxide dismutase and superoxide reductase
are assigned to PsbO. It has been shown that proteins with
similar folds can perform completely different functions,32
and so the use of a T-lymphocyte adhesion glycoprotein
(1hnf) as a template to model an oxygen-evolving protein
(PsbO) should not be very surprising. However, the struc-
tural correlation found in this work may also include some
unknown functional relation. In this respect, it was very
interesting to find that PsbO (involved in the production of
oxygen from water) has the same fold as that of superoxide
reductase (an enzyme that reduces superoxide to hydrogen
peroxide) and superoxide dismutase (an enzyme that
catalyses a similar reaction with superoxide, producing
hydrogen peroxide plus oxygen). If this predicted struc-
tural correlation includes any functional meaning, we
would suggest that PsbO might be involved in the quench-
ing of oxygen radicals produced during oxygen evolution.
This is a totally new but sensible hypothesis that should be
tested experimentally.
Proposal of a 3-D Remote Model for PsbO
The construction of a good alignment between template
and query proteins is essential for obtaining sensible
models. The global alignment of spinach PsbO against the
best template selected, 1hnf, was obtained by the
THREADER2 method, with some small regions taken
from an alignment obtained with the HFR method (Fig. 3).
Some small modifications were carried out because of the
FSSP structural alignments of the two groups of proteins
that correspond to each domain.
Most of the 1hnf region aligned with PsbO corresponds
to the structural core of its FSSP structural families.
Therefore, the main secondary structure features (mostly
the ?-strands) of 1hnf and those predicted for PsbO are in
good agreement. This is particularly clear for the first
domain because the second domain of PsbO has some
specific regions (mainly loops) that could not be aligned
(Fig. 3).
The alignment was prepared with SWISS-PDBviewer,33
Version 3.6 (http://www.expasy.ch/spdbv/mainpage.htm)
andsubmittedtoSWISS-MODEL33(http://www.expasy.ch/
swissmod/SWISS-MODEL.html) for the automatic build-
ing of a protein model. The quality of the model obtained
was fairly good according to the scores of the evaluation
programs(WhatCheck34),takingintoaccountthatitcorre-
sponds to a remote homology modeling case. For example,
the first-generation packing quality was ?3.733, Ram-
achandran plot appearance was ?3.883, inside/outside
distribution was 1.142. The structural model generated by
SWISS-MODEL can be visualized with any representation
software. As expected, given the low level of sequence
identity between the query and the framework, the core of
each domain was fairly well formed, but other regions,
such as surface loops, presented serious modeling difficul-
ties. The regions corresponding to insertions and deletions
(see alignment in Fig. 3) corresponded mainly to loops in
the structure, an indication that the structural core of the
template matches the PsbO predicted structure. Long
loops in the PsbO second domain were impossible to model,
including the regions from residues Gly152 to Gly163 and
from Gly177 to Asn198 (boxes in Fig. 3). The first loop
corresponds to a very conserved PsbO region12and the
second to a RGD cell-attachment motif (residues 178–
180), followed by a short predicted ?-helix, which does not
have any counterpart in 1hnf. The size of the insertions
confirms that PsbO second domain is less similar to the
template than the first domain.
The structure of the PsbO model is represented in
Figure 4; the corresponding coordinates are available from
http://gredos.cnb.uam.es/pazos/PsbO.Giventhatthemodel
corresponds to a two-domain protein linked with a flexible
region, it is possible that the relative positions of both
domains in PsbO would be different from those of 1hnf.
PsbO Model Based on 1hnf Is Congruent With
Known Structural Data
Biophysical and biochemical data on the structure of
PsbO are available from several sources. A few residues of
significant importance for the PsbO structure have been
identified, such as two Cys (Cys28 and Cys51) close to the
N-terminus that form a disulfide bridge essential to main-
taining native PsbO tertiary structure.35,36Trp241 is the
only Trp in the protein; it is closed to the C-terminus, and
its fluorescence emission spectrum indicates it is buried in
a nonpolar niche, probably inside the PsbO core.7,10
The proposed model, based on 1hnf (Fig. 4), locates the
cysteines in the first ?-domain in a way that could interact
to form the disulfide bridge and stabilize the two ?-sheets
that compose such a domain. In the model Trp241 is in the
last ?-strand in the interior of the second domain, where it
forms part of the hydrophobic core. This is in good agree-
ment with the previous experimental data. In addition,
reports on refolding kinetics experiments based on the
intrinsic fluorescence of Trp have indicated that PsbO
presents an unfolding pattern similar to the characteristic
pattern of inmunoglobulins,7providing additional indirect
support for the proposed structural model.
