Magnesium(II)-ATP Complexes in 1-Ethyl-3-Methylimidazolium Acetate Solutions Characterized by 31 Mg β-Radiation-Detected NMR Spectroscopy.

The complexation of MgII with adenosine 5'-triphosphate (ATP) is omnipresent in biochemical energy conversion, but is difficult to interrogate directly. Here we use the spin- 1/2 β-emitter 31 Mg to study MgII -ATP complexation in 1-ethyl-3-methylimidazolium acetate (EMIM-Ac) solutions using β-radiation-detected nuclear magnetic resonance (β-NMR). We demonstrate that (nuclear) spin-polarized 31 Mg, following ion-implantation from an accelerator beamline into EMIM-Ac, binds to ATP within its radioactive lifetime before depolarizing. The evolution of the spectra with solute concentration indicates that the implanted 31 Mg initially bind to the solvent acetate anions, whereafter they undergo dynamic exchange and form either a mono- (31 Mg-ATP) or di-nuclear (31 MgMg-ATP) complex. The chemical shift of 31 Mg-ATP is observed up-field of 31 MgMg-ATP, in accord with quantum chemical calculations. These observations constitute a crucial advance towards using β-NMR to probe chemistry and biochemistry in solution.

Introduction

A key difficulty in elucidating the (bio)chemical function of MgII is the limited sensitivity of spectroscopic techniques capable of directly probing the closed shell ion. Consequently, studies of its coordination chemistry are rather sparse (see e.g., ref. [2, 3]), despite the physiological importance of MgII. Having an experimental technique effective at resolving these details would greatly benefit all fields concerned with understanding the chemistry of MgII. While nuclear magnetic resonance (NMR) spectroscopy is used ubiquitously to this end for many elements, magnesium has only a single (stable) NMR isotope, 25Mg (nuclear spin I=5/2; gyromagnetic ratio γ/(2π)=−2.60793(9) MHz T−1; electric quadrupole moment Q=199(2) mb; 10 % natural abundance),[ , ] whose utility as a probe suffers from its non‐zero quadrupole moment and low receptivity.[ , ] For example, the salient feature of MgII binding to a ligand such as adenosine 5′‐triphosphate (ATP) is typically line broadening, obscuring fine structural signatures. To circumvent these limitations, we instead use the short‐lived β‐emitter 31Mg (nuclear spin I=1/2; gyromagnetic ratio γ/(2π)=−13.4699(23) MHz T−1; half‐life T 1/2=236 ms)[ , , ] as our NMR probe and monitor its resonance through the anisotropic property of its β‐decay—a technique known as β‐radiation‐detected NMR (β‐NMR) spectroscopy.[ , , ]

The principles of β‐NMR are nearly identical to “conventional” NMR (see e.g., ref. [13]), with differences originating from the use of an unstable probe (e.g., NMR detected via radioactive decay products).[ , , ] This approach affords a nearly ≈1010‐fold increase in sensitivity, enabling spectra to be acquired under conditions which cannot be attained by any other method—including very low probe concentrations (see e.g., ref. [14, 15, 16]). In this sense, β‐NMR is quite similar to muon spin spectroscopy (μSR). While μSR is known for its utility in chemistry (see e.g., ), β‐NMR's uses are traditionally rooted in nuclear and solid‐state physics, with chemical applications being relatively unexplored. Some progress to this end has been made recently, with 8Li β‐NMR being used to study the glassy phase of polymers,[ , , , ] small molecules, and room temperature ionic liquids (RTILs). Similarly, several groups have now implemented setups capable of measurements in liquids,[ , , , , , , ] greatly expanding the scope of possible experiments.

Our primary interest here is applying 31Mg β‐NMR to study MgII complexation in solution. Distinct from “conventional” NMR, nuclear spin polarized 31Mg is introduced into solution by ion‐implantation in an accelerator beamline under ultra‐high vacuum (UHV).[ , ] A question that naturally arises is: does 31Mg, following implantation, attain an equilibrium configuration within its radioactive lifetime? This is indeed the case when the solvent molecules are the ligands in the coordination complex that forms, as was demonstrated in two imidazolium based RTILs. With the current work, we aim to progress to the more interesting situation of 31Mg binding to a foreign solute molecule. To this end, we implanted 31Mg into a series of 1‐ethyl‐3‐methylimidazolium acetate (EMIM‐Ac) solutions containing the prototypical MgII ligand in biochemistry, ATP. A priori, it was not obvious if the probe would associate with the biomolecule both within its radioactive lifetime and before its spin polarization was lost (via spin‐lattice relaxation). As we shall show below, both of these conditions are fulfilled.

