Redox-linked protonation state changes in cytochrome bc1 identified by Poisson-Boltzmann electrostatics calculations.

Abstract

Cytochrome bc(1) is a major component of biological energy conversion that exploits an energetically favourable redox reaction to generate a transmembrane proton gradient. Since the mechanistic details of the coupling of redox and protonation reactions in the active sites are largely unresolved, we have identified residues that undergo redox-linked protonation state changes. Structure-based Poisson-Boltzmann/Monte Carlo titration calculations have been performed for completely reduced and completely oxidised cytochrome bc(1). Different crystallographically observed conformations of Glu272 and surrounding residues of the cytochrome b subunit in cytochrome bc(1) from Saccharomyces cerevisiae have been considered in the calculations. Coenzyme Q (CoQ) has been modelled into the CoQ oxidation site (Q(o)-site). Our results indicate that both conformational and protonation state changes of Glu272 of cytochrome b may contribute to the postulated gating of CoQ oxidation. The Rieske iron-sulphur cluster could be shown to undergo redox-linked protonation state changes of its histidine ligands in the structural context of the CoQ-bound Q(o)-site. The proton acceptor role of the CoQ ligands in the CoQ reduction site (Q(i)-site) is supported by our results. A modified path for proton uptake towards the Q(i)-site features a cluster of conserved lysine residues in the cytochrome b (Lys228) and cytochrome c(1) subunits (Lys288, Lys289, Lys296). The cardiolipin molecule bound close to the Q(i)-site stabilises protons in this cluster of lysine residues.

Full text

Page 1

Redox-linked protonation state changes in cytochrome bc1identified by
Poisson–Boltzmann electrostatics calculations
Astrid R. Klingena, Hildur Palsdottirb,1, Carola Hunteb, G. Matthias Ullmanna,⁎
aStructural Biology/Bioinformatics Group, University of Bayreuth, Germany
bDepartment of Molecular Membrane Biology, Max Planck Institute of Biophysics, Frankfurt am Main, Germany
Received 19 September 2006; received in revised form 15 January 2007; accepted 17 January 2007
Available online 31 January 2007
Abstract
Cytochrome bc1is a major component of biological energy conversion that exploits an energetically favourable redox reaction to generate a
transmembrane proton gradient. Since the mechanistic details of the coupling of redox and protonation reactions in the active sites are largely
unresolved, we have identified residues that undergo redox-linked protonation state changes. Structure-based Poisson–Boltzmann/Monte Carlo
titration calculations have been performed for completely reduced and completely oxidised cytochrome bc1. Different crystallographically observed
conformations of Glu272 and surrounding residues of the cytochrome b subunit in cytochrome bc1from Saccharomyces cerevisiae have been
considered in the calculations. Coenzyme Q (CoQ) has been modelled into the CoQ oxidation site (Qo-site). Our results indicate that both
conformationalandprotonationstatechangesofGlu272ofcytochromebmaycontributetothepostulatedgatingofCoQoxidation.TheRieskeiron–
sulphur cluster could be shown to undergo redox-linked protonation state changes of its histidine ligands in the structural context of the CoQ-bound
Qo-site.Theprotonacceptor roleoftheCoQligandsintheCoQreductionsite(Qi-site)issupportedbyourresults.Amodifiedpathforprotonuptake
towards the Qi-site features a cluster of conserved lysine residues in the cytochrome b (Lys228) and cytochrome c1subunits (Lys288, Lys289,
Lys296). The cardiolipin molecule bound close to the Qi-site stabilises protons in this cluster of lysine residues.
© 2007 Elsevier B.V. All rights reserved.
Keywords: Protonation probability; Titration behaviour; Respiratory chain; Membrane protein; Cardiolipin; Rieske iron–sulphur cluster; Poisson–Boltzmann
electrostatics calculation
1. Introduction
The cytochrome bc1complex (cytochrome bc1) is a key
enzyme of biological energy conversion in bacteria and
mitochondria. It is a multi-subunit transmembrane protein
complex that transfers electrons from reduced coenzyme Q
(CoQ) to a mobile redox-active protein and translocates protons
across the membrane. The free energy of the catalysed redox
reaction is converted into the energy of a transmembrane proton
motive force. In mitochondria, cytochrome bc1spans the inner
mitochondrial membrane and transfers electrons from ubiquinol
to cytochrome c. Mitochondrial cytochrome bc1 represents
complex III of the respiratory chain.
The coupling between electron transfer and proton transloca-
tion in cytochrome bc1is based on the so-called modified Q-
cycle mechanism (Fig. 1) [1–3]. The mechanism requires two
CoQ-binding sites and a site for cytochrome c reduction,
connected by chains of protein-bound redox cofactors. Three
different subunits of the complex form the catalytic sites and
bindthe cofactors: cytochrome b,cytochrome c1,andthe Rieske
iron–sulphur protein(ISP). Theoxidation ofCoQ iscatalysed in
the so-called Qo-site of the complex: the reduced and protonated
quinol form of CoQ is converted into the oxidised and
deprotonated quinone form. The two electrons of this reaction
are transferred to two different electron acceptor groups, namely
Biochimica et Biophysica Acta 1767 (2007) 204–221
www.elsevier.com/locate/bbabio
Abbreviations: CoQ, coenzyme Q; ISP, iron–sulphur protein; FTIR, Fourier
transform infrared spectroscopy; PDB, protein data bank (www.rcsb.org/pdb);
HDBT, hydroxydioxobenzothiazole; PB, Poisson Boltzmann; MC, Monte
Carlo; CDL, cardiolipin; Specific residues are denoted by their single letter
amino acid code, their residue number and a subunit identifier; CYB,
cytochrome b subunit; CYC1, cytochrome c1subunit; ISP, Rieske iron–sulphur
protein subunit; SU9, small subunit 9
⁎Corresponding author. Tel.: +49 921 553545; fax: +49 921 553544.
E-mail address: matthias.ullmann@uni-bayreuth.de (G.M. Ullmann).
1Present address: Lawrence Berkeley National Laboratory, Berkeley,
California, USA.
0005-2728/$ - see front matter © 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.bbabio.2007.01.016

