Báo cáo Y học: Exploring the primary electron acceptor (QA)-site of the bacterial reaction center from Rhodobacter sphaeroides Binding mode of vitamin K derivatives pptx

To investigate the binding mode of these qui-nones to the QAbinding site we have determined the binding free energy and charge recombination rate from QA–to D+ kAD of 29 different 1,4-na

Exploring the primary electron acceptor (QA)-site of the bacterial

Binding mode of vitamin K derivatives

Oliver Hucke, Ralf Schmid and Andreas Labahn

Institut fu¨r Physikalische Chemie, Albert-Ludwigs-Universita¨t Freiburg, Germany

The functional replacement of the primary ubiquinone (QA)

in the photosynthetic reaction center (RC) from

Rhodo-bacter sphaeroideswith synthetic vitamin K derivatives has

provided a powerful tool to investigate the electron transfer

mechanism To investigate the binding mode of these

qui-nones to the QAbinding site we have determined the binding

free energy and charge recombination rate from QA–to D+

(kAD) of 29 different 1,4-naphthoquinone derivatives with

systematically altered structures The most striking result

was that none of the eight tested compounds carrying methyl

groups in both positions 5 and 8 of the aromatic ring

exhibited functional binding To understand the binding

properties of these quinones on a molecular level, the

structures of the reaction center-naphthoquinone complexes

were predicted with ligand docking calculations All protein–

ligand structures show hydrogen bonds between the

carbonyl oxygens of the quinone and AlaM260 and

HisM219 as found for the native ubiquinone-10 in the X-ray structure The center-to-center distance between the naph-thoquinones at QAand the native ubiquinone-10 at QB(the secondary electron acceptor) is essentially the same, com-pared to the native structure A detailed analysis of the docking calculations reveals that 5,8-disubstitution prohibits binding due to steric clashes of the 5-methyl group with the backbone atoms of AlaM260 and AlaM249 The experi-mentally determined binding free energies were reproduced with an rmsd of  4 kJÆmol)1in most cases providing a valuable tool for the design of new artificial electron accep-tors and inhibiaccep-tors

Keywords: ligand docking; structure activity relationship; bacterial reaction centers; Rhodobacter sphaeroides; infrared spectroscopy

The photosynthetic reaction center (RC) of the purple

bacterium Rhodobacter sphaeroides (R sphaeroides) is an

intrinsic membrane protein complex that performs the

conversion of light energy into chemical energy A

complex framework of redox cofactors is buried in the

protein matrix The cofactors are arranged in two

branches, the active A-branch and the inactive B-branch,

showing nearly twofold symmetry (reviewed in [1,2])

Following the absorption of a photon, an electron is

transferred within 200 ps from the bacteriochlorophyll

dimer, the primary donor D, via a bacteriochlorophyll

monomer (BA) and a bacteriopheophytin (FA) to a tightly

bound ubiquinone molecule (QA), forming the first stable

charge separated state D+QA– The subsequent electron

transfer step from QA– to QB proceeds on a slower time

scale ( 200 ls) After rereduction of the photooxidized primary donor by a soluble cytochrome c2 the light-induced electron transfer leads to the formation of the doubly reduced QB and concomitant binding of two protons from the cytoplasmic side of the membrane The ubiquinol dissociates from the RC and is reoxidized by the cytochrome bc1complex releasing two protons to the periplasmic side of the membrane The net result of these reactions is a transmembrane pH difference that drives ATP synthesis

In vitro, in the absence of both, an external reductant for the primary donor and the secondary quinone, the charges

on QA–and D+recombine with the rate kAD This reaction was subject to numerous spectroscopic studies One important approach to investigate this electron transfer mechanism in wild-type reaction centers was pioneered by Okamura et al [3,4] They developed a method of ubiquinone removal and readdition of ubiquinone or any other synthetic quinone Gunner et al [5] measured the temperature dependence of kAD in RCs with different anthra-, benzo- and naphthoquinone derivatives at QA Most of these compounds display a low midpoint redox potential in situ compared to the native UQ-10, leading to

an increase of the free energy difference between the D+QA–

and the ground state DQA Evaluating the free energy dependence of kAD Gunner et al [5] deduced that the charge recombination from QA– in native RCs is an activationless process More recently, the quinone replace-ment method was used to derive thermodynamic param-eters for that reaction [6] and to study the forward electron

Correspondence to A Labahn, Institut fu¨r Physikalische Chemie,

Albert-Ludwigs-Universita¨t Freiburg, Albertstr 23a, D-79104

Freiburg, Germany Fax: + 49 761 203 6189,

Tel.: + 49 761 203 6188, E-mail: labahn@uni-freiburg.de

Abbreviations: UQ-10, ubiquinone-10; NQ, 1,4-naphthoquinone; RC,

reaction center; B, Blastochloris; R, Rhodobacter; Q A , primary electron

acceptor; Q B , secondary electron acceptor; LDAO,

lauryldimethyl-amine-N-oxide; DAD, diaminodurene.

