Spontaneous Formation of an Fe/Mn Diamond Core: Models for the Fe/Mn Sites in Class 1c Ribonucleotide Reductases

Abstract

Enzymes like the Class 1c ribonucleotide reductases found in human pathogenic bacteria have recently been found to contain Fe/Mn active sites that are closely related to their better characterized diiron cousins. These enzymes are proposed to form high-valent intermediates with Fe–O–Mn cores. Herein we report the first examples of synthetic Fe/Mn complexes that mimic the doubly-bridged intermediates proposed for enzymatic oxygen activation. Fe K-edge EXAFS analysis has been used to characterize the structures of each of these compounds. The linear compounds accurately model the Fe•••Mn distances found in the Fe/Mn proteins in their resting states, and the doubly bridged diamond core compounds accurately model the distances found in high-valent biological intermediates. Unlike their diiron analogues, the paramagnetic nature of the Fe/Mn compounds can be analyzed by EPR, revealing S = ½ signals that reflect antiferromagnetic coupling between the high-spin Fe(III) and Mn(III) units of the heterobimetallic centers. These compounds undergo electron-transfer with various ferrocenes, the linear compounds being capable of oxidizing diacetyl ferrocene, a weak reductant and the diamond core compounds being capable of oxidizing acetyl ferrocene. The diamond core compounds can also perform HAT reactions from substrates with X-H bonds with bond dissociation free energies (BDFEs) up to 75 kcal/mol and are capable of oxidizing TEMPO-H at rates of 0.32 – 0.37 M⁻¹s⁻¹, which are comparable to those reported for some mononuclear Fe^III-OH and Mn^III-OH compounds. However, such reactivity is not observed for the corresponding diiron compounds, a difference that Nature may have taken advantage of in evolving enzymes with heterobimetallic active sites.

Graphical Abstract

Introduction

In Nature, oxygen activation, reduction, and formation are performed at a variety of metallo-cofactors and active sites. These include mono- and diiron sites,^1,2 mono- and dicopper sites,³ the CaO₅Mn₄ oxygen evolving complex,⁴ and the iron-copper active site of heme-copper oxidases.⁵ In the past 15 years, two more proteins with heterobimetallic sites that activate oxygen have been discovered.^6,7 Ribonucleotide reductase 1c (RNR1c) is found in the human pathogen Chlamydia trachomatis, and is responsible for reducing RNA to DNA.^6,8,9 R2-like ligand binding oxidase (R2lox) is found in the human pathogen Mycobacterium tuberculosis and while its function is currently unknown, it has been correlated with virulence.⁷

These two enzymes are proposed to follow analogous mechanisms of oxygen activation. O₂ binds to a Fe^II/Mn^II site to form an Fe^III/Mn^III-peroxy intermediate.¹⁰ The O–O bond is then cleaved to form high-valent Fe^IV/Mn^IV intermediates with bis-μ-O^2- “diamond core” moieties.^10–12 In RNR_1c this high-valent intermediate is then reduced by one electron and protonated to form a Fe^III(μ-O)(μ-OH)Mn^IV intermediate that is responsible for initiating the radical translocation that forms the catalytically active thiyl radical.^9,13,14 In contrast, the proposed high-valent intermediate in R2lox performs a two-electron oxidation on the protein scaffold and forms an ether crosslink between valine and tyrosine residues.^10,15

In other biological systems that utilize high valent metals, synthetic mimics have been helpful for understanding the structure and reactivity of short-lived biological intermediates. While much work has been done synthetically for mono- and diiron systems,^2,16 the OEC,¹⁷ and mono- and dicopper systems,¹⁸ currently there has been little work done to model RNR_1c and R2lox.

Early synthetic work on Fe/Mn compounds was performed by Wieghardt et al. who were interested in the physical and electronic structures of heterobimetallic complexes. These complexes were formed by self-assembly following the hydrolysis of a 1:1 mixture of FeCl₃(TACN) (TACN = 1,4,7-triazacyclononane) and MnCl₃(Me₃TACN) (Me₃TACN = 1,4,7-trimethyl-1,4,7-triazacyclonane) in the presence of sodium acetate, resulting in the formation of [(TACN)Fe(μ-O)(μ-OAc)₂Mn(Me₃TACN)]^2+.19

Early synthetic efforts of Que and coworkers used the symmetric dinucleating ligand BPMP (BPMP = 2,6-bis[(bis(2-pyridylmethyl)-amino)methyl]-4-methylphenolate) to form Fe^III/Mn^II complexes in an effort to model purple acid phosphatase (PAP) active sites.²⁰ These efforts were followed up more recently by Latour and Nordlander, who used unsymmetric dinucleating ligands to incorporate the open coordination site on the Mn ion found in PAP.^21,22

While previously studied systems have effectively modeled the PAP active site, none had an open coordination site on each metal that would allow the structures proposed in RNR_1c and R₂lox to be effectively modeled. To this end, we have demonstrated that the reaction of a high valent Fe^IV(O) compound with Mn^II complexes forms the oxo-bridged Fe^III-O-Mn^III products.²³ One of these complexes, [(TPA)Fe^III(μ-O)Mn^III(TPA)]⁴⁺, was shown to be able to bind exogenous ligands such as acetate. We also now show that three linear [(L)Fe^III(μ-O) Mn^III(L’)] (L = tris(pyridin-2-ylmethyl)amine, TPA, tris((5-methylpyridin-2-yl)methyl)amine, 5Me₃TPA, or 1-(4-methoxy-3,5-dimethylpyri-din-2-yl)-N-((4-methoxy-3,5-dimethylpyridin-2-yl)methyl)-N-((5-methoxy-4,6-dimethylpyridin-2-yl)methyl)methanamine, TPA*; L’ = TPA) complexes can be converted to their respective conjugate bases [(L)Fe^III(μ-O)(μ- OH)Mn^III(L)]³⁺ by the addition of one equivalent of base, representing the first synthesis of such compounds. Depending on the supporting ligand on the ferric ion, the strength of base required to fully convert the linear compound to the doubly-bridged complex changes, which allows for the pK_a of each compound to be determined. In sharp contrast to the corresponding diiron complexes, the Fe/Mn compounds spontaneously form Fe(μ-O)(μ- OH)Mn diamond cores, a unique result that highlights the difference between homo- and heterobimetallic complexes. These Fe/Mn compounds also exhibit electron transfer and hydrogen atom transfer reactivity that is unobserved in diferric complexes.