Twootherstructuralcharacteristicsexperimentallyiden-
tified for PsbO are its elongated form, with a length-to-
depth ratio of 4:1,6and its flexible architecture,8,9which
corresponds to its composition being rich in Gly and Pro
pairs (Fig. 1). The proposed model, based on 1hnf, is in
378
F. PAZOS ET AL.

Page 8

good agreement with both features because it corresponds
to an elongated protein structure and contains a consider-
able proportion of loops rich in GG, GP, and PP pairs. The
location of the conserved PGG triplet in the middle of two
?-domains also provides additional flexibility to the pro-
tein.
Most recent structural information on PsbO has been
provided by X-ray data at 3.8 Å resolution about cyanobac-
terial PSII,11which showed that a 35-Å section of PsbO
has a ?-sheet with at least five strands. Our model agrees
with these data because both the domains in the PsbO
model are 32–40 Å long and include two ?-sheets with
more than five strands. From an examination of the
matching of the strands in size and position, we also
suggest that the region presented in the X-ray low-
resolution model corresponds to the N-terminal region of
our PsbO model.
CONCLUSION
The use of computational methods to analyze, compare,
and predict protein structure is a great aid in guiding
experiments in protein biochemistry. Fold recognition is
part of the protein structure prediction methodology, and
it was used in this work to gain new insight into a protein
with an unknown detailed structure. The structure of
PsbO has been a controversial subject in recent years, with
some researchers claiming that it is a “natively unfolded”
protein,8,9and others suggesting that it has a “molten
globule”–type structure.10Our results indicate that PsbO
has a well-defined ? core, surrounded by long flexible
loops. This proposition is well supported by the new X-ray
data about PSII.11Our model also agrees well with the
idea that the protein has a elongated flexible structure,
including a possible hinge point between two domains.
This proposed PsbO architecture, with two domains con-
Fig. 3.
shown above its sequence, and the actual 1hnf secondary structure is indicated bellow its sequence (E corresponds to ?-structure and H to ?-helix). The
residues in ?-strands are shown in shadow if the reliability of the secondary structure prediction is above 90%. Some significant PsbO residues are
marked in gray boxes on the sequence, including the two Cys (C28 and C51), the PGG motif (from residue 118 to residue 120), the RGD cell attachment
domain (from residue 178 to residue 180), and the only Trp present in PsbO (W241). The numbering corresponds to the spinach sequence.
Sequence to structure alignment of PsbO and 1hnf. Vertical bars mark the conserved residues. The predicted PsbO secondary structure is
STRUCTURAL MODEL OF PSBO PROTEIN
379

Page 9

nected by a linker region, would be ideal for providing
conformational flexibility required to close and give stabil-
ity to the manganese cluster in the thylakoid membrane or
to be released free in the lumen when PSII is inactivated.5
Some of the ideas derived from the model can now serve as
a guide for additional experimental approaches that may
validate it and may provide new insights into the detailed
mechanism of PsbO function. Some proposals for future
experiments are: (1) the cloning and expression of each
separated PsbO domain to check their independent fold
and folding kinetics and to test how it is their particular
docking or interaction works with the oxygen-evolving
complex; (2) construction of mutants in the highly con-
served P148–P174 loop to see how they affect the oxygen-
evolving activity; and (3) testing the ability of the PsbO
protein to react with hydrogen peroxide or with other
oxygen radicals. Some of these lines of research are
currently being undertaken in our laboratory.
The modeling steps followed in this work can be also
seen as a general strategy for the selection of templates for
Fig. 4.
based on a remote similarity with the 1hnf template. Consequently, large deviations are expected from the real structure of PsbO.37The representation
was made with RASMOL.38The critical residues C28, C51, and W241 are presented as balls. Different regions are indicated corresponding to the two
possible structural domains, the linker region (PGG motif) and the position of the two long insertions that were not included in the structural model.
