Environment- and Conformation-Induced Frequency Shifts of C–D Vibrational Stark Probes in NAD(P)H Cofactors

Abstract

NAD(P)H cofactors are found in all forms of life and are essential for electron and hydrogen atom transfer. The linear response of a carbon–deuterium (C–D) vibration based on the vibrational Stark effect can facilitate measurements of electric fields for critical biological reactions including cofactor-mediated hydride transfer. We find both inter- and intramolecular electric fields influence the C–D frequency in NAD(P)H and nicotinamide-like models where the reactive C4-hydrogen has been deuterated. Hence, the C–D frequency can report both environmental electrostatics and conformational changes of the nicotinamide ring. Conformation-dependent effects are mediated through space as electrostatic effects, rather than through-bond. A Stark tuning rate of ~0.57 cm⁻¹/(MV/cm) was determined using both experimental and computational approaches, including vibrational solvatochromism, molecular dynamics simulations, and in silico Stark calculations. The vibrational probe’s Stark tuning rate is shown to be robust and suitable for measuring fields along hydride transfer reaction coordinates in enzymes.

Graphical Abstract

Among the most important biochemical reactions involving C–H bonds is redox chemistry, particularly that coupled to a biological cofactor such as nicotinamide adenine phosphate (NAD⁺) and nicotinamide adenine diphosphate (NADP⁺) or their reduced forms, NADH and NADPH. These cofactors play key roles in over two thousand known enzymes including such key drug targets as aldose reductase, estradiol dehydrogenases, glucose dehydrogenase, lactate dehydrogenase, retinal and retinol dehydrogenases, and nitric oxide dioxygenase.¹ Nicotinamide cofactors arguably share a similar level of significance to ATP, heme, retinoids, and chlorophyll in biochemistry,² making a physical description of their catalytic function in redox chemistry a central goal. A role of electrostatics in NAD(P)H-coupled oxidation/reduction reactions in enzymes has been postulated, yet until now quantitative electrostatic discussions have been primarily limited to the theoretical realm or to axes other than the hydride transfer itself.^3–10 In this work we demonstrate how the introduction of a vibrational Stark probe in the form of a C–D bond at the transferable hydride position can serve as an effective reporter of local electrostatics. In the course of investigations of several enzymes where NAD(P)H plays a key role, the use of this method first requires primary information on the origin(s) of variations of the C–D stretch frequency for the H(D) involved in hydride transfer. This proves to be a complex and interesting story in itself, shedding new light on the emerging science of C–D Stark probes and detailing the role of intramolecular electric fields on vibrational frequency shifts.

Vibrational Stark probes are sensitive reporters of electrostatics at the molecular scale, mapping the strength of local electric fields to experimentally observable frequency shifts.^10–13 Carbon-deuterium vibrational probes offer several advantages as vibrational Stark probes, but only one prior study of a quantitatively calibrated C–D probe as a ruler for measuring electrostatics has been reported to our knowledge.⁸ Due to their low dipole moments, C–Ds are minimally perturbative to their environments, oscillate in a relatively clean vibrational region (2000–2400 cm⁻¹), and can be seamlessly incorporated into chemical systems for understanding environmental influences on C–H(D) bond chemistry, including the myriad of roles they play in biochemistry. A major drawback is that the C–D oscillator strength is quite weak, an aspect for which Raman spectroscopy holds promise as a detection method.^14,15 Electrostatics on substrate molecules containing C–D probes have been investigated recently using the vibrational Stark effect;¹⁰ however, a generalizable strategy for measuring electric fields on the hydride-transfer axis near the cofactor in cofactor-coupled reactions has been lacking. Building off isotope labeling methods developed for studying kinetic isotope effects in oxidoreductases,^16–19 we noted that a deuterium substitution at the reactive hydride position of the nicotinamide ring generates a localized vibrational mode whose direction approximates the hydride transfer reaction coordinate in many NAD(P)H-coupled oxidoreductases.^14,20 In the following, we demonstrate how this C–D stretch can be measured in both NAD(²H) and NADP(²H) using vibrational spectroscopy and used as a reporter of local, catalytically relevant electric fields.