Note: web page available at http://pc2-em6.physchem.uni-freiburg.de/

Andreas/homepage

(Received 13 August 2001, revised 14 November 2001, accepted

23 November 2001)

transfer to the primary ubiquinone [7–9] Ka´lma´n &

Maro´ti used the quinone reconstitution method to study

the proton binding kinetics and stoichiometry associated

with the reduction of QA[10,11] It was shown that both

processes are controlled by protonatable residues in the

interior of the protein Measuring the delayed fluorescence

of the excited dimer, D*, in RCs with different quinones as

QATurzo´ et al [12] derived an empirical relation between

the in situ free energy of the primary quinone and the

charge recombination rate, providing the free energy levels

of the corresponding charge separated states Similar

quinone reconstitution experiments have been performed

for the QBsite [13–15] Palazzo et al [14,15] incorporated

the reaction centers into lipid vesicles and measured the

temperature dependence of the charge recombination rate

from QB– to D+ Based on a detailed analysis they

determined the binding free energy, enthalpy and entropy

of the ubiquinone to the QB-site

Methyl substituted 1,4-naphthoquinones have gained

substantial interest as they bind tightly to the QA site,

enabling its functional reconstitution while retaining the

native ubiquinone at QB [16,17] The difference in the

semiquinone anion spectra allows the direct monitoring of

the electron transfer from QAto QBin the VIS region with

transient absorption spectroscopy In addition, these

sub-stitutions change the driving force of the electron transfer reaction from QA–to QB(and thus the electron transfer rate and equilibrium) but they affect neither conformational changes nor the protonation rates or protonation equilibria near QB Graige et al [18] and Li et al [19,20] applied this method to study the effect of the driving force on the first electron transfer to QB Similarly, the mechanism of the proton-coupled electron transfer reaction [QA–QB–+ H+ fi

QA(QBH)–] was elucidated by Okamura and coworkers They showed that this reaction is a two-step process in which fast protonation precedes rate-limiting electron transfer [17,21,22] Moreover, methyl substituted naphtho-quinones play an important role in reaction centers of other photosynthetic organisms For instance, 2-methyl-3-(isoprenyl)(7)9)1,4-naphthoquinone (menaquinone) was identified as the primary electron acceptor in the reaction centers from Blastochloris viridis [23], Chloroflexus auran-tiacus[24] and in the photosystem I of green plants [25–27] Hence, a systematic variation of the redox potential of the naphthoquinone compounds and a detailed knowledge of their binding properties are of critical importance for the investigation of electron transfer reactions in photosynthetic reaction centers

In this work, we investigated the binding properties of 29 vitamin K derivatives with respect to the QA site of the reaction center from R sphaeroides (see Table 1, Scheme 1) Their midpoint redox potentials were altered by varying both, the number and the position of methyl groups at the ring system In some cases, additionally a hydrocarbon tail was introduced in position 3 to improve the binding affinity to QAin analogy to ubiquinone [28] Light-induced FTIR difference spectroscopy was used to detect structural changes of the binding pocket upon binding of the different quinones We measured the dissociation constants

Kd of these compounds and compared our results with those from ligand docking calculations The calculated structures of the quinone-protein complexes provide insights into the aspects that govern the binding of quinones to the QA site, allowing to test whether their positions at QA are identical with that of ubiquinone-10

A preliminary account of this work has been presented elsewhere [29]

M A T E R I A L S A N D M E T H O D S

Quinone-depleted reaction centers Reaction centers from the strain R sphaeroides R-26 were isolated and purified in lauryldimethylamine-N-oxide (LDAO) from photosynthetically grown cells following the procedure of Feher & Okamura [30] The ratio of absorbance (A280/A802) of the purified RCs was < 1.25 QA and QBwere removed from RCs according to the method

of Okamura et al [3] The residual quinone content was 0.05–0.1 mol Q per mol RC

Quinones 1,4-Naphthoquinone, 2-methyl-1,4-naphthoquinone and 2-methyl-9,10-anthraquinone were purchased from Aldrich 5,6,7,8-Tetramethyl-1,4-naphthoquinone was newly synthe-sized from the Diels–Alder adduct of 2,3,4,5-tetramethyl-thiophene-1,1-dioxide and 1,4-benzoquinone according to a

Table 1 Overview of the 1,4-naphthoquinone compounds used in this

work The structure of the naphthoquinones is shown in Scheme 1.

Quinones without undecyl tail

Quinones with undecyl tail

2,5,8-Trimethyl-3-undecyl-NQ 258TM3UNQ

2,6,7-Trimethyl-3-undecyl-NQ 267TM3UNQ

2,5,6,7,8-Pentamethyl-3-undecyl-NQ PM3UNQ

literature procedure [31]

5,6,7,8-Tetramethyl-1,4-naphtho-quinone formed yellow needles, 1H-NMR (250 MHz,

CDCl3, dH): 2.28, s, 6H; 2.53, s, 6H; 6.71, s, 2H; mp: 184–

185°C; elemental analysis for C14H14O2 requires: C,

78.48% H, 6.57%; found: C, 78.43% H, 6.74% The

synthesis of all other quinones was as described previously

[31] Stock solutions of the quinones (1 lM)20 mM) were

prepared in dioxan and stored at 4°C

Determination of dissociation constants,Kd,

and charge recombination rates,kAD

Reconstitution of the quinones into the QA-site was

accomplished by adding small amounts of the stock solution

(1 lM)20 mM) to quinone-depleted RCs (20–300 nM)

sus-pended in 10 mMMops [3-(M-morpholino)propanesulfonic

acid], 50 mM KCl, 0.04% dodecyl-b-D-maltoside at

pH¼ 7.2 The system was equilibrated for at least 25 min

[T¼ (295 ± 2) (K)] Transient absorbance changes were

recorded on a spectrometer of local design [32] Charge

recombination kinetics were measured by monitoring the

change of the donor absorbance at 865 nm following a

single laser flash The rate constant kADwas obtained from

a single exponential fit to the data using the software

package PEAKFIT (version 4.0, SPSS Inc.) on an

IBM-compatible PC The occupancy of the QA-site corresponds

to the amount of RCs in the D+QA– state Its value was

determined from the amplitude of the charge recombination

kinetics at t¼ 0 measured on the time scale of kADÿ 1relative

to the amount of bleaching of RCs with a fully occupied QA

site (see [4] for details) To account for the 5–10% RCs

where ubiquinone remained at QAafter quinone removal

the amplitude of the signal was corrected accordingly

The binding affinity of quinones to the QA-site of the

reaction center can be described with the model of ligand

binding to a population of single, noninteracting QA-sites

As naphthoquinone compounds bind functionally only to the QAsite [5,16] the apparent dissociation constant Kdcan

be obtained from Eqn (1) in case of [Q]0 [QRC]:

