Multistate CASPT2 Study of Native Iron(III)-Dependent Catechol

ARTICLE pubs.acs.org/JPCB

Multistate CASPT2 Study of Native Iron(III)-Dependent Catechol Dioxygenase and Its Functional Models: Electronic Structure and Ligand-to-Metal Charge-Transfer Excitation Naoki Nakatani,† Yutaka Hitomi,‡ and Shigeyoshi Sakaki*,§ †

Department of Chemistry, Graduate School of Science, Kyoto University, Kitashirakawa-Oiwakecho, Sakyo-ku, Kyoto 606-8502, Japan ‡ Department of Molecular Chemistry and Biochemistry, Faculty of Science and Engineering, Doshisha University, Kyotanabe, Kyoto 610-0321, Japan § Institute for Integrated Cell-Material Sciences (iCeMS), Kyoto University, Yoshida, Sakyo-ku, Kyoto 606-8501, Japan

bS Supporting Information ABSTRACT: We theoretically investigated the ligand-to-metal charge-transfer (LMCT) excitation of the native iron(III)-dependent catechol dioxygenase and its functional model complexes with multistate complete active space second-order perturbation theory (MS-CASPT2) because the LMCT (catecholate-to-iron(III) charge-transfer) excitation energy is believed to relate to the reactivity of the native enzyme and its functional model complexes. The ground state calculated by the MS-CASPT2 method mainly consists of the iron(III)
catecholate electron configuration and moderately of the iron(II)
semiquinonate electron configuration for both of the enzyme active centers and the model complexes when the active center exists in the protein environment and the model complexes exist in the solution. However, the ground-state wave function mainly consists of the iron(II)
semiquinonate electron configuration for both the enzyme active site without a protein environment and the model complexes in vacuo. These results clearly show that the protein environment and solvent play important roles to determine the electronic structure of the catecholatoiron(III) complex. The LMCT excitation energy clearly relates to the weight of the iron(III)
catecholate configuration in the ground state. The reactivity and the LMCT excitation energy directly relate to the ionization potential of the catecholate (IPCAT) in the model complex. This is because the charge transfer from the catecholate moiety to the dioxygen molecule plays a key role to activate the dioxygen molecule. However, the reactivity of the native catechol dioxygenase is much larger than those of the model complexes, despite the similar IPCAT values, suggesting that other factors such as the coordinatively unsaturated iron(III) center of the native enzyme play a crucial role in the reactivity.

’ INTRODUCTIONS Oxygenase enzymes perform various biodegradation and biosynthetic reactions with the dioxygen molecule (O2). It is wellknown that the dioxygen molecule is not reactive with most organic molecules because of its triplet spin state in the ground state. Thus, the activation of the dioxygen molecule is essential to the function of oxygenase enzymes. Catechol dioxygenase, found in wide range of soil bacteria, activates the dioxygen molecule to catalyze aromatic bond cleavage of catechol (1,2-hydroxybenzene) and its derivatives.1 This oxygenation is important as the last step of biodegradation of aromatic compounds. Catechol dioxygenase includes either a nonheme iron(II) or iron(III) complex in its active site, which plays a key role in the dioxygen activation process. In this regard, a lot of studies were reported, as summarized in recent reviews.2
6 In particular, iron(III)-dependent catechol dioxygenase is of considerable interest because its oxygenation reaction is against our general understanding that a high-spin iron(III) center does not react with the dioxygen molecule. r 2011 American Chemical Society

Protocatechuate 3,4-dioxygenase (PCD), the most typical iron(III)-dependent catechol dioxygenase, contains a high-spin iron(III) center coordinated with two histidines (His460 and His462), two tyrosines (Tyr408 and Tyr447), and one water molecule in its active site.7 The substrate protocatechuate 3,4dihydroxybenzoate (PCA) is bound with the active center to form a PCD
PCA complex 1 with concomitant dissociation of Tyr447 and water molecule,8 as shown in Scheme 1. In 1, the active center is a five-coordinate catecholatoiron(III) complex. Because a high-spin iron(III) center does not react with a dioxygen molecule in general, the iron(III)
catecholate (iron(III)
CAT) complex is considered to be inactive for dioxygen activation. However, the PCD rapidly catalyzes the oxygenation reaction of PCA with a large rate constant of 5  105 M
1 s
1.9 Received: October 20, 2010 Revised: March 20, 2011 Published: April 04, 2011 4781

dx.doi.org/10.1021/jp110045f | J. Phys. Chem. B 2011, 115, 4781–4789

The Journal of Physical Chemistry B Scheme 1

ARTICLE

Scheme 3

Scheme 2

On the other hand, large numbers of biomimetic and bioinspired catecholatoiron(III) complexes have been synthesized for two decades. Interestingly, some of them exhibit the catalytic activity for the oxygenation reaction of catecholate.2,10 They are called functional model complexes. These are of considerable importance because the dioxygen activation process by iron(III)dependent catechol dioxygenase is often discussed based on those functional model complexes. Because the six-coordinate catecholatoiron(III) model complex reacts with the dioxygen molecule despite the absence of the available binding site,11 it has been proposed that the dioxygen molecule directly reacts with the catecholate moiety. This proposal is called the “substrate activation” mechanism.11 Considering that the closed-shell singlet catecholate anion is not reactive with the dioxygen molecule, it has been proposed that an iron(II)
semiquinonate (iron(II)
SQ) complex is formed through the ligand-to-metal charge transfer (LMCT) from the catecholate moiety to the iron(III) center and that the radical center is induced in the semiquinonate moiety to react with the dioxygen molecule (Scheme 2). Another reaction mechanism called the “oxygen activation” mechanism was proposed, in which the dioxygen molecule directly binds to the iron center first.5,12 In this mechanism, the charge transfer (CT) from the catecholate anion to the iron(III) center occurs to afford the high-spin iron(II) center which can easily react with the dioxygen molecule because the high-spin iron(III) center does not react with the dioxygen molecule as mentioned above. Previous theoretical works supported the oxygen activation mechanism in enzymatic reaction.13,14 Also, it has been discussed that the presence of an unsaturated five-coordinate catecholatoiron(III) species is crucial for the oxygenation catalysis because the vacant site is necessary for binding with dioxygen molecule.8,15 In both substrate activation and oxygen activation mechanisms, the CT from the catecholate to the iron(III) center is considered to be crucial for the dioxygen activation reaction, as mentioned above. Actually, the LMCT absorption is observed

