A Quantitative Raman Spectroscopic Signal for Metal–Phosphodiester Interactions in Solution

Abstract

Accurate identification and quantification of metal ion–phosphodiester interactions are essential for understanding the role of metal ions as determinants of three-dimensional folding of large RNAs and as cofactors in the active sites of both RNA and protein phosphodiesterases. Accomplishing this goal is difficult due to the dynamic and complex mixture of direct and indirect interactions formed with nucleic acids and other phosphodiesters in solution. To address this issue, Raman spectroscopy has been used to measure changes in bond vibrational energies due to metal interactions. However, the contributions of inner-sphere, H-bonding, and electrostatic interactions to the Raman spectrum of phosphoryl oxygens have not been analyzed quantitatively. Here, we report that all three forms of metal ion interaction result in attenuation of the Raman signal for the symmetric vibration of the nonbridging phosphate oxygens (ν_sPO₂⁻), while only inner-sphere coordination gives rise to an apparent shift of ν_sPO₂⁻ to higher wavenumbers (ν_sPO₂⁻M) in solution. Formation of ν_sPO₂⁻M is shown to be both dependent on metal ion identity and an accurate measure of site-specific metal ion binding. In addition, the spectroscopic parameter reflecting the energetic difference between ν_sPO₂⁻ and ν_sPO₂⁻M (ΔνM) is largely insensitive to changes in phosphodiester structure but strongly dependent on the absolute electronegativity and hardness of the interacting metal ion. Together, these studies provide strong experimental support for the use of ν_sPO₂⁻M and ΔνM as general spectroscopic features for the quantitative analysis of metal binding affinity and the identification of metal ions associated with phosphodiesters in solution.

Metal–phosphodiester interactions play central roles as determinants of the three-dimensional structure of nucleic acids and as cofactors in the active sites of both RNA and protein phosphodiesterases (1–4). These interactions include electrostatic charge–charge interactions between the positively charged metal and the negatively charged phosphodiester backbone, hydrogen bonding (H-bonding) via coordinated water molecules, and inner-sphere coordination of one or more nonbridging oxygens. These chemical interactions give rise to two general classes of metal ion binding: binding in a stable chelated mode with geometric specificity that often involves inner-sphere coordination and binding in a diffuse mode in which numerous ions interact weakly, primarily via electrostatic interactions with the negatively charged phosphodiester backbone (5–9). The extent to which chelated and diffuse metal ion binding occurs is determined by local nucleic acid structure and is likely to include a combination of the three distinct chemical forms of metal–phosphodiester interaction. Quantitative assessment of the distribution of inner-sphere coordination, H-bonding, and electrostatic interactions, however, is still difficult to conduct experimentally. A significant part of this difficulty is due to the dynamic nature of metal–phosphodiester interactions, particularly in solution, which continues to provide a barrier to a complete understanding of the linkage between ion binding and function. In addition to these challenges, functional RNAs often require the binding of multiple ion species (e.g., Mg²⁺ and Na⁺) with distinct and overlapping levels of inner-sphere coordination, H-bonding, and electrostatic effects, making the analysis of individual ion interactions difficult to deconvolute.

An emerging strategy for the direct probing of specific chemical interactions between metal ions and phosphodiesters is based on the principle that these interactions will necessarily perturb the vibrational properties of the interacting atoms of the phosphodiester (10–12). Changes in the vibrational properties of bonded atoms have long been monitored by Raman spectroscopy, which measures the exchange of energy between photons and vibrating bonded atoms in a sample (11). Thus, when the vibrational properties of individual or groups of energetically coupled bonds are altered (e.g., through the binding of a metal ion), such changes are reported directly as a change in the energy of scattered photons. Indeed, changes in bond vibrations due to the presence of metal ions are readily observed in the Raman spectrum of phosphodiesters, including nucleotides and nucleic acids, and have been used as a semiquantitative method for monitoringmetal ion binding (12–15). Furthermore, because the time frame of photon scattering is fast relative to molecular motion, the Raman signal is largely immune to the distortions of the experimental signal caused by changes in the molecular structure or type of metal ion interaction while the experimental signal is being generated (11). Thus, under equilibrium conditions, Raman spectroscopy enables the assessment of the distribution of individual atomic interactions in a population of molecules.

One of the largest metal-dependent changes in the Raman spectra of phosphodiesters occurs in the group frequency associated with the symmetric stretch of the nonbridging phosphate oxygens (ν_sPO₂⁻) (13–15). Experimentally, the Raman peak for ν_sPO₂⁻ becomes attenuated and undergoes a large apparent shift to higher wavenumbers in the presence of numerous metal ions common to biological systems (13–15). The apparent shift in the position of the ν_sPO₂⁻ peak produces an inflection in Raman difference spectra where metal-dependent changes in the Raman spectrum are typically analyzed. The apparent shift of ν_sPO₂⁻ to higher wavenumbers has been proposed to reflect an altered vibrational mode involving the nonbridging oxygens (here termed ν_sPO₂⁻M) that is induced upon metal ion binding (14, 16). Consistent with this interpretation, formation of ν_sPO₂⁻M has been observed to correlate with the loss of fully hydrated magnesium ion and the formation of magnesium penta- or tetrahydrate in crystals of the HDV ribozyme (12). Furthermore, quantification of difference spectra has been used to approximate the same number of stable site-bound metal ions in HDV crystals as previously determined by other biochemical and biophysical methods (12, 17–19).

Application of Raman spectroscopy to analyzing phosphodiester–metal ion interactions, however, still faces several important challenges. Electrostatic interactions, for example, are known to make thermodynamically large and topologically complex contributions to ion affinity. Thus, while crystallography provides a significantly enhanced spectroscopic signal relative to that in solution, the closely packed geometry of the phosphodiester backbone in crystal lattices and the solvent conditions required for Raman crystallography may result in binding geometries and thermodynamics that may not quantitatively reflect ion binding in solution. In addition, the interpretation of metal-induced changes in ν_sPO₂⁻ is complicated by ambiguities in the extent to which inner-sphere coordination, outer sphere H-bonding, and electrostatic interactions contribute to changes in phosphodiester vibrational modes. Resolving this ambiguity and extending the method to quantitative analysis of metal–phosphate interactions in solution are thus essential to the development of this otherwise powerful spectroscopic approach.

To accomplish these goals, we examined ion-induced changes in the nonbridging oxygen vibrational modes in dimethyl phosphate (DMP), nucleotides, and nucleic acids in solution. In addition, we compared the effects of ions that differ in their ability to interact with nonbridging phosphate oxygens by inner-sphere coordination, H-bonding, or electrostatic interactions. These and other studies reveal that all three forms of metal interaction can significantly attenuate the magnitude of ν_sPO₂⁻, while the metal-induced vibrational mode (ν_sPO₂⁻M) is attributed to purely inner-sphere coordination. Quantitative analysis of the intensity of ν_sPO₂⁻M formation as a function of metal ion concentration is further shown to accurately monitor saturable binding of Mg²⁺ to ATP and ADP. In addition, the degree of the shift of ν_sPO₂⁻ to higher wavenumbers by Mg²⁺ and other metal ions is shown to correlate strongly with the absolute electro-negativity and absolute hardness of the interacting metal ion, providing a potential means of establishing the identity of interacting metal ions. The spectral characteristics of inner-sphere, H-bonding, and electrostatic metal ion binding described above are observed for both simple and structurally complex phosphodiesters over a broad range of ionic strength and metal ion type and thus are likely to reflect the behavior of phosphodiesters in general.

