ORIGINAL PAPER

The structure of the periplasmic nickel-binding protein NikAprovides insights for artificial metalloenzyme design

Mickael V. Cherrier • Elodie Girgenti • Patricia Amara • Marina Iannello •

Caroline Marchi-Delapierre • Juan C. Fontecilla-Camps • Stephane Menage •

Christine Cavazza

Received: 18 January 2012 / Accepted: 23 March 2012 / Published online: 21 April 2012

� SBIC 2012

Abstract Understanding the interaction of a protein with

a relevant ligand is crucial for the design of an artificial

metalloenzyme. Our own interest is focused on the syn-

thesis of artificial monooxygenases. In an initial effort, we

have used the periplasmic nickel-binding protein NikA

from Escherichia coli and iron complexes in which N2Py2

ligands (where Py is pyridine) have been varied in terms of

charge, aromaticity, and size. Six ‘‘NikA/iron complex’’

hybrids have been characterized by X-ray crystallography,

and their interactions and solution properties have been

studied. The hybrids are stable as indicated by their Kd

values, which are all in the micromolar range. The X-ray

structures show that the ligands interact with NikA through

salt bridges with arginine residues and p-stacking with a

tryptophan residue. We have further characterized these

interactions using quantum mechanical calculations and

determined that weak CH/p hydrogen bonds finely modu-

late the stability differences between hybrids. We empha-

size the important role of the tryptophan residues. Thus,

our study aims at the complete characterization of the

factors that condition the interaction of an artificial ligand

and a protein and their implications for catalysis. Besides

its potential usefulness in the synthesis of artificial mono-

oxygenases, our approach should be generally applicable in

the field of artificial metalloenzymes.

Keywords Artificial metalloenzyme � Ligand binding �X-ray crystallography � Iron chemistry � CH/p bonds

Introduction

In both natural and artificial metalloenzymes, the molecu-

lar recognition between the metal/metalloligand and the

binding pocket involves the second coordination sphere of

the metal, via hydrogen bonding and electrostatic and/or

hydrophobic interactions [1]. In the case of artificial

metalloenzymes, which combine the diversity of organo-

metallic or inorganic complexes with the selectivity of

proteins, it is essential to determine the structural factors

that stabilize the two components of the hybrids in order to

adapt the topology of the interaction and tune the reactivity

of the complex. Enantioselective reactions strongly depend

on these weak interactions [2].

Three main noncovalent binding strategies are currently

being developed to design artificial metalloenzymes: (1)

the chemical modification of natural ligands or cofactors,

such as heme in an engineered myoglobin [3, 4], (2) the use

M. V. Cherrier � P. Amara � M. Iannello �J. C. Fontecilla-Camps � C. Cavazza (&)

Institut de Biologie Structurale Jean-Pierre Ebel,

Groupe Metalloproteines,

UMR 5075, CEA, CNRS,

Universite Joseph Fourier Grenoble 1,

41 rue Horowitz,

38027 Grenoble Cedex 1, France

e-mail: christine.cavazza@ibs.fr

M. V. Cherrier

Spanish CRG Beam Line BM16,

European Synchrotron Radiation Facility,

6 rue Jules Horowitz,

BP 220, 38043 Grenoble, France

E. Girgenti � C. Marchi-Delapierre � S. Menage (&)

Laboratoire Chimie et Biologie des Metaux,

CEA, iRTSV, CNRS, UMR 5249,

38054 Grenoble, France

e-mail: stephane.menage@cea.fr

E. Girgenti � C. Marchi-Delapierre � S. Menage

Universite Joseph Fourier Grenoble 1, UMR 5249,

38041 Grenoble, France

123

J Biol Inorg Chem (2012) 17:817–829

DOI 10.1007/s00775-012-0899-7

of metal complexes that mimic natural ligands, illustrated

by the introduction of iron/Schiff base complexes into the

myoglobin’s active site [5], and (3) the anchoring of cat-

alytically active inorganic complexes to native substrates,

which is called the Trojan horse strategy. Ward and col-

laborators have extensively developed the third strategy,

using biotin–(strept)avidin technology [6] to optimize the

enantioselectivity of hydrogenation reactions [7]. They also

used the second strategy to insert a vanadyl complex into

the biotin-binding pocket of streptavidin. In this case,

evidence for substrate control by the protein environment

was found as these hybrids displayed the best enantiose-

lectivity achieved so far for sulfoxidation reactions cata-

lyzed by artificial metalloenzymes, with up to 93 %

enantiomeric excess [8]. This example emphasizes the

importance of supramolecular interactions for the activity

of enzymes.

To design artificial metalloenzymes, transport proteins

can be part of an interesting host/guest system because they

are rich in potential supramolecular interactions. Therefore,

we have used the periplasmic nickel-binding protein NikA

from Escherichia coli as a suitable host [9]. Like other

periplasmic binding proteins, NikA is composed of two

lobes (lobe I, residues 1–245 and 471–502; lobe II, residues

246–470) connected by a hinge region, in which the ligand

binding site is located [10]. The two domains can adopt

two different conformations: a ligand-free open form and a

ligand-bound closed form.

A few years ago, we determined the as-isolated NikA

structure and showed that this protein is capable of binding

Fe-EDTA [11]. Subsequently, we determined the structure

of NikA complexed with a metal-bound, putative natural

nickel-specific chelator, which appeared to be similar to

EDTA and that we modeled as butane-1,2,4-tricarboxylate

[12]. More recently, we characterized the intramolecular

arene dihydroxylation in a NikA-bound Fe-EDTA-like

catalyst [FeL, where L is N-benzyl-N0-(2-hydroxybenzyl)-

N,N0-ethylenediaminediacetic acid, complex 1 in Fig. 1)

during reductive dioxygen activation. The structures of

four intermediates and the end product of the reaction led

to the description of the reaction in the crystal [13].

We concluded from these studies that (a) NikA is able to

bind both natural and nonphysiological inorganic com-

pounds, (b) NikA is an appropriate host for the design of

Fig. 1 Inorganic complexes used in this study. BPMEN N,N0-dimethyl-N,N0-bis(2-pyridylmethyl)ethane-1,2-diamine, BPMCN N,N0-bis(2-

pyridylmethyl)-N,N0-dimethyl-trans-1,2-diaminocyclohexane, salen N,N0-bis (salicylideneiminato)ethylene

818 J Biol Inorg Chem (2012) 17:817–829

123

artificial monooxygenases, and (c) the interaction with the

carboxylate groups present in the metal complexes seems

critical for their binding to NikA, which is mediated by

arginine residues. We have also determined that our metal

complexes bind to NikA via a hydrogen-bond network.