DA865 ¼ DA

max

865½QŠ0

½QŠ0 þ Kd

ð1Þ

where DA865 corresponds to the concentration of bound

Q at the RC, [QRC] Its value was determined from the absorbance changes of the donor recovery due to the formation of D+QA– The dissociation constant, Kd, was determined by fitting the absorbance change at 865 nm as a function of the initial quinone concentration, [Q]0 with

DAmax865 and Kdas adjustable parameters

The condition of [Q]0 [QRC] essentially limits the applicability of the assay to Kd>100 nM For smaller values the amplitude of the charge recombination kinetics can not be determined accurately with the experimental set

up described above In this case 2-methyl-9,10-anthraqui-none (0.01 mM) with the inhibition constant Ki¼ 20 nMwas used as a competitive inhibitor similar as described previ-ously [28] This is suitable because charge recombination for this anthraquinone occurs in the microsecond range and, hence, does not interfere with the observation of the D+QA–

formation of the naphthoquinones In this case the disso-ciation constant was determined from a two parameter least squares fit of the absorbance change vs [Q]0according to Eqn (2):

max

865½QŠ0

½QŠ0 þ ½IŠ0Kd=Ki

ð2Þ where [I]0is the initial concentration of the inhibitor

The binding free energies Based on the work by Warncke & Dutton [33], we applied a correction method to determine the true binding free energy,

DG0bind, as a measure for the direct interactions between the quinones and the protein at the QAsite The dissociation constant, Kd, is correlated to the apparent binding free energy:

DG0

where T is the temperature and R the gas constant The apparent binding free energy, DG0app, contains contribu-tions from specific interaccontribu-tions between the quinone and the quinone binding site as well as unspecific hydrophobic interactions between the quinone and the nonpolar protein detergent micelles Hence, this energy can be represented as:

DG0app ¼ DG0

bind þ DG0

trans ð4Þ The transfer free energy, DG0

trans, describes the free energy change of the quinone transfer from the aqueous bulk phase

to the nonpolar protein/detergent micelles DG0

trans can be approximated by the distribution of the quinone between water and an apolar solvent, e.g cyclohexane, which is given

by the partition coefficient Pcw

DG0trans ÿRT lnPcw ð5Þ

Scheme 1 The structure of the naphthoquinone compounds The

rota-tional axis used for the description of the predicted

RC-naphthoqui-none complexes are indicated.

Hence, the true binding free energy, DG0bind, is given by

Eqn (6):

DG0bind DG0app þ RT lnPcw ð6Þ

Pcwvalues of quinones for the system cyclohexane/water

were estimated using the software package MOLECULAR

MODELING PRO(Chem SW, Fairfield, CA, USA) based on

the method of Ghose & Crippen [34,35]

Ligand docking calculations

The docking calculations were performed with FLEXX, a

program designed for the docking of small to medium sized

organic molecules into protein binding sites [36] During the

docking procedure, the protein is considered as rigid,

whereas the ligand conformation is flexible This is realized

through allowing rotations around acyclic single bonds of

the ligand structure Bond lengths and angles are kept

constant as given in the input structure A relatively soft

atom model is used byFLEXXto compensate for the rigidity

of the binding site, i.e small overlap of the ligand with the

receptor is tolerated by the program

The docking algorithm incorporated inFLEXXis based on

the chemical interactions of ligand and receptor: For the

computation of ligand placements geometrically restrictive

interactions are used (mainly hydrogen bonds and salt

bridges) Interaction geometries were deduced from the

analysis of crystallographic data The computed placements

were optimized with respect to the empirical scoring

function ofFLEXX, which estimates the binding free energy,

DG0bind, of the ligand receptor complex:

DG0bind ¼ DG0

translat þ DG0

rotNrot

þ DG0 hb

X

neutral Hÿbonds

fðDR; DaÞ

þ DG0 io

X

ionic interactions

fðDR; DaÞ

þ DG0 aro

X

arom interactions

fðDR; DaÞ

þ DG0 lipo

X

lipoph: interactions

fðDRÞ ð7Þ

Here, the terms DG0hb, DG0io, DG0aro and DG0lipo are the

interaction energies for neutral hydrogen bonds, ionic

interactions, aromatic interactions (aromatic interactions

as considered byFLEXXare interactions of the electrostatic

quadrupole of aromatic rings with permanent dipoles (for

example of amide bonds), the quadrupoles of other

aromatic rings and the induced dipoles of methyl groups)

and lipophilic contacts between ligand and receptor,

respectively, if ideal interaction geometries are assumed

The functions f (DR,Da) and f *(DR) penalize deviations

from these geometries (DR, deviation from ideal distance;

Da, deviation from ideal angular geometry) The values of

DG0

translat and DG0

rotNrot consider the loss of translational and rotational freedom of the entire ligand molecule and the

freezing of rotational degrees of freedom of the ligand

structure upon binding, respectively (DG0 , loss of binding

energy due to fixation of rotation around one rotatable bond; Nrot, number of rotatable bonds)

Depending on the number of possible ligand confor-mations, the docking calculations resulted in a set of up to

 200 possible protein ligand complexes per ligand These solutions were ranked according to the calculated binding free energy Unless stated otherwise, only the best placement (Ôplacement 1Õ), displaying the smallest value

of the binding free energy, was considered for further analysis

The binding site of the primary quinone in the reaction center protein from R sphaeroides was determined with the molecular modeling packageWHATIF[37] based on the X-ray structure from Stowell et al [38] (RCSB PDB code 1AIJ) It included all amino acids with at least one atom lying within a distance of 6.5 A˚ from the ubiquinone-10 molecule in the X-ray structure (a larger binding site had

no effect on the results of the docking calculations) In addition, four water molecules (numbers 64, 71, 409, 410

in the PDB file), the nonheme iron atom, parts of the bacteriochlorophylls and the bacteriopheophytins located

in the active branch of the reaction center are found within this cutoff distance and were therefore considered in the calculations