Results and Discussion

As previously reported, the formation of (L)Fe^III-O-Mn^III(L) complexes can be achieved by the reaction of the appropriate ferryl species and a Mn^II precursor.²³ In the interest of expanding this library of compounds and investigating ligand effects on their properties, two compounds related to the parent [(TPA)Fe^III-O-Mn^III(TPA)]⁴⁺ (1) complex have been synthesized by the same method, namely [(5Me₃TPA)Fe^III–O–Mn^III(TPA)]⁴⁺ (2) and [(TPA*)Fe^III–O–Mn^III(TPA)]⁴⁺ (3). The UV-Vis spectra of these compounds are similar, each with a sharp feature at ~500 nm, and a broader feature centered at ~550 nm (Figure S1). The shape of this second feature seems to vary between the compounds, but the basis for these relatively minor differences is not currently established.

Upon titration of 1, 2, or 3 with triethylamine, a strong base in acetonitrile (pK_a in MeCN = 18.82), a new species is observed to form with λ_max = 455 nm. This species maximally forms upon the addition of 1 equivalent of TEA and is designated 1-OH, 2-OH, or 3-OH, respectively, the conjugate bases of 1, 2, or 3 (Figure 1, top). Further addition of base causes these conjugate base complexes to decay. On the other hand, the addition of one equivalent of acid to 1-OH, 2-OH, or 3-OH fully converts these species back to 1, 2, or 3, respectively.

Figure 1.

Top – Titration of 2 (blue) with TEA to form 2-OH (red) and the decay of 2-OH upon the addition of excess TEA (dotted black) Middle – Addition of HClO4 to 2-OH (red) to regenerate 2 (blue). Bottom – Spectra of 1 mM 2 (blue) in a 0.5 cm pathlength cuvette and its conversion to 2-OH upon base addition (red). Note the difference in absorbance circa 350 nm.

Similar to the well established diiron precedents,^24–26 resonance Raman studies of the six Fe–O–Mn complexes in this study show Fe–O–Mn vibrations that are sensitive to the Fe–O–Mn angle.^2,23,24 The resonance Raman spectra of 1 – 3 obtained at 77 K or 233 K with 405-nm excitation (Figure 2 and S6; Table 1) show resonance-enhanced features in the range of 853–862 cm^-1. Of the three complexes, only the asymmetric Fe–O–Mn vibration for 1 appears as a Fermi doublet. Issues encountered with sample fluorescence in the case of 2 required experiments to be performed in liquid solution at 233 K.

Figure 2.

Resonance Raman data obtained at 77 K for 1 mM 1 in MeCN with 405-nm excitation and for 1 mM 1-OH in MeCN with 457-nm excitation. # denotes solvent peaks

Table 1.

UV-Vis and vibrational features of Fe–O–Mn and related diferric complexes

Compound	λmax (nm) (ε, M−1cm−1)	ν (cm−1)a	λexc (nm)	Ref.
Synthetic Fe/Mn Compounds
1	350 (4600) 500 (500)	853 79523	407 457	c
2	350 (4600) 500 (500)	853 862b	407	c
3	350 (4600) 500 (500)	856	407	c
1-OH	455 (1000)	651 611	457	c
2-OH	455 (1000)	651 611	457	c
3-OH	455 (1000)	651 611	457	c
Fe III 2 (μ-O)(μ-O 2 H 3 ) Complexes
TPA	322, 360 480, 610	-	-	27
5Et 3 TPA	360, 485 605	462	-	27
BPEEN	320, 365 480, 650	448	-	27
Fe III 2 (μ-O)(μ-OH) Complexes
TPA*	370, 550	-	-	28
6Me 3 TPA	340, 396, 550	675	514.5	27
BQPA	308, 396, 554	668	514.5	27
BPEEN	378, 430 475, 510, 558	-	-	27
BPMEN	377, 430 479, 512, 555	-	-	27

^aData obtained in frozen MeCN solution at 77 K unless otherwise noted.

^bData obtained in liquid MeCN solution at 233 K.

^cThis work.

The strong enhancement of 1 with 405-nm excitation suggests that the oxo-to-M(III) ligand-to-metal charge transfer (LMCT) transfer is at a lower wavelength than where the λ_max for 1 was initially assigned.²³ Indeed, taking UV-Vis spectra of 1, 2 and 3 at low concentrations in 0.5-cm pathlength cuvettes reveals a strong absorption near 350 nm (Figure 1). This band decreases intensity upon conversion to 1-OH, 2-OH, or 3-OH (Figures 1 and S3).

Essentially identical resonance Raman spectra with features at 651 and 611 cm^-1.are obtained for 1-OH, 2-OH, and 3-OH with 457-nm excitation (Figures 2 and S6). Based on Raman data for (μ-oxo)diiron complexes,²⁹ vibrations in this region are indicative of rather acute Fe–O–Fe angles approaching 90°, which are characteristic of compounds with M₂(μ-O)₂ ^30–35 or M₂(μ-O)(μ-OH)^27–29 cores. The stoichiometry observed for the conversion between 1, 2, and 3 to their respective conjugate bases shows that forming any potential “diamond core” species is a one-proton equilibrium. For this reason, we assign 1-OH, 2-OH and 3-OH as having (μ-O)(μ-OH) diamond cores.

The excitation profiles of all six compounds allow us to assign the oxo-to-M³⁺ charge transfer band more conclusively for the linear and doubly bridged compounds (Figures 3 and S5). Not surprisingly, the vibrations for 1, 2 and 3 are most enhanced with 405-nm excitation and likely associated with the 350-nm absorption band, a spectral feature typically assigned to oxo-to-M³⁺ charge transfer bands in Fe^III–O–Fe^III complexes.^26,34 On the other hand, 1-OH, 2-OH and 3-OH show the strongest enhancement of the 651 and 611 cm⁻¹ vibrations with 457-nm excitation, suggesting that the 455-nm feature may be assigned to an oxo-to- M³⁺ charge transfer band for the diamond core complexes.