Because the plasticity of the cavity seems to allow the

binding of different complexes, we decided to investigate

the nature of their interactions with Arg97, Arg137,

Trp100, Trp398, Tyr402, and His416, which line the

binding pocket, described for the first time in 2003 [10]

(Fig. 2). Subsequently, we have explored the ability of this

cavity to bind inorganic complexes in order to design

artificial monooxygenases [14].

Here, we report a series of EDTA-like mononuclear iron

complexes mimicking the active sites of nonheme iron

oxygenases [15], which have been designed to determine

the minimum motif necessary for NikA binding. We varied

the number and the position of the carboxylate groups as

well as the number and the nature of aromatic or cyclic

groups in order to favor extra p-stacking interactions with

tryptophan residues. The different hybrids were charac-

terized in solution, and the affinity of the metalloligand for

the protein was determined by fluorescence quenching

experiments. The binding of these complexes to NikA was

determined by X-ray crystallography. Here, we show that,

in addition to the expected strong N–H���O or O–H���Obonds, and van der Waals contacts, very weak CH/pinteractions also play a significant role in the binding of

inorganic complexes to NikA.

This finding will help the design of artificial metallo-

enzymes through better prediction of the binding of

inorganic complexes in the binding pocket of the protein

and their reactivity.

Materials and methods

Purification of NikA and synthesis of ‘‘NikA–metal

complex’’ hybrids

Complexes 1–6 were synthesized as previously described

[13, 16, 17]. Cytoplasmic apoNikA was overproduced and

purified as already reported [12]. The synthesis of hybrids

was performed by incubating 600 lL of a 270 lM protein

solution with a tenfold excess of complex in 40 mM

tris(hydroxymethyl)aminomethane (Tris)–HCl pH 7.5 at

4 �C overnight. The complex was first solubilized in 5 lL

dimethyl sulfoxide. Excess complex was removed using a

desalting column (PD10), which led to the separation of the

hybrid from the free complex in solution, followed by

dialysis using a Centricon filter with a 30-kDa cutoff (four

cycles).

Analytical methods

UV–vis absorption spectra were recorded with a Shimadzu

UV-1800 spectrophotometer. The protein concentration

was determined by quantitative amino acid analysis using a

model 7300 Beckman amino acid analyzer. Protein-bound

iron was determined under reducing conditions with bath-

ophenanthroline disulfonate after acidic denaturation of the

protein [18]. Manganese concentrations were measured by

Fig. 2 NikA crystal structure in

the EDTA-bound closed

conformation (Protein Data

Bank ID code 1ZLQ). The

protein is colored according

to the sequence from cyan(N-terminal) to yellow(C-terminal). The residues

constituting the binding site

considered in this paper are

colored in purple. A zoom of

the binding site is shown in the

inset

J Biol Inorg Chem (2012) 17:817–829 819

123

inductively coupled plasma optical emission spectrometry:

6 mL of a 0.4 mg/mL hybrid solution was acidified with

100 lL ultrapure nitric acid before measurement.

Fluorescence measurement and data analysis

Fluorescence spectra were measured with a JASCO FP-

6500 spectrofluorimeter equipped with a thermostatic cir-

culator keeping the temperature constant at 25 ± 1 �C. The

excitation wavelength was 280 nm, and the emission

spectra were recorded at 295–410 nm. The binding of

complexes to NikA was studied by the fluorescence

quenching titration method, using the intrinsic fluorescence

of the protein as a probe. Complexes were dissolved either

in dimethyl sulfoxide or in 40 mM Tris buffer pH 7.4. Zero

to 20 equiv of complex was added to 5 lM NikA in

40 mM Tris–HCl pH 7.4 in a final volume of 1.5 mL. After

each addition, the mixture was incubated for 15 min under

agitation at 25 �C, with the excitation lamp turned off to

avoid photobleaching, before monitoring the fluorescence.

The sample was stirred during the whole procedure. The

fluorescence intensities at 340 nm—average of three

measurements—were corrected first for background fluo-

rescence from the buffer and then for ligand adsorption,

before being plotted against the total complex concentra-

tion [19].

The binding titration data were fitted, by using a non-

linear least-squares routine, to the following quadratic

equation described by Murphy and Spudich [20], in which

a single type of binding site is assumed:

F ¼ F0

� ðF0 � F1Þ½C�t þ ½P�t þ Kd �

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

ð½C�t þ ½P�t þ KdÞ2 � 4½C�t½P�tq

2½P�t

6

6

6

4

7

7

7

5;

where F is the observed fluorescence after correction, F0 is

the initial fluorescence in the absence of quenching agent,

F? is the final amplitude of fluorescence quenching, [C]t

and [P]t are the concentrations of complex and protein,

respectively, and the equation is solved for Kd and F?.

The emission intensities are affected by the dilution

resulting from introducing aliquots of complex solution

during the experiment. Consequently, [P]t and [C]t were

adjusted for dilution over the whole curve:

½P�t ¼ ð½NikA�0 � V0Þ=Vtot;

where [NikA]0 is the starting NikA concentration, V0 is the

starting volume, and Vtot is the total volume for each

titration, and

½C�t ¼ ð½C�0 � VaddedÞ=Vtot;

where [C]0 is the complex stock solution concentration,

Vadded is the total added volume of complex solution, and

Vtot is the total volume for each titration. The fits indicated

the presence of one high-affinity binding site per NikA

molecule. An additional low-affinity site was found, but it

was not considered in this study.

Hybrid crystallization

Prior to crystallization, the protein was incubated with a

twofold molar excess of Fe(III)-EDTA. Hanging drops

were prepared by mixing 2 lL of 11.5 mg/mL protein

solution in 40 mM Tris–HCl pH 7.6 with 2 lL of 1.8 M

ammonium sulfate, 100 mM sodium acetate pH 4.7 reser-

voir solution. Reproducible orthorhombic crystals appeared

after 1 week. They were soaked in 4 lL of a mother liquor

solution containing 2.1 M ammonium sulfate, 100 mM

Tris–HCl pH 7.0 mixed with 5 mM or 10 mM complexes

5RR and 6SS, respectively. After one night, the crystals

were transferred for 4 h to the mother liquor (4 lL) to

remove the excess of free complex. All crystals were

cryoprotected using a solution obtained by adding 25 % (v/

v) glycerol to the mother liquor and flash-cooled in liquid

nitrogen.

Data collection and structure determination

Data were collected at beamlines BM30A and ID23-2 of

the European Synchrotron Radiation Facility (Grenoble,

France) for 5RR,NikA and 6SS,NikA, respectively. Data

indexation and scaling were carried out using XDS [21].