FLEXX uses an united atom model for all nonpolar hydrogen atoms, whereas polar protons are explicitly taken into consideration Where unambiguously clear, the posi-tions of the protons in the protein binding site were automatically assigned by FLEXX In cases of ambiguities (hydrogens of the hydroxyl groups, the N-bound proton of the histidine side chain) they were determined with WhatIf The geometries of the ligand structures were optimized with the MM+ force field of theHYPERCHEM(Hypercube Inc.) software prior to use as input files for the docking calculations

Sample preparation for FTIR difference spectroscopy

To reconstitute the reaction center with 1,4-naphthoqui-none derivatives as primary electron acceptor, 3.5 nmol of quinone-depleted RCs were dissolved in 3 mL buffer containing 10 mM Mops, 50 mM KCl and 0.04% dode-cyl-b-D-maltoside, pH¼ 7.0 followed by the addition of

20 lL of a 10-mM stock solution of the corresponding quinone in ethanol The samples were incubated at room temperature for a minimum period of 2 h and then concentrated with Microcon YM-100 centrifugal filter devices (Millipore Corp., cut-off molecular mass 100 kDa)

at 3000 g and 4°C to two aliquots of  50 lL volume each

To avoid partial denaturation of the RCs due to high ionic strength and detergent concentration upon the final concentration step, the samples were diluted by a factor

of 10 with a buffer without detergent containing 1 mM

Mops and 5 mMKCl at pH¼ 7.0 The second concentra-tion step yielded two highly concentrated samples of reconstituted RCs with a final volume of about 10 lL each One of them was used for FTIR difference spectro-scopy (see next paragraph) whereas the remaining sample was taken to determine the occupancy of the QA-site and the charge recombination kinetics with transient absorption spectroscopy under the same conditions except that the redox mediator diaminodurene (DAD) and sodium ascor-bate were omitted

FTIR difference spectroscopy

FTIR difference spectroscopy was performed as described

by Breton et al [39] with minor modifications The sample

( 10 lL) containing about 1.5 nmol reconstituted RC was

placed onto the depression of a CaF2 window After

addition of 5 lL of the aqueous solution of the redox

mediator DAD (2.5 mM) and sodium ascorbate (1.25 mM)

as reductant for the primary donor, the droplet was dried

under a smooth stream of nitrogen Before complete

dryness, the RC film was sealed with a second CaF2

window, yielding a sample with a thickness of a few

microns, minimizing the water absorption The two

win-dows were fixed by a metal mounting and placed in the

sample chamber (T¼ 2 ± 0.2 °C) FTIR spectroscopy

was performed with a Bruker IFS 66 V/S FTIR

spectrometer Light-minus-dark FTIR difference spectra

were obtained by recording the spectrum of the sample

under continuous illumination at a wavelength of 590 nm

and subtracting the spectrum measured in the dark For

each difference spectrum 1920 interferograms were

accu-mulated To improve the signal to noise ratio,  20

illumination cycles of the sample were averaged The FTIR

difference spectra were normalized with the vector

normal-ization method based on the spectral regions between 1500

and 1560 cm)1 and between 1670 and 1750 cm)1 The

difference spectra obtained by this method (i.e in the

presence of DAD/sodium ascorbate, leading to a fast

reduction of the primary donor after electron transfer to

QA) show exclusively the absorption changes upon the

reduction of the primary quinone, designated ƠQA–/QA

difference spectraÕ

R E S U L T S

Experimentally determined binding free energies

The methyl substituted naphthoquinones were

character-ized in terms of the midpoint redox potential E1/2, charge

recombination rate kADand dissociation constant Kd(see

Table 2) The in vitro midpoint redox potentials decrease

with increasing number of alkyl substituents due to their

inductive effect Similarly, the RCs with naphthoquinones

at QAdisplayed low in situ redox potentials leading to an

increase of the free energy difference between the states

D+QA– and DQA compared to RCs with native UQ-10

as QA Therefore, the charge recombination rates kAD

were significantly faster for all RCs with at least two

methyl groups in the aryl ring due to the charge

recombination by a thermally activated route via the

D+/A– state [5,40]

As previous studies revealed that even nonquinonic

compounds bind to the QAsite [33] indicating less specific

interactions between ligand and protein, the dissociation

constants of methyl substituted quinones are expected to be

dominated by the quinone polarity which determines the

transfer free energy However, the binding affinity is

strongly affected by the substitution pattern of the aryl ring

(Table 2) The most striking result of our study was

obtained for all naphthoquinones with methyl groups in

position 5 and 8 These substituents drastically weaken the

association by at least a factor of 400 which represents the

current limit for K determination

To determine the actual interaction energy between quinone and RC, the apparent binding free energies were corrected for the transfer free energy (Eqn 6) by using calculated values for the corresponding partition coefficients between cyclohexane and water (Table 2) In case of the naphthoquinone compounds without an undecyl chain in position 3 the results agree fairly well with the experimental values However, the Pcwvalues of the long-tail-derivatives are most likely overestimated by the Ghose–Crippen method as for each CH2group of the aliphatic chain the same increment was added to log P The corresponding binding free energies ranged from)13.1 to +6.2 kJỈmol)1

for the compounds with an undecyl chain whereas the other quinones exhibited values from)32.0 to )21.0 kJỈmol)1 Rigid quinone binding site

