Enzyme Control Over Ferric Iron Magnetostructural Properties

Abstract

Fe³⁺ complexes in aqueous solution can exist as discrete mononuclear species or multinuclear magnetically coupled species. Stimuli-driven change to Fe³⁺ speciation represents a powerful mechanistic basis for magnetic resonance sensor technology, but ligand design strategies to exert precision control of aqueous Fe³⁺ magnetostructural properties are entirely underexplored. In pursuit of this objective, we rationally designed a ligand to strongly favor a dinuclear μ-oxo-bridged and antiferromagnetically coupled complex, but which undergoes carboxylesterase mediated transformation to a mononuclear high-spin Fe³⁺ chelate resulting in substantial T₁-relaxivity increase. The data communicated demonstrate proof of concept for a novel and effective strategy to exert biochemical control over aqueous Fe³⁺ magnetic, structural, and relaxometric properties.

Graphical Abstract

We introduce a first-in-class stimuli responsive Fe³⁺ complex that is switched between discrete antiferromagnetically coupled and high-spin paramagnetic species through the action of a carboxylesterase enzyme. We also demonstrate how enzyme-mediated switching of Fe³⁺ magnetic properties offers a mechanistic foundation for new bioresponsive magnetic resonance sensor technology.

Complexes of Fe³⁺ in aqueous media exhibit diverse chemical speciation which imparts a rich array of magnetic properties.^[1] Depending on the ligand(s), solution pH, temperature, and concentration, the complex may exist in predominantly a paramagnetic mononuclear form or as a multinuclear magnetically coupled system bridged by μ-oxo or μ-hydroxide ligands.^[2] Thus, similar to other classes of metal complexes which can change structure or electron spin configuration upon external perturbation, one could picture Fe³⁺ complexes capable of O-bridging interactions as a versatile foundational molecular unit for design of molecular sensors,^[3] catalysts,^[4] drug delivery systems,^[5] and bio-responsive contrast agents for medical imaging.^[6] However, strategies to exert precision control over aqueous Fe³⁺ speciation are surprisingly underexplored.

We hypothesized that we could rationally design a system that switches from entirely μ-oxo-bridged dimeric to high-spin monomeric Fe³⁺ speciation in response to enzyme activity. We envision this chemistry as a basis for precision magnetic resonance (MR) imaging or sensor technology. Prior work using a system termed Fe-PyCy2AI (Figure S1) demonstrated how discrete antiferromagnetically coupled [(FePyCy2AI)₂O] and high-spin [FePyCy2AI]⁺ species exhibit an 18-fold difference in T₁-relaxivity (r₁),^[7] enabling a switch between r₁ values correspondingly low (0.072 mM⁻¹s⁻¹) and high (1.4 mM⁻¹s⁻¹) enough to be virtually MR silent and conspicuously MR visible at concentrations consistent with those achieved in vivo after a clinically relevant dose of MR contrast.^[7–8] We anticipate that enzyme mediated switching between antiferromagnetically coupled and paramagnetic Fe³⁺ species will enable technology to non-invasively detect, map, and monitor pathology at the cell and molecular level.

Our design strategy is premised on solution thermodynamics data available for Fe-PyCy2AI and Fe³⁺ complexes of structurally related hexadentate ligands (Figure S1, Table S1), which indicate that monomer vs. dimer speciation is governed in part by competition for Fe³⁺ coordination between the weakest pendant donor arm and the μ-oxo bridging ligand. We desire a ligand to support μ-oxo dimer speciation but featuring a pendant donor arm masked by an enzymatically-removed protecting group that upon biotransformation outcompetes the μ-oxo ligand for Fe³⁺ coordination (Figure 1A). We hypothesized that the ligand L^Pro, in which a pendant 2-hydroxybenzyl donor arm is concealed by a 4-(methyl)phenylpivolate ester protecting group, would form a μ-oxo bridged dimeric complex of Fe³⁺ ((FeL^Pro)₂O), Figure 1B). We also anticipated that carboxylesterase mediated pivolate ester hydrolysis, followed by spontaneous 1,6-elimination of the resultant 4-hydroxybenzyl group,^{[6b, 9]} would liberate the 2-hydroxybenzyl donor arm to form two equivalents of corresponding monomeric complex FeL (Figure 1C,D). The synthesis of (FeL^Pro)₂O and FeL are outlined in Scheme S1 and discussed in detail in the Supporting Information.