Phaser [22] was used to solve the structures by the

molecular replacement method with the atomic coordinates

of the NikA–Fe(III)-EDTA(H2O)- structure as the search

model [Protein Data Bank (PDB) ID code 1ZLQ). Crys-

tallographic refinements were conducted using Phenix [23]

and the three-dimensional models were examined and

modified using the graphics program COOT [24]. Crys-

tallographic statistics are summarized in Table 1. Fig-

ures 1, 4, and 5 were prepared using PyMOL (version 1.4,

Schrodinger, New York, USA). Figure 6 was prepared

using Maestro (version 9.2, Schrodinger, New York, USA).

The PDB ID codes are 4DCX and 4DCY for 5RR,NikA

and 6SS,NikA, respectively.

Electron densities corresponding to the ligands in crystal

structures of 3,NikA and 4,NikA were not modeled

owing to the presence of several complex conformations in

the two NikA molecules in the crystal asymmetric unit

(called A and B).

Interaction energies

To visualize nonbonded interactions involved in the bind-

ing of the complexes to NikA, we used a method that was

developed very recently by Johnson et al. [25] and was

820 J Biol Inorg Chem (2012) 17:817–829

123

subsequently implemented in the Jaguar program [26].

Briefly, this method consists in calculating the reduced

density gradient and the interaction strength. The gradient

isosurfaces show the positions of the noncovalent interac-

tions. They are color-coded according to the values of the

interaction strength. Negative values of the interaction

strength (blue) indicate attractive interactions such as

hydrogen bonds, whereas positive values (red) represent a

nonbonding interaction; negative values near 0 (green)

denote weak van der Waals interactions.

From each hybrid X-ray model, we extracted the metal

complex (Fig. 1), residue side chains in the first coordi-

nation shell, namely, Arg97, Arg137, Trp100, Trp398,

Tyr402, and His416 (Fig. 2), and water molecules hydro-

gen-bonded to the metal complex. For our model of com-

plex 5RR in trans topology, we replaced for simplicity a

hydrogen bond due to a glycerol molecule by that of a

water molecule. Finally, dangling bonds were saturated

with hydrogen atoms. We performed single-point energy

calculations and calculated the reduced density gradient

using density functional theory with the B3LYP functional.

The lacvp** basis set [27, 28] was used for the iron atom,

and the rest of the atoms were modeled with 6-31G**.

(Initially, we optimized the geometry of one of our models

but, because we did not find significant differences in the

resulting isosurfaces compared with the ones obtained from

the X-ray data-derived model, we decided not to optimize

the models for this study.)

Results

Hybrid synthesis

To determine the structural parameters that stabilize an

inorganic complex bound to NikA, the EDTA ligand was

modified by replacing carboxylate moieties with methyl

pyridines and benzyl and 2-hydroxybenzyl moieties. These

changes affect the charge, the metal Lewis acidity, and the

bulkiness of the ligand (complexes 1–6; Fig. 1). Ulti-

mately, chiral recognition was achieved through the use of

the chiral diamine bridge of the ligand. To achieve this

goal, N,N0-dimethyl-N,N0-bis(2-pyridylmethyl)ethane-1,2-

diamine (BPMEN) [29] and N,N0-bis(2-pyridylmethyl)-

N,N0-dimethyl-trans-1,2-diaminocyclohexane (BPMCN)

[30] were modified to EDTA-like ligands by addition of

one or two acetate arms. Indeed, the global charge of the

cavity is ?2 owing to Arg97 and Arg137 and is thus

suitable for the binding of negatively charged complexes.

Their iron complexes were described previously (complexes

Table 1 Data collection and

refinement statistics

See Fig. 1 for the structures

of the complexes

ESRF European Synchrotron

Radiation Facility, RMSD root

mean square deviation

6SS,NikA 5RR,NikA

ESRF beamline BM30A ID23-2

Wavelength (A) 0.9796 1.2798

Space group P212121 P212121

Unit cell parameters (A) a = 86.88 a = 86.35

b = 95.19 b = 95.64

c = 125.15 c = 125.19

No. of protein molecules per asymmetric unit 2 2

Resolution range (A) 44.5–2.0 (1.9–2.0) 35.0–2.0 (1.9–2.0)

Completeness (%) 99.9 (100.0) 98.7 (99.6)

I/r(I) 19.31 (6.24) 17.32 (6.37)

No. of measured reflections 508,639 (68,115) 493,650 (65,834)

Redundancy 7.1 (7.2) 7.0 (7.0)

Rsym (%) 7.0 (32.1) 6.7 (30.7)

Rwork/Rfree factor (%) 14.62/19.58 15.49/20.27

No. of protein atoms 8,144 8,042

No. of water molecules 973 859

Average B factor (A2) 30.5 36.1

RMSD

Bonds (A) 0.007 0.0068

Angles (�) 1.055 1.086

Ramachandran (%)

Favored region 97.5 96.9

Outlier region 0.0 0.1

J Biol Inorg Chem (2012) 17:817–829 821

123

2–6 in Fig. 1) [17]. All complexes were then incubated

aerobically with NikA to obtain the corresponding hybrids

for their further characterization.

Solution characterization of hybrids

The first characterization of the hybrids was based on the

determination of the metal/NikA ratios reported in Table 2.

Since the metal content is directly correlated with binding

of the complex, we note that the absence of carboxylate

groups prevents the complex binding to NikA, as illustrated

by complex 2. The position of the carboxylate moiety is

also important because in the case of N,N’-(bis-4,4’acetato

salicylideneiminato)ethylene manganese (Mn salen di-

COOH) (Fig. 1), both carboxylate groups in para positions

prevent the binding probably owing to clashes with the

binding pocket. Indeed, if one carboxylate group interacts

with Arg137, like for Fe-EDTA and 1, the cavity is too

narrow to accommodate this complex in an axial planar

position. All the other complexes, namely, 1 and 3–6, bind

to NikA with Fe/NikA ratios close to 1. At first glance, it

seems that one carboxylate group is sufficient for binding

of the complex and that the presence of aromatic groups

does not prevent it. Even the change of the ethylenedia-

mine motif to the bulkier cyclohexanediamine (5RR and

6SS) did not preclude the binding.

An example of UV–vis spectra for 5RR is shown in

Fig. 3a. The UV–vis spectrum of the hybrid displayed a

broad shoulder around 340 nm (e = 4.2 ± 0.5 mM-1

cm-1), putatively attributed to charge transfer transitions

arising from the iron chromophore, and a more intense

band at 280 nm arising from p–p transitions of the aro-

matic amino acids residues. The low-energy transitions are

observed in the same region in the case of the spectrum of

5RR, but differ in terms of intensity and broadness. This

indicates that the iron complex is interacting with the

protein framework and that no diiron species, in particular

loxo–diiron complexes, are present. All complexes, except

1, showed a similar trace. For 1, an Fe–Oph ligand-to-metal

charge transfer band at 530 nm was observed, which was

not affected by its binding to the protein.