A main assumption in the ligand-docking calculations was that the structure of the binding pocket does not change upon binding of the different vitamin K derivatives compared to the X-ray structure determined with UQ-10

as QA The FTIR QA–/QAdifference spectra show signals of both, the quinone and the adjacent region of the RC With isotope labeled ubiquinone and vitamin K1 as QABreton

et al [41] showed that the spectra in the range of 1750–

1670 and 1560–1500 cm)1are dominated by nonquinonic contributions resulting from the response of the protein to the QA-reduction Hence, these absorption bands are indicative for the interaction between the quinone and the protein binding pocket Therefore, structural changes upon ligand binding are expected to alter specifically the vibrational frequencies and/or intensities, making FTIR difference spectroscopy an attractive method for probing the effect of the different quinones on the structure of the binding site

Figure 1 shows the QÿA=QA difference spectra obtained for three of the naphthoquinones which are objects of this work compared to those of UQ-10 and vitamin K1 In the nonquinonic regions, the band shapes and vibrational frequencies exhibit a high similarity for all quinones No evidence for a change in the response of the protein due to

QA– formation was found in these spectra supporting the assumption of a rigid binding site

Docked structure with UQ-10

To test the docking algorithm ofFLEXXwith our system, we calculated the UQ-10/RC complex, as shown in Fig 2 With respect to the quinone head group the calculated structure agrees very well with the X-ray structure [38] (rmsd

of the head groups, 0.29 A˚) In contrast, the positions of the isoprenoid chains differ significantly beyond the first two isoprene units As both, the binding affinity and the midpoint redox potential of the quinone are mainly determined by the head group [28,33], these results encour-aged us to extend the calculations to the naphthoquinone compounds

Docked structures with naphthoquinone compounds Using our set of 29 naphthoquinone compounds (Table 1, Scheme 1) we computed the quinone-reaction center complexes For most naphthoquinones the ligand was

successfully docked into the binding pocket, except for 7 of

the 8 compounds containing methyl groups in both

positions 5 and 8

Quinone orientation within the binding pocket

In all predicted structures the 1,4-naphthoquinones share

essentially the same orientation of the naphthoquinone ring

system, i.e the calculated protein–ligand complexes show

two hydrogen bonds between the carbonyl oxygens of the

quinones and the amide nitrogen of AlaM260 and the

imidazole nitrogen of the HisM219 side chain as found for

UQ-10 in the X-ray structure The position and the

orientation of the quinone rings are similar to those of UQ-10 (Fig 3) The aromatic rings of the naphthoquinones are directed towards the interior of the binding pocket (i.e towards MetM262 and AlaM245) Up to eight specific interactions of the aromatic rings with side chain methyl groups (AlaM248, AlaM249, AlaM260, Cc2of ThrM222 and IleM265), aromatic rings (TrpM252) and backbone amide bonds (AlaM248, AlaM260, ThrM261) of surround-ing amino acids were assigned byFLEXX

In case of the quinone oriented with two hydrogen bonds

of the carbonyl oxygens and the aromatic ring directed towards the interior of the binding site, the ring system can assume two orientations They can be matched by a rotation

Table 2 Comparison of the physicochemical properties of naphthoquinone compounds with different number and position of alkyl substituents in the ring system Charge recombination rates from D+Q A

–

to DQ A , k AD , were determined in one-quinone RCs monitored via the absorbance change of the rereduction of D + following a single laser flash The dissociation constants K d were derived from a plot of the amount of bleaching vs the initial quinone concentration according to Eqns (1,2) From an error analysis the accuracy of –log K d was estimated to ± 0,2 Using the experimental values for K d the apparent binding free energies, DG0app, were calculated following Eqn 3 To account for the distribution of the quinone between the water and the apolar protein detergent micelles the cyclohexane water partition coefficients, P cw , were estimated with the Ghose–Crippen method [34,35] and used for correcting the apparent binding free energies (Eqn 6) The corresponding binding free energies DG 0

bind (Exp.) were compared to the data from ligand docking calculations [designated DG 0

bind ( FLEXX )] The predicted complexes were analyzed in terms of the distance differences along the axis from the primary donor D to the primary electron acceptor Q A with respect to that of 1,4-naphthoquinone [Dr(NQ)] Experimental conditions: 20–300 n M quinone-depleted RCs, 30 n M –100 l M naphthoquinone, 10 m M Mops, 50 m M KCl, 0.04% dodecyl-b- D -maltoside, pH ¼ 7.2 (T ¼ 293 K) See Table 1 for abbreviations NF, no formation of Q A

–

was detected; ND, not determined; NP, no acceptable placement was found by FLEXX

Quinone

E 1/2 a

(mV)

k AD

(s)1)

Dr(NQ) (A˚) –log K d

DG 0 app

(kJÆmol)1) log P cw log P cwb

DG 0 bind (kJÆmol)1) Exp FLEXX D d

a In vitro midpoint redox potentials for the redox couple Q/Q – reported against ferrocene as internal standard measured in DMF [31,53].

b

Experimental data taken from [54].cOnly one hydrogen bond between the carbonyl oxygen of the quinone and surrounding amino acid residues was found.dD  DG 0

bind ðFLEXXÞ ÿ DG 0

bind ðExp:Þ.