We used a deuterium-labeled derivative of 1-methylnicotinamide (Figure 1A) as a model compound for probing local electric fields in enzyme-catalyzed NAD(P)H-coupled oxidation–reduction reactions. By employing a combination of two approaches—specifically, experimental vibrational solvatochromism with molecular dynamics simulations to extract a vibrational Stark tuning rate, and independent in silico Stark calculations based on quantum mechanical treatment of a model compound under an in silico applied electric field along the C–D bond—we demonstrate that the C–D vibrational probe at the donor carbon serves as a sensitive indicator of local electrostatics. The relative electric field on the C–D can be mapped to a vibrational frequency shift with a Stark tuning rate of 0.57 cm⁻¹/(MV/cm). Moreover, the C–D also responds to local electrostatic perturbations in the form of varying conformations of the small molecule, where the orientation of substituent groups yields through-space electrostatic-induced frequency perturbations of the C–D frequency. Similar in some ways to the conformational flexibility of the side-chain of penicillin G which can alter the field on a carbonyl by as much as 20–25 MV/cm upon large-scale rotation,²¹ or how various peptide conformations alter the Cα–D stretch mode by 20–40 cm⁻¹ in deuterated amino acids,^22,23 the molecular conformation here likewise changes the frequency of the C–D, which in this case we attribute to an electrostatic mechanism. Interestingly, very slight changes in conformation, including subtle inversions of amines, can significantly alter the local electrostatic environment of the transferable hydride. These field changes can be up to twice as large as in the previous case of penicillin’s side chain, with potential implications to the role of the cofactor binding conformation in enzymes on the kinetics of hydride transfer.

Figure 1.

(A) Distinct conformations within nicotinamide-containing molecules such as 1,4-dihydro-1-methylnicotinamide (DHMN, shown here deuterated at the 4R-position) are possible. These are determined by cis–trans rotation about the amide bond and pyramidal inversions at the tertiary amine N_T and primary amine N_P. (B) Heterogeneity is observed in the IR spectrum for 200 mM deuterated DHMN in water. Four major spectral features a–d are attributed to distinct conformations of the DHMN molecule, which affect both frequency and transition dipole moment tuning via intramolecular electric fields in accordance with the vibrational Stark effect (see text). (C) Various conformers (a-d) of (4-deuterated) DHMN lead to varying through-space intramolecular electric fields on the C–D oscillator dipole. These intramolecular electrostatic interactions can explain the spectral heterogeneity shown in panel B. Here and throughout this paper, 4R-deuterated isomers of DHMN are depicted, while mirrored conformations also exist for the 4S-deuterated DHMN (as DHMN was synthesized as a racemic mixture) with identical vibrational properties to their 4R-mirrored counterparts. Conformers a-b are shown to have a trans-oriented amide, compared to cis in conformers c-d. Inversions about N_T (conformer a to b) and N_P (conformer c to d) also change local electric fields projected along the C–D. Approximate rates of conversion are calculated from reported calculations of energy barriers²⁴ or as determined from the Berny-TS optimization using B3LYP/6–311++G(2d,2p) in Gaussian 16, starting from the planar amine structures with force constants calculated at the initial structure.

The 1,4-dihydro-1-methylnicotinamide (DHMN, Figure 1A) deuterated at the 4-position (see SI section 1 for synthesis)^25,26 was chosen as a model compound because it is soluble in a wide range of solvents necessary for the solvatochromism studies discussed below. The IR spectrum in the C–D stretch region is shown in Figure 1B and exhibits significant heterogeneity. The existence of two predominant bands plus smaller, blue-shifted shoulders indicates the existence of multiple populations. The ¹H NMR spectrum of this molecule in CD₃OD or D₂O shows that the sample is pure (>94%, Figures S1–S2; also for the other isotopically labeled molecules, Figures S3–S6), with the largest impurity being the presence of oxidized form, whose C–D would be substantially shifted to ~2300 cm⁻¹ based on frequency calculations at the B3LYP/6–311++G(2d,2p) level. The very close frequencies and consistent solvatochromic shifts ($\Delta {\bar{\nu }}_{\text{C}-\text{D}}$ from one solvent to another consistently within a few cm⁻¹ for each of the major bands) in a variety of solvents suggested strongly that the populations were due to various conformations of the same product molecule in solvents. Indeed, DFT optimizations revealed the presence of several distinct energy minima of the molecule each within 2 kcal/mol, which are differentiated by a few key conformational differences in dihedral angles (Figure 1A,C). We note that calculations by Deng et al. at a primitive level of theory qualitatively predict a role of conformation in C–D frequency shifts in nicotinamide,¹⁴ yet experimental and electrostatic analyses have been lacking until now. In the following sections we set out to categorize these conformational differences and the physical basis for their effects on the C–D.