The insertion of 5RR into the protein was also demon-

strated by X-band EPR spectroscopy by the presence of

two transitions around g = 4 (4.3 and 4.8 associated with a

g = 3.8 signal), attributed to the signature of high-spin

hexacoordinated ferric species with close but distinct

E/D values (0.33 and about 0.25 as estimated from

S = 5/2 rhombograms) [31]. These signatures at 9 K in

40 mM Tris pH 7.4 may be related to the presence of two

different conformations for the complex and/or by a change

in the coordination sphere (Fig. 3b). It has to be noted that

ferric complex 5RR in the buffered solution under the same

Table 2 Solution characterization of hybrids

Metal content

(M/NikA)

Charge Kd (lM)

Fe-EDTA,NikA 1.2 ± 0.2 -1 2.5 ± 0.3

Mn salen diCOOH,NikA 0.2 ± 0.05 – –

1,NikA 1.1 ± 0.2 0 0.64 ± 0.02

2,NikA 0.2 ± 0.05 – –

3,NikA 0.8 ± 0.2 ?1 1.4 ± 0.2

4,NikA 1.3 ± 0.2 ?2 5.6 ± 0.2

5SS,NikA ND ?1 14.9 ± 1.2

5RR,NikA 0.8 ± 0.2 ?1 31.4 ± 2.7

6SS,NikA 0.7 ± 0.2 ?2 30.2 ± 1.6

6RR,NikA ND ?2 62.3 ± 8.8

See Fig. 1 for the structures of the complexes.

Mn salen di-COOH N,N’-(bis-4,4’-acetato salicylideneiminato)eth-

ylene manganese, ND not determined

Fig. 3 a UV–vis spectra of NikA (10 lM; black line), 5RR,NikA

(10 lM; red line), and 5RR (10 lM; green line) in tris(hydroxy-

methyl)aminomethane (Tris)–HCl buffer (40 mM, pH 7.4). b EPR

signal for 100 lM 5 RR,NikA in Tris HCl buffer (pH 7.4) at 11 K

822 J Biol Inorg Chem (2012) 17:817–829

123

recording conditions displayed only the g = 4.3 signal,

suggesting the presence of two different binding configu-

rations in NikA. A similar behavior was observed in the

case of 6RR and 6SS. Conversely, the systems containing 1

[13] or Fe-EDTA [11] displayed only one signal in their

EPR spectrum.

Structural characterization

Crystal structures were obtained with Fe-EDTA and com-

plexes 1 and 3–6 to better characterize their binding to

NikA. The Fe-EDTA,NikA and 1,NikA crystal struc-

tures have been previously described [11, 13]. The high-

resolution structure determination of 3,NikA and

4,NikA could not be accomplished because of the pres-

ence of a mixture of complex topologies in the binding site,

but the structural analyses confirm the high affinity of the

binding pocket of NikA for different iron complexes.

The crystal structures of 5RR,NikA and 6SS,NikA

were solved at 2.0-A resolution (Fig. 4). As previously

observed, two different molecules are present in the

asymmetric unit (called A and B). The main difference

between them is the relative position of the two lobes of

NikA. Molecule A is in a more open conformation than

molecule B, owing to crystal packing. Relative to molecule

A, the hinge angle between the two lobes [32] in molecule

B is smaller by 4.2�, 4.9�, and 4.8� for Fe-EDTA,NikA,

5RR,NikA, and 6SS,NikA, respectively. Replacement of

Fe-EDTA by 5RR and 6SS at the binding site of NikA has a

limited effect on the protein structure, in part due to the

packing restraints: the hinge motion is, on average,

1.2� ± 0.2� and 1.2� ± 0.5� for molecules A and B,

respectively. Average B factors (Bav) of the structures show

that molecule A is better ordered than molecule B: the

B factors for molecule B are 24, 38, and 35% higher than

those for molecule A for Fe-EDTA,NikA (molecule A,

17.2 A2; molecule B, 21.4 A2), 5RR,NikA (molecule A,

24.3 A2; molecule B, 33.5 A2), and 6SS,NikA (molecule

A, 29.6 A2; molecule B, 40.0 A2), respectively. In contrast

to Fe-EDTA,NikA, in 5RR,NikA and 6SS,NikA, lobes I

Fig. 4 Crystal structures of NikA binding sites of 5RR,NikA and

6SS,NikA. The omit electron density map (F0 - Fc) is display at the

3r level (green). For each hybrid, the two molecules (A and B) of the

asymmetric unit are presented. In molecules A and B, the complexes

adopt cis-b and trans topologies, respectively

J Biol Inorg Chem (2012) 17:817–829 823

123

and II of molecule B are either anisotropically agitated or

disordered (Fig. 5). The atomic motions correspond to a

modification of the interlobe opening, and, consequently, to

a modification of the NikA binding site. Overall, the

replacement of Fe-EDTA by 5RR or 6SS in NikA crystals

destabilizes the protein, but in the case of molecule A, the

increase of the B factor was limited owing to the high

number of interactions with neighboring protein molecules

in the crystal (72 noncovalent interactions, of which 47 are

hydrogen bonds) (Fig. 5) [33]. The residues involved in

these bonds are also almost equally distributed between the

two lobes. In the case of molecule B, the number of

interactions with the neighboring protein molecules is

about half of that found for molecule A (38 noncovalent

interactions, of which 20 are hydrogen bonds) and these

interactions involved mainly lobe II residues.

As expected from their similarities with Fe-EDTA, once

bound to NikA, both complexes display the salt bridge

between Arg137 and one carboxylate moiety of the ligand

(Fig. 4). The iron ion of 5RR is hexacoordinated by the two

nitrogen atoms of the pyridine rings, the two carboxylate

groups, and the two nitrogen atoms from the ethylenedia-

mino moiety. The iron ion of 6SS is also hexacoordinated,

but the missing carboxylate group appears to have been

replaced by a water molecule in molecule B. No clear

electron density for this region was found in molecule A.

The structures of 5RR,NikA and 6SS,NikA reveal that

the inorganic complexes adopted two conformations in our

study, in agreement with the findings of the solution EPR

experiments. In molecule A, both complexes have an

unexpected trans topology around the iron with both pyr-

idine rings cis to each other in the equatorial plane and both

nitrogen substituents oriented anti to each other. This

topology is energetically unfavorable [34, 35] and has only

been observed twice in the BPMEN family, for the

BPMPN (N,N’-bis(2-pyridylmethyl)-1,3-diaminopropane)

[36] and the L8py2 (N,N’-bis(2-pyridylmethyl)-1,5-diaza-

cyclooctane) [37] ligands, in which one or two propylene

straps, respectively, connect the two 2-pyridylmethylamine

moieties to enforce the topology. In this study, the trans

topology could be stabilized thanks to the presence of the

more apicophilic carboxylate moiety. In contrast, in mol-

ecule B, 5RR and 6SS adopt the cis-b topology where the

two pyridine groups coordinate cis to each other but in

perpendicular planes and the two nitrogen substituents are

in a syn conformation previously illustrated only by the

crystal structure of [FeII(b-5-Me2-BPMCN)(OTf)2] (where

OTf is trifluoromethanesulfonate).