of 180° on the x-axis (Fig 3) but are only distinguishable for

asymmetrical methyl substitution patterns We arbitrarily

chose the naphthoquinone orientation with the hydrogen

bonds between the C1carbonyl and HisM219 and the

C4carbonyl and AlaM260 as reference (designated

reference orientation) In the calculated structures, for all

naphthoquinones with this reference orientation the methyl

groups at a specific position of the naphthoquinone ring

system have very similar environments which we used for

the definition of methyl group positions within the QA

binding site (Table 3) Each position is defined by the

contacts that are observed in the predicted structures

between a specific methyl group and the RC atoms It is

named according to the number of the C-atom of the

naphthoquinone to which the methyl group is bound For

instance, the¢position 5¢ is formed by all RC atoms showing

contacts to the 5-methyl group A 180° rotation on the

x-axis moves a methyl group from position 5, 6 and 2 to

position 8, 7 and 3, respectively

The detailed analysis of the placements of the tailless

naphthoquinones reveals that the presence of specific methyl

substituents favors one of the two possible orientations

with respect to the x-axis: In all placements with 5- or

8-substitution (5MNQ, 25DMNQ, 28DMNQ, 235TMNQ)

the corresponding methyl group was found at position 8 The best placements (with regard to theFLEXXscore of the binding free energy) of these quinones with the methyl group at position 5 in the binding pocket show binding free energies which deviate from the optimal value correspond-ing to placement 1 by 3.9, 1.7, 6.9 and 3.5 kJÆmol)1, respectively In the best placements of quinones with 6- or 7-methyl substitution (6MNQ, 26DMNQ, 27DMNQ, 236TMNQ) the methyl group was found at position 6 Compared to these placements structures with the methyl group at position 7 of the binding site exhibit less favorable energies of 3.1, 6.4, 0.1 and 2.4 kJÆmol)1, respectively The favored placement of 2MNQ shows the methyl group at position 2 and the corresponding binding free energy is 4.1 kJÆmol)1lower than that of the best placement of this compound, rotated by approximately 180° on the x-axis (which places the 2-methyl group at position 3) This also accounts for the small energy differences of 1.7 and 0.1 kJÆmol)1 between the two rotated orientations of 25DMNQ and 27DMNQ, respectively Here, the placement

of the 5-methyl group to position 8 and the 7-methyl group

to position 6 leads to the unfavorable position 3 of the 2-methyl group

Evaluating the docked structures of the naphthoquinones with an undecyl tail we found for 2UNQ the rotated orientation of the quinone head group with hydrogen bonds between the C-carbonyl group and AlaM260 and

Fig 1 QÿA=Q A FTIR difference spectra of quinone-depleted reaction

centers from R sphaeroides reconstituted with UQ-10, vitamin K 1

(Vit K 1 ), 2,3,5-trimethyl-1,4-naphthoquinone (235TMNQ),

2,8-dimethyl-3-undecyl-1,4-naphthoquinone (28DM3UNQ) and

2,3,6-trimethyl-1,4-naphthoquinone (236TMNQ) A total of 40 000

interferograms were averaged for each spectrum As shown for vitamin

K 1 and UQ-10, the regions of 1750–1670 cm)1and 1560–1500 cm)1

are free of quinonic contributions (indicated as nonquinonic) [41].

Differences in the structure of the binding site due to structural

differences of the primary quinone at Q A are expected to alter the

vibrational frequencies and intensities of the spectra in these regions.

See text for details of conditions a.u absorbance units.

Fig 2 Comparison of the ubiquinone-10 (UQ-10) position at Q A in the photosynthetic reaction center from R sphaeroides obtained from docking calculations with the X-ray structure Carbon atoms of the UQ-10 from [38] and the docked quinone are depicted in black and green, respectively Amino acids (blue) surrounding UQ-10 were taken from the crystal structure ([38], PDB file 1AIJ) The nitrogen, oxygen and hydrogen atoms are drawn in light blue, red and white, respec-tively For sake of clarity the isoprenoid chains were truncated The dashed lines indicate the hydrogen bonds between the carbonyl oxy-gens of the quinones and AlaM260 and HisM219 The quinone head groups fit with an rmsd of 0.29 A˚.

the C4-carbonyl group and HisM219 However, in all six

2-methyl-3-undecyl-1,4-naphthoquinone derivatives the same

quinone head group orientation was observed as in the

reference orientation of the tailless quinones For these

compounds the rotated reference orientation is prohibited

as the presence of the 2-methyl group prevents a necessary

adjustment of the undecyl tail conformation directing the

hydrocarbon tail through the opening of the binding pocket

towards the protein exterior For the same reason, in the

predicted structure with 25DM3UNQ the 5-methyl group is

not found at position 8 of the binding site (as found for all

5-methylated tailless naphthoquinones, see above) but in the

less favorable position 5

It should be mentioned that for most of the tailless

quinones placements with the aromatic rings directed

towards the opening of the binding pocket were also

computed This orientation is obtained by a rotation of

 180° on the y-axis relative to the reference orientation,

designated Ôantireference orientationÕ) However, the

bind-ing free energies are significantly larger by an average of

+5.4 kJÆmol)1 compared to the highest ranked

place-ments indicating that this orientation is less favorable The

analysis of the different free energy terms of the scoring

function reveals that the main contribution ( 47%) of

this increase in the binding free energy results from a loss

of aromatic interactions Orientation of the aromatic

naphthoquinone rings to the opening of the binding site

reduces the average number of these interactions from 5.4

to 1.9 suggesting that these contacts play an important

role for naphthoquinone binding to the Q binding site

Other contributions to the change in the binding free energy include the loss of lipophilic contact area ( 15%) and deviations from the ideal hydrogen bond geometry ( 38%)

Distances from the naphthoquinone at QAto the secondary ubiquinone QBand the primary donor D According to the Marcus theory the rate of electron transfer between two molecules depends on three factors: the overlap of the electron densities (wavefunctions) of the two molecules, the difference in redox potential of the molecules (corresponding to the free energy difference) and the reorganization energy (reviewed in [42]) The most critical parameter in determining the electron transfer rate

is the electron density overlap which was found to depend exponentially on the distance of the reactants [43] The center-to-center distances are measured from the middle of the quinone rings and the center of the special pair (defined as the middle of the line connecting the Mg atoms) Due to different methyl substitution patterns the use of the edge-to-edge values produces incomparable values The distances between the different naphthoqui-none compounds at QAand the native ubiquinone at QB

range from 19.4 A˚ to 20.0 A˚ compared to 19.6 A˚ as found

in the X-ray structure of native RCs with ubiquinone at

QA Thus, within the resolution of the X-ray diffraction data both structures are identical However, in terms of the

Table 3 Possible orientations of methylated 1,4-naphthoquinones in the

Q A binding site as determined with FLEXX The methyl groups of the different methylated 1,4-naphthoquinones occupy distinct positions within the Q A binding site as defined by their contacts with amino acid atoms of the reaction center Each position number refers to the C atom number of the naphthoquinone ring (see Scheme 1) to which the corresponding methyl group is bound It was assumed that the naphthoquinone is located in the most common orientation with the two hydrogen bonds between the C 1 carbonyl and the C 4 carbonyl formed to HisM219 and AlaM260, respectively (arbitrarily defined as the Ôreference orientationÕ).