Cis and Trans Isomerization of the Amide on DHMN.

One of the distinctions between the optimized structures is the rotamer state of the conjugated amide substituent, whether cis (the conformation shown in Figure 1A) or trans (~180° rotation of the amide to swap positions of oxygen and nitrogen) with respect to the relative position of the amide-NH₂ to the C–D bond. The cis state is consistently found to be lower in energy by 1–2 kcal/mol using a variety of computational techniques²⁷ (Figure S7) and is also the form observed in X-ray crystallographic structures of several related derivatives.^28,29 Interestingly, this contrasts with structures of NADH or NADPH bound to enzymes which consistently demonstrate the trans form.^30–32 This has possible mechanistic implications for hydride transfer—it has been proposed, for instance, that the trans orientation of the amide dipole facilitates hydride transfer by decreasing the enthalpy of the hydride transfer transition state,³³ via polarization of the hydride acceptor by the cofactor’s carbonyl.²⁴ The small ~7 kcal/mol rotational barrier for cis–trans interconversion of DHMN²⁴ and even smaller thermodynamic difference in energy between the two states suggests both may be observed in a given solution-phase environment.

Puckering about N_P and N_T Atoms on DHMN.

In addition to the cis and trans forms of the amides, there is also puckering at the two nitrogen centers associated with the primary and tertiary amines (N_P and N_T) based on the positioning of the lone pair on either side of the ring, leading to pyramidalization.^30,34 We note that the pyramidal inversion across the nitrogen centers produces small energetic differences of <0.2 kcal/mol between different pyramidally inverted forms when optimized and calculated at the B3LYP/6–311++G(2d,2p) level. Perhaps more importantly, these inversions also lead to differences in electrostatic environment on the 4-deuterium (the transferable hydride position in NAD(P)H). For the cis conformation of the amide, pyramidal inversion of the amide nitrogen (N_P) will change the electrostatic environment of the transferable hydride—a lone pair on the same side as the transferable hydride will increase the electric field on the bond to the transferable hydride due to the asymmetry in charge localization across the molecule’s electrostatic map. Calculations of the electric field magnitude projected along this C–H bond show that this quantity is indeed affected significantly by both cis–trans rotation and amide pyramidal inversion. Capping the amide with a hydrogen and placing dummy atoms at the positions of the carbon and hydrogen as reporters of the electric field in the B3LYP/6–311++G(2d,2p) electrostatic potential calculation,³⁵ we calculate a through-space, intramolecular field arising from the amide alone as varying between +45.1 MV/cm from carbon to hydrogen in the case of a trans-amide (conformer-b), and 1.2 MV/cm in the case of the cis amide with hydrogens syn to the C–H bond in question (conformer-d; Figure S8). Pyramidalization of the amide nitrogen also changes the field significantly in the cis form, between 1.2 MV/cm (conformer-d) and 19.9 MV/cm (conformer-c) when the hydrogens are puckered syn or anti to the C–H bond in question (Figure S8). We note that these field differences are significant and comparable in magnitude to previously observed solvent-induced electric fields on C–D bonds.⁸ They are sufficient to cause significant perturbation of the C–D frequency for the deuterium-labeled molecule and perturb the energetics of the C–H(D) bond dipole, potentially playing a role in hydride transfer for NAD(P)H cofactors bound to enzymes.

Stark Tuning Rate of the C–D Bond in DHMN.