around the near-infrared region both in the native catechol dioxygenase and in the functional models. For instance, the PCD
PCA complex 1 exhibits the LMCT excitation in the visible light region (1.65 eV).16 In functional models, the modification of the ligand and the introduction of a substituent on the catecholate ring induce a considerable shift of the LMCT absorption spectrum and a significant change of the reactivity for the dioxygen molecule. On the basis of those results, Que and coworkers suggested that the reactivity for the dioxygen molecule depends on the LMCT excitation energy.10a Later, Hitomi, Funabiki, and their co-workers found a linear relationship between the LMCT excitation energy and the reactivity.17 For instance, [Fe(TPA)(CAT)]þ 2, [Fe(TPA)(DMC)]þ 3, and [Fe(TPA)(4CC)]þ 4 exhibit the LMCT absorption at 1.53, 1.39, and 1.58 eV, respectively, and the rate constants for the dioxygen activation reaction are 1.3, 54, and 0.16 M
1 s
1, respectively, in acetonitrile (MeCN) solution,17 where TPA = tris(pyridylmethyl)amine, CAT = catecholate, DMC = 3,5-dimethylcatecholate, and 4CC = 4-chlorocatecholate (see Scheme 3 for structures). Apparently, the reactivity increases as the LMCT excitation energy decreases. This result is important experimental support that the iron(II)
SQ electron configuration participates in the ground state of the catecholatoiron(III) complex. Several experimental works presented discussion that the ground state mainly consists of the resonance structure between iron(III)
CAT and iron(II)
SQ electron configurations and that the reactivity for the dioxygen molecule is mainly determined by the content of the iron(II)
SQ configuration.18 On the other hand, previous EPR and M€ossbauer spectroscopic studies showed that the active site of the enzyme kept high-spin iron(III) character and the iron(II)
SQ character was not observed in the ground state.19 These experimental observations provide a question of how the iron(II)
SQ configuration participates in the ground state of the catecholatoiron(III) complex and the dioxygen activation reaction. Thus, it is worth theoretically investigating the relations between the electronic structure of the iron center, the LMCT excitation energy, and the reactivity for the dioxygen molecule of both the enzyme and functional models. Theoretical knowledge of such relations is indispensable to discuss the dioxygen activation by the iron(III)-dependent catechol dioxygenase, to understand the origin of high reactivity of the native enzyme, and to design a reactive bioinspired oxygenation catalyst. 4782

dx.doi.org/10.1021/jp110045f |J. Phys. Chem. B 2011, 115, 4781–4789

The Journal of Physical Chemistry B

ARTICLE

Scheme 4. Definition of QM and MM Regionsa

a

Atoms described by the sphere model are the QM region, and the other atoms are the MM region.

Figure 1. (A) Geometry of native iron(III) catechol dioxygenase (PCD
PCA 1) (optimized by the two-layer ONIOM(DFT(B3LYP): AMBER) method) and geometries of functional model complexes (optimized by the DFT(B3LYP) method); (B) [Fe(TPA)(CAT)]þ 2, (C) [Fe(TPA)(DMC)]þ 3, and (D) [Fe(TPA)(4CC)]þ 4. Important bond lengths (unit: Å): In 1, Fe
O3 = 2.05, Fe
O4 = 2.37, Fe
OTyr408 = 1.94, Fe
NHis460 = 2.17, Fe
NHis462 = 2.18, and O3
HR = 1.68. In 2, Fe
O1 = 2.01, Fe
O2 = 2.05, Fe
N1 = 2.29, Fe
N2 = 2.23, Fe
N3 = 2.19, and Fe
N4 = 2.23. In 3, Fe
O1 = 2.01, Fe
O2 = 2.06, Fe
N1 = 2.28, Fe
N2 = 2.24, Fe
N3 = 2.19, and Fe
N4 = 2.24. In 4, Fe
O1 = 2.01, Fe
O2 = 2.04, Fe
N1 = 2.28, Fe
N2 = 2.24, Fe
N3 = 2.18, and Fe
N4 = 2.23.

In this work, we theoretically investigated the LMCT excitation energies of the native iron(III)-dependent catechol dioxygenase and several model complexes, their electronic structures of both ground and excited states, and the relation between the LMCT excitation energy and the reactivity for the dioxygen molecule. The resonance between iron(III)
CAT and iron(II)
SQ states occurs when these two states exist close to each other in energy. This is known as the quasi-degenerate electronic structure. In such a case, the wave function is represented by a mixture of these states. This means that catechol dioxygenase and its functional model complexes should be investigated with multireference electronic structure theory to incorporate the nondynamical correlation effect. We employed here the multistate (MS) CASPT2 method20 because this method can incorporate well nondynamical correlation effects as well as dynamical correlation effects.