Figure 3.

Excitation profile of 2-OH at 4 different laser wavelengths. The signal is most enhanced with excitation at 457 nm, very close to the charge transfer band observed in its visible spectrum. The red box highlights the Fe–O–Mn vibrations of interest.

To verify that these vibrations derived from an oxo moiety, H₂¹⁸O was added to a solution of 2. The UV-Vis spectrum of 2 in the presence of excess water resembled that of 2-OH (Figure S4), and the resonance Raman spectrum of this species indeed revealed the vibrations at 651 and 611 cm⁻¹ to be sensitive to ¹⁸O substitution, showing a downshift of ~30 cm⁻¹ (Figure S8) Furthermore the vibration at 853 cm⁻¹ from 2 had disappeared. These results indicate that the 853 cm⁻¹ vibration belongs to one species, most likely 2, while the vibrations at 651 and 611 cm⁻¹ belong to a different species that is formed from 2 in the presence of excess water (2-OH). Given the labeling results, we assign the ~850 cm⁻¹ vibrations to the Fe–O–Mn asymmetric mode of a nearly linear Fe–O–M moiety, consistent with results from diiron and other heterobimetallic complexes.^23,24,27 The two vibrations of comparable frequency associated with diamond core moieties can be respectively assigned as the asymmetric and symmetric Fe–O–Mn vibrations.²⁹ The more intense feature has been assigned as the asymmetric vibration, following the pattern found in the chemistry of (μ-oxo) diiron complexes, which provides a precedent for the metal-site inequivalence leading to a greater enhancement of the ν_asym mode at 651 cm⁻¹ than the ν_symm mode at 611 cm^-1.²⁶

These results also show that the presence of excess water converts the Fe/Mn open-core complex 2 spontaneously to the closed-core complex 2-OH. Resonance Raman spectra of 1 and 2 formed in wet acetonitrile (Figure S7) show a mixture of the open and closed core compounds, lending further support to the notion that Fe/Mn diamond cores spontaneously form from the open core complexes in the presence of water. Thus water is a strong enough base to deprotonate the bound water and generate the oxo-hydroxo diamond core in 2-OH.

Fe K-edge EXAFS analysis corroborates the conclusions derived from the resonance Raman data. The EXAFS results for 1, 2, and 3 clearly show that each complex has a Fe•••Mn distance of 3.63–3.65 Å, corresponding to a nearly linear Fe–O–Mn unit. The Mn scatterer is clearly visible in the EXAFS spectra (Figures 4 and S9) as an intense peak near R + Δ = 3.5 Å. All three compounds have primary coordination spheres typical of oxo-bridged high-spin diferric compounds, with an O scatterer at 1.81–1.84 Å and 5 N/O scatterers at 2.12–2.14 Å.^27,28,34,36

Figure 4.

Top – k-space (inset) and Fourier Transformed EXAFS data for 1 (black dots) and best fit (solid red) Bottom-k-space (inset) and Fourier Transformed EXAFS data for 1-OH (black dots) and best fit (solid red).

In contrast, Fe K-edge EXAFS analysis of 1-OH, 2-OH and 3-OH shows a dramatic contraction of the Fe•••Mn distances to ~2.8 Å, indicative of a significant bending of the Fe–O–Mn unit into a diamond core.³⁵ This contraction is noticeable in the Fourier-transformed EXAFS spectra, where the Mn scatterer is now found below R + Δ = 3 Å (Figures 4 and S9). Like the corresponding open core compounds, 1-OH, 2-OH, and 3-OH exhibit primary coordination sphere features of high-spin ferric compounds. Also, both 1-OH and 2-OH have Fe–O scatterers at 1.83 and 1.99 Å, which are respectively typical bond lengths for Fe^III–μ-O and Fe^III–μ-OH units.^27,37 In contrast, 3-OH is best fit with one shell with two Fe–O scatterers at 1.89 Å, suggesting that the Fe–O and Fe–OH distances in this complex differ by less than 0.12 Å, the limit of the resolution for these data. Taken together, these results lend support to our assignment of 1-OH, 2-OH and 3-OH as Fe(μ-O)(μ-OH) Mn complexes.

The observed Fe•••Mn distances in 1-OH, 2-OH and 3-OH are comparable to those found for Fe^III₂(μ-O)(μ-OH) complexes supported by TPA* and BPEEN ligands (Table 2). These ligands are similar to those used in the current study, with amine and pyridine-based donors with no α substituents on the pyridines. In contrast, the Fe^III₂(μ-O)(μ-OH) complexes supported by 6Me₃TPA and BQPA (Table 2) have longer Fe•••Fe distances. We attribute this outcome to the greater steric interactions imparted by the introduction of α substituents on the pyridine moieties of 6Me₃TPA ligand and the quinolines on the BQPA ligand – steric factors not shared by the Fe/Mn compounds in the present study.

Table 2.

Structural properties of synthetic Fe/Mn compounds and their biological and diferric analogs.

Synthetic Fe/Mn Compounds*
	Fe–O	Fe–N(ave)	Fe–O(H)	Fe-M	Ref.
1	1.81	2.13	-	3.65	This work
2	1.82	2.14	-	3.63	This work
3	1.84	2.12	-	3.64	This work
1-OH	1.82	2.14	1.99	2.79	This work
2-OH	1.84	2.15	1.99	2.77	This work
3-OH	1.89	2.10	1.89	2.81	This work
Biological Fe/Mn Species
RNR1c FeIV(μ-O)2MnIV	1.81	-	-	2.74	12
RNR1c Fe III (μ-O)(μ-OH)Mn IV	Mn-O1.74	-	-	2.91	42
Mt R 2 lox resting state	2.1	-	-	3.60	15
FeIII 2(μ-O) Complexes TPA (no add’l bridges)	1.79	2.15	2.09 a 2.14 c	3.57	34
5Et3TPA† (μ-H3O2)	1.80	2.18	2.05 a 1.91 b	3.35	34
TPA (μ-H3O2)	1.81	2.17	2.04 a 1.91 b	3.39	34
BPMEN † (μ-H3O2)	1.82	2.23	2.05 a 1.99 b	3.39	36
Fe III 2 (μ-O)(μ-OH) Complexes
TPA * †	1.88	2.15	1.93	2.79	28
6Me3TPA †	1.82	2.20	1.99	2.95	27
BQPA †	1.89	2.19	1.94	2.89	27
BPEEN †	1.85	2.20	1.97	2.84	27
FeIII2(μ-O)2 Complexes
6Me 3 TPA	1.84 1.92	2.24	-	2.72	27

