Fast proton-coupled electron transfer observed for a high-fidelity structural and functional [2Fe-2S] Rieske model.

Rieske cofactors have a [2Fe-2S] cluster with unique {His2Cys2} ligation and distinct Fe subsites. The histidine ligands are functionally relevant, since they allow for coupling of electron and proton transfer (PCET) during quinol oxidation in respiratory and photosynthetic ET chains. Here we present the highest fidelity synthetic analogue for the Rieske [2Fe-2S] cluster reported so far. This synthetic analogue 5(x-) emulates the heteroleptic {His2Cys2} ligation of the [2Fe-2S] core, and it also serves as a functional model that undergoes fast concerted proton and electron transfer (CPET) upon reaction of the mixed-valent (ferrous/ferric) protonated 5H(2-) with TEMPO. The thermodynamics of the PCET square scheme for 5(x-) have been determined, and three species (diferric 5(2-), protonated diferric 5H(-), and mixed-valent 5(3-)) have been characterized by X-ray diffraction. pKa values for 5H(-) and 5H(2-) differ by about 4 units, and the reduction potential of 5H(-) is shifted anodically by about +230 mV compared to that of 5(2-). While the N-H bond dissociation free energy of 5H(2-) (60.2 ± 0.5 kcal mol(-1)) and the free energy, ΔG°CPET, of its reaction with TEMPO (-6.3 kcal mol(-1)) are similar to values recently reported for a homoleptic {N2/N2}-coordinated [2Fe-2S] cluster, CPET is significantly faster for 5H(2-) with biomimetic {N2/S2} ligation (k = (9.5 ± 1.2) × 10(4) M(-1) s(-1), ΔH(‡) = 8.7 ± 1.0 kJ mol(-1), ΔS(‡) = -120 ± 40 J mol(-1) K(-1), and ΔG(‡) = 43.8 ± 0.3 kJ mol(-1) at 293 K). These parameters, and the comparison with homoleptic analogues, provide important information and new perspectives for the mechanistic understanding of the biological Rieske cofactor.

Introduction

Rieske-type [2Fe–2S] clusters are unique biological electron transfer (ET) cofactors that feature a heteroleptic ligand environment distinct from that of common [2Fe–2S] ferredoxins, with one of the Fe atoms ligated by two cysteine thiolates and the other by two histidine imidazoles.[1] Rieske clusters serve as structural gates in bacterial oxygenase enzymes that catalyze oxidative hydroxylation of aromatic compounds, and they play an important role in the bifurcated Q-cycle of the quinol-oxidizing cytochrome complexes in respiratory and photosynthetic ET chains.[2,3] Histidine ligation is functionally relevant, since it enables coupling of electron and proton transfer upon reaction of the diferric Rieske cluster with hydroquinone substrates, producing the mixed-valent cluster that is protonated at the His ligands.[3,4] Redox potentials of Rieske [2Fe–2S] clusters in bc-type proteins have indeed been found to be pH-dependent and coupled to the protonation state of the Fe-bound histidines.[5] NMR investigations on a 15N-labeled Rieske protein have revealed a change of the histidines’ pKa values from around 12.5 in the reduced mixed-valent cluster to around 7.4/9.1 in the oxidized diferric cluster.[6] However, mechanistic details of the hydroquinone oxidation mediated by Rieske proteins, such as the sequence or synchronism of proton and electron transfer, have remained a topic of debate.[6−9]

For many years, synthetic analogues have contributed significantly to elucidating the properties and electronic structures of biological iron–sulfur cofactors,[10] but models for the Rieske cluster have remained elusive until recently. In 2008 we reported the first (and so far only) synthetic [2Fe–2S] cluster that emulates the heteroleptic {N2/S2} ligation characteristic for the biological site, (NEt4)21 (Figure 1, left).[11] While (NEt4)21 is an excellent structural and spectroscopic Rieske model in both the diferric and mixed-valent states (Mössbauer and EPR, respectively), the reduced mixed-valence species proved quite unstable and the lack of peripheral N atoms at the bis(indole) ligand in (NEt4)21 precluded any functional studies toward proton-coupled electron transfer (PCET).

Figure 1

The first (and so far only) structural Rieske model 1 (left) and functional homoleptic Rieske models 2, 3, and 4 offering the possibility of protonation at the backside of the N-ligand (right).

Several robust diferric [2Fe–2S] complexes, 22– to 42–, with homoleptic bis(benzimidazolate) ligation have been published.[12−15] In two cases these have allowed, just recently and for the first time, structural characterization of synthetic [2Fe–2S] analogues in their mixed-valent state (33–, 43–),[13−16] and even in the super-reduced diferrous state (44–).[16] For mixed-valent 23– and 43– significant valence delocalization was inferred from Mössbauer and EPR spectroscopy (about 20%, class II according to Robin and Day),[13,15] whereas the EPR spectrum of mixed-valent 13– did not reflect delocalization to such an extent. Thus, the unique Rieske-type heteroleptic coordination of 13– seems to promote enhanced valence localization. Homoleptic analogues 22– to 42– furthermore provide peripheral N atoms akin to the histidine ligands in Rieske cofactors, and their reversible protonation and their redox activity toward PCET have thus been investigated during the past two years. The twice-protonated neutral diferric cluster 4H2 could even be isolated and structurally characterized,[15] and for both systems 3 and 4 thermodynamic parameters of the PCET square scheme have been elucidated.[14,15] The protonated mixed-valent species, upon reaction with TEMPO, were found to undergo concerted proton–electron transfer with rather similar rate constants on the order of 103 M–1 s–1 under pseudo-first-order conditions at 20 °C.