’ COMPUTATIONAL DETAILS We investigated here a substrate-bound complex of native iron(III)-dependent catechol dioxygenase (PCD
PCA) 1 and several functional model complexes, [Fe(TPA)(CAT)]þ 2, [Fe(TPA)(DMC)]þ 3, and [Fe(TPA)(4CC)]þ 4, as shown in Figure 1. In 1, we added hydrogen atoms based on the experimental structure from the Protein Data Bank (PDB code: 3PCA)8 and determined the positions of added hydrogen atoms with partial molecular dynamics simulation in which all atoms except for the hydrogen atoms were fixed to those of the experimental structure; see Supporting Information page S3 for details. Then,

we performed geometry optimization by employing the two-layer ONIOM(QM:MM) method.21 We involved Fe, PCA, Tyr408, Arg457, His460, and His462 in the QM region and the others in the MM region, as shown in Scheme 4. We employed density functional theory with the B3LYP functional (DFT(B3LYP))22 for the QM calculations and the AMBER96 force fields for the MM calculations. In the QM calculation, we employed Stuttgart/ Dresden (311111/22111/411/1) ECP10MDF basis sets for Fe,23 6-31G* basis sets for H and C, and 6-31þG* basis sets for N and O. In the functional model complexes, 2, 3, and 4, we performed the geometry optimization by the DFT(B3LYP) method with the same basis sets as those employed for the QM calculation of 1. To evaluate the excitation energy, we employed teh multistate CASPT2 (MS-CASPT2) method,20 considering large nondynamical electron correlation in the Fe atom and quasi-degenerate states of the catecholatoiron(III) complex. For Fe, we used the same basis set as that used in the DFT calculation and atomic natural orbital (ANO) basis sets for the others, [4s2p/2s1p] for the H in the CAT moiety, [4s/2s] for the other H, [10s6p4d/ 3s2p1d] for the C in the CAT moiety, [10s6p/3s2p] for the other C, and [10s6p4d/4s3p1d] for N and O.24 Though we tried to use the ANO basis set for Fe in the CASPT2 calculation of the realistic compound, we could not perform the calculation due to huge computational cost. Here, we carried out an MS-CASPT2 calculation of a small model system using the ANO basis set for Fe and found that the results from this calculation are similar to those from the MS-CASPT2 computation with the Stuttgart/ Dresden (311111/22111/411/1) basis set and ECP, which suggests that the use of the Stuttgart/Dresden basis set for Fe is reasonable; see Supporting Information page S4. In 1, we employed point charge (PC) approximation with AMBER96 charges to evaluate the protein electrostatic environment that is involved in the MM region. In the functional model complexes of 2, 3, and 4, we employed the polarized continuum model (PCM)25 to evaluate solvation effects of MeCN, which were used in the experimental study.17 In the PCM calculation, the dielectric constant of 36.64 (MeCN) was employed. In MS-CASPT2 calculations, the CASSCF method26 was employed to obtain a reference wave function. The active space of the CASSCF calculation includes nine electrons in nine orbitals 4783

dx.doi.org/10.1021/jp110045f |J. Phys. Chem. B 2011, 115, 4781–4789

The Journal of Physical Chemistry B

ARTICLE

Figure 2. Averaged orbitals of 1 calculated by the 11SA-CASSCF(9 in 9) method.

such as five d orbitals of the Fe and two sets of π and π* orbitals of the CAT moiety; see Figure 2 for the natural orbitals of 1. This active space is enough to represent iron(III)
CAT and iron(II)
SQ configurations that are likely to appear in the ground and the LMCT excited states, as will be discussed below. The CASSCF calculation with this active space is represented by the CASSCF(9 in 9) method hereafter. If the resonance structure between the iron(III)
CAT and the iron(II)
SQ configurations exists, these states must exist in a very close energy region. Thus, we carried out 11 stateaveraged (11SA) CASSCF(9 in 9) calculations to obtain the reference wave function; see Supporting Information Figure S2 for relative stabilities. In MS-CASPT2 calculations, we employed the imaginary shift of 0.20 au to avoid the intruder problem.27 In the 3d transition-metal complexes, the importance of the double-shell effect was reported.28 Here, we also carried out CASSCF(9 in 14)/CASPT2 calculations in which the 4d-like orbitals were involved in the active space. In these calculations, we employed small model compounds because a realistic compound is too large to perform CASSCF(9 in 14)/ CASPT2 calculation. The character of the wave function varies little between the CASSCF(9 in 9)/CASPT2 and the CASSCF(9 in 14)/CASPT2 calculations. Though the excitation energy by the former calculation is somewhat larger than that by the latter one, the same trend is presented by both of the calculations. These results suggest that the CASSCF(9 in 9)/ CASPT2 calculations present correct understanding about the electronic structure of iron(III)-CAT systems and qualitatively correct results about the excitation energy; see Supporting Information pages S7
8. In 1, we analyzed the electrostatic interaction between the active site and each amino acid residue, which is approximately represented by classical electrostatic interaction energy EES

defined by eq 1 EES

¼ ¼

i ∈ MM

∑i

EES, i

i ∈ MM

∑i

A ∈ iB ∈ QMQ

∑A ∑B

A QB

!

ð1Þ

rAB

where index i denotes each residue and QA is the charge on atom A. We further introduced the energy difference, ΔEES, between the iron(III)
CAT and the iron(II)
SQ states defined by eq 2 ΔEES

¼ ¼ ¼ ¼

i ∈ MM

∑i

i ∈ MM

∑i

i ∈ MM

∑i

i ∈ MM

∑i

ΔEES, i ironðIIÞ

ðEES, i

ironðIIIÞ


EEX, i

A ∈ iB ∈ QMQ

∑A ∑B

Þ

ironðIIÞ A QB

rAB




A ∈ iB ∈ QMQ ðQ ironðIIÞ A B

∑A ∑B

A ∈ iB ∈ QMQ Q ironðIIIÞ A B

∑A ∑B

!

!

rAB

ironðIIIÞ
QB Þ

rAB

ð2Þ where the negative ΔEES value represents that the electrostatic interaction stabilizes the iron(III)
CAT state more than the iron(II)
SQ state. We focused on the sextet spin state here, which corresponds to the high-spin iron(III) center, because the other possible spin states such as the quartet and doublet spin states are much more unstable than the sextet spin state by 10.4 and 40.8 kcal/mol, respectively, as reported by our previous work.14 All geometry optimizations were performed by the Gaussian 03 (revision C02) 4784

dx.doi.org/10.1021/jp110045f |J. Phys. Chem. B 2011, 115, 4781–4789

The Journal of Physical Chemistry B

ARTICLE

Table 1. Calculated LMCT Excitation Energies (ΔE in eV), Calculated Oscillator Strength (f in au), and Experimental Values of Complexes 1, 2, 3, and 4 cplx.