^*distances indicated in italics are derived from x-ray crystallography, distances in plain text are derived from EXAFS

^†-TPA* = tris((4-methoxy-3,5-dimethylpyridin-2-yl)methyl)amine; 6Me₃TPA = tris((6-methylpyridin-2-yl)methyl)amine; BQPA = 1-(quinolin-2-yl)-N,N-bis(quinolin-2-ylmethyl)methanamine; BPEEN = N,N′-diethyl-N,N′-bis(2-pyridylmethyl)ethane-1,2-diamine; 5Et₃TPA = tris((5-ethylpyridin-2-yl)methyl)amine; BPMEN = N,N′-dimethyl-N,N′-bis(2-pyridylmethyl)ethane-1,2-diamine;

^a– indicates Fe–OH₂ bond length;

^b– indicates Fe–OH bond length;

^c– indicates Fe–OClO₃ bond length

Mn³⁺ may also play a role in modulating the metal-metal distance. Interestingly, dimanganese complexes with diamond cores have almost exclusively favored bis(µ-oxo) cores, leading to metal-metal distances of ~2.7 Å, which are shorter than those of their diiron counterparts.^38–40 This is despite the fact that high-spin Fe³⁺ and Mn³⁺ in octahedral geometries have identical effective ionic radii.⁴¹ A direct comparison between 6-Me-substituted TPA diamond cores – [(6Me₂TPA)₂Mn₂(µ-O)₂]²⁺ and [(6Me₃TPA)₂Fe₂(µ-O)₂]²⁺ - shows a 0.05 Å contraction for the dimanganese complex – clear evidence that Mn³⁺ favors a contraction in metal-metal distance.^27,40

When the structures of the synthetic complexes reported herein are compared to those of the biological intermediates, it is clear that the metal-metal distances reported for 1-OH, 2-OH and 3-OH are comparable to the two found in for biological Fe/Mn intermediates,^12,42 despite differences in metal oxidation states. Indeed, synthetic diiron examples show that there is only a variation of around 0.2 Å in the metal-metal distances among complexes in three different oxidation states (Fe^III₂, Fe^IIIFe^IV, and Fe^IV₂).^{27,28,32,33,43} These results on synthetic Fe^IIIMn^III complexes represent the first confirmation that such structures can exist and represent a structural basis to better understand these fleeting biological intermediates.

It should be noted that the data we have collected for the series of open core complexes 1 - 3 differ from data we previously reported for 1.²³ The earlier data reported for 1 showed an Fe•••Mn distance of 3.34 Å, which is intermediate between those now associated with 1 and 1-OH and corroborated by the corresponding complexes 2 and 3 and their conjugate bases. We believe this lends support for our new assignment of the structure of 1.

The resonance Raman data for 1 and 2 both show closed core impurities in the presence of water (Figure S7). Therefore, we speculate that the initial preparation of 1 reported in 2018 may have been performed in wet solvent, which may have led to the formation of an “aquamer” of 1 where the Fe–O–Mn moiety is bridged by an H₃O₂⁻ ligand (vide infra). The Fe•••Mn distance of 3.34 Å originally reported for 1 has not been reproduced in this work but is consistent with the Fe•••Fe distance found in diferric compounds bridged by O^2- and H₃O₂⁻ ligands (Table 2). Given the UV-Vis and resonance Raman data that point towards an acid-base equilibrium between the open and closed core species, we propose the possibility that a third intermediate may be formed after the deprotonation of a water ligand on either the iron or manganese ion to create an H₃O₂⁻ bridge (Scheme 2). The loss of water from this intermediate then gives rise to the diamond core compounds. Unfortunately, we have not been able to reproduce the conditions for making this third intermediate. However, the precedent set in diiron chemistry where all three species have been characterized with various ligand sets shows that such a hypothesis is plausible.^27,28,34,36

Scheme 2.

Acid-base and aquation equilibria proposed for open core Fe/Mn complexes

The aquation equilibria of synthetic Fe/Mn and diferric complexes highlight the differences in the thermodynamic properties that result from incorporating a second metal. The preference for the closed core confirmation in synthetic systems could shed light on why nature chooses to use heterobimetallic active sites. It is possible that such sites allow for a more diverse array of core structures, or encourage the formation of different core structures, along the oxygen activation pathway.

The X-band EPR spectra of 1, 2, 1-OH and 2-OH show signals near g = 2, indicative of an S = ½ species arising from of an antiferromagnetically coupled high-spin Fe^III and Mn^III pair (Figure 5). The six-line splitting arises from hyperfine splitting from the I = 5/2 Mn^III ion in the complex. Complexes 1 and 2 both exhibit EPR signals, with g = 2.039, 2.015, 2.015 and g = 2.038, 2.015, 2.015, respectively, and ⁵⁵Mn nuclear hyperfine coupling of different magnitudes along the three principal axes (Table 3). In contrast, 1-OH and 2-OH exhibit rhombic EPR signals, with g = 2.03, 2.022, 2.015 and g = 2.030, 2.027, and 2.014 respectively, with different ⁵⁵Mn hyperfine splittings from those of 1 and 2 (Table 3). No additional ¹H superhyperfine splitting is observed in the hydroxo-bridged compounds, nor are the EPR signals affected upon labeling the hydroxo bridges with deuterium. (Figure S10).

Figure 5.

EPR spectra of 1, 2, 1-OH and 2-OH (data are colored spectra, simulations are black spectra). Simulation parameters are listed in Table 3.

Table 3.