An important open question now remains regarding the effect of the heteroleptic {N2/S2} ligation on the PCET reaction. Here we present the first synthetic Rieske model, 52–, that comprises all beneficial features of 12– to 42–, namely, the characteristic {N2/S2} donor set that leads to enhanced valence localization in the one-electron-reduced state and a potential protonation site at the {N2} ligand backbone that is similar to the His-ligated subunit in the natural archetype. 52– thus represents the highest fidelity Rieske model so far and allows for the effect of the electronic structure on the PCET reaction to be evaluated. The thermodynamic square scheme (Figure 2) is fully established, and three of the four species involved are characterized by single-crystal X-ray diffraction.

Figure 2

Square scheme of protonation and reduction reactions involving [2Fe–2S] Rieske model 5. The subscripts denote the {N2}- and {S2]-ligated Fe sites.

Results and Discussion

Diferric Cluster FeNIIIFeSIII

Rieske model 52– was designed with the same bis(benzimdazolate) capping ligand as was used for homoleptic cluster 42–, because this ligand proved advantageous with respect to solubility and crystallization properties. Diferric (NEt4)25 was synthesized via a stepwise ligand exchange reaction starting from the tetrachloro-coordinated [2Fe–2S] cluster (NEt4)2[Cl2FeS2FeCl2], in close analogy to the synthesis of the first structural Rieske model, (NEt4)21.[11] To this end, phenylbis(benzimidazolyl)methane was first deprotonated with KH and then added to a solution of (NEt4)2[Cl2FeS2FeCl2] in MeCN at −30 °C to furnish the {N2} cap. 1,1′-Biphenyl-2,2′-dithiolate, after deprotonation with KH, was subsequently attached as the {S2} capping ligand. The integrity of 52– in solution has been supported by ESI-MS, and no ligand scrambling was observed (see the Supporting Information, Figure S1). Diffusion of diethyl ether into a solution of diferric (NEt4)25 in MeCN led to growth of crystals, but of rather low quality. Better quality crystals suitable for X-ray analysis could be obtained by diffusion of diethyl ether into a solution of (CoCp*2)25 in MeCN at 4 °C; the molecular structure of the diferric cluster anion is shown in Figure 3.

Figure 3

Left: schematic view of diferric cluster 52–. Right: molecular structure of the anion of (CoCp*2)25 in the crystal (thermal displacement ellipsoids set at 30% probability). For clarity all hydrogen atoms have been omitted.

(CoCp*2)25 crystallizes in the monoclinic space group P21/c. Selected metric parameters are listed in Table 1, together with selected data for three different biological Rieske clusters for comparison.[17−19] The Fe···Fe distance in 52– (2.687 Å) is slightly shorter than in diferric 12– (2.703 Å) and the biological systems (2.71–2.72 Å), but overall geometric parameters are in good agreement. Further discussion is provided below.

Table 1

Selected Bond Lengths (Å) and Angles (deg) of Diferric Clusters (CoCp*2)25 and (NEt4)25H, Mixed-Valent (NEt4)35, and Biological Rieske Clusters[17−19],a

	(CoCp*2)25b	(NEt4)5H	(NEt4)35	SOFX[17]	RIE[18]	RFS[19]
d(Fe···Fe)	2.687(1)	2.694(1)	2.682(1)	2.719	2.71	2.72
d(FeN–μ-S)	2.191(1)/2.205(1)	2.189(1)/2.194(1)	2.241(2)/2.248(1)	2.258/2.259	2.23/2.25	2.28/2.31
d(FeS–μ-S)	2.200(2)/2.206(2)	2.200(1)/2.216(1)	2.210(2)/2.212(2)	2.267/2.263	2.24/2.25	2.35/2.34
d(FeN–N)	1.984(4)/1.988(4)	1.985(2)/1.988(2)	2.057(4)/2.074(4)	2.100/2.083	2.13/2.16	2.19/2.23
d(FeS–S)	2.22(2)/2.44(2), 2.37(2)/2.17(2)b	2.297(1)/2.298(1)	2.335(2)/2.345 (2)	2.348/2.332	2.22/2.29	2.24/2.31
∠(N–FeN–N)	91.65(16)	92.61(10)	86.65(17)	92.12	90.78	90.52
∠(S–FeS–S)	102.8(6)/106.8(5)b	104.45(3)	102.23(6)	109.73	105.61	110.19
∠(μS–FeN–μ-S)	104.88(6)	104.90(3)	104.68(6)	106.24	105.62	109.14
∠(μS–FeS–μ-S)	104.57(6)	103.80(3)	106.95(6)	105.81	105.64	105.70

SOFX = Rieske protein II from Sulfolobus acidocaldarius, RIE = soluble domain of Rieske protein from bovine mitochondrial bc1 complex, and RFS = soluble domain of Rieske protein from spinach chloroplast b6f complex.

Disordered {S2} ligand (see the Supporting Information for details); the true FeS–S bond length likely is an average value, 2.30 Å.