The X-ray crystallography was helpful to unambigu-

ously determine the helicoidal chirality at the metal center

of the reported complexes, which was generated by the

distribution of chelate rings about the central ferric ion

[38]. In the case of Fe-EDTA, only the K configuration was

observed, indicating the stereoselective binding of the

complex to NikA [11]. A similar result was observed in the

case of complexes 5RR and 6SS, with only (RR, K) and (SS,

K) diastereoisomers, respectively, regardless the topology

of the reported complexes, which could be cis-b or trans.

Binding of complexes to NikA in solution

Fluorescence spectroscopy is related to the emissive

properties of tryptophan and tyrosine residues. In our case,

Trp100 and Trp398 line the binding cavity of NikA, so the

quenching and the emission energy change will be due to

the proximity of the iron complex. Quenching was gener-

ally observed, with the exception of 2 and Mn salen

diCOOH, which do not interact with NikA. The resulting

Fig. 5 Anisotropic B factors of Fe-EDTA-,NikA, 5RR,NikA, and

6SS,NikA. The anisotropic B factor of each atom is displayed as an

ellipsoid and is colored according to its value from blue (4.7 A2) to

red (119.2 A2). The purple surfaces localize the residues involved in

the interaction with the packing neighboring molecules. Double

arrows highlight the motion of domain I (B-DI) and domain II (B-DII)of molecule B. Asterisks localize the cavity formed between the two

domains, and where the inorganic complex (black) is bound. A-DIdomain I of molecule A, A-DII domain II of molecule A

824 J Biol Inorg Chem (2012) 17:817–829

123

Kd values are shown in Table 2. In some cases, a slight

blueshift of the maximum emission wavelength was

observed, showing a higher hydrophobicity of the trypto-

phan environment [39], attributed to direct interactions

with the metalloligand.

The Kd values at pH 7.4 indicate affinities increasing in

the following order: 6RR \ 6SS & 5RR \ 5SS \ 4 \ 3

\ Fe-EDTA \ 1. Even though all the Kd values are, on

average, in the micromolar range, there were about two

orders of magnitude differences between the lower and

higher values. In 2007, through use of isothermal titration

calorimetry, NikA was shown to bind Ni-EDTA with an

affinity of 30 lM [40]. This value is about 30 times higher

than the affinity measured in this study for Fe-EDTA. This

may be related either to the difference in charge of the two

complexes or to the use of different methods which often

lead to dissimilar Kd values. On the basis of the crystal-

lographic studies and the EPR measurements, it is likely

that mixtures of two conformations were present in solu-

tion in the case of 5 and 6 (and probably also 3 and 4).

Because their ratios and respective affinities are unknown,

the Kd values for these complexes cannot be unambigu-

ously determined. However, some information about the

binding of complexes to NikA was obtained:

1. The number of negative charges in the bound complex

has a significant effect on the affinity owing to the

overall positive charge of the binding pocket. The high

affinity of Fe-EDTA is likely partly due to the

interaction of two of its four carboxylates with

Arg97 and Arg137; this is not the case for the other

complexes studied, where Arg97 is never involved in

their binding.

2. In 1 and 3, the aromatic moieties in the ligand interact

with aromatic side chains, thereby allowing its binding

in spite of its neutral or positive charge (Table 2).

3. The presence of two carboxylates in 3 instead of one

carboxylate in 4 improved the affinity of the complex.

4. In the presence of the cyclohexyl substituent, the

affinity of the complexes decreased most probably

because of steric hindrance.

5. The chirality of the cyclohexyl substituent in 5 and 6

seems to affect their interaction with NikA, leading to

similar Kd values for 6SS and 5RR, despite the ?2

charge of the former. The Kd values for 5RR and 6RR

were twice as high as those for 6SS and 5SS,

respectively. We cannot explain this difference.

Interaction energies

All the designed complexes bind to NikA via supramo-

lecular interactions. Some of them were observed by X-ray

crystallography (like the salt bridge with Arg137), some

others were suspected (like p-stacking with Trp398), and it

is likely that some weak interactions went unnoticed. To

precisely define the potential for binding of the complex

to NikA and to determine the residues involved in this

process, we evaluated the interaction energies for the

following complexes: Fe-EDTA, 1, 1Fe(II) reduced by

dithiothreitol (1red) [13] (Fig. 6), 5RRtrans, 5RRcis-b, and

6SScis-b. However, standard methods to calculate interaction

energies (such as in silico alanine scanning) did not appear

well adapted to compare the binding of these complexes to

NikA as (a) they do not have the same charge, and

(b) metal ions are not well parameterized in these com-

putational methods. Thus, in addition to analyzing the

hybrid structures, we employed a new method to detect

weak noncovalent interactions [25, 41].

All the complexes we studied interact with NikA through

a network of hydrogen bonds, salt bridges, p-stacking, and

CH/p interactions as described below (Table 3; Fig. 6). In

addition to the salt bridges formed with Arg137 and Arg97,

and p-stacking with Trp398 observed by protein crystal-

lography, many other interactions were visualized thanks to

this approach.

p-stacking

Relative to 1,NikA, in 1red,NikA, the phenyl ring

moved to the position previously occupied by the phenolate

ligand and the latter became an apical ligand to iron [13].

In 1red,NikA, in addition to Trp398, the phenyl moiety

also stacked with Tyr402. In this case, the phenol group, in

an apical position, formed a third p-stacking interaction

with the His416 ring. Unexpectedly, in 5,NikA and

6,NikA, the pyridyl group did not interact with Trp398

and Tyr402. Instead, p-stacking interactions were observed

with Trp100: in 5RRtrans,NikA, one p-stacking interaction

was present between one pyridyl group and the Trp100 side

chain. In 5RRcis-b,NikA, the pyridyl group present in the

same plane as the cyclohexyl moiety forms two p-stacking

interactions with Trp100 and His416. In 6SScis-b,NikA,

there was a p-stacking interaction between its pyridine

ring, perpendicular to the only carboxylate group of the

molecule, and the Trp100 side chain.

CH/p interactions

The CH/p interactions were very different in the various

complexes.

The binding of Fe-EDTA to NikA involves six CH/pinteractions: three between hydrogen atoms of the

N(CH2)2N moiety of EDTA and the Trp100 side chain and

three between hydrogen atoms of the CH2NCH2 moiety of

EDTA and Trp398.