Page 2

a b-type haem group (haem bL, bound to the cytochrome b
subunit of the complex) and a Rieske [Fe2S2] iron–sulphur
cluster (bound to the ISP subunit). By movement of its mobile
head domain [4], the Rieske ISP then transfers an electron to the
haem c1 group bound to the cytochrome c1 subunit of
cytochrome bc1. Cytochrome c1 finally reduces the c-type
haem group of the substrate cytochrome c. The second electron
of the oxidation of CoQ is transferred via haem bLand haem bH
of cytochrome b to the so-called Qi-site. In the Qi-site, oxidised
CoQ gets reduced to semiquinone and finally to quinol by two
electrons sequentially arriving from the Qo-site. Since CoQ
oxidation and reduction are coupled to proton release and
uptake, respectively, the location of the two CoQ-binding sites
relative to the membrane links electron transfer to the
translocation of protons across the membrane.
Although the overall scheme of the Q-cycle mechanism is
widely accepted, details of the reactions in the catalytic sites
remain largely unresolved [5,6]. Among the most heavily
debated aspects of cytochrome bc1catalysis are the nature of
proton acceptor and donor groups [7–11], the nature of the rate-
limiting steps [12,13], the sequence of single proton and
electron transfer steps [12,14], the mechanistic basis of the
bifurcation of electron transfer pathways in the Qo-site [13–18],
and the control of harmful bypass reactions [19–21]. Consider-
able effort has been made to clarify these mechanistic details.
However, this effort resulted in the formulation of different
mechanistic models that are based on often conflicting in-
terpretation of the available experimental data.
To evaluate the conflicting models of Qo- and Qi-site
catalysis, it is necessary to identify redox-linked protonation
state changes in cytochrome bc1, because redox-linked protona-
tion state changes are at the very heart of any mechanism
coupling electron and proton transfer reactions. Several recent
Fourier transform infrared (FTIR) spectroscopy studies [22–27]
have successfully embarked on identifying redox-linked proto-
nation state changes in cytochrome bc1. In this work, we report
redox-linked protonation state changes of cytochrome bc1that
have been identified from Poisson–Boltzmann/Monte Carlo
titration calculations. The calculations are based on the crystal
structures of cytochrome bc1from the yeast Saccharomyces
cerevisiae[28–30].TheconformationalvariabilityoftheQo-site
observed in different crystal structures has been taken into
account. For different redox states of the complex that
correspond to the experimental conditions of the FTIR studies
weobtainprotonationprobabilitiesfor all titratablegroupsin the
protein complex.
In the following, we give some basic theoretical background
and describe the setup of our calculations. We then present and
discuss the redox-linked protonation state changes revealed by
our calculations. A small number of titratable residues were
identified that have markedly different protonation probabilities
in the fully oxidised and fully reduced state of the complex. Our
resultssupporttheideathatconformationalvariability oftheQo-
site plays a role during CoQ oxidation. We observe a coupling