Position within the

TrpM252 C d1 , N e1

IleM265 C d1 , C c1 , C a

AlaM260 C, C a , N, O

HOH 64, 409 O

MetM262 C a , C b , C c , S d , C e

MetM262 S d , C e

Fig 3 Calculated position of

2,6-dimethyl-3-undecyl-1,4-naphthoqui-none (26DM3UNQ) in the Q A binding pocket as an example of the

predicted structures with 1,4-naphthoquinones as primary acceptors of

the RC from R sphaeroides The ring systems of both compounds

show a high similarity in terms of the position and orientation x and y

denote rotational axis used for the description of the quinone

orien-tation within the Q A binding site See Fig 2 for coloring.

distance between the primary donor and QAsome

naph-thoquinones are further apart from the donor (up to 1.1 A˚

in case of 5-methyl-1,4-naphthoquinone) than the native

ubiquinone affecting significantly the electron transfer rate

kAD(see Discussion)

Evaluation of binding free energies

The binding free energies of the protein-ligand complexes

were estimated with theFLEXXscoring function (Eqn 7)

For all functionally binding naphthoquinones without an

undecyl tail the values range from)25.8 to )20.0 kJÆmol)1

The contributions of the hydrogen bonds, aromatic

inter-actions and lipophilic contacts amount to 32,  13 and

 55%, respectively In case of 58DMNQ a protein-ligand

complex was predicted although with the charge

recombi-nation assay no binding was observed However, the

binding free energy yielded a significantly higher value of

)16 kJÆmol)1 mainly due to the loss of one of the two

hydrogen bonds This compound was therefore disregarded

in all further analysis

The calculated binding energies of the substituted

undecyl naphthoquinone derivatives range from)22.1 to

)14.9 kJÆmol)1 Based on the different terms in the scoring

function we deduced that the major component is the

lipophilic contact energy ( 67%) due to the large

hydrophobic surface of the alkyl chain Smaller

contribu-tions arise from the hydrogen bonds ( 24%) and

interac-tions between the aromatic rings of the naphthoquinones

and the residues of the binding site ( 9%) A detailed

analysis reveals that the binding free energies of the quinone

head groups (methyl substituted ring systems without

undecyl chain) lie only slightly (on average 1.3 kJÆmol)1)

above the energies found for comparable naphthoquinones

without an alkyl chain, showing that the presence of the

undecyl tail has no significant effect on the interaction

between the quinone head group and the protein binding

site

D I S C U S S I O N

Binding or nonbinding?

The experimental values for the dissociation constants

manifested that 5,8-disubstitution of the naphthoquinone

system prohibits binding to the QA site even in case of a

long tail in position 3 This result was rather surprising as

Warncke and Dutton found for 3-decyl substituted

ubi-quinone-0 and 2-methyl-1,4-naphthoquinone a decrease in

the dissociation constants by more than two orders of

magnitudes compared to the corresponding analogues

with a hydrogen at this position [28] Our findings agree

with previously published results obtained for 58DMNQ

[44] As only functional binding is detected with the

charge recombination assay it cannot be decided whether

5,8-dimethyl-1,4-naphthoquinone compounds were not

bound at QAor the structure of the RC-quinone complex

prohibits photoreduction However, the experimental data

coincide strikingly with the results of the docking

calcu-lations: For none of these compounds an acceptable

protein–ligand complex was found which strongly

sup-ports the idea of nonbinding to QA This can be pinned

down to steric reasons:

As described above, the methyl groups of all 5-substi-tuted quinones except that of 25DM3UNQ were found at position 8 within the binding pocket in the calculated complexes (Table 3) In these structures, the coordinates of the methyl group were practically identical leading to the same protein environment formed predominantly by the side chains of HisM219, IleM223, MetM262 and IleM265 (Table 3) We have constructed hypothetical protein-ligand complexes of the nonbinding compounds 58DMNQ, 258TMNQ, 2358TEMNQ and 258TM3UNQ based on the predicted structures with the corresponding binding analogues 5MNQ, 25DMNQ, 28DMNQ, 235TMNQ, 25DM3UNQ and 28DM3UNQ For this purpose, the ring systems of each of the 5,8-disubstituted naphthoqui-nones and the corresponding monosubstituted quinone in the computed protein-ligand complex were superimposed All resulting hypothetical structures share the same features (Fig 4) The additional methyl group shows an intolerable van der Waals overlap with either the backbone atoms of AlaM260 and AlaM249 in case of 58DMNQ, 258TMNQ, 2358TEMNQ whereas for 258DM3UNQ steric clashes with the side chains of HisM219, IleM223 and MetM262 are found This can not be avoided by a displacement of the quinone head group within the binding site, as the methyl group of the 5- or 8-monosubstituted quinone

is already in close contact with the adjacent part of the binding pocket restricting the positional freedom of the quinone

Naphthoquinone positions: implications for the charge recombination rates

A main assumption for comparing the different naphtho-quinone positions is that the original structure of the QA binding pocket remains unchanged in view of the drastic methods including the application of high concentrations of