Protonated Diferric Cluster H–FeNIIIFeSIII

To investigate the left part of the PCET square scheme (Figure 2), protonation and subsequent deprotonation experiments with diferric (NEt4)25 and varying acids and bases (see the Supporting Information for an overview) were followed by UV–vis spectroscopy. Addition of 1 equiv of 2,6-DMPH(BF4) (2,6-DMP = 2,6-dimethylpyridine) to a solution of (NEt4)25 in MeCN at −20 °C led to evolution of a prominent absorption band at 385 nm (4.7 × 104 M–1 cm–1) (Figure 4, left), in analogy to what has been observed upon protonation of the related homoleptic [2Fe–2S] cluster (NEt4)24.[15] This band has been found to indicate a tautomerism process at the particular bis(benzimidazolate) ligand, where initial protonation at one of the benzimidazolate N atoms induces migration of the bridgehead methine proton to the other peripheral benzimidazolate N atom. As a consequence, the ligand backbone becomes roughly planar and both peripheral N atoms of the benzimidazolate groups finally carry a proton (Figure 5). A similar tautomerism has previously been described for simple bis(imidazolium) compounds.[20] Furthermore, the band at 541 nm (6300 M–1 cm–1) for 52– is shifted to 568 nm (6100 M–1 cm–1) in 5H–, whereas the band at 450 nm (10000 M–1 cm–1) is shifted to lower wavelengths at 433 nm (11000 M–1 cm–1). Clean conversion is indicated by four isosbestic points at 556, 442, 293, and 265 nm. Protonation to give 5H– proved to be reversible, since the original spectrum of 52– was restored upon addition of base (either diazabicycloundecane (DBU) or phosphazene base 1-tert-butyl-2,2,4,4,4-pentakis(dimethylamino)-2λ5,4λ5-catenadiphosphazene (t-BuP2); see the Supporting Information for formulas and abbreviations and Figure S5 in the Supporting Information for the back-titration).

Figure 4

Left: addition of 0.5 (dark blue) and 1.0 (blue) equiv of 2,6-DMPH(BF4) to 52– (black) in MeCN at −20 °C, generating 5H–. Right: further addition of 0.5 (purple) and 1.0 (red) equiv of 2,6-DMPH(BF4) to 5H– (blue), generating 5H2.

Figure 5

Protonation of diferric 52– leading to the reversible formation of 5H– and 5H2. Lower right: molecular structure of the anion of (NEt4)5H·2DMF·Et2O in the crystal (thermal displacement ellipsoids set at 30% probability). For clarity all hydrogen atoms except the N–H atoms, which are hydrogen bonded to the two DMF molecules, have been omitted.

In contrast to 2,6-DMPH(BF4), addition of 1 equiv of 2,2,6,6-TMPH(BF4) (2,2,6,6-TMP = 2,2,6,6-tetramethylpiperidine) to a solution of (NEt4)25 in DMF did not lead to full conversion to 5H–, but 4 equiv is needed. We conclude that the pKa of 2,2,6,6-TMPH(BF4) in DMF (≥19) is in the same range as the pKa of 5H– and thus is too low to achieve full protonation. Interestingly, λmax of the absorption characteristic for the tautomerized ligand is shifted by about 5 nm if the reaction is carried out in DMF instead of MeCN, suggesting the involvement of solvent molecules in H-bonding. This, as well as the tautomerism discussed above, has been confirmed by the X-ray diffraction analysis of singly protonated diferric cluster (NEt4)5H, which could be crystallized via DMF/Et2O diffusion at 4 °C. The solid-state structure clearly shows hydrogen bonds between the protonated benzimidazolate N and DMF solvent molecules included in the crystal lattice (Figure 5, bottom right). After crystallographic characterization of twice-protonated 4H2,[16] this represents the second synthetic [2Fe–2S] cluster that could be isolated in protonated form and the first that also emulates the protonated Rieske cluster with its heteroleptic ligation. Geometric parameters of the cluster core remain almost unchanged upon protonation (see Table 1), which will be discussed below.

Addition of a second equivalent of 2,6-DMPH(BF4) (Figure 4, right) led to disappearance of the band at 385 nm characteristic for 5H– that had emerged during the first protonation event. This observation led us to conclude that binding of a second proton to give 5H2 is possible and restores the original situation at the C atom bridging the two benzimidazolate moieties, disrupting conjugation within the {N2} capping ligand just as in 52– (Figure 5). Clean conversion is indicated by four isosbestic points at 537, 447, 283, and 258 nm. Shifts of the other absorptions are relatively minor. However, the product 5H2 seemed to be unstable in MeCN at −20 °C, and upon addition of DBU minor decomposition was detected by the appearance of a new band at 424 nm concomitant with broadening of all bands, especially in the region at about 540 nm (see Figure S6 in the Supporting Information).

Following the protonation events by 1H NMR in MeCN-d3 at room temperature showed the formation of a new signal at about 14 ppm which has been attributed to the formation of an NH group. Furthermore, the double set of signals for the {N2} capping ligand turned into a single set, clearly revealing the flattening of the bis(benzimidazolate) scaffold with resulting 2-fold symmetry in 5H– (see Figures S3 and S4 in the Supporting Information for NMR spectra). The process is almost reversible upon addition of t-BuP2, though formation of minor amounts of free ligand and an unknown paramagnetic species evidence the limited stability of 5H– in solution at room temperature. Addition of a second equivalent of 4-DMAPH(OTf) (4-DMAP = 4-(dimethylamino)pyridine) did not lead to spectral changes at −30 °C, suggesting that 4-DMAPH+ (pKa = 17.95)[21] is a weaker acid than 5H2, in contrast to 2,6-DMPH+ (pKa = 14.13).[21]

The pKa value of diferric 5H–, relevant for establishing the square scheme in Figure 2, has been determined by protonation of 52– with 1 equiv of 2,6-DMPH(BF4) and back-titration with DBU (pKa = 24.34)[21] in MeCN followed by UV–vis spectroscopy under inert conditions at room temperature. According to mass balance, a pKa of 23.6 ± 0.3 was thus derived (see the Supporting Information for details).