between protonation reactions and conformational transitions
that may be the basis of the recently discussed gating of the Qo-
site reaction [6,19,20]. Concerning the Qi-site, our results
propose a modified path for proton uptake to the active site.
2. Material and computational setup
2.1. Preparation of the crystal structures
Our calculations are based on two different crystal structures of cytochrome
bc1fromS.cerevisiae.ThesestructurescontaintheQo-siteinhibitorsstigmatellin
(PDB code 1KB9, 2.3 Å resolution, Ref. [29]) and hydroxydioxobenzothiazole
(HDBT, PDB code 1P84, 2.5 Å resolution, Ref. [30]), respectively. The HDBT-
inhibited structure contains an additionally refined second cardiolipin molecule
[31].Bothcrystalstructurescompriseninedifferentsubunits.Twocopiesofeach
subunit form the dimeric state of the complex, which is considered to be the
catalytically active state of cytochrome bc1[32].
In both crystal structures, the Rieske head domain is found in its so-called
b-position, forming the Qo-site together with the cytochrome b subunit.
The crystal structures were prepared for the Poisson–Boltzmann (PB)
calculations using the molecular modelling package CHARMM [33]. All lipids,
detergent and water molecules were retained during the structure preparations.
Thedesiredredoxstatesofthecomplexwereintroducedbyassigningappropriate
partialchargestotheatomsoftheredoxcofactors.Hydrogenatompositionswere
constructed and subsequently energy-minimised using the steepest decent (SD)
and conjugate gradient (CG) techniques and short molecular dynamics
simulations as implemented in CHARMM. The minimisation consisted of
1000SDsteps,500MDstepsof0.2 fsat100 K,500MDstepsof 0.5fs at200K,
500 MD stepsof 1 fs at 300 K, 500 MD stepsof 1 fs at 100 K,1000SD steps and
2000CG steps. The procedurewas followedfor all states of the system that were
treated in separate PB-calculations. The crystal structures contain antibody
fragments, which bind to the Rieske protein head domain and were used for
cocrystallisation. These antibody fragments were not included in our
Fig. 1. The Q-cycle mechanism of cytochrome bc1. Protons from the oxidation
of quinol in the Qo-site are set free to the intermembrane space. One electron
from the oxidation of quinol is transferred via the high-potential chain of redox
cofactors ([Fe2S2] cluster of the Rieske ISP subunit and haem c1of cytochrome
c1) to cytochrome c. The second electron is recycled back into the CoQ pool: it
is transferred to the Qi-site via the low-potential chain of redox cofactors (haem
bLand bHof cytochrome b). Two molecules of CoQ have to get oxidised in the
Qo-site to providethe twoelectronsfor completereductionof oneCoQ molecule
in the Qi-site. QH2+2 cyt c(ox)+2 Hmatrix
overall reaction catalysed by cytochrome bc1. Electron transfer is indicated by
dashed arrows, proton translocation by bold arrows. Q—oxidised CoQ,
quinone. QH2—reduced CoQ, quinol. SQ—stable semiquinone intermediate
in the Qi-site. ISP—iron–sulphur protein. IMS—mitochondrial inter-membrane
space, IMM—inner mitochondrial membrane.
+
→Q+2 cyt c(red)+4 HIMS
+
is the
205A.R. Klingen et al. / Biochimica et Biophysica Acta 1767 (2007) 204–221