Figure 1.

(A) Pictorial representation of enzyme driven Fe³⁺ magnetostructural change. PG = protecting group, D = pendant donor arm. (B) The L^Pro ligand was rationally designed to support μ-oxo-bridged dimer down to μM concentrations at pH 7.4. (C) Ligand L supports speciation of high-spin Fe³⁺ monomer up to mM concentrations at pH 7.4 (D) (FeL^Pro)₂O is converted to FeL via carboxylesterase mediated pivolate hydrolysis.

Single crystals of (FeL^Pro)₂O and FeL suitable for X-ray diffraction were grown from saturated solutions of the complexes in 1:1 water: MeCN and neat water, respectively (Figures 2, S4–5, Tables S2–9). The crystal structure of (FeL^Pro)₂O reveals the presence of two nearly equivalent octahedral Fe³⁺ centers bridged by a linear mono-μ-oxo bridge. Antiferromagnetic coupling in mono-μ-oxo bridged diferric systems is believed to occur predominantly through a π-bond superexchange pathway,^{[1a, 10]} and in this regard the Fe-O bond distances (1.768(2) Å and 1.775(2) Å) and near linear Fe-O-Fe bond angle (176.01(15)°) are considered optimal to promote magnetic bridge orbital overlap. The structural features of the (FeL^Pro)₂O μ-oxo bridge are consistent with those of previously reported mono-μ-oxo bridged, antiferromagnetically coupled complexes.^{[2f, 7, 11]} In contrast, The FeL crystal structure confirms the presence of a monomeric, distorted octahedral Fe³⁺ complex.

Figure 2.

ORTEP diagram of (A) (FeL^Pro)₂O and (B) FeL showing 30% probability ellipsoids for all non-H atoms (C = grey, N = blue, O = red, and Fe = brown). Solvent molecules are omitted for clarity.

UV-visible, EPR, and NMR magnetic susceptibility measurements indicate that the solution structures of (FeL^Pro)₂O and FeL at pH 7.4 reflect those observed in the solid state. For example, the UV-vis spectrum for the orange colored (FeL^pro)₂O complex exhibits distinct features with λ_max at 312 nm (ε = 6,970 M⁻¹cm⁻¹) and 342 nm (ε = 5,230 M⁻¹cm⁻¹) consistent with μ-oxo-Fe charge transfer transitions,^{[1a, 2g]} whereas solutions of FeL exhibit a broad transition with λ_max 518 nm (ε = 1,990 M⁻¹cm⁻¹) typical of phenolate-Fe charge transfer transition and giving rise to purple color (Figure 3A). Continuous wave X-band EPR spectroscopy of (FeL^Pro)₂O does not yield substantial signal across the entire X-band range and at varied temperature and power (Figures 3B, S6). On the other hand, the EPR spectrum on an equivalent concentration sample of FeL results in strong signal consistent with the presence of a single species that can be simulated as a rhombic S = 5/2 system of E/D = 0.32 (Figures 3B, S7–8). Bulk magnetic susceptibility NMR measurements at 298K yield μ_eff values of 2.4 μ_B and 5.8 μ_B per Fe for (FeL^Pro)₂O and FeL, also consistent with an antiferromagnetically coupled complex and a mononuclear S = 5/2 complex, respectively.^{[1a, 2d]}

Figure 3.

(A) UV-vis spectra of (FeL^Pro)₂O (black trace) and FeL (purple dashed trace) in 9:1 mixture of pH 7.4 100 mM Tris buffer: MeCN at RT. (B) Continuous wave X-band EPR spectra of (FeL^Pro)₂O (black trace) and FeL (purple dashed trace) recorded on 250 μM Fe solutions in pH 7.4 100 mM Tris buffer (9.37 GHz, 30 K, 0.2 mW).