MS-CASPT2

CASPT2

TD-DFT

state ΔE/eV f/au state ΔE/eV state ΔE/eV f/au 1

6

2A 36A

2

0.01

0.63

0.00

0.74

46A

0.81

0.07

6

1.79

0.01 4 A

1.07

5A

0.93

0.00

2.02 1.08

0.09 56A 0.00 26A

1.17 0.37

66A 26A

1.22 0.41

0.07 0.00

36A

1.12

0.00 36A

0.44

36A

0.46

0.00

0.46

46A

1.41

0.12

6

1.45

6

0.47

4A

6

0.17 4 A

6

2A

0.85

0.00 2 A

0.19

2A

0.35

0.00

36A

0.86

0.00 36A

0.21

36A

0.42

0.00

46A

1.25

0.19 46A

0.23

46A

1.45

0.12

0.46

6

2A

0.43

0.00

0.59 0.62

6

0.48 1.36

0.00 0.15

6

2A 6

3A 46A a

0.06 36A

2A 36A

0.57

56A 26A 4A

4

1.57

0.01 2 A

6

6

6

3

1.40

6

1.21 1.24 1.54

6

6

0.00 2 A 6

0.00 3 A 0.17 46A

3A 46A

Scheme 5

exptl. ΔE/eV

1.65a 2.18a

1.53b

1.39b

1.58b

b

Reference 16. Reference 17.

program package.29 Molecular dynamics simulation was performed by the AMBER 8 program package.30 The MS-CASPT2 calculations were performed by the Molcas 6.4 program package.31

’ RESULTS AND DISCUSSIONS Absorption Spectra and Electronic Structures. MS-CASPT2calculated excitation energies and oscillator strengths are summarized in Table 1. These results agree well with the experimental values; the rms error is 0.09 eV for the first peak of the experimental data. As discussed later, the ground state mainly consists of the iron(III)
CAT electron configuration, and the low-lying excited states mainly consist of iron(II)
SQ electron configurations, indicating that the excitation is assigned to the LMCT transition. In 1, the excitation to the lowest-energy excited state (26A) exhibits a very small oscillator strength, though the intense absorption is experimentally observed as the first peak. Thus, the first peak is not assigned to the X6A f 26A excitation but to the X6A f 36A excitation, the oscillator strength of which is calculated to be considerably large. Also, the excitation to the 56A state of 1 exhibits a somewhat large oscillator strength, which corresponds to the second peak experimentally observed.16 These two excitation energies are calculated to be 1.57 and 2.02 eV, respectively, by the MS-CASPT2 method, which both agree well with their experimental values of 1.65 and 2.18 eV, respectively. In 2, 3, and 4, oscillator strengths of X6A f 26A and X6A f 36A excitations were calculated to be almost 0. Thus, the first peaks of 2, 3, and 4 are not assigned to these two excitations but to the X6A f 46A excitation because the latter excitation exhibits a considerably large oscillator strength. The calculated excitation energy increases in the order 3 < 2 < 4 < 1. This increasing order agrees with that of the experimentally observed first peak.16,17 These results indicate that the MS-CASPT2 method presents reliable results for the electronic structures of 1
4. We also evaluated the excitation energy with the CASPT2 and time-dependent DFT (TD-DFT)32 methods. As summarized in Table 1, however, the CASPT2-calculated excitation

energies are considerably smaller than the experimental values. The TD-DFT-calculated excitation energy increases in the order 1 < 4 < 2 < 3, which is reverse to the experimental trend. These results indicate that the multireference and quasi-degenerate problems must be taken into consideration to evaluate correctly the excitation energy of these complexes. The use of the MSCASPT2 method is crucial to discuss the ground and excited states of these catecholatoiron(III) complexes. We discuss LMCT excitation energies and electronic structures based on the MS-CASPT2 computational results hereafter. Interestingly, “dark states”, which possess either very small or no oscillator strength, are calculated in the lower-energy region rather than the first peak, as mentioned above (see also Table 1). The excitation energies to the lowest-energy excited states of 1, 2, 3, and 4 are calculated to be 1.40, 1.08, 0.85, and 1.21 eV, respectively, and their oscillator strengths are calculated to be less than 0.01 au. If their excitation energies are very small, these dark states should be considered to be in thermal equilibrium. However, these dark states still exist at high energies compared to the thermal energy and do not mix into the ground state (at most 4  10
15). To discuss the electronic structures of the ground and excited states, we must involve iron(III)
CAT and iron(II)
SQ configurations in the MS-CASPT2 calculation because these two electron configurations are expected to exist in the close energy region, as discussed above. In the iron(II)
SQ state, 10 kinds of electron configuration must be considered, as described in Scheme 5. Actually, 1 iron(III)
CAT and 10 iron(II)
SQ electron configurations are calculated at the low-energy region, but the other electron configuration (π
π* excitation, for instance) is calculated at a much higher energy; see Supporting Information Figure S2. We evaluated the weight of the iron(III)
CAT configuration wFe(III) and the weight of the sum of all of the iron(II)
SQ configurations wFe(II). In the ground states, the wFe(III) value is 0.80, 0.74, 0.68, and 0.76 for 1, 2, 3, and 4, respectively, and the wFe(II) value is 0.13, 0.19, 0.25, and 0.17 for 1, 2, 3, and 4, respectively. These results clearly indicate that the ground states of these catecholatoiron(III) complexes mainly consist of the iron(III)
CAT configuration (68
80%) and that the iron(II)
SQ configurations moderately contribute to the ground state (13
25%). The above results lead to the conclusion that the ground state is represented by the multireference wave function, which corresponds to the “resonance structure”. 4785

dx.doi.org/10.1021/jp110045f |J. Phys. Chem. B 2011, 115, 4781–4789

The Journal of Physical Chemistry B

Figure 3. Relationship between the MS-CASPT2-calculated LMCT energy and the weight of the iron(III)
CAT configuration wFe(III), in the ground states of complexes 1, 2, 3, and 4.

