Communication: Near edge x-ray absorption fine structure spectroscopy of aqueous adenosine triphosphate at the carbon and nitrogen K-edges

The hydrolysis of adenosine triphosphate (ATP) to the corresponding diphosphate is able to thermodynamically drive many enzymatic reactions in biological systems including energetically unfavorable molecular conversions.¹ ATP-utilizing enzymes have special pockets or grooves in their three-dimensional structures to bind the ATP and catalyze the loss of the phosphate, usually in the presence of a metal cocatalyst, such as magnesium $(Mg2+)$⁠.^2–4 To reliably model the uptake of ATP into these special binding sites, more accurate models of solvated ATP, as well as the protein crystal structures, are needed. This requires an understanding of the details of the solvation environment and how that environment is disrupted by association with proteins.

Near edge x-ray absorption fine structure (NEXAFS) spectroscopy can be used to study solvation because of the sensitivity of the extended antibonding orbitals and Rydberg states that comprise the final states of the core-level transitions to molecular environment.^5–8 Previous NEXAFS studies have examined adenine in the gas phase as well as in the form of various solid and thin films.^9–16 While these provided valuable insight into the electronic structure of the adenine ring, understanding the effects of hydration is clearly essential to discerning and modeling the observed biological activity. The liquid microjet apparatus and methodology developed by Wilson et al.¹⁷ overcome the practical difficulties of obtaining carbon, nitrogen, and oxygen K-edge NEXAFS spectra in aqueous systems, as well as minimizing radiation damage. Although this method typically requires concentrations of analyte larger than that allowed by adenine’s limited water solubility, the charge of the triphosphate chain and hydrophilicity of the ribose sugar in ATP allow sufficient aqueous concentrations for measurement of the core-level spectra.

In this work, we report the carbon and nitrogen K-edge NEXAFS spectra of aqueous adenosine triphosphate under conditions of varying concentration, $pH$⁠, and metal complement $(Na+,Mg2+,Cu2+)$⁠, and we compare these with published experimental spectra of the gaseous and solid states. Because of self-aggregation, concentration studies allow us to search for spectral changes resulting from adenine rings $π$-stacking in solution, while $pH$ variation allows us to observe both the protonated and unprotonated adenine ring of ATP. Additionally, the study of complementary metal ions allows us to examine the effects of inner- and outer-sphere metal ion association with the adenine ring. Comparing these spectra with those of the gas phase and solid state spectra helps elucidate the chemical nature of the aqueous system.

Disodium adenosine triphosphate was purchased commercially as a powder from Sigma-Aldrich (USA) with a stated purity of 99.9% and was used without further purification. $MgCl2$ and $CuSO4$ were purchased from Fischer Scientific (USA) as crystalline powders with a stated purity of 99% and used without further purification. Solution samples were prepared by dissolving specific amounts of ATP in $18MΩ$ water (Millipore). For $Mg2+$ and $Cu2+$ solutions, the corresponding salt was dissolved in the ATP solution to give a 1:1 metal to ATP molar ratio. The $pH$ was adjusted by addition of $1M$ reagent grade HCl or NaOH (Fischer) until the desired $pH$ was obtained. Final dilution yielded solutions that ranged from $0.1M$ to $0.3M$ in ATP. Solutions were used immediately to avoid sample degradation.

Carbon and nitrogen K-edge x-ray absorption spectra were recorded over the relevant energy ranges for each solution (C: 285–310 eV; N: 390–420 eV). All measurements were obtained at Beamline 8.0.1 of the Advanced Light Source in Berkeley, CA. The experimental apparatus has been described in detail by Wilson¹⁷ and the instrumental details are essentially identical to those contained therein. Briefly, an intense $(1012photonss−1)$⁠, high energy resolution $(5000E/∂E)$ beam of x-ray radiation is focused onto an approximately $30μm$ diameter jet of the sample solution. The jet is produced by a syringe pump pressurizing the liquid through a pulled glass capillary tip. The signal is collected as total electron yield by means of a positively biased copper electrode located within 0.5 mm of the jet/beam intersection. Because of the differential pumping of the apparatus, the experiment operates without a window, allowing for spectral acquisition without window-based interference. The obtained spectra were area-normalized and calibrated to $CO2$ and $N2$⁠.