Both (FeL^Pro)₂O and FeL are the dominant species formed across a concentration range spanning over 3 orders of magnitude. The Fe-μ-oxo charge transfer bands at 312 nm and 342 nm offer a spectroscopic handle to monitor the concentration dependent equilibrium between μ-oxo bridged Fe³⁺ and the corresponding monomeric species.^[7] For (FeL^Pro)₂O, ε at 312 nm and 342 nm remained unchanged down to concentrations as low as 3.6 μM Fe (Figure S9A), indicating that no substantial dissociation to monomeric FeL^Pro occurs down to at least the low μM concentration range. Similarly, no concentration dependent changes on the FeL absorption profile were observed within the concentration range spanning the detection thresholds of our spectrophotometer, and the EPR spectra showed no changes in lineshape or signal down to 10 μM (Figure S10). We also monitored 1/T₁ as a function of FeL concentration in the 0.5–7.5 mM range, and here too we observe a linear dependence on Fe concentration, i.e. r₁ is unchanged, indicating that no significant degree of self-association to the low-r₁ magnetically coupled species occurs at concentrations up to at least 7.5 mM (Figure S9B).

Antiferromagnetically coupled (FeL^Pro)₂O is an inefficient relaxation agent compared to paramagnetic FeL. The r₁ values for (FeL^Pro)₂O and FeL at 1.4T and 310 K are 0.11 mM⁻¹s⁻¹ and 0.59 mM⁻¹s⁻¹, respectively. The r₁ value of (FeL^Pro)₂O mirrors that of the previously reported antiferromagnetically coupled dimeric complex (FePyCy2AI)₂O, which possesses a nearly identical primary Fe³⁺ coordination sphere, whereas the r₁ value of FeL is consistent with complexes of high-spin Fe³⁺ that do not possess a rapidly exchanging water co-ligand.^{[6c, 12]} The (FeL^Pro)₂O vs. FeL r₁ differential is largely field independent between 1.4T to 11.7T and is maintained in solutions of 4.5% wt. volume bovine serum albumin (Figures S11–12, Table S10). The phantom test tube images shown in Figure 4 demonstrate how the large (FeL^Pro)₂O vs. FeL r₁ differential can be leveraged to generate a high degree of image contrast. Figure 4A shows T₁-weighted images of test tubes containing asolutions of 75 μM Fe as either (FeL^Pro)₂O or FeL. The space between the test tubes comprises buffer without Fe present. Note how the test tube containing (FeL^Pro)₂O is nearly isointense with neat buffer whereas the test tube containing FeL is hyperintense. The data in Figure 4B show that the percentage signal increase over neat buffer in the FeL sample is roughly 4-fold greater than than in the (FeL^Pro)₂O sample. Figure 4C shows a T₁ values of recorded in test tubes containing 10 μM Fe as either (FeL^Pro)₂O or FeL and of the surrounding buffer mix. A concentration of ≥10 μM is oft cited as the lower concentration threshold for in vivo detection for clinical T₁-relaxation agents and thus FeL can be expected to generate only very modest signal enhancement in a T₁-weighted image acquired using the imaging parameters from Figure 4A. The mean T₁ values recorded in the (FeL^Pro)₂O phantom and in neat buffer solution are very similar (2700 ± 20 ms and 2710 ± ms, respectively), whereas mean T₁ in the FeL phantom is a meaningful 70 ms shorter than water (2640 ± 20 ms). Figure 4D shows how even at 10 μM Fe concentration the substantially greater T₁-shortening properties of FeL vs. (FeL^Pro)₂O can be capitalized on to generate strong contrast in T₁-weighted images through application of a 1810 ms inversion pre-pulse to null signal from neat buffer or samples of comparable T₁. Figure 4E shows how application of the pre-pulse inversion results in xx-fold greater signal generating capacity of FeL compared to (FeL^Pro)₂O (1.7% vs. 25.5% signal intensity increase over surrounding water). We note that it is common in both clinical imaging and non-clinical molecular imaging research to null T₁ values of native tissue and sensitive contrast agent enhancement through application of an inversion pre-pulse.^[13] Taken together, the images of Figure 4 show how (FeL^Pro)₂O vs. FeL r₁ difference can be leveraged to achieve an MR signal turn on effect using any number of commonly employed T₁-weighted acquisitions and even at low concentrations down to 10 μM Fe.

Figure 4.