The contribution of the iron(II)
SQ configurations is considerably larger than the thermal contribution of the dark states. This is consistent with experimental result previously reported.18a Though it was previously reported from the ESR and the M€ossbauer spectroscopic studies that the iron(II)
SQ character was not observed in the active site of native enzyme,19 our result indicates that the ground state involves a non-negligible contribution of the iron(II)
SQ configuration in 1. However, this result is not inconsistent with experimental facts; see ref 33 for detailed information. Because the DFT-optimized Fe
O distances of 3 were somewhat longer than the experimental Fe
O distances (∼1.9 Å) of [Fe(TPA)(DBC)]BPh4 (DBC = 3,5-di-tert-butylcatecholate),10a we investigated 3 having those experimental Fe
O distances. The ground state involves ∼20% of the iron(II)
SQ configurations at the experimental Fe
O distances. This result is not very different from that with the DFT-optimized Fe
O distances, suggesting that our results based on the DFT-optimized geometries are qualitatively correct, at least; see Supporting Information page S6. Furthermore, it should be mentioned here that the weight of the iron(II)
SQ configuration decreases as the Fe
O distance increases, indicating that the Fe
O distance is important for the discussion of the extent of the iron(II)
SQ configuration in the catecholatoiron(III) complex. Interestingly, the wFe(III) value is proportional to the calculated and the observed LMCT excitation energies, as shown in Figure 3, indicating that the LMCT excitation energy relates to the ground-state electronic structure; as the iron(III)
CAT character decreases and the iron(II)
SQ character increases in the ground state, the LMCT excitation energy becomes smaller. Also, the wFe(II) value relates to the reactivity of functional models for dioxygen activation; for instance, the reactivity increases in the order 4 < 2 < 3, which is the same as the increasing order of the iron(II)
SQ character in the ground state. However, this relation is not kept in the native enzyme 1, as follows; though the wFe(II) value of 1 is calculated to be the smallest in all of these complexes examined here, 1 is the most reactive for the dioxygen molecule. This indicates that other factors are responsible for the reactivity of 1, which will be discussed below. Environmental Effects on the Electronic Structure. Here, we wish to make a comparison among the catecholatoiron(III)

ARTICLE

Figure 4. Active site model that corresponds to the QM region in the protein (see also Computational Details), where MePhO = p-methylphenoxide (model of tyrosine), MeIm = 3H-4-methylimidazole (model of δ-protonated histidine), and PrG = guanidinopropane (model of arginine). The geometries were taken to be the same as that optimized by the two-layer ONIOM(DFT(B3LYP):AMBER) method.

Table 2. Environmental Effects on the LMCT Excitation Energy and Weight of the Iron(III)
CAT Configuration wFe(III).a ΔE/eV

wFe(III)

Native Enzyme in vacuo

1vac

1.31

0.19

þ electrostatic interaction

1

1.57

0.80

in vacuo

2vac

1.22

0.35

þ solvation (MeCN)

2

1.45

0.74

[Fe(TPA)(CAT)]þ

a

Calculated by the MS-CASPT2 method.

complex in vacuo, protein, and solvent. First, we investigated the active site of native enzyme 1 and the functional model complex 2 without any environmental effects; for brevity, 1 and 2 in vacuo are named 1vac and 2vac, respectively, but the names of 1 and 2 are used to represent 1 in the protein and 2 in acetonitril solution, respectively. In 1vac, only the QM region, [Fe(MePhO)(MeIm)2(PrG)(PCA)], was considered, where p-MePhO (p-methylphenoxide), MeIm (3H4-methylimidazole), and PrG (guanidinopropane) were employed as models of tyrosine, δ-protonated histidine, and arginine, respectively, as shown in Figure 4. The excitation energies of 1vac and 2vac are calculated to be 1.31 and 1.22 eV, respectively, which are much smaller than those of 1 and 2 (Table 2). Interestingly, the wFe(III) value is 0.19 and 0.35 for 1vac and 2vac, respectively, which are much smaller than those of 1 and 2 (Table 2). These results indicate that the ground state mainly consists of the iron(II)
SQ configuration in vacuo, which is completely different from those of 1 and 2. Also, the absorption spectra of 1vac and 2vac are assigned as metal-to-ligand charge-transfer (MLCT) excitations from the iron(II) center to the semiquinonate moiety. Thus, it is concluded that the environmental effects are of considerable importance, and the electronic structure of the catecholatoiron(III) complex is completely changed by the environment. Here, we investigate two important factors about protein environmental effects; one is the electrostatic interaction by a protein, and the other is hydrogen bonding with an amino acid residue. 4786

dx.doi.org/10.1021/jp110045f |J. Phys. Chem. B 2011, 115, 4781–4789

The Journal of Physical Chemistry B

ARTICLE

Figure 5. (A) Difference in the electrostatic interaction energy between the iron(III)
CAT state and the iron(II)
SQ state for each residue i (ΔEES,i as defined by eq 3). The positive value represents that the residue i stabilizes the iron(III)
CAT state compared to the iron(II)
SQ state and vice versa. (B) Residues that remarkably contribute to ΔEES,i.