The effect of an electric field on the vibrational frequency of a bond oscillator, known as the vibrational Stark effect (VSE), typically gives rise to a linear change in infrared absorption frequency.^36,37 This was demonstrated in the case of a carbonyl bond³⁸ to arise from the oscillator dipole moment varying between ground and first-excited vibrational states by the quantity $\Delta \overrightarrow{\mu }$, whose magnitude gives the Stark tuning rate. The field-induced energy shift is greater for the excited state since its dipole moment is typically larger than the ground state, thus causing a shift in vibrational frequency depending on the field from either environmental (solvent, protein) or intramolecular conformational factors. In the case of 4-deuterated DHMN, the C–D dipole direction is best modeled as running from deuterium to more positively charged carbon when using electrostatic potential fitting via the Merz–Singh–Kollman approach³⁹ (Table S1). The IR frequency shift of the C–D bond relative to the gas phase frequency ${\bar{\nu }}_{\text{C}-\text{D}}^{\text{gas}}$ is given by $\Delta {\nu }_{\text{C}-\text{D}}={\bar{\nu }}_{\text{C}-\text{D}}^{\text{gas}}=-\Delta {\overrightarrow{\mu }}_{\text{C}-\text{D}}·\overrightarrow{F}=-\left|\Delta {\overrightarrow{\mu }}_{\text{C}-\text{D}}\right|F_{\text{C}-\text{D}}$ due to the interaction of the C–D difference dipole $-\Delta {\overrightarrow{\mu }}_{\text{C}-\text{D}}$ with the field projection onto the bond $F_{\text{C}-\text{D}}$. As discussed below, it is important to distinguish frequency shifts due to intramolecular factors that change the shape of the oscillator potential well, considered to be “through-bond” effects, from those due to parts of the molecule whose orientation affects the local electric field experienced by the probe oscillator, considered to be “through-space” effects.

We argue that conformational changes of both amide cis–trans isomerization as well as pyramidalization of the amines should largely be considered through-space effects that influence the electric field experienced by the C–D (Figure 1A), with minimal changes to the bond Stark tuning rate. Our DFT calculations demonstrate negligible donor–acceptor interactions between the C–D and amide nitrogen or oxygen using a natural bond orbital (NBO) analysis⁴⁰ (SI section 2b), and spatial dependence in that the amide pyramidalization affects the C–D frequency in the cis form more than the trans form as a result of the nitrogen being about 1 Å closer to the C–D bond (Figure S9). DFT also demonstrates that the tertiary amine in the nicotinamide ring matters more in determining the frequency of the C–D bond when in the trans form (Figure S10). Even this latter case could be substantially governed by a through-space electrostatic effect at the C–D itself: while charge density stemming from the tertiary nitrogen lone pair can flow through the conjugated bonds in the ring (a through-bond effect), this will have the additional result of more charge density on one side of the ring versus the other, affecting the local electrostatic environment of the C–D oscillator via through-space Coulombic effects. A similar Stark tuning rate for all possible conformational minima in the molecule would strongly suggest a dominance of through-space electrostatic effects on the C–D oscillator frequency.

Electrostatic differences caused by the cis–trans rotation and pyramidalization also affect the polarization of the C–D bond and its transition dipole moment, changing the intensity of the IR bands as also indicated in DFT calculations (Figures S9 and S10).⁴¹ In a wide variety of different solvents, the experimental spectrum consistently shows two major bands (Figure 1b, Figure S11) corresponding to the amide pyramidalization conformers (c and d) of the cis form. The higher-intensity, red-shifted band (peak-d) corresponds to the form where the N_P hydrogens are syn to the deuterium (conformer-d, in Figure 1C). Peak-c corresponds to the cis form where the N_P hydrogens are anti to the deuterium (Figure 1). The smaller blue shoulders near 2130–2160 cm⁻¹ are indicative of the trans-form of the amide (conformers a and b), smaller in terms of both population and absorptivity while still observed in all solvents. Peak-b corresponds to the trans form where the methyl group on N_T is anti to deuterium, while peak-a corresponds to the trans form where the N_T methyl group is syn to deuterium.