(A) T₁-weighted gradient-echo images at 4.7T and RT of test tubes filled water containing 75 μM Fe as (FeL^Pro)₂O or FeL in 90:10 pH 7.4 Tris buffer: MeCN. The space in between the test tubes is filled with neat 90:10 buffer: MeCN. (B) Percentage signal intensity increase over buffer without Fe recorded in the test tubes containing 75 μM (FeL^Pro)₂O or FeL. The SI increase achieved with FeL is 4-fold greater than that from (FeL^Pro)₂O. (C) T₁-maps recorded for solutions containing 10 μM Fe as (FeL^Pro)₂O or FeL. Notice how mean T₁ in the sample containing (FeL^Pro)₂O is very close to neat buffer but the mean T₁ in the FeL sample is 70 ms shorter. (D) T₁-weighted gradient-echo images acquired with a 1810 ms inversion prepulse to null signal from buffer with Fe present. (E) Percentage signal intensity increase over neat buffer recorded in the test tubes containing (FeL^Pro)₂O or FeL. The SI increase achieved with FeL is 15-fold greater than that from (FeL^Pro)₂O.

Dimeric (FeL^Pro)₂O is converted to FeL in the presence of porcine liver esterase. Reaction monitoring by UV-vis or T₁ change of 95 μM (FeL^Pro)₂O mixed with 12 U/mL enzyme at 310K (Figure 5A,B) indicated that FeL formation proceeds with k_obs = 2.2 ± 0.5 s⁻¹ (t_1/2 = 19 ± 4 min). Control experiments in the absence of enzyme indicate (FeL^pro)₂O is stable against hydrolysis over 24h (Fig S14). The UV-vis changes comprise simultaneous absorption decrease at 312 and 342 nm and increase at 518 nm, respectively, marked by a clear isosbestic point at 365 nm. Reaction monitoring by HPLC at RT (Figure 5C) indicated that (FeL^Pro)₂O is consumed in under 1 min resulting in substantial buildup of intermediates (FeL^Pro)(FeL^Int)O and (FeL^Int)₂O that are formed en route to FeL (Figure 1D). We note that experimental readouts in Figures 5A–B should be largely insensitive to the presence of (FeL^Pro)(FeL^Int)O and (FeL^Int)₂O, as we do not anticipate either intermediate to contribute to absorbance at 518 nm, nor should we antipate that the intermediate r₁ values will differ meaningfully from (FeL^Pro)₂O. Because (FeL^Pro)₂O is near completely consumed before acquisition of the first post-mixing UV-vis or T₁ data point and (FeL^Int)₂O is not expected to absorb at 518 nm or to possess substantially different r₁ than (FeL^Pro)₂O, the spectroscopically determined k_obs predominantly reflects the rate of the (FeL^Int)₂O elimination reaction. We were unable to identify any μ-oxo-bridged species containing ligand L via HPLC-MS, indicating rapid FeL formation following 1,6 quinone methide elimination from (FeL^Int)₂O. Fast generation of FeL upon pendant donor arm ‘unmasking’ indicates that the kinetics of enzyme-triggered dimer formation may be optimized as needed for future systems, as established strategies exist to increase the rates of 4-hydroxy or 4-aminobenzyl linker group 1,6-elimination reactions.^[14]

Figure 5.

Reaction monitoring of 95 μM (FeL^Pro)₂O (190 μM Fe) mixed with 12 U/mL porcine liver esterase in 98:2 pH 7.4 100 mM Tris buffer: MeCN mixture. (A) Change in UV-vis spectra over the course of 10 min to 130 min at 310K. Arrows show decrease in absorbance at 312 nm and 342 nm and increase in absorbance at 518 nm. Inset shows isosbestic point at 365 nm consistent with direct conversion of μ-oxo-bridged intermediate to FeL. (B) Change in 1/T₁ (R₁, black circles) and ε at 518 nm (red circles). Solid lines represent mono-exponential fits to the data. The measurements were performed in triplicate, the error bars correspond to standard deviation. (C) HPLC traces (Method A3, SI) recorded prior to and out to 120 min after mixing at RT. HPLC retention times for (FeL^Pro)₂O is 12.6 min, [(FeL^Int)(FeL^Pro)O] and (FeL^Int)₂O elute at 11.1 and 9.6 min, respectively; the minor and major interconverting diastereomers of FeL elute at 8.8 and 7.8 min.