Mixed-Valent Cluster FeNIIFeSIII

Electrochemical properties of 52– were studied by cyclic voltammetry in MeCN/0.1 M NBu4PF6 at various scan rates and at room temperature (Figure 6, left). The cluster undergoes two cathodic redox processes: the first chemically reversible reduction occurs at E1/21 = −1.43 V and the second quasi reversible reduction at E1/22 = −2.19 V (vs Fc/Fc+). From the separation of the two redox waves a conproportionation constant Kc = 7.1 × 1012 can be derived. While Kc is 4 orders of magnitude smaller than the value calculated for 43–,[13] the stability of mixed-valent 53– against disproportion is still relatively high.

Figure 6

Left: cyclic voltammogram of 52– (c = 1.0 mM) in MeCN/0.1 M Bu4NPF6 at rt vs Fc/Fc+. E1 = −1.43 V and E2 = −2.19 V at various scan rates (v = 100, 200, 300, 500, and 1000 mV s–1). Right: electrochemical reduction of 52– in MeCN/0.1 M Bu4NPF6 at rt at a potential of −1.6 V. The course of reduction was followed by UV–vis spectroscopy.

Mixed-valent 53– was thus generated by bulk electrolysis starting from diferric 52– in MeCN/0.1 M Bu4NPF6 at room temperature at an applied potential of −1.6 V. The course of reduction was followed by UV–vis spectroscopy (Figure 6, right), and clean conversion is reflected by an isosbestic point at 353 nm. Reduction led to an overall decrease of intensity in the visible region of the spectrum, the band at 543 nm dropping to 4900 M–1 cm–1. Only a band at 330 nm assigned to a ligand to metal charge transfer (LMCT) deriving from the {S2} ligand increased in intensity (16300 M–1 cm–1). Both bands at 374 and 447 nm, assigned to CT transitions from the {N2} ligand by comparison with homoleptic 52–,[13,15] almost vanished, suggesting that the reduction is localized at the {N2}-ligated iron atom.

In another experiment after 50% of the reduction was completed (to ensure that side products had not been formed yet), a sample was taken and investigated by EPR spectroscopy (Figure 7). The total spin of the mixed-valent species ST = 1/2 caused by antiferromagnetic coupling of FeIII and FeII gives rise to a characteristic rhombic EPR spectrum. Simulation of the spectrum gave g values of 2.017, 1.934, and 1.854, with an average value gav = 1.935.

Figure 7

EPR spectrum of 53– in MeCN/0.1 M Bu4NPF6 measured as frozen glass at 20 K. The red line is a powder simulation with g = 2.017, 1.934, and 1.854 and Gaussian line widths Γ = 8.5, 14, and 26 G.

In accordance with the UV–vis results, the wide g anisotropy of the EPR spectrum indicates that the unpaired electron in 53– is largely localized at the {N2}-coordinated iron atom,[22] in analogy to what has been observed for reduced Rieske cofactors. Table 2 compares g values for a series of natural and synthetic [2Fe–2S] clusters. Biological Rieske clusters usually show an average value gav of 1.90–1.91,[23] whereas higher values are observed for common all-cysteinato-ligated [2Fe–2S] ferredoxins (gav = 1.945–1.975).[23] As Mouesca has pointed out, electronic delocalization in [2Fe–2S] clusters tends to increase the average g value, gav = 1/3∑g, toward the free electron value (g = 2.0023).[24] Therefore, the value gav = 1.935 reflects increased valence localization in 53– compared to the synthetic homoleptic {N2/N2} analogues 23– (gav = 1.940 in DMF/0.25 M n-BuNClO4, 77 K)[12] and 43– (gav = 1.951 in DMF, 6 K),[13] but electron localization is less pronounced than in 13– (gav = 1.918).[11]

Table 2

EPR Data of Model Complexes 53–, 13–, and 43– and Selected Rieske Proteins[25−27],a

	53–	13–	43–	Tt(25)	ISP bc1[26]	Cyt b6f[27]
g1	2.017	2.015	2.015	2.02	2.024	2.03
g2	1.934	1.936	1.937	1.90	1.89	1.90
g3	1.854	1.803	1.900	1.80	1.79	1.76
gav	1.935	1.918	1.951	1.91	1.90	1.90

Tt = Rieske protein of Thermus thermophilus, ISP bc1 = bovine mitochondrial Cyt bc1, and Cyt b6f = cytochrome b6f complex from spinach.

Increased valence localization in the new Rieske model 53–, if compared to previous homoleptic Rieske models 23– and 43– with two bis(benzimidazolate) capping ligands in their mixed-valent states, reflects the reduced symmetry of the complex and the heteroleptic {S2/N2} ligation, which leads to site preference of the unpaired electron. Furthermore, the negatively charged thiolate ligand is a σ- and π-donor, which stabilizes the higher oxidation state (FeIII). Hence, the new model 53– more closely emulates the electronic situation of the biological antetype than previous models 23– and 43–.