Figure 1 shows the molecular structure of ATP. The spectral features of the carbon and nitrogen edges reported herein are compared with published data for adenine in gaseous and solid phases. Oxygen spectra were recorded, but there was insignificant signal above the water background for analysis. Figure 2 shows the carbon edge spectra of aqueous ATP at $pH$ 7.5. In the $π∗$ region, there are two major peaks with maxima at 286.5 and 287.3 eV, along with shoulders at 286.0 and 288.1 eV. These peaks match well with the gas phase⁹ and solid state¹⁰ spectra of adenine, although the gas phase spectrum exhibits a peak at 286.8 eV that is not observed in either the solid or aqueous ATP data presented here. Zubavichus¹⁰ attributed the two peaks to two different carbon environments, viz., carbons directly bonded to either one or two nitrogens. The $π∗$ transitions should be relatively unchanged from those of isolated adenine due to the absence of $π$ bonds in the ribose structure. Ukai et al.¹⁸ studied core-level spectra of ribose-monophosphates of guanine (GMP), a purine-based nucleoside structurally similar to ATP, and found no significant shifts due to the addition of the sugar and phosphate in the $π∗$ region. Thus, the $π∗$ regions of adenine and ATP should be directly comparable. Additionally, the relative peak intensities of the solid and aqueous systems are similar, suggesting that the aqueous state more resembles that of the solid than the gas phase, as one might expect.

Figure 3 shows the nitrogen NEXAFS spectra of aqueous ATP at $pH$ 7.5. The ATP spectrum exhibits major peaks at 400.3 and 402.5 eV, along with a minor feature at 401.5 eV. The major peak in both adenine spectra falls at 399.5 eV and is attributed to the transitions of core electrons of ring nitrogens to the lowest unoccupied molecular orbital (LUMO). The apparent blueshifting of the major aqueous ATP peak compared to gas- and solid-phase data is likely due to hydrogen bonding with the solvent.

The lesser feature at 401.5 eV and minor peak at 402.5 eV in the ATP spectrum are harder to directly compare to the adenine spectra. The gas phase data include a shoulder at 401.5 eV, which is assigned via calculations to the amino nitrogen 1s LUMO transition.⁹ The solid-phase data have a nearby peak at 401.3 eV, which is attributed to ring nitrogen 1s $LUMO+1$ by Zubavichus.¹⁰ However, this peak at 401.3 eV compares better with the larger peak at 402.5, given the relative intensities in the spectra as well as the overall blueshifting of the spectrum. In either case, there is a feature that does not directly compare with either gas or solid-phase data, suggesting that the solvation is playing a role in altering the LUMO. We note that gas phase calculations do not show this peak for adenine.⁹ We expect that lone pairs and $NH$ bonds interact more strongly with solvent through hydrogen bonding, as compared to carbon and its less polar electronic features.⁶ 

Increasing the concentration of ATP from $0.1M$ to $0.3M$ produces no significant changes in the carbon or nitrogen spectra. This is an interesting result in light of the well-known self-association of nucleotides. In solution, the nucleotides will agglomerate through $π$-stacking interactions at high concentrations.^19,20 Using self-association values from Sigel,²¹ increasing concentrations from $0.1M$ to $0.3M$ ATP at $pH$ 7 should constitute a change of free ATP of around 33% and increasing the relative amount of doubly and higher associated ATP stacks. NEXAFS is probably insensitive to $π$-stacking of nucleotides, in accordance with what is seen in GMP.¹⁸ 