METHODS

Reagents

ATP, ADP, MnCl₂, and cobalt hexamine were obtained from Sigma. CaCl₂, CoCl₂, and ZnCl₂ were purchased from Fisher Scientific. CdCl₂ was obtained from Acros Organics. MgCl₂ was purchased from Ambion Inc. DMP and DMTP were synthesized by hydrolysis of dimethyl chlorophosphate and dimethyl chlorothiophosphate (Aldrich). Specifically, individual chlorophosphates were incubated overnight in a 40-fold molar excess of water and dehydrated to near dryness under vacuum. The resulting phophodiesters were then brought to a stock concentration of 0.4 M with water, and the solution was adjusted to pH 5.0 with formate. Eleven-nucleotide single-stranded RNA 5′-UCAAGUACCGA-3′, the 12-nucleotide GAAA tetraloop (5′-GGGCGAAAGUCC-3′), and the 27-nucleotide bulged stem–loop sequence derived from the P4 helix/stem-loop portion of RNase P RNA (20) (5′-GGAAGUCCGGUCUUCGGACCGGCUUCC-3′) were made by solid phase synthesis (Dharmacon Inc.). Yeast tRNA^PHE and Escherichia coli RNase P RNA were prepared by in vitro transcription using standard methods (Ambion). All RNAs were purified on 8–22.5% 19:1 acrylamide/bisacrylamide gels. RNA bands were identified by UV shadowing, cut out of the gel, and eluted overnight at room temperature in 10 gel volumes of 10 mM Tris-HCl (pH 8.0), 300 mM NaCl, and 1 mM EDTA. The eluent was extracted twice in an equal volume of a 1:1 phenol/chloroform mixture and once with an equal volume of pure chloroform. The resulting RNAs in the aqueous phase were precipitated in 2.5 volumes of ethanol and recovered by centrifugation. RNA pellets were resuspended in 1 mL of glass-distilled water and dialyzed against 2 L of 1 mM EDTA overnight in a Float-A-lyzer G2 (Spectrum Laboratories) micro dialysis tube. RNAs were subsequently dialyzed against water for 24 h prior to storage at −20 °C.

Raman Spectroscopy

Raman spectra were recorded using a HoloLab Series 5000 Raman microscope (Kaiser Optical Systems). Individual samples (4 µL in the form of a hanging drop from a siliconized coverslip) were exposed to 100 mW of 647.1 nm laser excitation passed through the microscope’s 20× objective lens for 300 s. Calibration of the Raman microscope was done using neon and tungsten lamp standards, which indicate that the variation in the positions of individual Raman bands is less than 1 cm⁻¹. The measurement of all spectra was conducted at ambient temperature, which varied between 20 and 25 °C. Variation of the Raman signals between 20 and 25 °C for the model compounds used in these studies was indistinguishable from the observed experimental error at constant temperature. Spectral data were analyzed using GRAMS/AI (Thermo Galactic Corp.). RNAs (20 mg/mL) and DMP (200 mM) were measured in 200 mM cacodylate (pH 6) in the absence or presence of different ions as indicated. ATP and ADP were measured at a reduced pH (3.8, 200 mM formate) to ensure a single negative charge per phosphate to parallel that of the other phosphodiesters compared in this work.

The effect of ion binding on the Raman signal for nonbridging phosphate oxygens was examined by quantitative analysis of Raman difference spectra in which the Raman spectrum of a phosphodiester model compound in the absence of ion (subtrahend) is subtracted from that in its presence (minuend). Raman peaks that exhibited no detectable perturbation (<2%) from ion binding were used as intensity standards that varied depending on the phosphodiester analyzed. Specifically, minuend and subtrahend spectra were normalized using the parent (raw) Raman peaks centered at 1466 and 1453 cm⁻¹ for DMP, 842 cm⁻¹ for ATP and ADP, 812 cm⁻¹ for the GAAA tetraloop, and ~726 cm⁻¹ for all other RNAs. For ATP and all RNAs studied, the concentrations of the same samples used to generate minuend and subtrahend spectra were also analyzed by UV absorption in parallel to control for potential changes in concentration due to evaporation during Raman analysis. Magnesium hexahydrate [Mg²⁺ (H₂O)₆] was quantified from the intensity (photon counts) of the Raman peak for the Mg²⁺−O symmetric stretch (ν_sM−O) centered at 360 cm⁻¹ in the raw spectral data. Quantitative assessment of inner-sphere coordination was accomplished by measuring the observed Raman intensity in the positive node (PN) of the metal-induced inflection of ν_sPO₂⁻ in Raman difference spectra [ν_sPO₂⁻M^PN (see the shaded region in the inset of Figure 2B)]. Raman difference spectra underestimate the total signal for ν_sPO₂⁻M by ~20% due the overlap of ν_sPO₂⁻ and ν_sPO₂⁻M peaks in the Raman spectrum. The intensity of ν_sPO₂⁻M^PN, however, varies directly within experimental error with the total signal for ν_sPO₂⁻M and allows a more accurate measure of changes in the Raman spectrum at lower concentrations of metal ion than peak fitting of small changes in the raw spectral data. All Raman difference spectra were derived from data collected during the same experiment.

Figure 2.

Influence of electrostatics, H-bonding, and inner-sphere coordination interactions on ν_sPO₂⁻. Raman difference spectra (1200–1000 cm⁻¹) of ν_sPO₂⁻ from DMP (A), HATP³⁻ (B), and yeast tRNA^PHE (C), in the presence of MgCl₂ (black line), NaCl (gray line), NH₄Cl (black dashed line), Co(NH₃)₆Cl₃ (gray dashed line), or N(CH₃)₄Cl (black dotted line) at an equal ionic strength of 0.45 M. The ability of individual metal ions to interact with nonbridging phosphate oxygens by electrostatic interactions (E), hydrogen bonding (H), or direct coordination (C) is noted by the aforementioned letters adjacent to the ions shown in the legend of panel A. The right inset in panel B shows an enlargement of all metal-induced difference spectra with the exception of that from MgCl₂ to facilitate comparison of weaker ion-induced changes to the Raman signal of HATP³⁻. The left inset in panel B shows all metal-induced difference spectra, including that from MgCl₂. The shaded area in the positive node (PN) of metal ion difference spectra (Mg²⁺, left inset; Na⁺, right inset) reflects the shift of the ν_sPO₂⁻ signal to higher wavenumbers due to inner-sphere coordination defined by ν_sPO₂⁻M^PN. With the exception of metal ion concentration, experimental conditions are identical to those described in the legend of Figure 1.

Changes in the energetic difference between the Raman peaks for ν_sPO₂⁻ and ν_sPO₂⁻M were monitored via the distance between the inflection points of the positive and negative nodes of the metal-induced inflection of ν_sPO₂⁻ in Raman difference spectra, which we define as ΔνM [ΔνM = ν_sPO₂⁻M − ν_sPO₂⁻ (Figure 1D)]. Unlike the intensity of the positive node of the metal-induced inflection of ν_sPO₂⁻ in difference spectra, ΔνM overestimates the absolute difference between the peaks for ν_sPO₂⁻ and ν_sPO₂⁻M due to their overlap in the Raman spectrum. Metal-induced changes in the position of ν_sPO₂⁻M relative to ν_sPO₂⁻ are nevertheless reported directly through changes in the inflection points of the positive and negative nodes of the metal-induced inflection of ν_sPO₂⁻, yielding the value of ΔνM. Measurement of ΔνM is thus utilized as a rapid means of comparing the relative differences in the spectroscopic position of ν_sPO₂⁻ and ν_sPO₂⁻M due to the coordination of distinct metal ion species.

Figure 1.