EPR parameters of synthetic and biological Fe/Mn compounds

Complex	g	AMn
1	2.039, 2.015	190, 287, 313
2	2.038, 2.015	190, 270, 311
1-OH	2.030, 2.022, 2.015	209, 368, 280
2-OH	2.030, 2.027, 2.014	200, 370, 277
Mt R2lox 44	2.034, 1.968, 1.953	282, 249, 257
RNR 1c Fe IV (μ-O) 2 Mn IV 12	2.028, 2.021, 2.013	221, 243, 246

A comparison of the EPR spectra of the reported complexes to those of related biological intermediates shows that the synthetic compounds are able to model some aspects of the electronic structures of the corresponding biological species. The g tensor of the resting state of Mt R2lox, which has a singly bridged Fe^III/Mn^III center, is rhombic, in contrast to the axial signals found for 1 and 2. However, both 1 and 2 as well as R2lox show clear evidence for ⁵⁵Mn nuclear hyperfine splitting along all three principal axes. Besides differences between the histidine and carboxylate ligands found in the enzyme active sites versus the pyridine donors in the synthetic complexes, the Fe–O_bridge distance reported for R2lox is 2.1 Å, significantly longer than the Fe/Mn–O_bridge distances found in 1 and 2. This long distance suggests that the bridging ligand is unlikely to be an oxo group, but rather a protonated derivative thereof instead.¹⁵

Perhaps unexpectedly, 1-OH and 2-OH reproduce the EPR signals of biological Fe/Mn diamond cores, despite the fact that the trapped high-valent intermediate for RNR 1c has been shown to have an Fe^IV(μ-O)₂Mn^IV center. The differences in oxidation state and bridging ligand do not appear to affect the g values for this set of compounds. The Mn^III ion is most likely in the high-spin state in 1-OH and 2-OH as well as in RNR 1c, which has an Fe^IV(μ-O)₂Mn^IV center. For mononuclear high-spin Mn complexes, Mn^II, Mn^III, and Mn^IV complexes all exhibit small g anisotropy with g values that do not vary much from g = 2.⁴⁵ Similarly, high-spin ferric ions also exhibit an isotropic g tensor with values close to 2. Even for the few mononuclear high-spin Fe^IV species reported, the g values do not significantly deviate from 2.^46,47 Therefore, the Fe/Mn diamond cores that generally exhibit strong antiferromagnetic coupling interactions between the two metal ions, such as 1-OH, 2-OH and the RNR 1c Fe^IV(μ-O)₂Mn^IV center have g values in the coupled system that are all close to 2. The biggest differences arise in the hyperfine coupling. A_Mn(y) is much larger for the synthetic compounds. The reasons for this pattern are unclear, and beyond the scope of this work, but these results clearly illustrate that synthetic models can be useful for mimicking the electronic structures of their biological cousins.

Based on the rR and UV-Vis results that show there is an equilibrium in solution between open and closed core compounds, it was of interest to explore the ligand effect on the pK_a of the open-to-closed-core conversion. To this end, each compound was titrated with a variety of bases. Plotting the yield of the hydroxo-bridged compounds versus the pK_a of the protonated base generates a sigmoidal plot (Figures 6 and S11 - S13). Fitting these data with a Boltzmann function and comparing the χo values for each compound gives the relative acidities of each open-core species. The pK_a values for these compounds are perhaps unexpected. Complexes 2 and 3 have essentially the same pK_a (within error), while 1 has a much higher pK_a. One might expect that the 5-Me, and 3,5-Me and 4-MeO substitutions on 2 and 3 might decrease the Lewis acidity of these compounds; however the opposite effect is observed. While the reason for this counterintuitive result is unclear, there must be a strong driving force to form an Fe/Mn diamond core. This is in stark contrast to the diiron counterparts of these compounds, which revert readily to open core compounds in the presence of water.²⁷

Figure 6.

Boltzmann fit of the base-dependent formation of 2-OH from 2

To explore the differences in reactivity between Fe^III/Mn^III and diferric complexes, the electron transfer reactivity of all compounds was assessed with various ferrocenes. Interestingly, 1 and 2 are both able to oxidize diacetyl ferrocene (Ac₂Fc, Ac₂Fc/Ac₂Fc⁺ E˚ = 0.49 V vs Fc/Fc⁺),⁴⁸ while 3 and the three corresponding (μ-O)(μ-OH) compounds do not. However, the latter are able to oxidize acetyl ferrocene (AcFc, AcFc/AcFc⁺ E˚ = 0.27 V vs Fc/Fc⁺).⁴⁸ This behavior is in stark contrast to the diferric compounds [(TPA)Fe(OH)(μ-O)(H₂O)Fe(TPA)]³⁺, [(5Me₃TPA)Fe(OH)-(μ-O)(H₂O)Fe(5Me₃TPA)]³⁺, and [(TPA*) Fe(OH)(μ-O)-(H₂O)Fe(TPA*)], none of which are able to oxidize ferrocene (Figure S5). These observations suggest that it is the Mn³⁺ ion that is responsible for oxidizing the derivatized ferrocenes.

Recently, Shafaat and coworkers published the protein electrochemistry of R2lox.⁴⁴ The cyclic voltammetry of the resting state of wild-type R2lox at pH 7 shows an anodic peak at 0.878 V vs NHE (0.237 V vs Fc/Fc⁺) and a cathodic peak at 0.612 V vs NHE (−0.029 V vs Fc/Fc⁺).The pH dependence for these electrochemical events shows that the oxidation/reduction is coupled to two proton transfers. While our results do not shed light on proton transfer events, they do show that synthetic compounds are more oxidizing than their protein counterparts by at least 250 mV in the cases of 1 and 2. The reduction potentials of 3, 1-OH, 2-OH and 3-OH are actually comparable to those found for WT R2lox. Based on our results, it also seems likely that the electrochemical processes observed in the protein arise from the Mn³⁺/Mn²⁺ couple.