Significant changes in the spectra occur as a result of protonation of the adenine ring in ATP. The $N∗$ (shown in Fig. 1) has a $pKa$ of 4.16 in ATP and the $sp2$ lone pair accepting a proton creates a positive charge. The spectra of aqueous ATP at $pH$ 7.5 and 2.5 are shown for both carbon and nitrogen K-edges in Fig. 4. In the carbon spectra, the peak at 287.3 eV splits into two peaks at 287.0 and 287.7 eV, while the peak at 286.5 eV remains unchanged, as $pH$ decreases. This can be described with the symmetry argument of Zubavichus.¹⁰ The peak at 287.4 eV in solid-phase adenine was attributed to the four carbons bound to two nitrogen atoms $(NCN)$⁠; the protonated nitrogen will directly affect the states of the two carbons bonded to it and have only secondary effects on the other two carbons. Tautomerization is a possible explanation, but the general abundance of adeninium tautomers is too low to be observed.^22,23 The nitrogen K-edge spectrum also undergoes changes with $pH$⁠. At $pH$ 2.5, there are now two clear peaks, and their peak values are at 399.9 and 401.8 eV, which match marginally with the gas-phase adenine data.⁹ There still appears to be a feature between these peaks at 400.1 eV. This would seem to be the same feature apparent in the $pH$ 7.5 data, although slightly redshifted.

Figure 5 shows the spectral changes associated with the addition of $Mg2+$ and $Cu2+$ ions to the ATP solution at the nitrogen K-edge. Both of these metals have large binding constants to the triphosphate chain $(K∼2000M−1–4000M−1)$ compared to sodium $(K∼43)$⁠,²⁰ but have differing modes of interaction with the adenine ring itself; the $Mg2+$ is reported to exhibit an outer-sphere interaction with the adenine ring, mediated by a water molecule, while $Cu2+$ can bind to $N∗∗$ (Fig. 1) of the adenine ring directly.^20,24 The MgATP spectra remain unchanged from the sodium ATP spectra, but CuATP exhibits a broadening of the 400.2 eV peak in the nitrogen spectrum and some minor relative intensity changes in the carbon spectrum (not shown). Intramolecular chelation in the $CuATP2−$ complex is reported to account for 67% of the total CuATP,²⁰ so we attribute the broadening of the nitrogen spectrum to the presence of this $Cu2+$ association. Minor changes in the carbon K-edge intensities (not shown) can be rationalized as secondary effects of the copper binding.

The data for the MgATP also support the conclusion of the concentration data with respect to the observation of ATP multiplexes. The MgATP complex has a self-association constant twice as large as that of free ATP (Ref. 21) and should lead to even larger concentrations of higher-order ATP complexes. The lack of any observable spectral changes further testifies to the insensitivity of NEXAFS spectra to the $π$-stacking of nucleosides.

We report the carbon and nitrogen K-edge NEXAFS spectra of aqueous adenosine triphosphate and discuss the spectral variation with concentration, $pH$⁠, and metal complement. While there are some similarities of the solvated ATP spectrum with the solid and gaseous forms of the adenine spectra, there are small but distinct changes in the hydrated form. The observations from the concentration variation and MgATP spectrum suggest that NEXAFS is insensitive to the $π$-stacking of nucleotides. The protonation of the adenine ring in ATP has marked effects on both the carbon and nitrogen spectra, some of which can be rationalized based on symmetry changes in the adenine ring environment. Finally, we observe the effect of inner-sphere binding of copper to the adenine ring of ATP, primarily manifested as peak-broadening in the nitrogen spectrum. First-principles calculations²⁵ of these aqueous spectra would be very useful for interpreting these data, but accurate calculations are beyond the scope of present computational capabilities for such a large system and the number of parameters involved.

This work was supported by the Director, Office of Basic Energy Sciences, Office of Science, U.S. Department of Energy under Contract No. DE-AC02-05CH11231 through the LBNL Chemical Sciences Division, the Molecular Foundry, and the Advanced Light Source. We wish to thank Wanli Yang and Jonathan Denlinger for excellent user support of Beamline 8.0.1.