Page 3

calculations: test calculations showed, that their removal from the structure has
negligible effect on the protonation probabilities of cytochrome bc1.
2.2. Modelling of undecylstigmatellin and CoQ into the Qo-site
In order to obtain results that can be compared to the FTIR data published
by Ritter et al. [25], we have performed calculations on cytochrome bc1
inhibited by undecylstigmatellin (UST). For these calculations, we have used
the stigmatellin-inhibited structure of cytochrome bc1 from S. cerevisiae.
Since the crystal structure contains stigmatellin A which has a different
hydrophobic tail than UST (Fig. 2A and B), we have changed the hydrophobic
tail of stigmatellin A into the simple alkyl chain of UST. The hydrophobic tail
of UST in the Qo-site has been energy minimised using CHARMM. The
details of the minimisation procedure are the same as for the minimisation of
the hydrogen atom positions described above. Since Ritter et al. [25] have
reported changes in the redox state of UST in the Qo-site of cytochrome bc1,
we consider two different redox states of UST in our calculations. The
probable chemical structures of oxidised and reduced UST are shown in Fig.
2B and C, respectively. For both redox states, we have calculated partial
charges using a density functional theory (DFT) approach (see Supplementary
Material).
To obtain mechanistically relevant results, we have performed calculations
on cytochrome bc1with the substrate CoQ modelled into the Qo-site. Since the
available crystal structures of cytochrome bc1do not contain CoQ in the Qo-
site, we have followed a modelling procedure that uses structural information
available for another CoQ-binding site, namely the QB-site of the photo-
synthetic reaction centre. The reaction centre has been crystallised with
stigmatellin (PDB code 4PRC) and with CoQ (PDB code 2PRC) in the active
site [34]. The binding mode of stigmatellin is very similar in the Qo-site and
the QB-site. In both sites, the carbonyl oxygen atom of the chromone ring
system of oxidised stigmatellin forms a hydrogen bond towards the nitrogen
atom of a histidine sidechain. The same histidine sidechain coordinates an iron
atom with its other nitrogen atom (the mononuclear iron centre in the reaction
centre and the Rieske cluster in cytochrome bc1). The hydroxy group of the
chromone ring system interacts with an oxygen-containing sidechain (serine in
the reaction centre, glutamate in cytochrome bc1. Using the relative orientation
of stigmatellin and CoQ observed in the crystal structures of the reaction
centre, and the position of stigmatellin in the Qo-site of cytochrome bc1, we
derive the position of CoQ in the Qo-site of cytochrome bc1 from yeast.
Starting from this similarity-based initial positioning, we have performed an
energy minimisation of the hydrophobic tail of CoQ using CHARMM, during
which only minor structural changes are observed. The details of the mini-
misation procedure are the same as for the minimisation of hydrogen atom
positions described above. The tail was modelled to contain one isopren unit.
In the framework of the single-occupancy model [35] we have by this pro-
cedure obtained a reasonable working model of CoQ in the Qo-site (Fig. 3). In
all calculations, the Qi-site contains CoQ as observed in both crystal structures
from S. cerevisiae.
2.3. Addition of a model membrane
To account for the effect of the membrane environment on the electrostatics
of cytochrome bc1we have added a model membrane around the crystal
structures prepared for the PB-calculations. The membrane is modelled by a