Graphical representation of the PsbO model based on the 1hnf framework. The PsbO backbone of the model is represented. Note that it is
380
F. PAZOS ET AL.

Page 10

proteins of unknown structure based on the combination of
results from different threading methods, sequence infor-
mation, and experimental data.
ACKNOWLEDGEMENTS
We thank members of the Protein Design Group for
continual support and stimulating discussions. We also
thank the Vasc Government for the predoctoral grant
awarded to P. Heredia.
REFERENCES
1. Xu Q, Nelson J, Bricker TM. Secondary structure of the 33 kDa,
extrinsic protein of photosystem II: a far-UV circular dichroism
study. Biochim Biophys Acta 1994;1188(3):427–431.
2. Ahmed A, Tajmir-Riahi HA, Carpentier R. A quantitative second-
ary structure analysis of the 33 kDa extrinsic polypeptide of
photosystem II by FTIR spectroscopy. FEBS Lett 1995;363(1–2):
65–68.
3. Sonoyama M, Motoki A, Okamoto G, Hirano M, Ishida H, Katoh S.
Secondary structure and thermostability of the photosystem II
manganese-stabilizing protein of the thermophilic cyanobacte-
riumSynechococcuselongatus.BiochimBiophysActa1996;1297(2):
167–170.
4. Shutova T, Irrgang KD, Shubin V, Klimov VV, Renger G. Analysis
of pH-induced structural changes of the isolated extrinsic 33
kilodalton protein of photosystem II. Biochemistry 1997;36(21):
6350–6358.
5. Hutchison RS, Betts SD, Yocum CF, Barry BA. Conformational
changes in the extrinsic manganese stabilizing protein can occur
upon binding to the photosystem II reaction center: an isotope
editing and FT-IR study. Biochemistry 1998;37(16):5643–5653.
6. Zubrzycki IZ, Frankel LK, Russo PS, Bricker TM. Hydrodynamic
studies on the manganese-stabilizing protein of photosystem II.
Biochemistry 1998;37(39):13553–13558.
7. Tanaka S, Kawata Y, Wada K, Hamaguchi K. Extrinsic 33-
kilodalton protein of spinach oxygen-evolving complexes: kinetic
studies of folding and disulfide reduction. Biochemistry 1989;
28(18):7188–93.
8. Lydakis-Simantiris N, Betts SD, Yocum CF. Leucine 245 is a
critical residue for folding and function of the manganese stabiliz-
ing protein of photosystem II. Biochemistry 1999;38(47):15528–
35.
9. Lydakis-Simantiris N, Hutchison RS, Betts SD, Barry BA, Yocum
CF. Manganese stabilizing protein of photosystem II is a thermo-
stable, natively unfolded polypeptide. Biochemistry 1999;38(1):
404–14.
10. Shutova T, Irrgang K, Klimov VV, Renger G. Is the manganese
stabilizing 33 kDa protein of photosystem II attaining a “natively
unfolded” or “molten globule” structure in solution? FEBS Lett
2000;467(2–3):137–140.
11. Zouni A, Witt H-T, Kern J, Fromme P, Krau? N, Saenger W, Orth
P. Crystal structure of photosystem II from Synechococcus elonga-
tus at 3.8 Å resolution. Nature 2001;409:739–743.
12. De Las Rivas J, Heredia P. Structural predictions on the 33 kDa
extrinsic protein associated to the oxygen evolving complex of
photosyntheticorganisms.PhotosynthesisResearch1999;61(1):11–
21.
13. Rost B. TOPITS: threading one-dimensional predictions into
three-dimensional structures. In: C Rawlings, D Clark, R Altman,
L Hunter, T Lengauer, & S Wodak, editors. The third interna-
tional conference on Intelligent Systems for Molecular Biology
(ISMB). Cambridge, U.K.: AAAI Press; 1995. p 314–321.
14. Sippl MJ. Calculation of conformational ensembles from poten-
tials of mean force. An approach to the knowledge-based predic-
tion of local structures of globular proteins. J Mol Biol 1990;213:
859–883.
15. Jones DT, Taylor WR, Thornton JM. A new approach to protein
fold recognition. Nature 1992;358:86–89.