Effect of Mg²⁺ on the Raman spectra of structurally distinct phosphodiesters. Raman spectral data (black) and Raman difference spectra (subtracting spectra at 0 M from 0.3 M MgCl₂, dark gray) for 200 mM DMP (A), 100 mM HATP³⁻ (B), and 0.81 mM (20 mg/mL) yeast tRNA^PHE (C). (D) Overlay of Mg²⁺ difference spectra (1200–1000 cm⁻¹) of the symmetric stretch for nonbridging phosphate oxygens (ν_sPO₂⁻) for DMP (gray dots), HATP³⁻ (gray line), and yeast tRNA^PHE (black). The arrow indicates the spectral distance between inflection points of the positive and negative nodes (marked by vertical black lines) of the Mg²⁺-induced inflection of ν_sPO₂⁻, which defines ΔνM. The inset shows a superposition of the metal-induced inflections shown in panel D to illustrate the relative similarity in the apparent shift of ν_sPO₂⁻ to higher wavenumbers. (E) Overlay of Mg²⁺ difference spectra (1200–1000 cm⁻¹) of ν_sPO₂⁻ for an 11-nucleotide single-stranded RNA (black dots), 12-nucleotide GAAA tetraloop (gray dots), 27-nucleotide bulged stem–loop sequence (black dashes), 76-nucleotide yeast tRNA^PHE (dark gray line), and 400-nucleotide E. coli RNase P RNA (black line). Dotted gray boxes indicate the metal-induced inflection of ν_sPO₂⁻ observed in difference spectra. DMP and RNAs were measured at pH 6 (200 mM cacodylate), while HATP³⁻ was measured at pH 3.8 (200 mM formate) to generate a single negative charge on the terminal phosphate.

Data Analysis

The dependence of binding of Mg²⁺ to ATP and ADP was determined from the area of the positive peak of Raman difference spectra measured for ATP and ADP at different Mg²⁺ concentrations and plotted using a standard binding equation appropriate for approximately equal concentrations of metal ion and PO₂⁻ ligand:

$$\begin{array}{cc}\text{ML}\hfill & =M_0{\text{FA}}^{\text{max}}\hfill \\ \hfill & =\left[L_0+M_0+K_{\mathrm{d}}-\sqrt{{(L_0+M_0+K_{\mathrm{d}})}^2-4M_0{L}_0}\right]/2\hfill \end{array}$$

where ML is the concentration of bound ligand, FA^max is fraction of the maximum ν_sPO₂⁻M signal observed, L₀ and M₀ are the initial concentrations of ligand and free metal, respectively, and K_d is the apparent dissociation constant. M₀FA^max values were normalized to unity to facilitate comparison. Areas of the positive peak of Raman difference spectra were determined using GRAMS/AI (Thermo Galactic Corp.). The experimental error of reported peak areas or position reflects the observed variation from at least three independent experiments.

RESULTS

Contribution of Inner-Sphere, H-Bonding, and Electrostatic Interactions to Metal-Induced Changes in the Raman ν_sPO₂⁻ Signal of Phosphodiesters

Comparative studies of phosphodiester–ion interactions were performed using three model systems representing a range of structural complexity from a simple phosphodiester, dimethyl phosphate (DMP), to multiple adjacent phosphates, protonated ATP (HATP³⁻, which maintains a single negative charge per phosphate), to a complex folded nucleic acid, yeast tRNA^PHE (Figure 1). In each case, ν_sPO₂⁻ is observed as a dominant spectroscopic feature at approximately the same position (~1100 cm⁻¹) and is relatively isolated from other vibrational modes in the Raman spectrum of nucleic acids such as those from nucleotide bases (~1200–1500 cm⁻¹) or the ribose–phosphate backbone [~600–900 cm⁻¹ (top black traces in Figure 1A–C)]. The strong similarity in these structurally distinct model systems is also observed for the intensity and shape of the ν_sPO₂⁻ peak. The observed ν_sPO₂⁻ signal thus is relatively insensitive to higher-order structure and is sufficient for quantitative analysis despite the relatively low signal-to-noise ratio compared to data obtained from crystals or surface-enhanced acquisition modes.

Metal-induced changes in the Raman spectrum were determined using difference spectra by subtraction of the spectrum of an individual phosphodiester from that in the presence of a metal ion (bottom gray traces, Figure 1A–C). When DMP, HATP³⁻, and tRNA were examined at the same concentration of Mg²⁺ (0.3 M) in solution, metal-induced changes throughout their respective Raman difference spectra are readily observed. Importantly, a similar metal ion-induced inflection near 1100 cm⁻¹ is observed for each phosphodiester and is consistent with the apparent shift of ν_sPO₂⁻ to higher wavenumbers upon metal ion binding to form an altered vibrational mode (ν_sPO₂⁻M) as observed in previous solution and crystal studies (12–14, 16). In this comparison of Raman spectra at a constant concentration of Mg²⁺, the total ionic strength (I) of the individual phosphodiesters was somewhat different (I^DMP = 1.1 M, I^HATP = 1.4 M, and I^tRNA = 1.1 M). Difference spectra compared at a constant ionic strength (Figure 2), however, are identical to those at a constant Mg²⁺ concentration.

An overlay of the Mg²⁺ difference spectra from DMP, HATP³⁻, and yeast tRNA^PHE (Figure 1D) indicates that structural differences influence the position and amplitude of the observed inflection in the ν_sPO₂⁻ portion of the spectra (1095 ± 1, 1131 ± 1, and 1106 ± 2 cm⁻¹ for DMP, HATP³⁻, and yeast tRNA^PHE, respectively). Superposition of the difference spectra, however, shows that there is little difference in the degree of metal-induced displacement of ν_sPO₂⁻ to higher wavenumbers (inset in Figure 1D). The difference in the relative position of ν_sPO₂⁻ and ν_sPO₂⁻M in the Raman spectrum defines the inflection points of the positive and negative nodes in the difference spectra. The apparent difference between the positive and negative nodes of difference spectra is defined as ΔνM [ΔνM = ν_sPO₂⁻M − ν_sPO₂⁻ (Figure 1D)]. Because of a systematic error inherent to subtracting peaks with overlapping intensity, the apparent ΔνM overestimates the intrinsic difference between the ν_sPO₂⁻ and ν_sPO₂⁻ M peaks by ~40%. Nonetheless, similar ΔνM values are observed for these structurally distinct phosphodiesters when they are compared using the same metal ion [ΔνM^Mg = 19 ± 1, 18 ± 1, and 24 ± 2 cm⁻¹ for DMP, HATP³⁻, and yeast tRNA^PHE, respectively (Figure 1D)]. This observation shows that structural differences have a much weaker effect on the differences in vibrational energy between ν_sPO₂⁻M and ν_sPO₂⁻ than the energy of the ν_sPO₂⁻ vibration itself.

In addition to these model systems, we also compared the Mg²⁺ difference spectra of RNAs differing in structural complexity (Figure 1E). Specifically, we compared an unstructured RNA oligonucleotide (11 nucleotides) with four structured RNAs including a GAAA tetraloop (12 nucleotides), a bulged stem–loop structure derived from the P4 helix of RNase P RNA (20) (27 nucleotides), yeast tRNA^PHE (76 nucleotides, from Figure 1D), and the complete RNA subunit of E. coli RNase P (400 nucleotides). The metal-induced changes in ν_sPO₂⁻ for polynucleotides are almost identical in position, amplitude, and degree of displacement of ν_sPO₂⁻ to higher wavenumbers to form ν_sPO₂⁻M. The similarity of metal-induced changes in ν_sPO₂⁻ in these structurally distinct RNAs suggests that structural context has little effect on either ν_sPO₂⁻ or the formation of ν_sPO₂⁻M in nucleic acids.