Protein Environment by Electrostatic Interaction. To analyze the electrostatic interaction between the active site and each amino acid residue, we employed the AMBER96 charges (QA) and the Mulliken charges of the iron(III)
CAT state ) and the iron(II)
SQ state (QFe(II) ), as appeared in (QFe(III) B B and QFe(II) were calculated by the 11SAeqs 2, where QFe(III) B B CASSCF(9 in 9) method.34 As shown in Figure 5A, Arg133 contributes to a positively large value of ΔEES, and Asp65 contributes to a negatively large value of ΔEES. This is because positively charged Arg133 and negatively charged Asp65 are close to the deprotonated carboxyl group of the protocatechuate (Figure 5B). These electrostatic interactions totally stabilize the iron(III)
CAT state. In other words, Arg133 plays a key role to increase the iron(III)
CAT character in the ground state. Solvation Effects. In the functional model complexes, solvation effects are important for the electronic structure. We employed the PCM method to consider the solvation effects of MeCN.17 As discussed above, the wFe(III) value is considerably larger in 2 than that in 2vac. Because MeCN is an aprotic polar solvent, the solvation effects mostly arise from the dipole
dipole interaction between the catecholatoiron(III) complex and solvent. From the 11SA-CASSCF wave function of 1vac, the dipole moment of the iron(III)
CAT state was calculated to be 18.7 D, but those of the iron(II)
SQ states were calculated to be 4.3
4.6 D. This result suggests that the dipole moment of the iron(III)
CAT state is approximately described by the interaction between þ3 and
2 centers, but those of the iron(II)
SQ states are approximately described by the interaction between þ2 and
1 centers. Thus, the polar solvent considerably stabilizes the iron(III)
CAT state, leading to the increase of the wFe(III) value in the ground state. Relation between the Electronic Structure and the Reactivity. On the basis of the CASPT2 computational results, we previously reported that the direct CT occurs from the π orbital of the CAT moiety to the dioxygen molecule in the dioxygen activation process.14 This means that the π orbital energy of CAT is an important factor for the activation of the dioxygen molecule because the CT from the CAT moiety to the dioxygen molecule becomes stronger as the π orbital energy becomes higher. However, the concept of orbital energy disappears in CASSCF and CASPT2 wave functions. We employed here the vertical ionization potential of the CAT moiety (IPCAT) as a measure of

Scheme 6. Electron Affinity of the Dioxygen Molecule (EAOX) and Ionization Potential of the Catecholate Moiety (IPCAT) versus the Fe
O2 Distancea

a

Data were taken from our previous study.24 The dashed red line represents the case in which the IPCAT decreases by some effects, for instance, a substituent on CAT ring, a ligand coordinated to the iron center, an environmental effect, and so forth.

the ability for the CT interaction instead of the orbital energy because the IPCAT directly relates to the π orbital energy of the CAT moiety.35 The IPCAT was defined by eq 3 IPCAT ¼ EðironðIIIÞ
CATÞ
EðironðIIIÞ
SQ Þ

ð3Þ

E(iron(III)
CAT) and E(iron(III)
SQ) represent the total energy of the catecholatoiron(III) system taking the iron(III)
CAT state and that of the semiquinonatoiron(III) system taking the iron(III)
SQ state, respectively. They are evaluated here with the CASPT2(9 in 9) and CASPT(8 in 9) methods, respectively, where the iron(III)
SQ state takes a septet spin 4787

dx.doi.org/10.1021/jp110045f |J. Phys. Chem. B 2011, 115, 4781–4789

The Journal of Physical Chemistry B

Figure 6. Optimized geometry of [Fe(TPA)(PCA)] 5 (optimized by the DFT(B3LYP) method). Important bond lengths (unit: Å): Fe
O1 = 1.96, Fe
O2 = 2.09, Fe
N1 = 2.37, Fe
N2 = 2.23, Fe
N3 = 2.21, and Fe
N4 = 2.22.

multiplicity.36 The calculated IPCAT decreases in the order 4 (6.04 eV) > 2 (5.99 eV) > 3 (5.71 eV), where in parentheses are IPCAT values. Because of this trend, the reactivity increases in the order 4 < 2 < 3, and the LMCT excitation energy decreases in the order 4 > 2 > 3, as experimentally observed.17 Thus, it is concluded that because the IPCAT decreases in the order 4 > 2 > 3, the reactivity for the dioxygen molecule increases, and the LMCT excitation energy decreases in this order (see Scheme 6). However, the IPCAT of 1 was calculated to be almost the same as that of [Fe(TPA)(DMC)]þ 3, though its reactivity is the highest and its LMCT excitation energy is calculated to be considerably higher than that of 3. These results indicate that not only the IPCAT but also other factors contribute to the highest reactivity of 1. Because not CAT but PCA is a substrate in the native enzyme, we investigated here the model complex [Fe(TPA)(PCA)] 5 to examine the difference between CAT and PCA (Figure 6). The IPCAT of 5 was calculated to be 5.82 eV, which is moderately larger than that (5.72 eV) of the PCD
PCA complex 1 but moderately smaller than that (5.99 eV) of [Fe(TPA)(CAT)]þ 2. This result indicates that PCA is moderately more reactive than the nonsubstituted CAT and that the dioxygen activation occurs easier in the native catechol dioxygenase than in 2. However, it is not enough to explain the highest reactivity of 1. The native enzyme has a five-coordinate iron center, in which the dioxygen molecule can easily approach the coordinatively unsaturated iron center to form a six-coordinate iron intermediate. Because the iron center of 2 is already six-coordinate, the dioxygen binding induces either considerably large structural change or the dissociation of some coordinating ligand. Both are unfavorable for the activation of the dioxygen molecule. In the previous experimental study, it was discussed that the ratedetermining step is O
O bond cleavage in the native enzyme but an O2 binding process in the six-coordinate model complex.37 The importance of the coordinatively unsaturated fivecoordinate metal center was also discussed by the Solomon group.15 Their previous discussion and our computational results suggest that the unsaturated five-coordinate iron(III) center is crucial to the high reactivity of native enzyme.

ARTICLE

CASPT2 method reproduced well the trend of experimental absorption energies of these catecholatoiron(III) complexes. The ground state of the catecholatoiron(III) complex consists mainly of the iron(III)
CAT configuration and moderately of the iron(II)
SQ configurations in the native enzyme 1 and model complexes 2
4 in polar solvent; the contribution of the iron(II)
SQ configuration is at most 25%. The ground-state electronic structure considerably changes by the environment; the ground states of 1
4 are calculated to mainly consist of the iron(II)
SQ configuration in vacuo. In the native iron(III)-dependent catechol dioxygenase, Arg133 plays important role to change the electronic structure of the ground state; Arg133 provides a positively large electrostatic field on the CAT to stabilize the iron(III)
CAT state. In 2
4, the polar solvent plays a crucial role, as follows. Because the dipole moment of the iron(III)
CAT state is considerably larger than that of the iron(II)
SQ state, the iron(III)
CAT state is considerably stabilized by a polar solvent such as MeCN. As the weight of the iron(II)
SQ configuration increases, the LMCT excitation energy decreases, indicating that the LMCT excitation energy is clearly reflected in the electronic structure of the ground state. The reactivity for dioxygen activation relates to the ionization potential (IPCAT) of the CAT moiety because the CT from the CAT moiety to the dioxygen molecule necessarily occurs in the dioxygen activation process. The reactivity of the functional model clearly relates to the IPCAT value. In the native iron(III)-dependent catechol dioxygenase, however, its high reactivity cannot be explained only by the IPCAT value, indicating that the unsaturated five-coordinate iron(III) center of the native enzyme is crucial for the high reactivity, as discussed before.15