torus-shaped belt of uncharged atoms, placed with HLINK from the TRIP
programpackage [36]. For the PDB depositedcoordinates,the membrane model
extends from yinterface matrix=−61 Å to yinterface intermembrane space=−39 Å along
the membrane normal of the complex. In the PB-calculations, a low dielectric
constant is assigned to the volume occupied by the uncharged membrane atoms,
whichthus adequately represent the hydrophobiccore of the membrane. Explicit
water molecules were removed from the structures for PB-calculation after
addition of the model membrane.
2.4. Treatment of conformational variability in the Qo-site
As obvious from the crystal structures from S. cerevisiae, the Qo-site of
cytochrome bc1 adopts different conformations in presence of the Qo-site
inhibitors stigmatellin and HDBT. In both crystal structures, the Rieske head
domain is found in its so-called b-position forming the Qo-site together with the
cytochrome b subunit [4]. The most obvious difference between the structures is
the orientation of the sidechainof E272CYB(Fig. 3). E272CYBpoints towards the
Rieske cluster in the stigmatellin-inhibited complex (referred to as conformation
Glu-FeS), and points towards haem bLin the HDBT-inhibited complex (referred
to as conformation Glu-b). Differences between the two conformations are
limited to residues 265 to 273 of cytochrome b and the sidechain of H253CYB.
We assume that the rest of the complex adopts the conformation observed in the
HDBT-inhibited crystal structure and we keep it fixed in this conformation
during the Monte Carlo calculations. The energy difference between the two
conformations (shown in green and purple in Fig. 3) in presence of CoQ in the
active site has been estimated by a combined molecular mechanics/Poisson–
Boltzmann approach that is detailed in the Supplementary Material. The value
for the conformational energy difference enters into the Monte Carlo
calculations that reveal to which extent the two conformations are populated
in the complex with CoQ in the Qo-site (see below).
2.5. Calculation of protonation probabilities
To characterise protonation probabilities as well as the probabilities of the
different Qo-site conformations, we have performed PB electrostatics and
Metropolis Monte Carlo (MC) titration calculations. The underlying theory is
described in detail in Ref. [37] and [38].
The PB-calculations were performed using multiflex from the MEAD
package [39]. The results describe the energetics of all N titratable groups in
cytochrome bc1in terms of N intrinsic pK-values and a symmetric N×N matrix
of pairwise interaction energies. The intrinsic pK-values are calculated as shifts
relative to experimentally determined pK-values of appropriate model reactions
in aqueous solution (see Supplementary Material). In our calculations, all
aspartate, glutamate, lysine, arginine, histidine, tyrosine and cysteine residues,
the propionate moieties of the haem groups and the lipid head-groups were
considered titratable. Apart from the Rieske ligand histidines, sidechains that are
involved in covalent or coordination bonds are not considered titratable.
Fig. 2. (A) Chemicalstructure of stigmatellin A contained in the crystal structure
of cytochrome bc1 from S. cerevisae. (B) Chemical structure of oxidised
undecylstigmatellin(UST).(C)ChemicalstructureofreducedUST.Theposition
ofthehighlightedoxygenatom(boldfont)hasbeenenergyminimisedduringthe
structure preparation procedure, in order obtain an appropriate tetrahedral
geometry at the corresponding ring carbon atom.
206A.R. Klingen et al. / Biochimica et Biophysica Acta 1767 (2007) 204–221