J Biol Inorg Chem (2012) 17:817–829 825

123

In 1,NikA, the network of interactions was completed

by CH/p bonds between four hydrogen atoms from

CH2N(CH2)2N and the Trp398 side chain, and three

hydrogen atoms from pyridyl-CH2NCH2 complex moieties

and the Trp100 side chain.

In 1red,NikA, five hydrogen atoms from CH2N

(CH2)2NCH2 interacted with Trp100, three hydrogen atoms

from phenyl-CH2NCH2 interacted with Trp398, and one

hydrogen from the phenyl-CH2 moiety interacted with

His416.

Fig. 6 Noncovalent interactions between 1Fe(II) reduced by dithio-

threitol (1red) and NikA. The model used to calculate the noncovalent

interactions is shown in a ball and stick representation. The color code

used for the atoms is as follows: orange carbon from 1red, turquoisecarbon from NikA, black hydrogen, red oxygen, and blue nitrogen.

The iron ion is shown as a dark-red sphere. The rest of the protein is

depicted with turquoise ribbons. Noncovalent interactions around the

ligand are represented by the gradient isosurfaces colored according

to the interaction strength (ranging from -0.02 to ?0.02 au) with a

blue–green–red scale. Blue, green, and red indicate strong attractive

interactions, weak van der Waals interactions, and strong nonbonded

overlaps, respectively. Examples of a salt bridge, a CH/p bond, and p-

stacking are enlarged in insets A–C

Table 3 Summary of the noncovalent interactions between the inorganic complexes and the protein residues

Fe-EDTA,NikA

(1ZLQ)

1,NikA

(3MVW)

1red,NikA

(3MVX)

5RR,NikA

trans5RR,NikA

cis-b6SS,NikA cis-b

Charge -1 0 -1 ?1 ?1 ?2

Arg 97–COO- ? ??

Arg 137–COO- ?? ?? ?? ?? ?? ??

H2O–COO- ?? ??? ? ???a ?? ?

Trp 100 p–p ? ? ?

Trp 398 p–p ? ?

Tyr 402 p–p ?

His 416 p–p ? ?

Trp 100 p/CH ??? ???? ????? ? ? ?

Trp 398 p/CH ??? ??? ??? ????? ?????? ?????

Tyr 402 p/CH ?? ??

His 416 p/CH ?

Each interaction is represented by a plus sign

1red 1Fe(II) reduced by dithiothreitola One glycerol present in the binding site was replaced by a water molecule (see ‘‘Materials and methods’’)

826 J Biol Inorg Chem (2012) 17:817–829

123

In 5RRtrans,NikA CH/p interactions were observed

between one hydrogen atom from pyridyl-CH2, five

hydrogen atoms from C6H10–NCH2–C6H4N and cyclo-

hexyl complex moieties, and Trp100, Trp398 and Tyr402

side chains, respectively. For the latter, the interaction was

weaker because Tyr402 is far from the cyclohexyl group.

5RRcis-b,NikA formed CH/p interactions similar to

those of 5RRtrans,NikA: one hydrogen atom from the

cyclohexyl and six hydrogen atoms from the C6H10–

NCH2–C6H4N moiety interacted with the Trp100 and

Trp398 rings.

In 6SScis-ß, CH/p interactions occurred between one

hydrogen atom from CH2COO, five hydrogen atoms from

C6H10–NCH2–C5H4N, and two hydrogen atoms from

cyclohexyl groups of the complex and Trp100, Trp398, and

Tyr402 side chains, respectively.

Structural water

Many water molecules were involved in the stabilization of

the complexes. The most important one was one that

established (a) one donating hydrogen bond with the oxy-

gen atom of the carboxylate group involved in the salt

bridge with Arg137, (b) one accepting bond with the OH

oxygen of Tyr402, and (c) one donating hydrogen bond

with the main chain oxygen of Arg137. This water mole-

cule is conserved in all the NikA-bound complexes studied

so far. When the complex has a second carboxylate group,

this moiety is called COOH2O in order to reflect its binding.

The COOH2O moiety in 1,NikA formed hydrogen bonds

with two water molecules. One of them also forms one

hydrogen bond with Trp398 Ne. In 1red,NikA, one water

molecule forms two hydrogen bonds with the second

oxygen of the carboxylate group and the Thr23 main chain

nitrogen atom. In 5RRtrans,NikA, one oxygen atom from

COOH2O forms two hydrogen bonds with two water mol-

ecules. One of the water molecules forms an additional

hydrogen bond with the Tyr382 OH group. Finally, the

oxygen of COOH2O of 5RRcis-b forms one hydrogen bond

with one water molecule.

Discussion

In the field of artificial metalloenzymes, the choice of the

host protein is crucial because the chiral environment in the

second coordination sphere of the metal complex deter-

mines the selectivity of the reaction and the stabilization of

both the inorganic complexes and the transition states.

Several binding modes have been developed in order to

insert a metal complex into a protein. We have chosen to

favor supramolecular interactions to modulate smoothly

the effect of the protein on the catalyzed reaction. Several

approaches can be used to transform the activity of a given

enzyme: the first metal coordination sphere can be fine-

tuned by modifying the ligand, and the second coordination

sphere can be modified by site-directed mutagenesis or

directed evolution of the protein. To design artificial

monooxygenases, we have chosen inorganic iron com-

plexes potentially capable of activating dioxygen and cat-

alyzing enantioselective oxidation reactions such as

sulfoxydation, hydroxylation, and epoxidation. Our first

goal was to characterize the different ‘‘NikA/iron com-

plex’’ hybrids we have synthesized. The study of the

affinity in solution showed that the Kd values spanned two

order of magnitude between the complexes. We have

shown that even positively charged inorganic complexes

can bind NikA with suitable affinities for the design of

artificial metalloenzymes. This indicates that the charge of

the ligand does not determine the interaction and demon-

strates the importance of nonionic interactions. Moreover,

we showed that the presence of two carboxylate groups in

the complex is not required for protein binding. Conse-

quently, we now favor the synthesis of complexes con-

taining only one carboxylate group, because in this case the

iron has a vacant position, which is potentially available for

dioxygen binding (in complexes 3 and 5, the presence of a

second carboxylate group can hinder both the formation of

the active oxo species and substrate binding).