’ ASSOCIATED CONTENT

bS

Supporting Information. Complete ref 29, MD-annealing calculation on the enzyme structure, basis set dependency of the MS-CASPT2 calculation, 11SA-CASSCF-calculated energies, MS-CASPT2 calculations on experimental Fe
O distances, and the CASSCF(9 in 14)/CASPT2 calculations of small model systems. This material is available free of charge via the Internet at http://pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected].

’ ACKNOWLEDGMENT The authors gratefully acknowledge Prof. T. Funabiki for fruitful discussion and helpful comments. N. Nakatani thanks to the Japan Society for the Promotion of Science (JSPS) for financial support. This work was supported, in part, financially by Grant-in-Aids for Specially-Promoted-Research (No. 22000009). ’ REFERENCES

’ CONCLUSIONS We reported here the MS-CASPT2 study on the ground and excited states of native iron(III)-dependent catechol dioxygenase 1 and its functional model complexes 2, 3, and 4. The MS-

(1) Hayaishi, O.; Hashimoto, K. J. Biochem. 1950, 37, 371–374. (2) Costas, M.; Mehn, M. P.; Jensen, M. P.; Que, L., Jr. Chem. Rev. 2004, 104, 939–986. (3) Gibson, D. T. In Microbial Degradation of Organic Compounds; Marcel Dekker: New York, 1984. 4788

dx.doi.org/10.1021/jp110045f |J. Phys. Chem. B 2011, 115, 4781–4789

The Journal of Physical Chemistry B (4) Lipscomb, J. D.; Orville, A. M. In Metal Ions in Biological Systems; Sigel, H., Sigel, A., Eds.; Marcel Dekker: New York, 1992; Vol. 28, pp 243
298. (5) Funabiki, T. In Oxygenases and Model Systems; Funabiki, T., Ed.; Kluwer Academic Publishers: Dordrecht, The Netherlands, 1997; pp 19
104. (6) Bugg, T. D. H. Tetrahedron 2003, 59, 7075–7101. (7) (a) Ohlendorf, D. H.; Lipscomb, J. D.; Webber, P. C. Nature 1988, 336, 403–405. (b) Ohlendorf, D. H.; Orville, A. M.; Lipscomb, J. D. J. Mol. Biol. 1994, 244, 429–440. (8) Orville, A. M.; Lipscomb, J. D.; Ohlendorf, D. H. Biochemistry 1997, 36, 10052–10066. (9) Bull, C.; Ballou, D.; Otsuka, S. J. Biol. Chem. 1981, 256, 12681–12686. (10) (a) Jang, H. G.; Cox, D. D.; Que, L., Jr. J. Am. Chem. Soc. 1991, 113, 9200–9204. (b) Raffard, N.; Carina, R.; Simaan, A. J.; Sainton, J.; Riviere, E.; Tchertanov, L.; Bourcier, S.; Bouchoux, G.; Delroisse, M.; Banse, F.; Girerd, J.-J. Eur. J. Inorg. Chem. 2001, 2249–2254. (11) Que, L., Jr.; Kolanczyk, R. C.; White, L. S. J. Am. Chem. Soc. 1987, 109, 5373. (12) Funabiki, T.; Konishi, T.; Yoshida, T.; Inoue, T.; Takano, M.; Yoshida, S. Recl. Trav. Chim. Pays-Bas 1987, 106, 242. (13) Borowski, T.; Siegbahn, P. E. M. J. Am. Chem. Soc. 2006, 128, 12941–12953. (14) Nakatani, N.; Nakao, Y.; Sato, H.; Sakaki, S. J. Phys. Chem. B 2009, 113, 4826–4836. (15) Davis, M. I.; Wasinger, E. C.; Westre, T. E.; Zaleski, J. M.; Orville, A. M.; Lipscomb, J. D.; Hedman, B.; Hodgson, K. O.; Solomon, E. I. Inorg. Chem. 1999, 38, 3676–3683. (16) Pau, M. Y. M.; Davis, M. I.; Orville, A. M.; Lipscomb, J. D.; Solomon, E. I. J. Am. Chem. Soc. 2007, 129, 1944–1958. (17) Hitomi, Y.; Yoshida, M.; Higuchi, M.; Minami, H.; Tanaka, T.; Funabiki, T. J. Inorg. Biochem. 2005, 99, 755–763. (18) (a) Funabiki, T.; Fukui, A.; Hitomi, Y.; Higuchi, M.; Yamamoto, T.; Tanaka, T.; Tani, F.; Naruta, Y. J. Inorg. Biochem. 2002, 91, 151–158. (b) Simaan, A. J.; Boillot, M.-L.; Carrasco, R.; Cano, J.; Girerd, J.-J.; Mattioli, T. A.; Ensling, J.; Spiering, H.; Gutlich, P. Chem.
Eur. J. 2005, 11, 1779–1793. (19) (a) Whittaker, J. W.; Lipscomb, J. D.; Kent, T. A.; M€unck, E. J. Biol. Chem. 1984, 259, 4466–4475. (b) True, A. E.; Orville, A. M.; Pearce, L. L.; Lipscomb, J. D.; Que, L., Jr. Biochemistry 1990, 29, 10847–10854. (c) Que, L., Jr.; Lipscomb, J. D.; Zimmerman, R.; M€unck, E.; Orme-Johnson, N. R.; Orme-Johnson, W. H. Biochim. Biophys. Acta 1976, 452, 320–334. (20) Finley, J.; Malmqvist, P.-Å.; Roos, B. O.; Serrano-Andres, L. Chem. Phys. Lett. 1998, 288, 299–306. (21) Dapprich, S.; Komaromi, I.; Byun, K. S.; Morokuma, K.; Frisch, M. J. J. Mol. Struct.: THEOCHEM 1999, 461
462, 1–21. (22) (a) Lee, C.; Yang, W.; Parr, R. G. Phys. Rev. B 1988, 37, 785–789. (b) Miehlich, B.; Savin, A.; Stoll, H.; Preuss, H. Chem. Phys. Lett. 1989, 157, 200–206. (c) Becke, A. D. J. Chem. Phys. 1993, 98, 5648–5652. (23) Dolg, M.; Wedig, U.; Stoll, H.; Preuss, H. J. Chem. Phys. 1987, 86, 866–872. (24) Pierloot, K.; Dumez, B.; Widmark, P.-O.; Roos, B. O. Theor. Chim. Acta 1995, 90, 87–114. (25) (a) Cances, M. T.; Mennucci, B.; Tomasi, J. J. Chem. Phys. 1997, 107, 3032–3041. (b) Cossi, M.; Barone, V.; Mennucci, B.; Tomasi, J. Chem. Phys. Lett. 1998, 286, 253–260. (c) Mennucci, B.; Tomasi, J. J. Chem. Phys. 1997, 106, 5151–5158. (26) (a) Roos, B. O.; Taylor, P. R. Chem. Phys. 1980, 48, 157–173. (b) Roos, B. O. In Ab Initio Methods in Quantum Chemistry II; Lawley, K. P., Ed.; Wiley: New York, 1987; p 399. (27) Forsberg, N.; Malmqvist, P.-Å. Chem. Phys. Lett. 1997, 274, 196–204. (28) (a) Anderson, K.; Roos, B. O. Chem. Phys. Lett. 1992, 191, 507–514. (b) Malmqvist, P.; Pierloot, K.; Shahi, A. R. M.; Cramer, C. J.; Gagliardi, L. J. Chem. Phys. 2008, 128, 204109. (c) Pierloot, K. Mol. Phys. 2003, 101, 2083–2094.