The mixed-valent cluster was generated chemically by reduction of (NEt4)25 with CoCp*2 in DMF at −20 °C, giving microcrystalline (CoCp*2)(NEt4)25. Crystalline material of mixed-valent (NEt4)35 could be obtained from DMF after addition of 1 equiv of NEt4Br and subsequent slow diffusion of Et2O at 4 °C into the solution. (NEt4)35 crystallizes in the monoclinic space group P21/n and represents the first exact Rieske model with heteroleptic {N2/S2} ligation that has been characterized by X-ray diffraction in the reduced state. This now allows a unique comparison of the molecular structures of a high-fidelity synthetic analogue for the Rieske cluster in three relevant forms, namely, in the diferric 52–, the diferric protonated 5H–, and the mixed-valent 53– states. Inspection of the central [2Fe–2S] core shows that only minor changes occur upon reduction or protonation (Figure 8), with the Fe···Fe distance showing negligible variations (<0.02 Å). This reflects the low reorganization energies of [2Fe–2S] clusters that make them favorable electron transfer cofactors in biology. Close comparison of the subtle structural changes upon reduction (52– versus 53–) is interesting, however, because it reveals that changes mainly occur around the {N2}-coordinated iron atom: bonds between FeN and the μ-S elongate by ∼0.05 Å upon reduction, while bonds between FeS and the μ-S remain essentially unchanged (<0.01 Å). Interestingly, in the case of homoleptic {N2}-capped 32–/33– and 42–/43–, the bonds between both Fe atoms and the μ-S lengthen by ∼0.03 Å upon reduction, showing that the unpaired electron is delocalized in 43– (on the crystallographic time scale), but is largely localized at the single {N2}-coordinated iron site in 53–. In line with these considerations reduction of 52– leads to a more pronounced lengthening of the bonds between FeN and the {N2} capping ligand (0.08 Å) than for the bonds between FeS and the {S2} capping ligand (0.04 Å). Fe–N bonds in the homoleptic {N2}-capped clusters show averaged elongations of 0.07 Å (32–/33–) or 0.05 Å (42–/43–). The most prominent structural difference was found for the N–FeN–N angle, which shrinks by around 5° in mixed-valent 53– compared to diferric 52– and 5H–. All these crystallographic findings, though subtle, corroborate that reduction of the Rieske model occurs at the FeN site in accordance with EPR and Mössbauer spectroscopy (see below). Selected geometric parameters are compiled in Table 1.

Figure 8

Overlay of the molecular structures of diferric 52– (red), mixed-valent 53– (blue), and protonated diferric 5H– (yellow).

The zero-field Mössbauer spectrum of diferric (NEt4)25 shows two quadrupole doublets at a ratio of 1:1 with isomeric shifts δ1 = 0.26 mm s–1 and δ2 = 0.28 mm s–1, as expected for two distinct ferric sites (Figure 9, left). Differences in quadrupole splitting allow an assignment to the all-sulfur-coordinated FeS (ΔEQ1 = 0.52 mm s–1) and the {N2}-capped FeN (ΔEQ2 = 1.16 mm s–1); the larger quadrupole splitting in the case of FeN reflects the increased electric field gradient resulting from the higher asymmetry of electronic charge distribution. Overall the Mössbauer data show good agreement with parameters found for (NEt4)21 and biological Rieske clusters (see Table S2 in the Supporting Information).

Figure 9

Zero-field Mössbauer spectra of solid (NEt4)25 (left) and (CoCp*)(NEt4)25 (right) at 80 K. Simulation of the data gave the following parameters: (left) δ1 = 0.26 mm s–1 and ΔEQ1 = 0.52 mm s–1 (red), δ2 = 0.28 mm s–1 and ΔEQ2 = 1.16 mm s–1 (blue); (right) δ1 = 0.35 mm s–1 and ΔEQ1 = 1.26 mm s–1 (red), δ2 = 0.69 mm s–1 and ΔEQ2 = 3.23 mm s–1 (blue).

Variable-temperature zero-field Mössbauer spectra of mixed-valent (CoCp*2)(NEt4)25 show two doublets at a ratio of about 1:1 in the range from 6 to 200 K, as expected for a mixed-valent [2Fe–2S] cluster with heteroleptic terminal coordination (spectrum at 80 K shown in Figure 9 (right), spectra at 6 and 200 K shown in the Supporting Information; see also Table S3 in the Supporting Information). The doublets can be assigned to the {N2}-coordinated FeN atom with an isomeric shift typical for FeII and a large quadrupole splitting (δ2 = 0.69 mm s–1, ΔEQ2 = 3.23 mm s–1, 80 K) and to the {S2}-coordinated FeIII atom featuring a smaller isomeric shift and smaller splitting (δ1 = 0.35 mm s–1, ΔEQ1 = 1.26 mm s–1, 80 K). These values are in good agreement with data for biological Rieske cofactors, though quadrupole splittings ΔEQ are somewhat smaller for the latter (Table 3). In contrast to mixed-valent (NEt4)32 and (NEt4)34, in which cases the two quadrupole doublets collapsed to a single quadrupole doublet at 200 K, electron hopping on the Mössbauer time scale cannot be observed for (CoCp*2)(NEt4)25.