One difficulty we encountered in the synthesis of

hybrids was the unpredictable behavior of the N2Py2

(where Py is pyridine) inorganic complexes in aqueous

solution. Initially, on the basis of spectroscopic data, we

suspected that two conformationally different complexes

coexisted in solution. Later and thanks to protein crystal-

lography, we confirmed and identified the nature of these

two different conformations, which were cis-b and trans in

the structures of 5RR,NikA and 6SS,NikA. In organic

solvents, a mixture of cis-a and cis-b topologies was pro-

posed on the basis of NMR spectroscopy data for com-

plexes 4 and 6 [17]. Conversely, no mixture was found for

3 and 5, as later confirmed by their X-ray structures that

showed a unique distorted cis-a topology [17]. In fact, the

trans topology has never been described before for N2Py2

complexes containing either a cyclohexylene or an ethy-

lenediamine backbone. So, the topology conversion of the

complex could be the result of either the aqueous solvent or

the protein environment. It is also essential to consider the

protein dynamics, especially in the case of transport pro-

teins containing cavities, which are potentially flexible

because of their function. One way of explaining the dif-

ferent topologies of the two NikA molecules in the crystal

structures is the higher flexibility of the interlobe angle in

molecule B, as indicated by its anisotropic B factor

(Fig. 5). The bulk of the inorganic complexes with cis-btopology should required a more open NikA conformation,

J Biol Inorg Chem (2012) 17:817–829 827

123

and, consequently, would be more compatible with mole-

cule B, leaving molecule A available for the trans

topology.

On the one hand, the presence of two different topolo-

gies in NikA complicated the hybrid characterization, but,

on the other hand, it allowed us to compile new data

concerning the interaction of metal complexes with NikA.

From this study, significant information has been obtained

about the contribution of each binding pocket residue to the

stabilization of metal complexes in this protein: Arg137 is

essential for the binding of inorganic complexes, whereas

the role of Arg97 seems to be dependent on the charge of

the complexes. This residue was only involved in the

binding of Fe-EDTA and 1red. Interestingly, when the

complex does not interact with Arg97, this residue can

adopt alternative conformations, sometimes far from the

binding site of the complex. In this case, the charge at the

binding pocket is ?1 and, consequently, is more favorable

for the binding of a positively charged complex. His416,

which is the only direct protein ligand involved in the

binding of the natural nickelophore [12], does not interact

with the metal in our nonphysiological complexes. The

coordination sphere of the iron ion represents a closed

shell, carboxylate ligands being better ligands than imid-

azole to a hard metal ion. Consequently, a dative bond was

not observed in our complexes and the His416 position is

very flexible. Its role is thus limited and it was only

observed for 1red and 5RRcis-b via p-stacking and/or CH/pinteractions (Table 3). However, these interactions have

been observed in the crystal structures and may be absent

in solution. Usually, some water molecules are displaced

by ligand binding in proteins. However, others play an

important role in protein–ligand contacts, forming hydro-

gen bonds with the two partners. This is illustrated in the

case of NikA by a strictly conserved water molecule that

interacts with Arg137, Tyr402, and all the ligands studied

to date. On the basis of the distances and angles in the

crystal structure, a cation–p interaction between Fe-

EDTA,NikA and Trp398 was proposed [42]. However, it

was not clearly detected in the present study. Both Trp100

and Trp398 participate significantly in ligand binding: the

role of the former was not previously described in crys-

tallographic studies and it was identified in this study for

the first time. Despite the presence of several aromatic

residues in the ligand-binding pocket, this study shows that

p-stacking interactions do not dominate the binding of

complexes containing pyridyl groups. Conversely, aro-

matic residues in NikA are involved in numerous CH/pinteractions and in p-stacking. The role of CH/p interac-

tions in biological molecules has been the subject of much

debate because the overall stabilization energy of a CH/pbond is only about 1.5–2.5 kcal/mol, which is in the range

of thermal fluctuations [43, 44]. Although the weak CH/p

hydrogen bond is not a conventional hydrogen bond, it has

been classified as such since 1989 [45]. Compared with

other types of hydrogen bonds, in the CH/p interaction, the

dispersion energy is very significant, whereas the contri-

bution of the electrostatic energy is minor. Their role in

biochemistry has become increasingly clear and it has been

shown experimentally in several proteins [46, 47]. Phen-

ylalanine, tyrosine, tryptophan, and occasionally histidine

are the most common p-acceptors in proteins, with tryp-

tophan being the most common one owing to the large

accessible area afforded by the two fused rings [48].

Moreover, these p-acceptors interact more frequently with

polarized donating CH groups, whereas NH–p or OH–pinteractions are seldom observed in proteins. It is now

accepted that CH/p hydrogen bonds are one of the major

attractive forces in protein–ligand interactions.

The interactions between the ligand and the protein

scaffold described above drive the helicoidal chirality at

the metal center [38, 49]. It is expected that both

(dia)stereoisomers coexist in solution since ferric com-

plexes are well known to be kinetically labile. Only one

protein-bound stereoisomer or diastereoisomer (when the

stereogenic cyclohexane is part of the ligands) is detected

in every complex. This selectivity is related to the orien-

tation of the diamine chelating moiety and the pivotal role

of the carboxylate/arginine hydrogen bonding. In the case

of Fe-EDTA, the diamine interacts with Trp100, stabilizing

the D stereoisomer. Owing to steric hindrance, the cyclo-

hexanediamine ring is rotated by 90�, interacting with Trp

398 and leading to the K stereoisomers for 5RR and 6SS. Its

influence on the stereoselectivity at the metal center is also

illustrated by the observation of the D stereoisomer of 3

and 4, which lack this motif. Conversely, the chirality of

the ligand has no influence on the helicoidal isomer (RR vs.

SS) and the topology of the ligand is not crucial, as cis-b or

trans complexes have equivalent stereoisomers. Thus, the

protein selects only one chiral complex, a property that

could be essential for enantioselective catalysis.

Conclusions

Generally, the strong/weak nature of hydrogen bonds in

natural enzymes is exploited to determine their structure

and function, and some recent studies have shown that, in

addition to classic ones, CH/p hydrogen bonds play a

determinant role. This study is a new example of the

implication of CH/p hydrogen bonds in the interaction of

proteins with ligands. In our case, CH/p hydrogen bonds

are involved in the interaction of nonphysiological metal

complexes with a protein. It is therefore important to

consider these interactions in the synthesis of artificial

metalloenzymes. In the case of NikA, where the interaction

828 J Biol Inorg Chem (2012) 17:817–829

123

of metal complexes within the binding pocket has been

fully described, we will be able to adapt the synthesis of

ligands to modify, for example, their binding modes,

adjusting negative charges and CH/p interactions. We have

confirmed the central role of Arg137 and the role of Arg97

in their interaction with carboxylate groups. We have also

shown the importance of CH/p interactions, which have

been found to play a key role in stereoselectivity in organic

chemistry as well as in biochemistry [44].

This study provides clues about the different factors to

consider for the synthesis of artificial metalloenzymes,

such as the behavior of inorganic complexes in aqueous

solutions, the role of weak interactions, and the protein

dynamics. It also underscores the need for collective

coordination modes in the design of a ligand binding site.

We will take these factors into account when designing

new artificial monooxygenases for enantioselective oxida-

tion reactions.