The MgII‐ATP complex was selected due to its ubiquitous function in biochemistry as an “energy currency”.[ , ] For example, MgII‐catalyzed ATP hydrolysis within enzymes is coupled directly to, and drives otherwise non‐spontaneous, biochemical processes. Moreover, though the RTIL solvent is quite different from an aqueous medium, it provides a coordination environment akin to MgII binding in proteins;[ , ] the abundant acetate ligands (≈6 M) resemble both glutamate and aspartate side chains, while EMIM‐Ac's dielectric constant is compatible with the range of values commonly reported for the interior of proteins (see e.g., ref. [38]). Thus, both electrostatically and in terms of ligand composition, EMIM‐Ac resembles MgII‐ATP binding sites in enzymes both within ATP hydrolysis and in phosphoryl transfer, though lacking specific optimized structures typically present in proteins.

Results and Discussion

Before discussing the observed 31Mg β‐NMR spectra, we digress briefly into the most essential experimental details. In these β‐NMR experiments, performed at TRIUMF's isotope separator and accelerator (ISAC) facility, 31Mg was extracted from an isotope production target (as a 40 keV 31MgI beam) by laser ionization, spin‐polarized in‐flight by optical pumping,[ , ] and implanted into the EMIM‐Ac solution. The solution, housed in an aluminum alloy holder, was suspended vertically in UHV (10−10 Torr), where EMIM‐Ac's virtually zero vapour pressure prevented evaporation,[ , , ] and its large viscosity (see e.g., ref. [44]) inhibited flow out of the container.[ , ] During implantation, the probe rapidly oxidizes to 31MgII, and its ensuing behaviour reflects the chemical properties of the closed shell ion. The β NMR measurements were performed at 295 K and 3.20 T (corresponding to a Larmor frequency of ≈43.1 MHz for 31Mg) using a dedicated high‐field spectrometer.[ , ] Resonances were acquired using a continuous wave (CW) radio frequency (RF) transverse magnetic field B 1 that was slowly stepped through 31Mg's Larmor frequency, integrating for 1 s at each frequency step. This approach is analogous to a CW NMR experiment using stable nuclei and similar to the approach adopted in μSR. Off resonance (or in absence of any RF field), the β‐decay asymmetry (proportional to the spin‐polarization of the 31Mg ensemble ) is constant, with this “baseline” value setting the maximum possible signal amplitude (see e.g., ref. [11]). On resonance, the 31Mg nuclei are rapidly depolarized, resulting in a reduction in the observed asymmetry, with a value of ≈0 corresponding to the complete 31Mg population visiting a particular “site” (i.e., coordination environment) during the measurement time window; however, if exchange dynamics occur (i.e., the probes occupy more than one structure within the duration of the RF pulse), they can contribute to each resonance peak, giving rise to double‐counting. We observe this “extra” amplitude in all spectra, highlighting the importance of chemical exchange in our measurements.

Prior to embarking on the β‐NMR measurements, it was important to first confirm the complexation of MgII and ATP in EMIM‐Ac. For this, we applied 31P NMR spectroscopy to explore the equilibrium chemistry of the binding process. These measurements established that MgII binds to ATP under our experimental conditions, allowing for estimates of equilibrium constants for the formation of Mg‐ATP and Mg2‐ATP. Next, 31Mg β‐NMR experiments were conducted (see Figure 1), as outlined below.

Figure 1

31Mg β‐NMR spectra in EMIM‐Ac with different amounts of solutes (MgCl2, Mg(Ac)2, and ATP), recorded at 295 K and 3.20 T (≈43.1 MHz). The striking differences in the spectra recorded with and without added ATP are a strong indication of MgII‐ATP complexation. The resonance at 0 ppm reflects the binding of MgII to the solvent anions (used as an in situ reference), the resonance at −6 ppm is assigned to a di‐nuclear 31MgMg‐ATP species, and the broad resonance at approximately −11 ppm is assigned to 31Mg‐ATP. The vertical scale is the same for all spectra. Each data point is drawn as a vertical black line, denoting the span of the (statistical) error bars. The solid coloured lines represent fits to a sum of Lorentzians and the baselines are indicated by dotted grey lines.