Given the high potentials and known pK_a’s of these compounds, the bond dissociation free energy (BDFE) of the O-H bond of a water bound to a one-electron reduced [(L)Fe(OH₂)(μ-O)Mn(OH₂)(L)]⁴⁺ can be calculated using equation 1, where C is a constant for a given solvent (C_MeCN = 52 kcal/mol).⁴⁹

$$\text{BDFE}=1.37{\text{pK}}_{\text{a}}+23.06E^o+\text{C}$$ Equation 1:

Based on equation 1, the BDFEs for the O–H bond in reduced 1 - 3 can be estimated to be between 68 and 75 kcal/mol. 1-OH, 2-OH and 3-OH should therefore be able to perform hydrogen atom transfer (HAT) from substrates with X–H BDFE’s below 75 kcal/mol. TEMPO-H (= 2,2,6,6-tetramethylpiperidin-1-ol, BDFE = 66 kcal/mol)⁴⁹ was the substrate of choice to probe this reactivity because of its suitable O-H BDFE. Its high potential (E˚ = 0.572 V vs Fc/Fc⁺)⁴⁹ and low acidity (pK_a = 40.1 in MeCN) make it unlikely for a stepwise mechanism to occur. Also, TEMPO-H has been used as a substrate for a wide range of mononuclear Fe^III–OH and Mn^III–OH compounds, which facilitates comparison between mono- and bi-metallic systems.^37,50–52

The reaction of 1-OH with TEMPO-H shows a linear dependence on TEMPO-H concentration, the analysis of which gives a second order rate constant of 0.33 M⁻¹ s⁻¹(Figure 8). With TEMPO-D as substrate, a rate of 0.24 M⁻¹s⁻¹ is obtained, corresponding to a kinetic isotope effect of 1.4 that is similar to that found for [(dpaq)Mn(OH)]⁺ (Table 5). 2-OH and 3-OH react with TEMPO-H with a nearly identical rate to 1-OH (0.32 M⁻¹s⁻¹ and 0.37 M⁻¹s⁻¹ respectively, Table 5), as well as similar KIEs (1.4 vs 1.7 and 2.7, Table 5). These congruent results strongly suggest that all three complexes react with TEMPO-H via the same mechanism. As previously mentioned, the potential and acidity of TEMPO-H make it highly unlikely that a stepwise mechanism (electron transfer-proton transfer or proton transfer – electron transfer) occurs. In fact, the hydroxo-bridged compounds are not oxidizing (0.49 V > E˚ ≥ 0.27 V) nor basic enough (pK_a1-OH = 8.9 pK_a2-OH = 8.2, pK_a3-OH = 7.9) to initiate either electron transfer or proton transfer. For these reasons, we propose that all three complexes oxidize TEMPO-H via hydrogen atom transfer.

Figure 8.

Top – UV-Vis spectra for the reaction of 1 mM 1-OH with 40 mM TEMPOH at −40 ˚C Bottom – Linear dependence of rate on TEMPO-H/D concentration.

TABLE 5.

Comparison of substrate oxidation rates for Fe^III–OH and Mn^III–OH complexes

Complex	Substrate	T (˚C)	k2	H/D KIE	Ref
1-OH	TEMPOH	−40	0.33	1.4	a
2-OH	TEMPOH	−40	0.32	1.7	a
3-OH	TEMPOH	−40	0.37	2.7	a
[(TMC-py)Fe(OH)]2+	TEMPOH	−40	7.1	6	37
(TMP)Fe(OH)	TEMPOH	25	76	-	53
[(PyPz)Fe(OH2)(OH)4+	xanthene	20	2.2 × 103	20	54
[(Py5)Fe(OH)]2+	DHA	25	4.3 × 10−4	6.3	55
[(dpaq)Mn(OH)]+	TEMPOH	25	0.13	1.8	56
[(dpaq2Me)Mn(OH)]+	TEMPOH	−35	3.9	2.7	50
[(dpaq5Cl)Mn(OH)]+	TEMPOH	−35	2.8	-	51
[(dpaq5NO2)Mn(OH)]+	TEMPOH	−35	7	-	51
[(SMe2N4(tren))-Mn(OH)]+	TEMPOH	25	2.1 × 103	3.1	52

^a.This work. Abbreviations used: TMC-py = 1-(pyridyl-2′-methyl)-4,8,11-trimethyl-1,4,8,11-tetrazacyclotetradecane, Py5 = (2,6-bis(bis(2-pyridyl)methoxymethane)pyridine, TMP = meso-tetramesitylporphyrinate; PyPz = quaternized tetra-2,3-pyridinoporphyrazine, dpaq = 2-(bis(pyridin-2-ylmethyl)amino)-N-(quinolin-8-yl)acetamidate anion.

When the rates of TEMPO-H oxidation for the hydroxo bridged Fe/Mn compounds are adjusted for temperature differences, they are comparable to that found for (TMP)Fe(OH),⁵³ but are much slower than the HAT rates for other mononuclear FeIII–OH complexes. The rates of reaction for all three with TEMPO-H are much faster than that for [(Py5)Fe(OH)]²⁺ with DHA (DHA = 9,10-dihydroanthracene), which can be rationalized by the fact the C–H bond in DHA with a C–H BDFE = 76 kcal/mol is harder to oxidize than the O–H bond in TEMPO-H with an O–H BDFE = 66 kcal/mol.⁴⁹

The reasons for these differences in reactivity can be explored by comparing the reduction potentials and pK_a’s of the mononuclear and binuclear complexes. The pK_a’s of the Fe^II-OH₂ complexes of PY5 and PyPz are both 8, which are comparable to values found for 1-OH and 2-OH (although these were measured in DMSO and H2O at pH 5.2, respectively)^54,55. In the case of the PyPz complex, the Fe^III/Fe^II potential is 0.477 V vs Fc/Fc⁺, significantly more oxidizing than the heterobimetallic complexes. Surprisingly, the redox potential of the PY5 complex is lower than those for 1-OH, 2-OH, and 3-OH (0.155 V vs. Fc/Fc⁺),⁵⁵ so the reason for the difference in observed reactivity is not as straightforward to rationalize, although the differences in substrate and solvent conditions make any comparison imperfect.

The HAT reactivity of the bimetallic complexes appears to be more similar to that of the mononuclear Mn^III-OH complexes than for the Fe^III-OH complexes. Indeed the rates of reaction with TEMPO-H for the series of [(dpaq)Mn(OH)]⁺ complexes are all within one order of magnitude of the rates of 1-OH, 2-OH, and 3-OH. Furthermore, the H/D KIE for the reaction of [(dpaq)Mn(OH)]⁺ is nearly identical to that of 2-OH. These similarities suggest that the Mn^III ion likely exerts a greater influence over the reactivity of the Fe/Mn complexes than the Fe^III ion.