Table 3

Mössbauer Parameters (mm s–1) of (CoCp*2)(NEt4)25 and Biological Rieske Clusters in the Reduced Statea

	(CoCp*2)(NEt4)25, 6 K	Tt,[25] 4.2 K	ISP,[29] 4.2 K	T4MOC,[30] 4.2 K
δ1	0.34	0.31	0.25	0.30
δ2	0.70	0.74	0.73	0.72
ΔEQ1	1.29	0.63	0.70	0.71
ΔEQ2	3.24	3.05	2.95	3.07
Δδ	0.36	0.43	0.48	0.42

Tt = Rieske protein of Thermus thermophilus, ISP = Rieske protein of Cyt bf complex from spinach, and T4MOC = Rieske protein from Pseudomonas mendocina in Escherichia coli.

An empirical correlation for δ and the oxidation number (x) of FeS4 units δ(x) = (1.43 – 0.40x) mm s–1 has been reported.[28] This would predict values of δ(III) = 0.23 mm s–1 and δ(II) = 0.63 mm s–1 and hence a difference of 0.4 mm s–1 for fully localized ferric and ferrous sites a and b. While for homoleptic mixed-valent 43– only half as much (Δδ = 0.22 mm s–1 at 4 K)[13,15] was observed, the present heteroleptic Rieske model 53– gives Δδ = 0.36 mm s–1 at 6 K (0.34 mm s–1 at 80 K), close to the expected value. Though this is still less than Δδ in the range 0.42–0.48 mm s–1 observed for biological Rieske sites featuring full valence localization (Table 3), it reflects the increased valence localization in 53– compared to previous homoleptic models, in accordance with EPR data.

Protonation of Mixed-Valent Cluster FeNIIFeSIII

Protonation of the mixed-valent cluster proved to be challenging because of low stability of the resulting species. Instantaneous degradation could be observed upon protonation at room temperature. At −30 °C in MeCN the addition of 1 equiv of 2,6-DMPH(BF4) could be followed by UV–vis spectroscopy (see Figure S7 in the Supporting Information), and only minor absorption changes were detected: The band at 328 nm (2.4 × 104 M–1 cm–1) drops by about 3000 m–1 cm–1, whereas absorption in the region between 350 and 650 nm rises by about 2000 M–1 cm–1. The maximum at 550 nm (4800 M–1 cm–1) broadens and experiences a slight red shift (5500 M–1 cm–1). If handled at −30 °C throughout, addition of t-BuP2 did lead to nearly full conversion back to the original spectrum of 53–. Since no intense band around 385 nm evolved upon protonation of 53–, it can be assumed that mixed-valent 5H2– does not undergo any tautomerism that was observed for diferric 5H– .

EPR spectra of the singly protonated species 5H2– showed a rhombic spectrum with g values of 1.994, 1.938, and 1.875 (gav = 1.937), but with rather large line widths (60, 33, and 55 G) even when spectra were recorded at low temperatures (10 K; see the Supporting Information). The reasons for this are unclear; a more detailed investigation and comparative analysis of the EPR spectra of 43– and 53– and their protonated forms is currently in progress.

The effect of protonation on the redox potentials was investigated by cyclic voltammetry, focusing on the first reduction wave that appears at E1/2 = −1.43 V vs Fc/Fc+ for 52– (Figure 10). Addition of 1 equiv of 4-DMAPH(OTf) led to the emergence of a new cathodic peak at Epc = −1.27 V, which is anodically shifted by +230 mV compared to the cathodic peak potential of the reversible couple for 52– (Epc = −1.50 V). This leads to an estimated redox potential of −1.20 V for protonated 5H– (assuming a similar shift Epc – E1/2 for 52– and 5H–). However, protonation obviously is not complete upon addition of 1 equiv of 4-DMAPH(OTf), since the original peak at Epc = −1.50 V is still discernible. Addition of a second equivalent of 4-DMAPH(OTf) caused the peak at Epc = −1.50 V to disappear, concomitant with a broadening of the peak at Epc = −1.27 V assigned to 5H– and emergence of a third redox event characterized by a peak at Epc = −1.02 V, shifted anodically by +480 mV compared to 52– (Epc = −1.50 V). The additional peak presumably reflects the presence of some twice-protonated species 5H2. When base was then added, the original redox wave characteristic of 52– (Epc = −1.50 V, E1/2 = −1.43 V) reappeared, confirming chemical reversibility of the protonation events (Figure 10).

Figure 10

Cyclic voltammogram of (NEt4)25 (c = 1.0 mM) in MeCN/0.1 M Bu4NPF6 at rt vs Fc/Fc+ at a scan rate of 500 mV/s (top). The redox potential is shifted upon addition of acid (second and third pictures from top). Subsequent addition of t-BuP2 proves the reversibility of the process (bottom).

The anodic shift of about +480 mV upon 2-fold protonation of 52– is more than twice as large as the shift observed for homoleptic clusters 32– and 42– after binding of two protons (+240 and +200 mV, respectively).[14,15] The difference reflects that the two protons are bound to the same bis(benzimidazole) ligand at the unique FeN site in 5H2, in close analogy to the situation in Rieske proteins, while the two protons are bound to two bis(benzimidazole) ligands at different Fe sites in homoleptic 3H2 and 4H2. Indeed, a shift of +300 to +440 mV upon going from very low to very high pH (hence going from the twice-protonated to the fully deprotonated form) has been reported for biological Rieske clusters depending on the type of protein;[31] redox potentials for the intermediate singly protonated forms have not yet been reported. Hence, the electrochemical response to protonation for the synthetic analogue 52– nicely emulates the properties of the biological antetype.