Having assigned the identities of the major bands in the IR spectrum, we next demonstrate that the C–D oscillator frequency varies linearly with respect to local electric field, and that the slope of this variation is relatively robust with respect to conformation and even identity of the amide substituent. Demonstrating these considerations is required to show that the nicotinamide C–D bond can be an effective Stark probe. To demonstrate these points and calibrate the vibrational Stark tuning rate $-\Delta {\overrightarrow{\mu }}_{\text{C}-\text{D}}$, we primarily relied on a hybrid technique that combines vibrational solvatochromism (Figures S11–S12, Table S2) with molecular dynamics simulations of the molecule in various solvent systems.³⁷ Because the two peaks of the cis-amide form were easier to deconvolve from the spectra, we focused on the correlation between these frequencies and the corresponding average electric fields (Table S3) on the C–D bond in various solvents determined using molecular dynamics with the general AMBER force field (GAFF).⁴² These peaks are fit reproducibly across replicate spectra with relatively small 95% confidence intervals, especially for more polar solvents. More polar solvents consistently blue shift these peaks over a range of up to 10 cm⁻¹, consistent with the magnitude of shifts observed for other C–D vibrational modes in a variety of condensed-phase environments.^15,43,44 Electric fields on C–D were calculated using the force field-derived Coulombic forces of the solvent molecules on the DHMN C–D bond, not including self-field contributions of the solute atoms. The net electrostatic forces on carbon and deuterium projected along the bond were then divided by their respective charges and averaged (see SI section 1 for details). Strong linear correlations $(R^2\ge 0.91)$ between field and frequency for both peaks indicate linear Stark behavior for this probe (Figure 2), in agreement with the linearity of a previously characterized aldehyde C–D bond.⁸ The vibrational Stark tuning rate $\left|\Delta {\overrightarrow{\mu }}_{\text{C}-\text{D}}\right|$ is consistently ~0.57 cm⁻¹/(MV/cm) for both pyramidalizations of the cis amide. Alternate methods of calculating solvent-induced electric field quantum mechanically using absolutely localized molecular orbitals (ALMO)^8,45 implemented in Q-Chem,⁴⁶ show a similar strong linear correlation with experimental frequencies for both bands, albeit with a slightly lower slope (0.46 cm⁻¹/(MV/cm)) (Figures S13–S14, Table S4). This difference is apparently a result of QM methods, like polarizable MD simulations, calculating higher fields by including polarization effects as has also been observed previously.^8,13

Figure 2.

Correlations between simulated average solvent electric field and experimentally observed frequency for the two cis-amide conformers demonstrate strong linearity with a very similar slope (Stark tuning rate) for both pyramidal conformations of N_P. Error bars shown are 95% confidence intervals for Voigt peak fits to three replicate FTIR spectra collections (Figure S11). For simplicity, only the 4R-deuterated stereoisomer is shown.

The consistency in Stark tuning rate for both pyramidal states of the cis amide suggests the robustness of $\left|\Delta {\overrightarrow{\mu }}_{\text{C}-\text{D}}\right|$ with respect to conformation. Independence of $\left|\Delta {\overrightarrow{\mu }}_{\text{C}-\text{D}}\right|$ with respect to conformation in turn suggests minimal changes to the electronic structure at C–D, so that conformation-induced frequency shifts are indeed through-space rather than through-bond effects. While the heterogeneous band structure can be deconvolved into the cis and trans forms, the smaller trans peaks are more difficult to resolve due to both their lower populations and lower transition dipole moments (<50% of the intensity of cis forms as calculated by DFT, Figures S9–10) resulting in substantial uncertainty in the peak positions (larger error bars of Figure S15). Combined with the lesser ability of fixed-charge force fields to capture hydrogen bonds⁴⁷ involving the carbonyl, which is closer to the deuterium in the trans form and induces secondary solvation effects on the C–D, it is difficult to generate a field-frequency correlation with the same degree of linearity (Figure S15; weighted $R^2=0.62$ and 0.86) However, alternative methods to be discussed in the following paragraphs do consistently indicate the robustness of $\left|\Delta {\overrightarrow{\mu }}_{\text{C}-\text{D}}\right|$ that extend even to the trans form.

In Silico Calculations of C–D Stark Tuning Rate in DHMN.