ARTICLE

(29) Frisch, M. J.; et al. Gaussian 03, revision C.02; Gaussian Inc.: Wallingford, CT, 2004. (30) Case, D. A.; Darden, T. A.; Cheatham, T. E., III; Simmerling, C. L.; Wang, J.; Duke, R. E.; Luo, R.; Merz, K. M.; Wang, B.; Pearlman, D. A.; Crowley, M.; Brozell, S.; Tsui, V.; Gohlke, H.; Mongan, J.; Hornak, V.; Gui, G.; Beroza, P.; Schafmeister, C.; Caldwell, J. W.; Ross, W. S.; Kollman, P. A. AMBER 8; University of California: San Francisco, CA, 2004. (31) Karlstr€om, G.; Lindh, R.; Malmqvist, P.-Å.; Roos, B. O.; Ryde, U.; Veryazov, V.; Widmark, P.-O.; Cossi, M.; Schimmelpfennig, B.; Neogrady, P.; Seijo, L. Comput. Mater. Sci. 2003, 28, 222–239. (32) Bauernschmitt, R.; Ahlrichs, R. Chem. Phys. Lett. 1996, 256, 454–464. (33) Because the spin state of the total wave function is a sextet (SZ = 5/2), the wave function of the iron(II)
SQ state consists of one doubly occupied d orbital and four R-spin singly occupied d orbitals on the Fe center and one R-spin singly occupied orbital on the semiquinonate moiety. It is likely that such a wave function does not provide the g value corresponding to the usual iron(II) species; in other words, it is not easy to experimentally characterize whether the catecholatoiron(III) complex contains the iron(II)
SQ electronic state or not based on the ESR spectrum. A M€ ossbauer spectrum might provide useful information about the oxidation state of the iron center. However, it was reported that several species are formed in the solution, including the dioxygenase and substrate under anaerobic conditions, and suggested that ESR and M€ossbauer spectra were recorded for the mixtures.19 It was not concluded that catecholatoiron(III) contains little iron(II) character under anaerobic conditions in those studies. (34) QFe(III) and QFe(II) were taken to be the same as the Mulliken B B charges of the iron(III)
CAT state and the most stable iron(II)
SQ state in 11SA-CASSCF(9 in 9) calculations. Note that the iron(III)
CAT state mainly consists of the iron(III)
CAT configuration (89%), and the most stable iron(II)
SQ state mainly consists of the iron(II)
SQ configurations (93%) in the protein environment. (35) In Koopmans theorem, the occupied orbital energy and the unoccupied one correspond to the ionization potential and the electron affinity, respectively. However, we considered here the ionization potential and the electron affinity because the CASSCF/CASPT2 wave function does not have an orbital energy picture. The discussion based on the ionization potential and electron affinity is essentially the same as the discussion with orbital energy, at least, or more accurate than that with the orbital energy because Koopmans theorem is not reliable in many transition-metal complexes. (36) The ground state of the catecholatoiron(III) complex mainly consists of the iron(III)
CAT configuration by ∼95%, and that of the ionized complex mainly consists of the iron(III)
SQ configuration by ∼95% when the CASPT2 method is employed. These results indicate that the energy difference between the iron(III)
CAT and iron(III)
SQ complexes corresponds well to the ionization potential of the CAT moiety in the catecholatoiron(III) complex. (37) (a) Walsh, T. A.; Ballou, D. P.; Mayer, R.; Que, L., Jr. J. Biol. Chem. 1983, 258, 14422–14427. (b) Merkel, M.; Pascaly, M.; Krebs, B.; Astner, J.; Foxon, S. P.; Schindler, S. Inorg. Chem. 2005, 44, 7582–7589.

4789

dx.doi.org/10.1021/jp110045f |J. Phys. Chem. B 2011, 115, 4781–4789