To determine the relative contribution of inner-sphere coordination, H-bonding, and electrostatic interactions to ion-induced changes in the ν_sPO₂⁻ region of the spectrum, we compared the Raman difference spectra of DMP, ATP, and yeast tRNA^PHE in the presence of metal and nonmetal ions that can interact variously via these three forms of chemical interaction (Figure 2). Specifically, we compared the effects of Mg²⁺ and Na⁺, which can form both inner- and outer-sphere interactions with nonbridging phosphate oxygens, to that of exchange inert cobalt hexamine [Co(NH₃)₆³⁺], a metal ion complex that mimics the overall structure of fully hydrated Mg²⁺ [Mg²⁺ (H₂O)₆] but is capable of only outer-sphere and electrostatic interactions (21). In addition, we examined the effect of ammonium (NH₄⁺) to monitor the influence of hydrogen bonding of an ion that differs from Co(NH₃)₆³⁺ in size, charge, and geometry, as well as the effect of tetramethylammonium [N(CH₃)₄⁺] to isolate electrostatic-induced changes in the νPO₂⁻ signal. To facilitate comparison of these distinct ions, we took initial measurements at equal ionic strengths (0.45 M). As expected from previous studies, ions capable of inner-sphere coordination (Mg²⁺ and Na⁺) produced characteristic inflections of ν_sPO₂⁻ in the Raman difference spectrum in DMP, ATP, and tRNA (Figure 2). In contrast, however, we found that ions restricted to outer-sphere and/or electrostatic interactions [Co(NH₃)₆³⁺, NH₄⁺, or N(CH₃)₄⁺] resulted only in the attenuation of the ν_sPO₂⁻ Raman signal near 1100 cm⁻¹.

Differences in the relative amplitude of Mg²⁺ - and Na⁺ - induced changes in ν_sPO₂⁻ in different model compounds appear to reflect differences in the relative affinity of individual ions. The Na⁺-induced changes in the ν_sPO₂⁻ region of the spectra of ATP are significantly smaller than those observed for Mg²⁺. ATP forms a tight (K_d ~ 10⁻⁵ M) stoichiometric complex with Mg²⁺ via interactions involving one or more of the phosphate oxygens, which strongly favors binding of the divalent ion (22–24). In contrast, Na⁺-induced inflections of ν_sPO₂⁻ were found to be nearly equal in magnitude to that of Mg²⁺ in DMP and only somewhat reduced in magnitude relative to the magnitude of changes in the spectrum due to Mg²⁺ in the context of yeast tRNA^PHE, where only a few of the nonbridging phosphate oxygens are involved in site-specific divalent metal ion binding.

The absence of the induction of the ν_sPO₂⁻M vibrational mode by Co(NH₃)₆³⁺, NH₄⁺, or N(CH₃)₄⁺ is consistent with their inability to form inner-sphere coordination interactions with phosphodiesters. Alternatively, the same negative result could be due to low relative affinities of these ions for diesters, nucleotides, and nucleic acids relative to Na⁺ and Mg²⁺. Cobalt hexamine, however, binds to nucleic acids with affinities similar to that observed for Mg²⁺ (25). To test whether this is the case under our experimental conditions, we measured the intensity (photon counts) of the positive node (PN) of the metal-induced HATP³⁻ ν_sPO₂⁻ difference spectra [ν_sPO₂⁻M^PN (shaded area in in the inset of Figure 2B)] as a function of Mg²⁺ concentration in the presence and absence of Co(NH₃)₆³⁺ (Figure S1 of the Supporting Information). We find that Co(NH₃)₆³⁺ can compete with Mg²⁺ for ATP binding with a concentration dependence that indicates a similar affinity. Binding of Co(NH₃)₆³⁺ is also implied by the attenuation of ν_sPO₂⁻ in the presence of Co(NH₃)₆³⁺ alone (Figure 2B). Indeed, for all phosphodiesters tested to date, binding of Co(NH₃)₆³⁺ results in attenuation of the intensity of the Raman signal from the ν_sPO₂⁻ vibrational mode without formation of the ν_sPO₂⁻M mode associated with inner-sphere coordination (E. L. Christian, unpublished observations). Similarly, ν_sPO₂⁻ attenuation in the absence of formation of ν_sPO₂⁻M is observed in the presence of NH₄⁺ or N(CH₃)₄⁺ at an ionic strength equivalent to that examined for MgCl₂ [450 mM (Figure 2)] and near the limit of NH₄⁺ solubility [4 M (data not shown)]. Thus, the ion-induced attenuation of ν_sPO₂⁻ appears to result from electrostatic interactions, H-bonding, and likely inner-sphere coordination, while the formation of ν_sPO₂⁻M reflects purely inner-sphere coordination.

Testing the Dependence of ν_sPO₂⁻M on Inner-Sphere Coordination by a Metal Ion Specificity Switch

To further understand the chemical basis for the attenuation of ν_sPO₂⁻ and formation of ν_sPO₂⁻M, we compared the Raman spectra of dimethyl phosphate (DMP) and dimethyl thiophosphate (DMTP) in the presence of either an oxophilic ion, Mg²⁺, or a thiophilic ion, Cd²⁺ (Figure 3). It is well-established that sulfur substitution of a nonbridging phosphate oxygen strongly inhibits inner-sphere coordination of oxophilic ions such as Mg²⁺ (31000-fold preference for oxygen over sulfur in ATPβS) but allows inner-sphere coordination of thiophilic ions such as Cd²⁺ (60-fold preference for sulfur over oxygen in ATPβS) (22, 23, 26). Thus, the extent to which Mg²⁺ and Cd²⁺ induce changes in the Raman spectrum of DMP and DMTP should provide an additional test for the linkage among inner-sphere coordination, attenuation of ν_sPO₂⁻, and formation of ν_sPO₂⁻M.

Figure 3.

Dependence of ν_sPO₂⁻M on direct metal ion coordination. (A) Raman spectrum (1140–1000 cm⁻¹) of ν_sPO₂⁻ from 0.2 M DMP in the absence of metal (black) or presence of 1 M Mg²⁺ (gray dots) or 1 M Cd²⁺ (gray line). (B) Difference spectrum subtracting DMP in the absence of metal from that in the presence of 1 M Mg²⁺ (gray line) or 1 M Cd²⁺ (dark gray). (C) Raman spectrum (680–540 cm⁻¹) of isolated νPOS⁻ from 0.2 M DMTP in the absence of metal (black) or presence of 1 M Mg²⁺ (light gray) or 1 M Cd²⁺ (dark gray). (D) Difference spectrum subtracting DMTP in the absence of metal from that in the presence of Mg²⁺ (light gray) or 1 M Cd²⁺ (dark gray). (E) Raman spectrum (1300–950 cm⁻¹) showing both symmetric (ν_sPO₂⁻) and asymmetric (ν_aPO₂⁻) stretches for non-bridging phosphate oxygens from 0.2 M DMP in the absence of metal (black) or presence of 1 M Mg²⁺ (light gray) or 1 M Cd²⁺ (dark gray). (F) Enlargement of the region of the Raman spectrum shown as a dotted box in panel E showing DMP in the absence of metal ion (black) or in the presence of 1 M Mg²⁺ (light gray) or 1 M Cd²⁺ (dark gray). The assignments of ν_sPO₂⁻, νPOS⁻, and νPO have been previously established in the literature (27, 43), and their associated metal-dependent shifts were confirmed by ¹⁸O substitution (data not shown). (G) Raman spectrum (1300–900 cm⁻¹) near the phosphoryl vibrational mode (νPO) from 0.2 M DMTP in the absence of metal (black) or presence of 1 M Mg²⁺ (light gray) or 1 M Cd²⁺ (light gray). Additional peaks of bridging O–P–O (νPO^3′5′) and C–O stretch (νCO) are shown for reference. (H) Difference spectrum of the region shown in panel G subtracting DMTP in the absence of metal from that in the presence of 1 M Mg²⁺ (light gray) or 1 M Cd²⁺ (dark gray). Additional vibrational modes in DMP and DMTP are provided for reference: ν_sCO and ν_aCO reflect the symmetric and asymmetric C–O stretch, respectively, and b_sCH₃ and b_aCH₃ reflect the symmetric and asymmetric bending modes of the methyl group, respectively (27, 43).