In examining the thermodynamics that drive HAT by Mn^III -OH compounds, the data show that the reduction potentials of the dpaq complexes, which range from −0.72 V to −0.57 V vs Fc⁺/Fc, are all much lower than what is found for the Fe/Mn complexes reported here.⁵¹ Similarly, the reduction potential of [(S^Me2N₄(tren))Mn(OH)]⁺ (−0.241 V vs Fc⁺/Fc) is much lower than those of the Fe/Mn compounds. These lower reduction potentials are perhaps not unexpected given the anionic ligands used to support these Mn^III-OH complexes. The lower reduction potentials however, are compensated for by a much higher basicity in the case of the dpaq compounds, with Mn^II–OH₂ pK_a’s that range from 27.8 to 29.5.⁵¹ [(S^Me2N₄(tren))Mn(OH)]⁺ has a much lower pK_a than the Fe/Mn complexes (pK_a = 5.3). This work suggests that [(S^Me2N₄(tren))Mn(OH)]⁺ would be the less competent oxidant. However, even when adjusted for temperature, it is still an order of magnitude faster than 1-OH and 2-OH at HAT. However, the HAT reactivity of [(S^Me2N₄(tren))Mn(OH)]⁺ was probed in water, which makes it an imperfect comparison.

Another possible reason for the differences in observed HAT reactivity between the newly reported heterobimetallic compounds and the mononuclear complexes is the nature of the hydroxo moiety. In the case of the mononuclear compounds, the hydroxide is terminal, while the hydroxide acts as a bridge in the Fe/Mn compounds. In high-valent diiron chemistry, the switch from a bridging oxo moiety to a terminal oxo unit can result in as much as a million-fold increase in HAT reactivity.⁵⁷ While the differences between mononuclear Mn^III–OH and Fe^III-OH complexes and the Fe/Mn complexes are not quite as stark, it could still be a key structural feature that leads to enhanced HAT rates for the mononuclear complexes.

Conclusions

This work demonstrates the synthesis of a novel series of Fe/Mn compounds. Importantly, these compounds serve as the first to model the diamond core structures of proposed intermediates along the oxygen activation pathway of RNR 1c and R2lox. The structures of the first synthetic complexes with Fe/Mn diamond cores have thus been characterized, and their spontaneous formation from their linear Fe^III–O–Mn^III precursors in the presence of excess water has been documented. In direct contrast to the diiron chemistry, Fe/Mn compounds prefer to form closed core compounds in the presence of water, an unexpected result. Furthermore, the electron transfer reactivity of Fe/Mn compounds was compared to their diiron counterparts. In all cases, the Fe/Mn compounds are more oxidizing than their diiron counterparts – especially in the case of 1 and 2, which are at least 490 mV more oxidizing than their diiron counterparts. Taken together, these results suggest that Nature uses heterobimetallic active sites to control the preference for diamond core compounds and to modulate the reduction potential of oxidative intermediates in key enzymatic reactions.

The series of synthetic Fe/Mn compounds exhibit an unexpected relationship between the basicity of the iron supporting ligand and the pK_a of the conversion from open core to closed core. While the reasons for this inverse relationship are unclear, it nonetheless illustrates the differences between diiron and Fe/Mn chemistry.

By using the electrochemical and pK_a data, we were able to estimate the BDFE of a water bound to the reduced form of the Fe/Mn compounds. This estimation led us to attempt HAT reactivity with TEMPO-H. All three complexes react with TEMPO-H via an HAT mechanism. The rates of these reactions are comparable to those for mononuclear Fe^III-OH and Mn^III-OH compounds for X-H (X = C, O) bond activation chemistry and serve as the first examples of Fe/Mn complexes that perform HAT reactions.

Five new Fe/Mn compounds have been characterized using a variety of spectroscopies. Three of these compounds are the first examples of synthetic Fe/Mn diamond cores. While it should be noted that none of these compounds model the high-valent oxidation states of the proposed enzymatic intermediates, they still serve as structural and spectroscopic mimics of RNR 1c and R2lox – the first of their kind.

Experimental Section

Commercially available chemicals such as all bases used, perchloric acid, etc. were purchased from commercial sources and used without further purification unless otherwise noted. TEMPO-H was prepared according to the literature procedure⁵⁸ and TEMPO-D was prepared following an analogous procedure using deuterated solvents. Mn^II(TPA)(OTf)₂ and [Fe(TPA)(MeCN)₂](OTf)₂ were prepared according to literature procedures.^59,60 The corresponding Fe^II(5Me₃TPA) and Fe^II(TPA) complexes were prepared using analogous procedures. 1-(tert-butyl-sulfonyl)-2-iodosylbenzene was prepared according to a literature procedure.⁶¹