Page 4

Histidines are treated as described by Bashford et al. [40]. N- and C-termini are
considered titratable as far as they are resolved in the crystal structures.
UnresolvedterminiaretheN-terminusofthesmallsubunits6,7,8and9,andthe
C-terminus of cytochrome c1and the small subunit 9. The effect of these
unresolved termini is considered marginal. Because of their structural flexibility
they can be expected to be well solvated. Their charges are consequently well
shieldedandhavelittleeffectonthetitrationbehaviourintherestofthecomplex.
Neutral blocking groups were attached to replace the unresolved termini.
In contrast to all other titratable groups, model pK-values for the ligand
histidines of the Rieske centre were not determined experimentally, but were
calculated by a combined DFT/PB approach (see Supplementary Material). The
underlying DFT calculations have been shown in previous work on Rieske
proteins to correctly reproduce experimental data in combination with PB-
calculations [41,42]. Since in MEAD titratable groups are considered to adopt
exactly two different protonation forms, the different one-proton equilibria of
the Rieske centre were treated in separate MEAD-calculations, and their relative
energies were then sampled by a subsequent MC analysis. Details of the
treatment of the Rieske centre in the PB/MC-calculations are given in the
Supplementary Material.
The following parameters were used in all PB-calculations: dielectric
constants ε=4 for the protein and the membrane and ε=80 for the aqueous
phase, ionic strength I=0.1 M for the aqueous phase and temperature T=300 K.
Standard partial charges from the CHARMM22 parameter set [33] were used for
the protein. Partial charges for other compounds (Rieske cluster, haem groups,
stigmatellin, CoQ and lipids) were derived from density functional theory
calculations performed with the ADF programme suite [43]. Details of these
calculations and resulting charges are given as Supplementary Material. The
calculation of partial charges for the Rieske cluster [41,42] and CoQ [44] has
been reported in previous publications from our group. Bondi radii [45] were
used for all atoms except for hydrogen (rH=1.0 Å).
FromtheresultsofthePB-calculations,thepH-dependentenergyofacertain
protonation and conformation state can be calculated as
Gðn;kÞ¼
X
þ1
N
i
RTln 10d xðnÞ
i
? xð0Þ
i
?
xðnÞ
i
?
?
pH ? pKintrðkÞ
i
?
xðnÞ
j
?
2
X
N
i
X
N
j
? xð0Þ
i
??
? xð0Þ
j
?
WðkÞ
i;jþ GðkÞ
conf
ð1Þ
with G(n,k)as the energy of conformation k in protonation state n. A certain
protonation state n is characterised by a protonation state vectorYx
components xi(n)=0 if group i is deprotonated, and xi
xi
oftheprotein,forwhichtheintrinsicpK-valueshavebeencalculated.pKi
intrinsicpKofgroupiinconformationk,whichcorrespondstothepK-valuegroup
i would have if all other groups were in their reference protonation form. Wi,j(k), is
the interaction energy between groups i and j in conformation k, with Wi,j=0 for
i=j. N is the total number of titratable groups, R the universal gas constant and T
ðnÞwith the
(n)=1 if group i is protonated.
(0)correspondstotheprotonationformofgroupiinthereferenceprotonationstate
intr(k)isthe
Fig. 3. The substrate CoQ modelled into the Qo-site of cytochrome bc1. The Qo-site is formed by two of the three catalytic subunits of the cytochrome bc1complex,
namely cytochrome b (CYB, coloured in green) and the Rieske iron–sulphur protein (ISP, coloured in ochre). Different conformations of the Qo-site as observed in the
presence of stigmatellin (coloured in purple: conformation Glu-FeS, Ref. [28]) and HDBT (coloured in green: conformation Glu-b, Ref. [31]) have been considered in
the calculation of protonation probabilities in presence of CoQ in the Qo-site. Differences between the two conformations are limited to residues 265 to 273 of
cytochrome b (the positions of the Cα-atoms in this fragment are labelled by their residue numbers), and the sidechain of H253CYB. Coordinates outside these regions
are virtually identical in the two structures, and coordinates from the HDBT-inhibited structure have been used for the rest of the complex. The two electron acceptors
of the CoQ oxidation reaction (haem bLand the Rieske cluster), the cysteine and histidine ligands of the Rieske cluster, and residues undergoing redox-linked changes
in their titration behaviour are highlighted. Distances between atoms are given to detail the orientation of CoQ modelled into the active site.
207A.R. Klingen et al. / Biochimica et Biophysica Acta 1767 (2007) 204–221