Acknowledgments We thank the Agence Nationale pour la

Recherche for grant ANR 08-CP2D-12, the CEA, Joseph Fourier

University, and the CNRS for institutional support. We also thank the

staff from the BM30A and ID23-2 beamlines of the European Syn-

chrotron Radiation Facility in Grenoble, France. We gratefully

acknowledge Pierre Richaud (CEA, LB3M) for performing induc-

tively coupled plasma optical emission spectrometry analyses on Mn

salen di-COOH,NikA and Jean-Pierre Andrieu (Institut de Biologie

Structurale) for quantitative amino acid analyses.

References

1. Davies CL, Dux EL, Duhme-Klair AK (2009) Dalton Trans

10141–10154

2. Ringenberg MR, Ward TR (2011) Chem Commun 47:8470–8476

3. Zhao X, Yeung N, Russell BS, Garner DK, Lu Y (2006) J Am

Chem Soc 128:6766–6767

4. Sato H, Hayashi T, Ando T, Hisaeda Y, Ueno T, Watanabe Y

(2004) J Am Chem Soc 126:436–437

5. Yokoi N, Ueno T, Unno M, Matsui T, Ikeda-Saito M, Watanabe

Y (2008) Chem Commun 229–231

6. Wilchek M, Bayer E (1990) Methods in enzymology. Academic,

San Diego

7. Ward TR (2011) Acc Chem Res 44:47–57

8. Pordea A, Creus M, Panek J, Duboc C, Mathis D, Novic M, Ward

TR (2008) J Am Chem Soc 130:8085–8088

9. Depina K, Navarro C, McWalter L, Boxer DH, Price NC, Kelly

SM, Mandrand-Berthelot MA, Wu LF (1995) Eur J Biochem

227:857–865

10. Heddle J, Scott DJ, Unzai S, Park SY, Tame JRH (2003) J Biol

Chem 278:50322–50329

11. Cherrier MV, Martin L, Cavazza C, Jacquamet L, Lemaire D,

Gaillard J, Fontecilla-Camps JC (2005) J Am Chem Soc

127:10075–10082

12. Cherrier MV, Cavazza C, Bochot C, Lemaire D, Fontecilla-

Camps JC (2008) Biochemistry 47:9937–9943

13. Cavazza C, Bochot C, Rousselot-Pailley P, Carpentier P, Cherrier

MV, Martin L, Marchi-Delapierre C, Fontecilla-Camps JC,

Menage S (2010) Nat Chem 2:1069–1076

14. Abu-Omar MM, Loaiza A, Hontzeas N (2005) Chem Rev

105:2227–2252

15. Costas M, Mehn MP, Jensen MP, Que L (2004) Chem Rev

104:939–986

16. Rousselot-Pailley P, Bochot C, Marchi-Delapierre C, Jorge-

Robin A, Martin L, Fontecilla-Camps JC, Cavazza C, Menage S

(2009) ChemBioChem 10:545–552

17. Oddon F, Girgenti E, Lebrun C, Marchi-Delapierre C, Pecaut J,

Menage S (2012) Eur J Inorg Chem 1:85–96

18. Fish WW (1988) Methods Enzymol 158:357–364

19. Birdsall B, King RW, Wheeler MR, Lewis CA, Goode SR,

Dunlap RB, Roberts GCK (1983) Anal Biochem 132:353–361

20. Murphy CT, Spudich JA (1999) Biochemistry 38:3785–3792

21. Kabsch W (2010) Acta Crystallogr Sect D 66:125–132

22. McCoy AJ, Grosse-Kunstleve RW, Adams PD, Winn MD, Sto-

roni LC, Read RJ (2007) J Appl Crystallogr 40:658–674

23. Adams PD, Afonine PV, Bunkoczi G, Chen VB, Davis IW, Ec-

hols N, Headd JJ, Hung LW, Kapral GJ, Grosse-Kunstleve RW,

McCoy AJ, Moriarty NW, Oeffner R, Read RJ, Richardson DC,

Richardson JS, Terwilliger TC, Zwart PH (2010) Acta Crystal-

logr Sect D 66:213–221

24. Emsley P, Cowtan K (2004) Acta Crystallogr Sect D

60:2126–2132

25. Johnson ER, Keinan S, Mori-Sanchez P, Contreras-Garcia J,

Cohen AJ, Yang WT (2010) J Am Chem Soc 132:6498–6506

26. Schrodinger (2011) Jaguar, version 7.8. Schrodinger, New York

27. Dunietz BD, Beachy MD, Cao YX, Whittington DA, Lippard SJ,

Friesner RA (2000) J Am Chem Soc 122:2828–2839

28. Hay PJ, Wadt WR (1985) J Chem Phys 82:299–310

29. White MC, Doyle AG, Jacobsen EN (2001) J Am Chem Soc

123:7194–7195

30. Costas M, Tipton AK, Chen K, Jo DH, Que L (2001) J Inorg

Biochem 86:188

31. Hagen WR (1992) Adv Inorg Chem 38:165–222

32. Hayward S, Berendsen HJC (1998) Proteins Struct Funct Genet

30:144–154

33. Tina KG, Bhadra R, Srinivasan N (2007) Nucleic Acids Res

35:473–476

34. Gillespie RJ (1960) Can J Chem 38:818–826

35. Holmes RR (1978) J Am Chem Soc 100:433–446

36. Mas-Balleste R, Costas M, van den Berg T, Que L (2006) Chem

Eur J 12:7489–7500

37. Bukowski MR, Halfen HL, van den Berg TA, Halfen JA, Que L

(2005) Angew Chem Int Ed 44:584–587

38. Herrero S, Uson MA (2008) Dalton Trans 4993–4998

39. Kamal JKA, Behere DV (2002) Biochemistry 41:9034–9042

40. Addy C, Ohara M, Kawai F, Kidera A, Ikeguchi M, Fuchigami S,

Osawa M, Shimada I, Park SY, Tame JRH, Heddle JG (2007)

Acta Crystallogr Sect D 63:221–229

41. Contreras-Garcia J, Yang WT, Johnson ER (2011) J Phys Chem

115:12983–12990

42. Zaric SD, Popovic DM, Knapp EW (2000) Chem Eur J

6:3935–3942

43. Bissantz C, Kuhn B, Stahl M (2010) Eur J Med Chem 53:6241

44. Nishio M (2011) Phys Chem Chem Phys 13:13873–13900

45. Nishio M, Hirota M (1989) Tetrahedron 45:7201–7245

46. Chakrabarti P, Samanta U (1995) J Mol Biol 251:9–14

47. Harigai M, Kataoka M, Imamoto Y (2006) J Am Chem Soc

128:10646–10647

48. Panigrahi SK, Desiraju GR (2007) Proteins Struct Funct Bioin-

form 67:128–141

49. Knof U, von Zelewsky A (1999) Angew Chem Int Ed

38:302–322

J Biol Inorg Chem (2012) 17:817–829 829

123