First, a 31Mg β‐NMR spectrum was recorded with 25 mM MgCl2 (anhydrous) dissolved in EMIM‐Ac, but in the absence of ATP (i.e., representing MgII binding to the solvent acetate anions). Note that without the added salt, the spectrum is broad with poorly resolved features; however, with the impurity sites saturated by stable “carrier” MgII, the spectra sharpen, revealing two characteristic “solvent” peaks. Further increasing the MgCl2 concentration had little effect on the signal. The most consequential feature of the “solvent” signal is the large amplitude peak, which is easily identified in all spectra and we assign it a chemical shift of 0 ppm, making it our in situ reference for the 31Mg spectra.

Next, a series of 31Mg β‐NMR experiments were conducted in solutions with 50 mM ATP and 25 mM of either MgCl2 or Mg(Ac)2⋅4 H2O (see Figure 1). It is evident that the spectra recorded in the presence and absence of ATP differ significantly, providing a strong indication for 31Mg binding to ATP. Note that the spectra recorded using different MgII salts are essentially identical, implying that the anion of the Mg‐salt does not affect the spectroscopic signature—further evidence of complexation by our probe. In the presence of ATP, three main peaks are easily distinguishable. These can be quantified by fitting to a sum of Lorentzians and a baseline,[ , ] identifying unique coordination environments at: 0 ppm (from the solvent), −6 ppm, and −11 ppm. To place an assignment on the remaining peaks, we consider their evolution with solute concentration, in conjunction with the species present in solution at equilibrium.

Analogous to the behaviour in aqueous solution, we expect MgII to form both mono‐ and di‐nuclear species when binding to ATP in EMIM‐Ac (see e.g., ref. [49, 52]). Consequently, we anticipate the presence of the following species in solution prior to introducing any 31Mg: MgII, ATP, Mg‐ATP, and Mg2‐ATP. The persistence of the “solvent” peak at all solute concentrations (see below) suggests that the binding of 31Mg is “staged”, with the implanted 31Mg initially forming complexes with the solvent acetate anions. This is reasonable, given their significant abundance (≈6 M). Subsequently, upon encounter, the solvent‐bound 31Mg may bind to either ATP or Mg‐ATP, forming 31Mg‐ATP or 31MgMg‐ATP, respectively. Any binding to Mg2‐ATP is, presumably, negligible. Thus, our focus in the following is on the presence of ATP and Mg‐ATP, which are disposed to form complexes with the implanted 31Mg.

31Mg β‐NMR spectra recorded at different MgCl2 concentrations (0 mM to 200 mM) with a constant ATP concentration (50 mM) are shown in Figure 2. Under these conditions, the main species present in solution prior to 31Mg implantation are controlled by the amount of added MgCl2. At high concentration (≥100 mM), all ATP are saturated by the “carrier” MgII (i.e., essentially only Mg‐ATP is present as a potential ligand for the implanted 31Mg), as confirmed by 31P NMR. The 31Mg β‐NMR spectra at these conditions, apart from the solvent peak, show only one additional resonance at −6 ppm, suggesting it corresponds to the di‐nuclear complex. We therefore assign this signal to 31MgMg‐ATP. As the MgCl2 concentration is decreased to ≤50 mM, the amplitude of the −6 ppm peak decreases systematically, coinciding with the emergence and growth of a signal at −11 ppm. At these conditions, the solutions contain both free and Mg‐bound ATP, with their respective populations increasing (decreasing) as the MgCl2 concentration is lowered. This consistency implies that the −11 ppm signal is due to the formation of a mono‐nuclear complex and we assign it to 31Mg‐ATP. Note that measurements using instead Mg(Ac)2⋅4 H2O as the “carrier” salt yielded identical results. Similarly, spectra recorded using ATP as the “titrant” (at fixed MgCl2 concentration) were also found to be consistent with the above interpretation (i.e., the −6 ppm peak grows with increasing ATP concentration). Together, these observations confirm that the observed behaviour for implanted 31Mg is intrinsic.

Figure 2

31Mg β‐NMR spectra in EMIM‐Ac at various MgCl2 concentrations (indicated in the inset) and 50 mM ATP, recorded at 295 K and 3.20 T (≈43.1 MHz). The resonance at 0 ppm reflects the binding of MgII to the solvent anions (used as an in situ reference), the resonance at −6 ppm is assigned to a di‐nuclear 31MgMg‐ATP species, and the broad resonance at approximately −11 ppm is assigned to 31Mg‐ATP. The vertical scale is the same for all spectra. Each data point is drawn as a vertical black line, denoting the span of the (statistical) error bars. The solid lines represent a fit to a sum of Lorentzians and the baselines are indicated by dotted grey lines.