To establish a thermodynamic square scheme for the new Rieske model, not only redox potentials but also the pKa value of mixed-valent 5H2– was determined (Figure 11). To this end 53– was protonated with 1 equiv of 4-DMAPH(BF4) in MeCN-d3 at low temperatures, and back-titration with phosphazene base (tert-butylimino)tris(1-pyrrolidinyl)phosphorane (t-BuP1(pyrr)) (pKa = 28.42) was followed by 1H NMR at room temperature (see the Supporting Information for details, Figures S11 and S12). A pKa of 27.9 ± 0.2 was thus determined for 5H2– in MeCN. In comparison to the homoleptic model complex 33– (pKa = 24.7 ± 0.4), the present heteroleptic Rieske-type cluster 53– is more basic, probably due to the increased valence localization and the more pronounced ferrous character at the protonation site. pKa values of reduced Rieske proteins (12.3–13.3; see above) are much lower than those of 53– and 33–. The comparability is limited, however, since those values for Rieske proteins have been determined in aqueous solution, and they are tuned by interactions with the surrounding protein environment comprising, for instance, several hydrogen bonds.[32]

Figure 11

Square scheme summarizing thermodynamic parameters for the second-generation Rieske model cluster in MeCN with potentials referenced against Fc/Fc+.

The bond dissociation free energy (BDFE) of the N–H bond of the protonated cluster 5H2– can be calculated from the available pKa and E1/2 data (see the Supporting Information),[33] giving BDFE = 60.2 ± 0.5 kcal mol–1 (252 ± 2 kJ mol–1). This BDFE for the NH bond is close to the value found for the related homoleptic model complex 3H2– (60.5 kcal mol–1),[14] but about 10 kcal mol–1 less than that of Rieske protein RsRp under basic conditions (71.5 kcal mol–1) and even 15 kcal mol–1 less than the value obtained for the protonated form of RsRp under acidic conditions (75.1 kcal mol–1).[5e,33] The lower BDFE, hence the weaker N–H bond in the model complexes, is likely related to the differently charged ligands in the mixed-valent state, namely, neutral histidines versus monoanionic bis(benzimidazolate). The very similar BDFE values for 3H2– and 5H2– evidence that heteroleptic ligation of the [2Fe–2S] core does not play any role in this respect.

With those values at hand, a pKa of about 23.8–24.2 could be calculated for diferric 5H– according to Hess’s law (see the Supporting Information). This is in good agreement with the value determined experimentally by UV–vis titration of 5H– with DBU, 23.6 ± 0.3 as described above. The difference in pKa for mixed-valent 5H2– and diferric 5H– of about 4 units is in accord with the change of the histidines’ pKa values from around 12.5 in the reduced mixed-valence forms to around 7.4/9.1 in the oxidized diferric forms of Rieske proteins.[5]

To examine its PCET reactivity, mixed-valent 5H2– (generated in situ by addition of 1 equiv of 4-DMAPH(BF4) to 53–) was treated with the nitroxyl radical TEMPO. Full conversion to deprotonated diferric 52– and 1 equiv of TEMPO-H was ascertained by 1H NMR spectroscopy. Because of the moderate N–H BDFE of 5H2–, calculation of the free energy of the reaction, ΔG°CPET (CPET = concerted proton and electron transfer), gives a sizable value, −26.4 kJ mol–1 (see the Supporting Information). To obtain mechanistic insight, double-mixing stopped-flow measurements were undertaken at varying temperatures under pseudo-first-order conditions using an excess of TEMPO, yielding kinetic parameters of the reaction (see the Supporting Information). At 20 °C, a second-order rate constant k = (9.5 ± 1.2) × 104 M–1 s–1 was determined, and the transition-state parameters ΔH‡ = 8.7 ± 1.0 kJ mol–1 and ΔS‡ = −120 ± 40 J mol–1 K–1 were derived from an Eyring plot (Figure S15, Supporting Information). For the free energy of the transition state, ΔG‡, at 293 K a value of 43.8 ± 0.3 kJ mol–1 was thus calculated (see the Supporting Information).

To verify that the reaction follows a concerted and not a stepwise pathway, the initial steps of the alternative pathways were examined. These are, starting from 5H2– and TEMPO, either proton transfer or electron transfer, leading to 53– (and TEMPO-H•+) or 5H– (and TEMPO–), respectively. To ascertain the favored pathway, the respective activation energies have been compared. Since the activation energies ΔG‡ must be at least as high as the free energies ΔG°CPET, these values are a conservative lower limit to ΔG‡ (for calculation of the free energies ΔG°PT and ΔG°ET, see the Supporting Information). Since ΔG°PT = 184 kJ mol–1 and ΔG°ET = 72.4 kJ mol–1 are larger than the activation energy for the concerted pathway (ΔG‡ = 43.8 kJ mol–1), the stepwise pathways can be excluded.

ΔG‡ for 5H2–is about 10 kJ mol–1 smaller compared to the value for cluster 3H2–;[14] therefore, the rate constant at room temperature is more than 1 order of magnitude higher for heteroleptic 5H2– than for homoleptic 3H2–. ΔH‡ is of the same order of magnitude (about 2 kJ mol–1 larger), while ΔS‡ for 5H2– is less negative than for 3H2–.[14] The transition-state parameter ΔH‡ obtained for homoleptic 4H2– is at least twice as large as the value found for 5H2– (Table 4),[15] because the second-order rate constant is about 2 orders of magnitude lower than the one derived for 5H2–.