In contrast to the strong Mg²⁺-induced attenuation of ν_sPO₂⁻ and formation of ν_sPO₂⁻M in the Raman difference spectrum of DMP, similar spectral changes are not induced by Mg²⁺ in the corresponding νPOS⁻ vibrational mode in DMTP, shown in previous studies to form two characteristic signals at 607 and 638 cm⁻¹ (27) and confirmed in the current work by ¹⁸O substitution (data not shown) (compare panels A and B of Figure 3 with panels C and D). Significant attenuation of the intensity of the νPOS⁻ modes in DMTP, however, is observed in the presence of Cd²⁺, but not for ν_sPO₂⁻ in DMP. In addition, new metal ion-dependent vibrational modes are observed down-field from the unperturbed νPOS⁻ in the presence of Cd²⁺, while no significant ν_sPO₂⁻M formation is evident in DMP. Consistent with the observations described above, Ca²⁺ and Mn²⁺, which show a strong preference for oxygen over sulfur (31000–39000-and 158–193-fold, respectively, in ATPβS), also produce metal-dependent shifts and attenuation of ν_sPO₂⁻ of DMP but not DMTP, while Co²⁺ and Zn²⁺, which have been shown to have only a weak preference for sulfur over oxygen (~1.2- and 4.6-fold, respectively, in AMPS), produce metal-dependent shifts and attenuation of ν_sPO₂⁻ in both DMP and DMTP (22, 28).

Metal ion specificity is also evident in the vibrational modes coupled to ν_sPO₂⁻ and νPOS⁻. In DMP, metal ion coordination is predicted to alter the asymmetric (ν_aPO₂⁻) and symmetric (ν_sPO₂⁻) vibrational modes of the nonbridging phosphate oxygens because the atoms involved are identical. Consistent with this prediction, ν_aPO₂⁻ is shifted to higher wavenumbers in the presence of Mg²⁺, but not in the presence of Cd²⁺ (Figure 3E,F). Similarly, in DMTP, metal ion coordination to sulfur is predicted to alter the bond order and vibrational properties of the phosphoryl (P═O, νPO) as well as the adjacent P–S bond since these vibrational modes share the same phosphorus atom. Indeed, significant perturbation of νPO is observed for DMTP in the presence of Cd²⁺, but not in the presence of Mg²⁺ (Figure 3G,H). Taken together, the dependence of the metal-induced shift of ν_sPO₂⁻, νPOS⁻, and their coupled vibrational modes on oxophilic or thiophilic ions strongly supports the interpretation that these spectroscopic signals arise from direct inner-sphere coordination, consistent with the findings described above.

Thermodynamic Relationship between Saturable Ion Binding and ν_sPO₂⁻M

The utility of ν_sPO₂⁻M as a quantitative probe for inner-sphere coordination interactions is directly related to the extent to which this spectroscopic feature can be used to accurately measure metal binding thermodynamics. A simple and well-characterized system for studying inner-sphere metal ion binding is provided by HATP³⁻ which binds a single Mg²⁺ with high affinity through interactions with the nonbridging phosphate oxygens and maintains a single negative charge per phosphodiester (22–24).We therefore determined the extent to which increasing concentrations of Mg²⁺ resulted in concentration-dependent and saturable changes in the intensity of ν_sPO₂⁻M^PN. We also compared changes in ν_sPO₂⁻M^PN to the intensity of the Raman signal for magnesium hexahydrate [Mg²⁺(H₂O)₆] which can be detected by its symmetric stretching frequency at 360 cm⁻¹ [ν_sM–O (Figure 4)]. As noted above, loss of the Raman signal for fully hydrated magnesium and formation of magnesium penta- or tetrahydrate have been shown to correlate with increases in the intensity of ν_sPO₂⁻M in DMP and crystals of HDV ribozyme RNA (12). If ν_sPO₂⁻M represents stoichiometric binding of Mg²⁺ to ATP in a site-bound or chelated mode, then its intensity should follow predictable saturation binding thermodynamics with respect to Mg²⁺ concentration. In addition, inner-sphere coordination of Mg²⁺ by ATP must necessarily lead to a stoichiometric decrease in ν_sM−O.

Figure 4.

Inner-sphere coordination induces ν_sPO₂⁻M formation and ν_sM–O attenuation. Raman difference spectrum of ATP (60 mM) in the presence of excess MgCl₂ (200 mM), indicating the relative positions of the Mg²⁺-induced changes in ν_sPO₂⁻ (~1100 cm⁻¹) and magnesium hexahydrate (ν_sM–O, 360 cm⁻¹, structure at the right). Black arrows reflect previously observed correlations of Mg²⁺ binding (structure at the left) with an increased ν_sPO₂⁻ inflection and a decreased ν_sM−O intensity in DMP solution studies and crystals of the HDV ribozyme (12).

As shown in Figure 5, saturable binding behavior can be observed in the Mg²⁺ concentration dependence of the intensity of ν_sPO₂⁻M^PN under conditions where the ionic strength varies with the concentration of the added metal ion. Importantly, the apparent equilibrium constant for the Mg²⁺–HATP³⁻ complex determined by measuring ν_sPO₂⁻M^PN (log K^a = 2.8 ± 1.0 M⁻¹) is equal to that observed by NMR under similar conditions (log K^a = 2.79 ± 0.15 M⁻¹) (22). Furthermore, because the nucleotide concentration is in excess of the dissociation constant for Mg²⁺ ion binding, the stoichiometry for ion binding detected by ν_sPO₂⁻M^PN can be estimated from the intersection of lines tangent to the initial and saturating phases of the binding curve. The intersection observed for ATP–Mg²⁺ binding extrapolates to 100 mM MgCl₂ on the x-axis (Figure 4A), and the apparent stoichiometric ratio of 100 mM Mg²⁺ to 300 mM phosphate or 100 mM ATP is consistent with the binding of a single metal ion by ATP as observed in previous studies.

Figure 5.

Use of ν_sPO₂⁻M^PN to monitor site-specific metal ion binding. (A) Dependence of 100 mM protonated ATP (HATP³⁻) ν_sPO₂⁻M^PN on MgCl₂ concentration (millimolar). Black circles reflect the normalized upshifted ν_sPO₂⁻ Raman signal (photon counts) derived from the positive node of the Raman difference spectra (ν_sPO₂⁻M^PN). The black curve reflects the fit of ν_sPO₂⁻M^PN data to the binding isotherm appropriate for approximately equal concentrations of HATP³⁻ and MgCl₂ (see Methods). Solid black lines reflect tangents to the initial and saturating phases of the binding curve. The dotted black line reveals the MgCl₂ concentration corresponding to the intersection of tangent lines above and the stoichiometry of Mg²⁺ binding to HATP³⁻. (B) Dependence of the HATP³⁻ adenine Raman peak at 1310 cm⁻¹ (νA₁₃₁₀) on MgCl₂ concentration. Black squares reflect the normalized upshifted νA₁₃₁₀ Raman signal derived from the positive node of the Raman difference spectra (νA₁₃₁₀^PN). The gray ovals and curve reflect the normalized Mg²⁺ dependence of the increase in νPO₂⁻M^PN as described for panel A. (C) Dependence of the Raman peak for fully hydrated Mg²⁺ [Mg(H₂O)₆²⁺, νMg–O, 360 cm⁻¹] on Mg²⁺ concentration in the absence (gray) and presence of 100 mM HATP⁻³ (black). Solid lines reflect the linear least-squares fit to all points in the absence (gray) or presence (black) of HATP³⁻. Dotted lines reflect the least-squares fit to points above 200 mM MgCl₂, where Mg²⁺ binding to HATP³⁻ is observed to be saturating. (D–F) Repeats of experiments depicted in panels A–C, respectively, but in the presence of 100 mM protonated ADP (HADP²⁻). Protonated forms of ATP and ADP were chosen to parallel the single negative charge per phosphate of the other phosphodiesters compared in this work.