With the major features of Figure 2 outlined above, we consider some of the spectral details further. First, we note that the main feature of the resonance at −11 ppm, assigned to 31Mg‐ATP, is its relatively large linewidth. This observation is consistent with several possible binding modes of MgII to ATP.[ , , ] Accompanying the appearance of this signal is a noticeable drop in intensity of the “solvent” peak, particularly when the MgCl2 concentration is ≤25 mM. This may indicate that the binding of 31MgII to free ATP is faster than to Mg‐ATP, resulting in quicker depopulation of the solvent‐bound complex. At all other conditions, the amplitude of the 0 ppm resonance is maximal, indicating that all 31Mg nuclei occupy this structure (for at least ≈1 ms ) during the 1s RF pulse. In contrast, our 31P NMR data demonstrate that, at equilibrium, the MgII binding to ATP is shifted significantly towards the Mg‐ATP complex. From this we conclude that the 31Mg β‐NMR spectra reflect a non‐equilibrium situation, wherein the implanted 31Mg remain complexed with the solvent acetate anions initially (for at least ≈1 ms) after implantation and subsequently form an ATP‐containing complex during the (rather long) RF pulse. This interpretation is supported by the fact that neither of the 31Mg ATP or 31MgMg‐ATP resonance amplitudes reach their maximum, though both scale according to the ratio of added ATP and MgII concentrations, which determines the abundance of ATP and Mg‐ATP in solution. Moreover, as alluded above, the sum over all resonance amplitudes always exceeds the spectrum's “baseline”, meaning that the observed coordination species undergo dynamic exchange during the measurement's time window, supporting the interpretation of “staged” binding. In the future, it would be interesting to follow this process directly (e.g., using spectral hole‐burning ).

A surprising result from Figure 2 is in the experiment with only trace amounts of implanted 31Mg (i.e., no added MgCl2). Here, only the mono‐nuclear complex was expected to form, but the spectrum also showed a minor peak at −6 ppm, implying the formation of 31MgMg‐ATP. This is likely due to the presence of a small amount of MgII (i.e., as an impurity) in either the commercial ATP salt (estimated to be 18(1) μM from inductively coupled plasma mass spectrometry (ICP‐MS) measurements ) or the solvent (up to ≈1 mM ).

At the high MgCl2 concentration limit, an increase from 100 mM to 200 mM gives rise to a drop in the intensity of the −6 ppm signal. Based on the 31P NMR data and the derived equilibrium constants for the formation of Mg‐ATP and Mg2‐ATP, it is predicted that the concentration of Mg‐ATP is decreased in the sample with 200 mM MgCl2, due to a shift in the equilibrium towards Mg2‐ATP. This is expected to also give rise to a decrease in 31MgMg‐ATP signal, in agreement with the observed trend (see Figure 2), confirming the interpretation of the β‐NMR data.

Thus far, we have not discussed the impact of any H2O present in our solutions. H2O is a common impurity in hygroscopic RTILs such as EMIM‐Ac and we previously considered it as an explanation for the “minor” solvent peak near −6 ppm (see Figure 1); however, a subsequent measurement with intentionally added water rules this out, eliminating it as a possible “contaminant” for the peak assigned to 31Mg‐ATP. Another important consideration is how H2O content influences the pH of our solutions, which is well‐known to affect the complexation of MgII and ATP in aqueous solution. In pure EMIM‐Ac, assuming that no groups exhibit proton dissociation equilibria, the pH of the solution is, by definition, undefined; however, this is not a practical issue as, based on Fourier transform infrared (FTIR) spectroscopy measurements, the water content of “fresh” EMIM‐Ac is up to ≈1 % H2O (v/v), placing our solutions in the extreme basic limit (i.e., pH≈14). This suggests that any MgII coordination to ATP's nucleobase, known to occur at low pH, is unlikely. Of greater importance though is the possibility of ATP hydrolysis occurring in EMIM‐Ac. Our 31P NMR data reveal that a small fraction of ATP is hydrolyzed, producing adenosine 5′‐diphosphate (ADP) and inorganic phosphate; however, the measurements indicate that this amount is minor (≈4 % to ≈8 % of the total ATP content). While we cannot completely exclude the possibility that minor signals appear in the 31Mg β‐NMR data due to these species, we note that the spectroscopic signature for MgII binding to ADP differs from ATP.