Table 4

Thermodynamic and Kinetic Parameters for the Reactions of TEMPO with 3H2–, 4H2–, and 5H2–

	4H2–[15] (in DMF)	3H2–[14] (in MeCN)	5H2– (in MeCN)
ΔH⧧ (kJ mol–1)	17.6 ± 0.6	6.7 ± 1.3	8.7 ± 1.0
ΔS⧧ (J mol–1 K–1)	–130 ± 2	–159 ± 10	–120 ± 5
k at 20 °C (M–1 s–1)	722 ± 14	2200 ± 350a	95000 ± 12000
ΔG⧧ at 20 °C (kJ mol–1)	55.7 ± 1.1	54.0 ± 1.2	43.8 ± 0.3
BDFE(Fe–H) (kJ mol–1)		253 ± 4	252 ± 2
ΔG°CPET (kJ mol–1)		–25.1	–26.4

At 25 °C.

The differences in activation parameters and in rate constant k for the closely related complexes 3H2– and 4H2–, both with homoleptic bis(benzimidazolate) ligation, can likely be attributed to the different solvents used in those studies.[14,15] Polar solvents such as DMF can interact with both reaction partners, presumably decelerating the reaction; H-bonding interaction between DMF and the N–H units of 4H2 had indeed been detected by X-ray crystallography and by IR spectroscopy,[15] similar to what is seen here for the structure of 5H– in the solid state (vide supra). The significantly different rate constants k for 3H2– and 5H2–, both measured in MeCN solution, may appear counterintuitive at first sight, since ΔG°CPET = −6.0 kcal mol–1 (−25.1 kJ mol–1) found for 3H2– is quite similar to the value determined for 5H2–. The much lower activation energy ΔG‡ that leads to accelerated PCET in the case of 5H2–, however, might be an effect of increased localization of electron density in the Rieske-type [2Fe–2S] core with its heteroleptic {N2/S2} ligation. It should be noted that Fe–N bonds, upon reduction, elongate almost equally in 52– and in homoleptic 32–, while the FeS–S bonds in 52– change much less. This might give rise to a smaller reorganization energy, λ, during PCET, which in turn leads to faster PCET between the cluster and TEMPO. It has been shown before that Marcus’s theory, which had initially been established for ET reactions, can also be applied to PCET reactions.[35,36] Thus, a small reorganization energy λ should be advantageous not only for fast electron transfer, but for fast PCET as well. However, at present it cannot be excluded that steric effects, viz., a different TEMPO accessibility of the backbone N–H groups caused by different substituents at the nearby bridging C atom (Ph/H in 5H2– versus n-Pr/n-Pr in 3H2–), may also play a role; see Figure S19 in the Supporting Information for illustrative space-filling models generated from the crystal structures of mixed-valent 33– and 53–.

Conclusions

In summary, we report a synthetic analogue for the Rieske cofactor that not only emulates the heteroleptic {His2Cys2} ligation of the biological antetype, but also represents a functional model undergoing fast concerted electron and proton transfer. All four species in the PCET square scheme have been thoroughly characterized, and three of them, namely, diferric 52–, protonated diferric 5H2–, and mixed-valent 53–, could be studied by single-crystal X-ray diffraction. This provides unprecedented structural information and reveals that the [2Fe–2S] core undergoes only minor structural changes upon protonation or reduction, which is in line with low reorganization energies upon PCET. However, subtle variations of the [2Fe–2S] core in the diferric 52– and mixed-valent 53– states reflect that the additional electron is largely localized at the N-coordinated Fe site, in accordance with EPR and Mössbauer evidence. It is somewhat surprising and counterintuitive though that the more pronounced valence localization caused by the heteroleptic {N2/S2} ligation does not lead to larger core structural changes compared to homoleptically coordinated [2Fe–2S] complexes.

The thermodynamics of the PCET square scheme have been fully elucidated. Both the difference in pKa for diferric 5H– and mixed-valent 5H2– of about 4 units and the anodic shift of the reduction potential of around +230 mV upon protonation are in very good agreement with data for the biological system. The BDFE of the N–H bond of the protonated cluster 5H2– (60.2 ± 0.5 kcal mol–1) is around 10–15 kcal mol–1 lower than values reported for biological Rieske clusters, but similar to the N–H BDFE for a recently reported homoleptic [2Fe–2S] cluster. However, despite these similar BDFEs (and hence similar free energies ΔG°CPET), the reaction of the protonated mixed-valent cluster with TEMPO, yielding TEMPO-H and the nonprotonated diferric cluster, is significantly faster for the present system 5H2– compared to the homoleptic complexes. While steric factors cannot be excluded at this point, it is an interesting perspective that this might be an effect of increased localization of electron density at the PCET site in 5H2–, which would suggest some further rationale for nature’s choice of the Rieske-type [2Fe–2S] core with its heteroleptic {N2/S2} ligation.

The present new model system is only the second synthetic analogue emulating the heteroleptic ligation of the Rieske cofactor,[11] and the first that features a biomimetic {His2Cys2}-like ligation amenable to PCET at the N-coordinated subsite of the [2Fe–2S] cluster. It thus represents an excellent structural, spectroscopic, and functional analogue, and it is the highest fidelity Rieske model known so far.