To test whether the metal dependence of ν_sPO₂⁻M^PN intensity was specific to the nonbridging phosphate oxygens and distinguishable from metal-dependent changes in other vibrational modes in the Raman spectrum, we monitored Mg²⁺-dependent changes in the vibrational modes of the adenine base. Base vibrational modes are significantly more sensitive to metal-induced changes in geometry than ν_sPO₂⁻ (13). Indeed, as shown in Figure 1B, the presence of Mg²⁺ produces significant perturbation of the adenosine ring of ATP modes between 1200 and 1500 cm⁻¹ in the Raman spectrum. Previous studies show that Mg²⁺ binding to ATP and ADP can involve simultaneous inner-sphere coordination with nonbridging phosphate oxygens and outer-sphere coordination with N7 of adenine, which raises the possibility of coupled metal ion dependence for the vibrational modes associated with these two functional groups (29–32). We therefore measured the positive node of the metal-induced shift in the Raman signal for the adenine ring mode at ~1310 cm⁻¹ for ATP (νA₁₃₁₀^PN) as a function Mg²⁺ concentration and compared it to that observed for ν_sPO₂⁻M^PN (Figure 5B). Changes in the intensity of νA₁₃₁₀^PN as a function of Mg²⁺ concentration are clearly distinct from that observed for ν_sPO₂⁻M^PN. In contrast to the saturable binding behavior of ν_sPO₂⁻M^PN, the metal-dependent change in νA₁₃₁₀^PN intensity shows at least two distinct phases with no sign of apparent saturable binding behavior over the same Mg²⁺ concentration range. Metal-dependent changes in ν_sPO₂⁻M^PN thus are only weakly coupled to changes in the adenosine base, and quantification of its intensity allows isolation of metal ion interactions of the nonbridging phosphate oxygens.

To further test the extent to which ν_sPO₂⁻M^PN correlates quantitatively with inner-sphere metal ion coordination, we determined the concentration dependence of the intensity of the Raman peak for fully hydrated Mg²⁺ (ν_sM–O) in the absence and presence of 100 mM ATP (Figure 5C). The high concentration of ATP relative to the dissociation constant for Mg²⁺ ensures that essentially all metal ions are bound at substoichiometric concentrations (e.g., < 50 mM) of Mg²⁺. Under these conditions ν_sM–O is readily and quantitatively detected in the absence of ATP. However, in the presence of 100 mM ATP, the expected linear increase in ν_sM–O is shifted to higher MgCl₂ concentrations with an x-axis intercept between 50 and 100 mM Mg²⁺. These data show that fully hydrated Mg²⁺ is undetectable between 0 and 50 mM MgCl₂ in the presence of 100 mM ATP, the same concentration range over which the intensity of ν_sPO₂⁻M^PN has an essentially linear concentration dependence. The x-axis intercept between 50 and 100 mM Mg²⁺ observed for ν_sM–O in the presence of 100 mM ATP also provides further evidence of a 1:1 stoichiometry of the Mg²⁺–ATP complex (22–24).

Previous isotope (¹⁸O) substitution experiments designed to detect interactions with individual phosphate oxygens in ATP indicate that metal ion binding α-, β-, or γ-phosphates all participate in the formation of the ν_sPO₂⁻M vibrational mode (33). Metal complexes with ATP, however, include a mixture of mono- or multidentate (α–β, β–γ, or α–β–γ) phosphate–metal complexes in which individual ν_sPO₂⁻M vibrational modes are likely to be coupled (33). This observation raises the possibility that the complexity of phosphate–metal complexes in ATP might preclude the use of this model system to monitor metal-dependent changes in ν_sPO₂⁻. To examine the extent to which complex heterogeneity and vibrational mode coupling affect the observed metal-dependent changes in ν_sPO₂⁻M^PN seen in site-specific ion binding, we repeated the Mg²⁺ titration studies above using the proton form of ADP (HADP²⁻) (Figure 5D–F). HADP²⁻ forms a more structurally homogeneous population with metal contacts to both α- and β-phosphates at sufficient affinity (log K^a = 2.94 ± 0.14 M⁻¹) to allow stoichiometric binding of Mg²⁺ (22). Increasing concentrations of Mg²⁺ resulted in an increasing intensity of ν_sPO₂⁻M^PN that followed saturable binding and stoichiometric behavior analogous to that observed for ATP (Figure 5D). The saturable binding behavior of ADP ν_sPO₂⁻M^PN is clearly distinct from the metal ion concentration-dependent changes in the Raman signal for the adenosine base at ~1310 cm⁻¹ [νA₁₃₁₀^PN (Figure 5E)]. We also find that fully hydrated Mg²⁺ is undetectable between 0 and 50 mM MgCl₂ in the presence of 100 mM ADP, the same concentration range over which the intensity of metal-induced ν_sPO₂⁻M^PN in ADP has an essentially linear concentration dependence (Figure 5D,F). Different mixtures of mutidentate phosphate–metal complexes that likely include different degrees of vibrational coupling between phosphates thus do not appear to contribute significantly to the observed metal-dependent changes in the Raman spectrum of ATP and ADP. Together, these data strongly support the interpretation that inner-sphere coordination between metal ion and nonbridging oxygen can be quantified by integration of the intensity of ν_sPO₂⁻M^PN and that this signal is likely to be applicable to the solution analysis phosphodiesters in general.

Effects of Ion Identity on ν_sPO₂⁻M

The thermodynamic correlations between the intensity of ν_sPO₂⁻M^PN and metal ion binding via inner-sphere coordination interactions predict that this spectroscopic signal should be strongly influenced by the physical properties of different metal ions. Absolute electro-negativity reflects an atom’s ability to attract electrons toward itself in a chemical bond and therefore is likely to strongly correlate with the vibrational properties of ν_sPO₂⁻M (34). We therefore measured the degree of the ν_sPO₂⁻ shift to higher wavenumbers (ΔνM) reflected in the metal-induced inflection observed in Raman difference spectra of DMP in the presence of different metal ions at a constant (3 M) ionic strength (Figure 6). Examples of individual metal-induced changes to ν_sPO₂⁻ in DMP Raman spectra are shown in Figures S2 and S3 of the Supporting Information.

Figure 6.

Dependence of ΔνM on absolute electronegativity. Magnitudes of the metal-induced shift of ν_sPO₂⁻ to higher wavenumbers (ΔνM) of 0.2 M DMP in the presence of alkali (▲), alkaline earth (◆), and transition (■) metals at equal ionic strengths (3 M) plotted vs measured values of metal ion absolute electronegativity as reported by Pearson (34).