Finally, to further substantiate the interpretation of the 31Mg β‐NMR spectra, we used DFT calculations to determine the optimum coordination geometry and corresponding chemical shielding tensor for the Mg‐ATP and Mg2‐ATP complexes. Specifically, starting from analogous structures in aqueous solution, we considered the species [Mg‐ATP(Ac)2]4− and [Mg2‐ATP(Ac)4]4−, whose geometries and shieldings were computed using B3LYP/pc‐2[ , , , , , ] and B3LYP/pcSseg‐2,[ , , , , ] respectively, each using an integral equation formalism of the polarizable continuum model (IEFPCM)[ , ] to account for the solvent. The fact that the geometry optimizations for these structures converged to energy minima demonstrates that they are stereochemically possible in EMIM‐Ac. In all cases, the MgII were found to be hexacoordinated by oxygen from the phosphate and acetate groups (see Figure 3), the latter being either mono‐ or bi‐dentate (within the first coordination sphere). Note that two configurations for the Mg‐ATP complex were found, both containing MgII coordinated to all three phosphates. It is conceivable that there are other (local) minima on the potential energy surface (i.e., several conformers of the Mg‐ATP complex may co‐exist at room temperature), in qualitative agreement with the large linewidth of this resonance in our data. Conversely, only a single structure for Mg2‐ATP was found, with one MgII coordinating to the α‐ and β‐phosphate, and the other to the β‐ and γ‐phosphate. The calculated (isotropic) shieldings for the Mg‐ATP (576.4 ppm and 578.8 ppm) and Mg2‐ATP (566.8 ppm and 568.0 ppm) species lead to a chemical shift difference of 8 ppm to 12 ppm. Noting that the computed 31Mg‐ATP shift is upfield from 31MgMg‐ATP, we obtain reasonable agreement (within the error of the calculations) with the experimental difference of ≈5 ppm, providing additional support for the assignment of the 31Mg β‐NMR resonances.

Figure 3

Summary of the 31Mg β‐NMR experiments probing MgII binding to ATP in EMIM‐Ac. Nuclear spin‐polarized 31MgI  was implanted into EMIM‐Ac solutions suspended vertically within an aluminum alloy plate inside an accelerator beamline under UHV.[ , , ] During implantation, the probe rapidly oxidizes to 31MgII, whereafter it binds (initially) to the solvent acetate anions and subsequently forms 31Mg‐ATP or 31MgMg‐ATP. The formation of either complex depends chiefly on the amount of free and Mg‐complexed ATP present prior to implantation. Using a CW resonance technique,[ , , ] our β‐NMR spectra reveal distinct chemical shifts, whose structural assignments (indicated in the inset) are derived from their systematic evolution with solute concentration. Here, the model spectrum corresponds to the experiment with 50 mM ATP and 25 mM MgCl2 (see Figures 1 and 2). The large resonance amplitudes indicate that all three species undergo dynamic exchange on the millisecond timescale. Structures for the mono‐ and di‐nuclear complexes (obtained from DFT calculations[ , ]) are also shown (drawn using PyMOL ).

Conclusion

In summary, we have shown the first instance of an ion‐implanted β‐NMR probe (31Mg) binding to a solute molecule (ATP) before the probe spin depolarizes. This is a necessary prerequisite for the general application of β‐NMR spectroscopy in solution chemistry and our result holds promise for future applications in biochemistry. For the case of MgII binding to ATP in the RTIL EMIM‐Ac, we were able to resolve distinct MgII‐ATP coordination environments using the short‐lived β‐emitter 31Mg. Based on their variation with solute concentration, the recorded resonances have been assigned to: solvent (acetate) bound 31Mg (0 ppm), 31MgMg‐ATP (−6 ppm), and 31Mg‐ATP (−11 ppm). From the persistence of the “solvent” signal across all measurements, the formation of these species was found to be “staged”: the implanted 31Mg initially binds to the solvent, then associates with either ATP or Mg‐ATP, depending on their (equilibrium) concentrations in solution. Using DFT calculations, structures for both the mono‐ and di‐nuclear coordination complexes were identified, whose computed isotropic shielding constants were found to be consistent with the measured 31Mg chemical shifts. These findings, along with a sketch of the β‐NMR experiments, are illustrated in Figure 3. As a final and important point, the 31Mg β‐NMR experiments allow for elucidation of MgII containing species at extremely low probe concentrations (<1 nM) and under conditions where no other experimental technique can provide useful data.

Conflict of interest

The authors declare no conflict of interest.

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supporting Information

Click here for additional data file.

Supporting Information

Click here for additional data file.

Supporting Information

Click here for additional data file.

Supporting Information

Click here for additional data file.