Low temperature UV-Visible absorption spectra were recorded on an HP 8453A diode array spectrometer fitted with a cryostat obtained from UNISOKU Scientific Instruments, Japan. Raman spectra were collected with an Acton AM-506 monochromator equipped with a Princeton LN/CCD data collection system, with excitation by 405/457/515/561 nm solid state lasers from Cobolt Lasers, Inc. Spectra in acetonitrile were collected at 77 K using a 135˚ backscattering geometry and at 233 K using a 90˚ backscattering geometry. The detector was cooled to −120˚ C prior to the experiments. Spectral calibration was performed on a 1:1 v:v mixture of acetonitrile and toluene. The collected data were processed using Spectra-gryph.⁶² A multi-point baseline correction was performed for all spectra. X-band EPR spectra were collected at 30 K on a Bruker Elexsys E-500 spectrometer equipped with an Oxford ESR-910 cryostat. EPR integrations were carried out using EasySpin (version 5.2.25). EPR simulations were carried out using SpinCount software developed by Prof. Michael Hendrich at Carnegie Mellon University.⁶³ Iron K-edge X-ray absorption spectra were collected on SSRL beamline 9−3 using a 100-element solid-state Ge detector (Canberra) with a SPEAR storage ring current of ∼500 mA at a power of 3.0 GeV. The incoming X-rays were unfocused using a Si(220) double crystal monochromator, which was detuned to 70% of the maximal flux to attenuate harmonic X-rays. Between 6 and 8 scans of the fluorescence excitation spectra for each sample were collected from 6882 to 8000 eV at a temperature (10 K) that was controlled by an Oxford Instruments CF1208 continuous-flow liquid helium cryostat. An iron foil was placed in the beam pathway prior to the ionization chamber (Io) and scanned concomitantly for an energy calibration, with the first inflection point of the edge assigned to 7112.0 eV. A 3, 6, or 9 μm Mn filter and a Soller slit were used to increase the signal-to-noise ratio of the spectra. Photoreduction was monitored by scanning the same spot on the sample twice and comparing the first derivative peaks associated with the edge energy during collection, but none was observed in the present study. The detector channels from the scans were examined, calibrated, averaged, and processed for EXAFS analysis using EXAFSPAK to extract χ(k). Theoretical phase and amplitude parameters for a given absorber−scatterer pair were calculated using FEFF 8.40 and were utilized by the “opt” program of the EXAFSPAK package during curve fitting. In all analyses, the coordination number of a given shell was a fixed parameter and was varied iteratively in integer steps, while the bond lengths (R) and mean-square deviation (σ2) were allowed to freely float. The amplitude reduction factor S0 was fixed at 0.9, while the edge-shift parameter E0 was allowed to float as a single value for all shells. Thus, in any given fit, the number of floating parameters was typically equal to (2 × number of shells) + 1. The k range of the data is 2−15 Å⁻¹.

Sample Preparation Procedures:

EPR and rR samples were prepared in a similar manner. A 1 mM solution of the Fe(II) precursor was cooled to −40˚ C in the UV-Vis cryostat. To this solution was added 1 equivalent of 1-(tert-butylsulfonyl)-2-iodosylbenzene to generate the corresponding Fe(IV)O complex. To the Fe(IV) compound was added 1 equivalent of (TPA)Mn(OTf)₂. In the case of 1, 2, and 3 the reaction was monitored until the Fe/Mn compound was maximally formed, then transferred with a pre-cooled pipette to an EPR tube and frozen in a liquid nitrogen bath. This procedure differed for rR spectra collected at 233 K. These samples were transferred to flatbottom NMR tubes and transferred to a −40˚ C bath before spectra were collected. 1-OH, 2-OH, and 3-OH were generated from solutions of 1, 2, and 3 by the addition of 1 eq. of triethylamine. These samples were frozen in a similar manner to 1, 2, and 3.

EXAFS samples were prepared by analogous methods to the EPR and rR samples, but from 5-mM Fe(II) solutions. These solutions of the Fe/Mn complexes were transferred to Mӧssbauer cups and frozen in liquid N2.

pK_a Determination:

The pK_a of each complex was determined using the following procedures. A 1 mM, 1.2 mL MeCN solution of the starting Fe^II complex was cooled to −40˚ C in the UV-Vis cryostat. To this was added 12 μL of 0.1 M sArIO (1 eq.) dissolved in 2,2,2-trifluoroethanol to generate the Fe^IV=O complex. To this was added 12 μL of 0.1 M Mn^II(TPA)(OTf)2 in MeCN (1 eq.). Once the Fe^III(μ-O)Mn^III complex had been fully formed, 12 μL of 0.1 M solution of a given base was added to the solution. The change in absorbance at 455 nm was noted. This was repeated in triplicate for each base. A plot of the change in absorbance at 455 nm with respect to the pK_a of the conjugate acid of the titrated base was generated for each trial. This was then fit using a sigmoidal Boltzmann function with a Levenberg Marquardt iteration algorithm in Origin 2016. The χo values generated from these fits were then averaged across the three trials for each compound.

Electron Transfer Experiments:

A 1 mM,1.2 mL MeCN solution of the starting Fe^II complex was cooled to −40˚ C in the UV-Vis cryostat. To this was added 12 μL of 0.1 M sArIO (1 eq.) dissolved in 2,2,2-trifluoroethanol to generate the Fe^IV=O complex. To this was added 12 μL of 0.1 M Mn^II(TPA)(OTf)₂ in MeCN (1 eq.). Once the Fe^III(μ-O)Mn^III complex had been fully formed, 12 μL of either Ac2Fc or AcFc were added to the open core complex and the reaction was monitored for the formation of the corresponding ferrocenium cation. In the case of the (μ-O)(μ-OH) complexes, 12 μL of 0.1 M TEA in MeCN were added first. Once the (μ-O)(μ-OH) complexes were formed in full yield, 12 μL of 0.1 M AcFc were added to the solution and the reaction was monitored for the formation of the corresponding ferrocenium cation.

Kinetics Experiments:

All kinetics measurements were performed using analogous procedures. A 1 mM,1.2 mL MeCN solution of the starting Fe^II complex was cooled to −40˚ C in the UV-Vis cryostat. To this was added 12 μL of 0.1 M sArIO (1 eq.) dissolved in 2,2,2-trifluoroethanol to generate the Fe^IV=O complex. To this was added 12 μL of 0.1 M Mn^II(TPA)(OTf)2 in MeCN (1 eq.). Once the Fe^III(μ-O)Mn^III complex had been fully formed, 12 μL of 0.1 M 2,4,6-trimethyl-pyridine was added (1 eq.) to form the corresponding Fe^III(μ-O)(μ-OH)Mn^III complex. To this solution was added 24 μL of either 0.5 M (10 eq.), 1 M (20 eq.), 1.5 M (30 eq.) or 2 M (40 eq) of TEMPO-H or TEMPO-D. The change in absorption of the 455 nm peak was monitored over time. Three trials of each concentration were performed, and the decay of the 455 nm peak was fit with a single exponential. The kobs for each concentration was averaged over the three trials and plotted with respect to substrate concentration to give a second-order rate constant plot.

Figure 7.

UV-Vis spectra in the oxidation of Ac₂Fc by 2

Scheme 1.

Spontaneous formation of a complex with an Fe^III/Mn^III diamond core upon addition of water

Table 4.

pK_a values for Fe/Mn compounds

Compound	pKa
1	8.9(1)
2	8.2(3)
3	7.9(1)