When ΔνM (in cm⁻¹) is plotted as a function of absolute electronegativity, the behavior of individual ions falls into two groups. Alkali and alkaline earth metals show a linear dependence of the magnitude of ΔνM on electronegativity, while transition metals do not follow this trend (Figure 6). Two transition metals (Cd²⁺ and Ni²⁺) produced no significant ν_sPO₂⁻ shift to higher wavenumbers, presumably due to low levels of inner-sphere coordination to DMP (data not shown). The same distinction between types of metal ions is observed when ΔνM is plotted versus the related physical property of absolute hardness (Figure S4 of the Supporting Information). In the current analysis of the comparison of ΔνM values for different metal ions at constant ionic strengths, the metal: phosphodiester molar ratio will differ by approximately 3-fold between mono- and divalent metal ions. The parameter ΔνM, however, is essentially insensitive to differences in the metal: phosphodiester molar ratio as no significant change in ΔνM is observed over a broad ion concentration range (e.g., 1–5 M Na⁺ or 0.15–2 M Mg²⁺) for all metal ions shown in Figure 6 (Figure S5 of the Supporting Information). A linear fit of ΔνM as a function of metal ion concentration yields little to no slope, indicating that the metal-induced shifts of ν_sPO₂⁻, and thus ΔνM, reflect distinct physical properties of the individual metal–phosphate interactions. In particular, the data given above indicate for alkali and alkaline earth metals, and thus many of the metals used in the study of phosphodiester chemistry and nucleic acids structure, that absolute electronegativity and absolute hardness contribute significantly to the vibrational energy of ν_sPO₂⁻M and are strong predictors of the observed degree of the ν_sPO₂⁻ shift to higher wavenumbers.

The differences in the magnitude of ΔνM for different metal ions relative to the observed experimental error (≤ 1 cm⁻¹) suggest that this physical parameter could be used to identify or distinguish between individual metal ion species. Na⁺ and Mg²⁺, for example, are clearly resolved in the structurally distinct model systems of DMP (ΔνM^Na = 12.8 ± 0.6 cm⁻¹, and ΔνM^Mg = 19.7 ± 0.8 cm⁻¹), ATP (ΔνM^Na = 13 ± 1 cm⁻¹, and ΔνM^Mg = 18 ± 1 cm⁻¹), and yeast tRNA^PHE (ΔνM^Na = 16 ± 2 cm⁻¹, and ΔνM^Mg = 24 ± 2 cm⁻¹) (Figure 2A–C). We therefore further tested the generality of this effect by measuring the extent to which Na⁺ and Mg²⁺ binding could still be resolved in RNAs differing in complexity. Table 1 compares ΔνM induced by Na⁺ and Mg²⁺ in DMP and ATP with that observed in a single-stranded RNA oligonucleotide (11 nucleotides), a GAAA tetraloop (12 nucleotides), a bulged stem–loop structure (27 nucleotides), yeast tRNA^PHE (76 nucleotides), and E. coli RNase P RNA (400 nucleotides). We find that the ability to distinguish ΔνM values induced by Na⁺ from that induced Mg²⁺ is maintained in all model systems tested. Although the magnitude of the difference in the ΔνM induced by Na⁺ relative to Mg²⁺ is somewhat variable between different model compounds, this difference is well within the magnitude of the somewhat larger experimental error (~2–3 cm⁻¹) observed for RNA samples (Table 1). These data indicate that ΔνM is likely to be broadly useful for the characterization or distinction between individual metal ion species interacting with the phosphoryl oxygens of nucleotides, nucleic acids, and other phosphodiester model compounds.

Table 1.

Effect of Structure on ΔνM^a

model system	no. of nucleotides	ΔνM for Na+	ΔνM for Mg2+
DMP		12.8 ± 0.6	19.7 ± 0.8
ATP	1	13 ± 1	18 ± 1
single-stranded RNA	11	16 ± 3	24 ± 2
GAAA tetraloop	12	18 ± 2	25 ± 2
P4 hairpin	27	15 ± 3	23 ± 2
yeast tRNAPHE	76	16 ± 2	24 ± 2
RNase P RNA	400	16 ± 3	24 ± 2

^aMeasured ΔνM induced by Na⁺ (3 M) or Mg²⁺ (1 M) in 0.2 M DMP, 100 mM ATP, or 20 mg/mL 11-nucleotide single-stranded RNA (ssRNA), 27-nucleotide RNA hairpin from the P4 NMR fragment of RNase P RNA (20), 76-nucleotide yeast tRNA^PHE, or 400-nucleotide RNase P RNA. All samples were analyzed as described in the legend of Figure 1.

DISCUSSION

Metal-induced changes to the symmetric stretch of nonbridging phosphate oxygens in the Raman spectrum (ν_sPO₂⁻) have long been used as an experimental signal for metal binding in phosphodiester model compounds and nucleic acids and have recently been proposed as a semiquantitative measure of inner-sphere coordination (12–15, 33). Missing from current methods, however, is an understanding of the extent to which phosphodiester structure and indirect forms of metal ion interaction such as H-bonding and electrostatic effects contribute to the observed metal-dependent changes in the Raman spectra of phosphodiesters. Such an understanding is necessary for quantitative interpretation of the Raman spectra in terms of ion interactions. In this work, we find that the attenuation of ν_sPO₂⁻ is due to all three forms of chemical interaction, while the characteristic shift of ν_sPO₂⁻ to higher wavenumbers in Raman difference spectra can be induced only by ions capable of inner-sphere coordination of the nonbridging phosphoryl oxygens. We also show that the shift of ν_sPO₂⁻ to higher wavenumbers is dependent on metal ion identity and may be used to estimate the distribution of ions binding in mixed metal ion solutions. These findings are observed in structurally distinct model compounds and thus appear to reflect the spectroscopic behavior of metal ion binding to phosphodiesters in general.

The observations described above provide insight into the physical basis for differences in the metal-dependent changes in ν_sPO₂⁻ by different metal ions in difference spectra from different phosphodiester model systems. In DMP, which lacks the structural complexity to form high-affinity complexes with individual ions, the extent of formation of ν_sPO₂⁻M^PN induced by Mg²⁺ is only several fold larger than that induced by Na⁺, while the extents of attenuation of ν_sPO₂⁻ induced by Mg²⁺ and Na⁺ are essentially equal (Figure 2A). Attenuation induced by Co-(NH₃)₆³⁺ and NH₄⁺ (H-bonding) and N(CH₃)₄⁺ (electrostatic interactions) are roughly equal in intensity with each other and are approximately half the intensity induced by Mg²⁺ or Na⁺. Although quantitative comparison of weakly bound ion pairs is problematic, qualitatively these data show that for a single phosphodiester ion-induced changes in ν_sPO₂⁻ can involve both shifts to higher wavenumbers from inner-sphere coordination and significant attenuation by electrostatic (and possibly H-bonding) interactions. The data also suggest that the relative contributions of direct and indirect interactions from Mg²⁺- and Na⁺-induced changes in ν_sPO₂⁻ are not equal. While similar levels of Mg²⁺- and Na⁺-induced attenuation of ν_sPO₂⁻ reflect similar total levels of metal–phosphate interaction with DMP, a proportionally higher intensity of ν_sPO₂⁻M^PN induced by Mg²⁺ suggests a greater fraction of inner-sphere coordination interactions than that with Na⁺.