Preliminary stability assays to monitor Fe release from (FeL^Pro)₂O and FeL in the presence of transferrin and excess Zn²⁺ ions indicate that both complexes are robust against Fe release in the presence of strong competing challenges likely encountered in vivo. For example, spectrophotometric monitoring of transferrin formation in 50 μM Fe³⁺ solutions of (FeL^Pro)₂O and FeL incubating with 100 μM apo-transferrin at 310K, indicate the Fe transchelation reactions proceed with k_obs = 0.028 h⁻¹ (t_1/2 = 25 h) and with k_obs = 0.008 h⁻¹ (t_1/2 = 90 h), respectively (Figure S15). Spectrophotometric monitoring of Fe release from 50 μM Fe solutions solutions of (FeL^Pro)₂O and FeL in the presence of 20 molar equivalents Zn²⁺ at pH 6.0 indicates that Fe is displaced from (FeL^Pro)₂O and FeL with k_obs = 0.047 h⁻¹ (t_1/2 = 15 h) and with k_obs = 0.120 h⁻¹ (t_1/2 = 6 h), respectively (Figure S16). The reaction between FeL and excess Zn²⁺ completed with 24 μM FeL remaining. Detailed analysis of thermodynamic stability and of Fe dissociation mechanisms and the corresponding rate laws extend beyond the scope of this study. We also note that there are no established benchmarks by which to relate in vitro dissociation assays to Fe release in vivo. However, given that a typical small molecular MRI contrast agents are eliminated from humans with half-life on the order or 1–2 hours (and on the order of minutes from rodents),^[15] the inertness of both (FeL^Pro)₂O and FeL to Fe release even under forcing metal ion challenge conditions indicate that this new class of Fe complex represents a viable option for development for in vivo imaging applications.

We note that (FeL^Pro)₂O was designed to generate mechanistic proof of principal for enzyme modulation of Fe³⁺ magnetostructural properties and not for in vivo imaging. However, even this unoptimized prototype exhibits an impressive >5-fold r₁ increase following enzymatic activation. The r₁ of a high-spin Fe³⁺ complex can be further optimized by engineering ligands that accommodate a rapidly exchanging Fe³⁺ water co-ligand, or through modifications to increase the Fe³⁺ rotational correlation time.^{[7, 16]} For example, one could envision hydrolytic unmasking of pendant arms containing O-donor functional groups such as carboxylate, phosphonate, or alkoxide as alternate means of μ-oxo-bridge splitting. Pendant arms which form 5-membered rings upon Fe binding (ie. acetate) result in more acute N-Fe-O bond angles than the 2-hydroxybenzyl arm of FeL, thus offering a higher likelihood supporting 7-coordinate mononuclear Fe³⁺ containing a water co-ligand.^{[8, 17]} Furthermore, more strongly basic donor groups like phosphonate and alkoxide have been shown to engage in prototropic exchange near pH 7.4 when coordinated to metal ions including Fe³⁺, providing an alternate means to enabling efficient dipolar ¹H relaxation in the absence of a water co-ligand.^{[12a, 18]} Alternately, incorporating metal complexes into higher molecular weight multimeric structures or ligand modifications to promote plasma protein binding represent established strategies by which to increase rotational correlation time and thus r₁.^{[15–16, 19]}

We also note that our ligand design approach is broadly applicable to hydrolytic enzymes. Future work building upon our proof concept data will focus on further optimizing the r₁ increase upon switching from antiferromagnetically coupled to paramagnetic Fe³⁺, on developing sensors for enzyme markers of human disease, and on evaluating and optimizing enzyme ‘turn-on’ kinetics, in vivo stability, and pharmacokinetics.

The preliminary results described above outline a ligand design strategy by which profound changes to magnetic and structural properties of an aqueous Fe³⁺ complex can be triggered through enzyme activity. The ligand L^Pro was rationally designed to promote formation of μ-oxo-bridged and antiferromagnetically coupled Fe³⁺ and supports (FeL^Pro)₂O formation down to at least low μM concentrations. Antiferromagnetic coupling in (FeL^Pro)₂O results in low T₁-relaxivity. Carboxylesterase mediated cleavage of the ancillary pivolate ester frees a pendant 2-hydroxybenzyl donor arm resulting in two molar equivalents of FeL, with ligand L supporting exclusively high-spin paramagnetic Fe³⁺ and high r₁ up to concentrations of least 7.5 mM. Switching from antiferromagnetically coupled Fe³⁺ to high-spin Fe³⁺ offers a powerful mechanistic basis for stimuli responsive molecules and for precision MR sensor technology.