Fig 4 Constructed placement of 5,8-dimethyl-1,4-naphthoquinone (58DMNQ) in the Q A binding site compared to the native structure For reasons of clarity, the part of the binding site formed by IleM223 and MetM262 is drawn schematically as blue curve The pink dashed lines symbolize steric clashes of the 5-methyl group with (mainly backbone) atoms of AlaM260 and AlaM249 prohibiting binding of this quinone See Fig 2 for color scheme.

ionic detergent (LDAO) and inhibitor (o-phenanthroline) to

remove the native ubiquinone from the binding site Breton

et al [45] measured the QA–/QAFTIR difference spectra with

native RCs having UQ-10 as QAand compared the result

with that of quinone–depleted RCs after reconstitution of

the QA-binding pocket with UQ-10 Within the noise level,

the two spectra were practically identical More recently,

Kuglstatter et al [46] determined the X-ray structure of the

photosynthetic RC from R sphaeroides reconstituted with

9,10-anthraquinone as QA to 2.4 A˚ resolution

Quinone-depleted RCs were prepared under the same conditions as

described in this work Within the resolution limit no

structural changes of the QAbinding pocket were observed

From our predicted placements it follows that the

position of 1,4-naphthoquinones within the QA binding

pocket of the reaction center varies depending on the

substitution pattern of the naphthoquinone This slightly

affects the distances of QA to other cofactors involved in

electron transfer reactions which may influence the rates of

these reactions The differences are neglectable for the

forward electron transfer from QA– to QB whereas the

deviations are more critical with respect to the distance

from the primary donor to QA According to the Marcus

theory, at room temperature the charge recombination rate

kADdepends on the reorganization energy, the standard

reaction free energy and the electronic coupling matrix

element (designated VR at the distance R between the

reactants) To estimate the effect of quinone relocation in

the calculated complexes on the rate kAD we use in a

simple model the expression for the distance dependence of

the electronic coupling matrix element (Eqn 8) by ignoring

any possible changes in the reorganization energy, the

driving force and the electron transfer pathway upon

substitution

V2R ¼ V2

Here, V0 is the maximum electronic coupling matrix

element, R is the distance between the reactants and b is

the transmissional coefficient For this quantity Moser et al

[43] have empirically determined a value of b¼ 1.4 A˚)1

The maximum difference with respect to the

donor-quinone distance was found between 5MNQ and 27DMNQ

(Table 2) The position of the latter compound was 1.1 A˚

closer to the donor leading to an approximate fivefold

increase in the charge recombination rate kAD With respect

to ubiquinone as found in the X-ray structure a mean

displacement of  0.6 A˚ away from the donor was

determined for the different naphthoquinone compounds

This displacement may account for a 2.4-fold decrease in

the rate kAD

Evidence for position-dependent influences on the rate

kADwas previously reported by Warncke et al [28] They

studied both menaquinone and ubiquinone compounds

with systematically altered hydrocarbon tail structures

Using the empirical relation for the distance dependence

of the electron transfer rate in proteins of Moser et al [43]

the relocation of the quinones was estimated to 0.8 A˚ and

0.6 A˚ along the line connecting the quinone and the primary

donor, respectively Similar values were derived fromFLEXX

calculations on quinones with systematically altered

hydro-carbon chain length (data not shown) Gunner et al [5]

proposed positional differences of about 1 A˚ to explain a

three to fourfold increased recombination rate for b- compared to a-substituted 9,10-anthraquinones More-over, in the X-ray structure of the reaction center from

R sphaeroides with 9,10-anthraquinone as QA [46] its position was found to be 1 A˚ displaced compared to that

of ubiquinone The docked anthraquinone-reaction center structure exhibits very similar results (data not shown)

Comparison of experimental and calculated binding free energies

1,4-Naphthoquinones without undecyl chain According

to our model, the free energies of the quinone transfer from the aqueous solution to the hydrophobic detergent and protein-detergent micelles were estimated by calculating the free energies for the transfer from water to cyclohexane The experimental binding free energies were corrected for these transfer energies to account for the tendency of the hydrophobic quinone compounds to accumulate within the hydrophobic micellar phase Although this correction is based on a relatively simple model, we achieved a reasonable agreement between theoretical and experimental binding free energy values (Table 2) The standard deviation

of the two data sets amounts to only 3.9 kJÆmol)1 1,4-Naphthoquinones with undecyl chain The experimen-tal binding energies of the naphthoquinones with a 3-undecyl chain display on average an offset of 14.4 kJÆmol)1 compared to the predicted values Warncke & Dutton [33] found empirically that the binding free energies (DG0

bind) of many quinones to the QAsite can be corrected with respect

to their hydrophobic transfer free energies (DG0trans) by applying a simple linear relationship yielding a measure for the direct interactions of the protein with the ligand In case

of ubiquinones with more than two isoprene unit tails, corresponding to a linear chain length of eight carbon atoms, this correction method failed [28] This was explained with the third and subsequent isoprene units being not completely removed from contact with the solvent There-fore, DG0trans is expected to be overestimated for naphtho-quinones with a hydrocarbon chain of 11 carbon atoms with our method as well Other possible sources of error include inaccuracies of the lipophilic contact energy by the scoring function ofFLEXXand of the calculated partition coefficients

Pcw Under our experimental conditions the apolar phase was not cyclohexan but consists of both, detergent micelles and mixed micelles of detergent and protein

Assuming that a systematic error in estimating the effective energy for the transfer of the undecyl-naphthoqui-none compounds from water to the mixed protein-detergent micelles accounts for the discrepancies, a simple offset correction based on the average values of the experimental and predicted values matched the two data sets with a standard deviation of 4.1 kJÆmol)1(data not shown)

Aromatic interactions of the naphthoquinone compounds with the QAbinding site

From our calculations it follows that the interactions between the aromatic rings of the tailless naphthoquinone derivatives and the protein play an important role with res-pect to the quinone orientation within the QAbinding site The binding free energy associated with these interactions