Differences in the relative contribution of inner-sphere versus H-bonding/electrostatic interactions are increased dramatically in the context of the high-affinity Mg²⁺ binding site created by HATP³⁻, where Mg²⁺-induced formation of ν_sPO₂⁻M^PN and attenuation of ν_sPO₂⁻ dwarf those observed by other ions (Figure 2B). In addition, the intensity of Mg²⁺-induced ν_sPO₂⁻M^PN relative to attenuation of ν_sPO₂⁻ is much larger than that observed in DMP. These data reveal significantly enhanced levels of inner-sphere coordination in the context of site-specific binding relative to that observed for an unstructured phosphodiester. In contrast, the relative levels of ion-induced changes caused by Na⁺, Co(NH₃)₆³⁺, NH₄⁺, and N(CH₃)₄⁺ in ATP are roughly similar to those observed in DMP (Figure 2B, right inset). The level of Na⁺-induced ν_sPO₂⁻M^PN induction relative to ν_sPO₂⁻ attenuation is nevertheless significantly larger in ATP than that observed in DMP, indicating a greater degree of inner-sphere coordination in the context of the polyphosphate. Similarly, the level of attenuation of ν_sPO₂⁻ by Co(NH₃)₆³⁺ is nearly equal to or somewhat reduced relative to that observed for NH₄⁺, and N(CH₃)₄⁺ in DMP, but larger than the observed levels of ν_sPO₂⁻ attenuation by NH₄⁺, and N(CH₃)₄⁺ in ATP, consistent with the measured similarity of Mg²⁺ and Co(NH₃)₆³⁺ binding to ATP observed in the competition studies described above (Figure S1 of the Supporting Information).

Interestingly, ion-dependent changes in ν_sPO₂⁻ in yeast tRNA^PHE indicate enhanced levels of inner-sphere metal ion coordination relative to DMP (compare panels A and C of Figure 2). Specifically, the intensity of Mg²⁺-induced ν_sPO₂⁻M^PN relative to attenuation of ν_sPO₂⁻ is much larger than that observed in DMP and is likely to reflect a higher binding free energy due to the higher charge density of the polyanion. Alternatively, the greater intensity of ν_sPO₂⁻M^PN for tRNA^PHE could correlate with site-specific metal ion binding. However, the same ratio of ν_sPO₂⁻M^PN to ν_sPO₂⁻ attenuation is observed for a short (11-nucleotide) single-stranded RNA oligonucleotide and other structurally distinct RNAs (Figure 1E). The insensitivity of the ratio of ν_sPO₂⁻M^PN to ν_sPO₂⁻ between tRNA^PHE and different RNA structures suggests that ν_sPO₂⁻M^PN largely reflects diffuse metal ion binding. Consistent with this observation, the relative intensities of ν_sPO₂⁻M^PN induced by Mg²⁺ and Na⁺ in tRNA^PHE are similar to that observed for DMP (Figure 2A) despite multiple site-bound Mg²⁺ ions in tRNA^PHE that will be fully occupied at 150 mM Mg²⁺ (35–37). The data described above suggest that diffuse interactions by mono- and divalent metal ions in nucleic acids are also involved in low but measurable levels of inner-sphere coordination. Thus, while important qualitative information can be gained by analysis of changes in ν_sPO₂⁻M^PN for RNA, significant additional investigation will be necessary to understand how to interpret these data quantitatively.

Nonetheless, the data presented here provide insight into the physical basis of the shift of ν_sPO₂⁻ to higher wavenumbers to form ν_sPO₂⁻M and suggest a potential means of determining the distribution of interacting ions in mixed metal ion environments. The inner-sphere coordination interaction that gives rise to ν_sPO₂⁻M predicts that the strength of the metal–nonbridging phosphate interaction should contribute significantly to the frequency of this vibrational mode. Different degrees of metal-induced shifts of ν_sPO₂⁻ to higher wavenumbers for different metal ions have been reported in the literature (13, 14); however, no systematic analysis or correlation to absolute electronegativity has been reported previously. As noted above, we observe a linear dependence of the degree of the ν_sPO₂⁻ shift to higher wavenumbers (ΔνM) on the absolute electronegativity and the absolute hardness of alkali and alkaline earth metals (Figure 6 and Figure S2 of the Supporting Information). Importantly, the differences in ΔνM values for individual metal ions are often in excess of experimental error and could be used to identify or distinguish between individual metal ions. With the exception of that of Li⁺, differences in ΔνM are not sufficient to distinguish between individual monovalent ions. However, differences in ΔνM are large enough to distinguish between some divalent ions (e.g., Mg²⁺ and Ca²⁺) and more generally between mono- and divalent ions. The ability to identify a specific metal ion type is particularly important under conditions where multiple metal ion species are present. Biologically active forms of many structural and catalytic RNAs, for example, involve the binding of both mono- and divalent metal ions, such as Na⁺ and Mg²⁺. We find that ΔνM features for Na⁺ and Mg²⁺ are easily resolved in all phosphodiester model systems tested from the simple phosphodiester model system of DMP through the full range of structurally distinct RNAs. The measured values of ΔνM in the model systems mentioned above, nevertheless, represent the average extents of metal-induced ν_sPO₂⁻ shifts of all phosphodiesters in the sample and do not provide information about an individual phosphate position. The Raman signal, however, can provide site-specific information when combined with other biochemical information and parallel Raman analysis of differently modified compounds (38).

The concentration of metal ion required to determine ΔνM varies with the metal affinity of the individual phosphodiester compound. High metal ion concentrations (~1–3 M) were required to measure ΔνM in DMP, while much lower ion concentrations (e.g., 5 mM Mg²⁺ or 100 mM Na⁺) can be used to measure ΔνM in polynucleotides such as the 12-nucleotide GAAA tetraloop (Figure 6, data not shown). This finding indicates that the Raman spectroscopic approach described above can be used to monitor metal–phosphate interactions at or near physical levels of ion concentration.

The data presented here also indicate the need for additional caution in interpretation of metal–phosphodiester interactions from spectroscopic features in addition to ν_sPO₂⁻M^PN. Attenuation of ν_sPO₂⁻, for example, is evident in the Raman spectra of all metal ions we have tested, including transition metals. Our findings, however, indicate that electrostatic and H-bonding interactions can contribute significantly to the attenuation of ν_sPO₂⁻ (Figure 2) and may limit quantitative interpretation of inner-sphere metal ion binding using this spectroscopic signal. In addition, we observed, like other studies, that metal ion binding could also be monitored by changes in the Raman spectrum of a nucleotide base (Figure 5B,E). We find, however, that metal-dependent changes in the Raman signal for the adenine base are far more complex than those observed for ν_sPO₂M. Specifically, they reveal additional influences of structural changes or electrostatic effects beyond saturable ligand binding and should be carefully interpreted even for simple model systems such as ATP or ADP.

In summary, the data presented above enhance the quantitative and interpretive power of Raman analysis of metal–phosphodiester interactions. In particular, these findings help to describe the physical basis for the observed metal-dependent changes in ν_sPO₂⁻ that occur in different phosphodiester compounds and in the presence of different metal ions in solution. Such an understanding increases the ability to interpret the extent to which direct and indirect interactions contribute to individual metal ion interactions and, in some cases, to identify or distinguish between individual metal ion species.

It is important to keep in mind that Raman spectroscopic analysis in solution typically requires the ability to obtain samples that are both concentrated (e.g., 20 mg/mL RNA) and highly purified. Practical application of Raman spectroscopy for the quantitative analysis of metal–phosphate interactions in solution is thus currently best suited for the study of relatively small and structurally well-defined model systems such as the ones used in this study. Within this experimental constraint, however, many important questions regarding the nature of metal–phosphate interactions can now be addressed quantitatively. In particular, the Raman spectroscopic method described above provides a potential means of characterizing the distribution of ion binding interactions to RNA and thus information relevant to the ion-dependent folding of RNA and the stabilization of its t hree-dimensional structure. Quantitative computational studies of ion binding using nonlinear Poisson–Boltzmann theory (8, 39) along with direct ion counting methods involving equilibrium dialysis in combination with dye binding fluorescence and atomic emission spectroscopy (40–42) have been used to define the counterion atmosphere as a diffuse continuum up to ≥ 10 Å from the nucleic acid surface. Raman spectroscopy provides an opportunity to test and refine these models by measuring the fraction of the counterion atmosphere directly bound to the negatively charged phosphate backbone.