Page 5

the temperature. Gconf
conformations k considered (see above). A separate PB-calculation is performed
for every conformation k.
To obtain the protonation probabilities of all groups i and the probabilities of
the different conformations k as a function of pH, the states (n, k) of Eq. (1) are
sampled by a Metropolis MC algorithm implemented in the programme cmct
[46].Anoutputensembleoflowenergystatesiscalculatedfor everypH(pH0to
14instepsof0.1pH-units).AtagivenpH,theprotonationprobability〈xi〉(pH)of
acertaingroupiisequivalenttotheprobabilitytofindthegroupprotonatedinthe
MC output ensemble. In the same way, the probability of a certain conformation
can be obtained from the composition of the MC output ensemble.
For every pH-step, the MC calculations consist of 500 equilibration scans
and 20,000 production scans at T=300 K. An MC scan consist of N MC steps,
with N as total number of titratable groups. In one MC step,the protonation form
of one group is changed, and the change in energy is evaluated. After every fifth
MC step, the conformation of the system is changed, and the change in energy is
evaluated. If the end state of a single MC step is accepted, this state is used as
starting state for the next MC step. At the end of each MC production scan, the
state of the system is added to the output ensemble. In double (triple) MC steps,
the protonation form of two (three) groups is changed simultaneously. Double
(triple) MC steps are applied to groups with an interaction energyWi,jlarger than
2 (3) pK-units.
(k)
accounts for the relative energies of the different
2.6. Correlation analysis of protonation probabilities
The electrostatic interaction among multiple titratable groups can lead to
highly unusual titration behaviour of the individual groups [47,48]. To
rationalise such unusual titration curves, the pairwise correlation ci,j(pH) of
protonation probabilities is a valuable analysis tool:
ci;jðpHÞ ¼ hxixjiðpHÞ ? hxiiðpHÞdhxjiðpHÞ:
〈xi〉(pH) is the probability to find group i to be protonated, irrespective of the
protonation state of group j, and vice versa for 〈xj〉(pH). In contrast, 〈xixj〉(pH) is
the probability to find both groups to be protonated at the same time. ci,j(pH)
takes values between −0.25 and +0.25. If ci,j(pH) is large and negative, the
groups i and j are anticorrelated, meaning that protonation of group i disfavours
protonation of group j, and vice versa.
ð2Þ
3. Results
3.1. Redox-linked protonation state changes in cytochrome bc1
In order to identify redox-linked protonation state changes in
cytochrome bc1we have calculated the titration behaviour of all
titratable groups in the complex, once for the completely
oxidised state, and once for the completely reduced state. In the
oxidised state, all redox cofactors are oxidised and the CoQ
molecules in the Qo- and Qi-site are in the oxidised and
deprotonated quinone form. In the reduced state, all redox
cofactors are reduced and the CoQ molecules in the Qo- and Qi-
site are in the reduced and protonated quinol form. In both redox
states, the system is allowed to adopt either of the two Qo-site
conformations.
Twelve titratable groups in cytochrome bc1display notice-
ably different protonation probabilities in the completely
reduced and the completely oxidised state of the system. For
all other groups, the rmsd between their protonation probabil-
ities in the reduced and oxidised state is below 0.2 for the pH-
range from 0to 14. The behaviour of most titratable groups does
thus not change between the completely reduced and com-
pletely oxidised state of the system. The catalytically active
dimeric cytochrome bc1complex from S. cerevisiae consists of
two copies each of nine different subunits. Since equivalent
titration behaviour is observed for all pairs of identical subunits
in the dimeric cytochrome bc1 complex, only the results
obtained for one copy of the respective subunits are discussed,
although all titratable residues in all subunits were considered in
the calculations.
3.2. Conformational variability in the Qo-site with bound CoQ
In the oxidised as well as in the reduced state of cytochrome
bc1with CoQ in the Qo-site, both Qo-site conformations are
populated. There are redox-linked differences in the population
of the two conformations: in the oxidised complex, almost only
the Glu-b conformation is populated in the physiological pH
range (Fig. 4A). In the reduced complex, in contrast, also the
Glu-FeS conformation is populated to a considerable degree,
namely to about 30% (Fig. 4B). These results indicate that the
two Qo-site conformations observed in the crystal structures of
cytochrome bc1from S. cerevisiae have sufficiently similar
energies to both be populated when CoQ is bound to the Qo-site.
Furthermore, the population of the two conformations depends
on pH as well as on the redox state of the complex.
In order to test our method for calculating the populations of
the two different conformations, we have performed calcula-
tions on the stigmatellin-inhibited and the HDBT-inhibited
complex. In these calculations, the inhibited complexes were
allowed to adopt either of the two Qo-site conformations. In
agreement with the available crystallographic data, we found
that the stigmatellin-inhibited complex populates exclusively
Fig. 4. Population of the different Qo-site conformations Glu-b and Glu-FeS. (A) Completely oxidised cytochrome bc1with oxidised and deprotonated CoQ in the Qo-
site. (B) Completely reduced cytochrome bc1with reduced and protonated CoQ in the Qo-site.
208 A.R. Klingen et al. / Biochimica et Biophysica Acta 1767 (2007) 204–221