16. Karplus K, Barrett C, Hughey R. Hidden Markov Models for
detecting remote protein homologies. Bioinformatics 1998;14(10):
846–856.
17. Jones DT. GenTHREADER: an efficient and reliable protein fold
recognition method for genomic sequences. J Mol Biol 1999;287:
797–815.
18. Sippl MJ, Lackner P, Domingues FS, Koppensteiner WA. An
attempt to analyze progress in fold recognition from CASP1 to
CASP3. Proteins 1999;3:226–230.
19. Olmea O, Rost B, Valencia A. Effective use of sequence correlation
and conservation in fold recognition. J Mol Biol 1999;295:1221–
1239.
20. Fischer D, Eisenberg D. Fold recognition using sequence-derived
predictions. Protein Science 1996;5:947–955.
21. Rice D., Eisenberg D. A 3D-1D substitution matrix for protein fold
recognition that includes predicted secondary structure of the
sequence. J Mol Biol 1997;267:1026–1038.
22. Kelley LA, MacCallum RM, Sternberg MJE. Enhanced genome
annotation using structural profiles in the program 3D-PSSM. J
Mol Biol 2000;299(2):501–522.
23. Fischer D. Hybrid fold recognition: combining sequence derived
properties with evolutionary information. Maun Lani, HI: Pacific
Symp Biocomputing, 2000. p 119–130.
24. Jones DT, Miller RT, Thornton JM. Successful protein fold
recognition by optimal sequence threading validated by rigorous
blind testing. Proteins 1995;23:387–397.
25. Pazos F, Helmer-Citterich M, Ausiello G, Valencia A. Correlated
Mutations contain information about protein–protein interaction.
J Mol Biol 1997;271(4):511–523.
26. Bernstein FC, Koetzle TF, Williams GJ, Meyer, Jr., EE, Brice MD,
Rodgers JR, Kennard O, Shimanouchi T, Tasumi M. The Protein
Data Bank: a computer-based archival file for macromolecular
structures. J Mol Biol 1977;112:535–542.
27. Murzin AG, Brenner SE, Hubbard T, Chothia C. SCOP: a struc-
tural classification of proteins database for the investigation of
sequences and structures. J Mol Biol 1995;247:536–540
28. Sonnhammer E, Kahn D. Modular arrangement of proteins as
inferred from analysis of homology. Protein Sci 1994;3:482–492.
29. Bateman A, Birney E, Durbin R, Eddy SR, Howe KL, Sonnham-
mer EL. The Pfam Protein Families Database. Nucl Acids Res
2000;28:263–266.
30. HolmL,SanderC.Mappingtheproteinuniverse.Science1996;273:
595–602.
31. Pazos F, Rost B, Valencia A. A platform for integrating threading
resultswithproteinfamilyanalyses.Bioinformatics1999;15:1062–
1063.
32. Russell RB, Sasieni PD, Sternberg MJ. Supersites within super-
folds. Binding site similarity in the absence of homology. J Mol
Biol 1998;282:903–918.
33. Guex N, Peitsch MC. SWISS-MODEL and the Swiss-PdbViewer:
an environment for comparative protein modeling. Electrophore-
sis 1997;18:2714–2723
34. Hooft RWW, Vriend G, Sander C, Abola EE. Errors in protein
structures. Nature 1996;381:272.
35. Burnap RL, Qian M, Shen JR, Inoue Y, Sherman LA. Role of
disulfide linkage and putative intermolecular binding residues in
the stability and binding of the extrinsic manganese-stabilizing
protein to the photosystem II reaction center. Biochemistry 1994;
33(46):13712–13718.
36. Betts SD, Ross JR, Hall KU, Pichersky E, Yocum CF. Functional
reconstitution of photosystem II with recombinant manganese-
stabilizing proteins containing mutations that remove the disul-
fide bridge. Biochim Biophys Acta 1996;1274(3):135–142.
37. Chothia C, Lesk AM. The relation between the divergence of
sequence and structure in proteins. EMBO J 1986;5:823–826.
38. Sayle R, Milner-White E. RASMOL: biomolecular graphics for all.
Trends Biochem Sci 1995;20(9):374–376.
STRUCTURAL MODEL OF PSBO PROTEIN
381