To corroborate the vibrational Stark tuning rate of the C–D oscillator for a given conformation of the nicotinamide ring, we also performed in silico Stark calculations of the C–D Stark tuning rate (Figures 3, S16, Table S5) in Q-Chem.⁴⁶ Geometry optimization and harmonic frequency calculations at the B3LYP/6–31+G(d) level were performed in the presence of an applied external, uniform dipolar electric field parallel to the (4R-deuterated) C–D bond, with varying starting conformations of the amide dihedral and both amines. The harmonic frequency corresponding to the C–D stretch mode was scaled according to a constant empirical scaling parameter,⁴⁸ 0.964 (SI section 1). We also briefly compared the harmonic in silico Stark calculations to those with anharmonic corrections to the C–D stretch mode for two conformations, finding no substantial differences from the harmonic calculations (SI section 2e, Figures S17–S18). In the case of the trans conformer of the molecule (Figure 3A), the relationship between field and frequency is primarily linear with a Stark tuning rate of 0.56 cm⁻¹/(MV/cm), close to the value of 0.57 cm⁻¹/(MV/cm) obtained based on solvatochromism for the cis conformer of the molecule. Note that a previous instance using this in silico Stark method showed an agreement within ~10% of the experimental Stark tuning rate value,⁸ making these new results suggestive of a close match between cis and trans Stark tuning rates. Hence, a combination of computational and experimental results for the 4R-deuterated DHMN and related analogues confirm a vibrational C–D Stark tuning rate near 0.6 cm⁻¹/(MV/cm) for all conformations of the molecule.

Figure 3.

(A) DFT-calculated field-frequency correlation for the trans-amide form of the deuterium-reduced 1-methylnicotinamide model compound demonstrates primarily first-order behavior with a Stark tuning rate close to the experimental value of the cis-form. (B) DFT calculations for the cis rotamer of the molecule demonstrate pyramidal inversions of both the tertiary (N_T) and primary (N_P) amines as the conformation changes in response to the applied field to minimize energy. After the N_P inversion, in the positive field regime analogous to solvatochromism conditions, the slope is again closer to that of the trans rotamer. Frequencies were determined via optimization and harmonic frequency analysis at the B3LYP/6–31+G(d) level, scaled by 0.964, with a uniform dipolar electric field applied onto the C–D bond direction as implemented in Q-Chem.

The in silico Stark results for the cis rotamer demonstrate unusual behavior manifesting a discontinuity in the frequency as a function of the field (Figure 3B). An initially positive field-frequency correlation becomes negative as the field becomes closer to zero, before a sudden drop between 2 and 3 MV/cm, after which point the correlation is positive again. Further analysis of the molecular conformation under these applied fields showed an intriguing conformational change of the optimized molecular structure near 0 MV/cm as pyramidal inversion about the amide nitrogen occurs in response to the applied field along the nicotinamide C–D bond changing from positive to negative. In accordance with the trends described previously, this causes a drop in frequency as the field created by the NH₂ dipole changes orientation to be parallel to (stabilizing of) the C–D oscillator dipole moment. As the conformational change begins to occur via distortion of the lone pair on the nitrogen, two electrostatic effects—both the applied in silico field and the intramolecular electric field arising from the changing amide conformation—compete with one another for influence on the C–D frequency. The amide conformation change ultimately makes a greater difference to the C–D frequency due to its close proximity and the relatively large electrostatic perturbation that such a conformational change creates. Once the amide inverts, the correlation subsequently becomes positive and linear again as no more major geometric distortions occur with increasing field. The slope is 0.52 cm⁻¹/(MV/cm), again within 10% of the experimental value.

Vibrational Stark Tuning Rate Validation for a Nitrile-Substituted Analogue.

Even more strikingly, we further demonstrated that the identity of the amide substituent can be changed entirely with minimal change to $\Delta {\overrightarrow{\mu }}_{\text{C-D}}$. We synthesized a related molecule by reducing 3-cyanopyridine with deuterium at the reactive C4 position, generating a similar C–D probe to that in DHMN (SI section 1).⁴⁹ Like the DHMN model compound, this molecule also shows two peaks in its vibrational spectrum. However, since no amide exists in this molecule to affect the C–D frequency through either pyramidalization or cis–trans rotation, the frequency shifts must be due to pyramidalization about the ring nitrogen. Unlike DHMN, the ring of reduced 3-cyanopyridine exists with a very slight boat-like pucker (Figure S19), such that pyramidalization causes flipping of the boat conformation. B3LYP/6–311++G(2d,2p) optimizations and frequency calculations reveal that two primary conformations of the ring exist, depending on whether the nonstereoselectively added deuterium is in the flagpole (scaled ${\bar{\nu }}_{\text{C}-\text{D}}=2136.1$ cm⁻¹) or bowsprit (scaled ${\bar{\nu }}_{\text{C}-\text{D}}=2118.9$ cm⁻¹) orientation. The IR spectrum similarly reveals two bands that, as in the case of deuterated DHMN, both blue shift in more positive electric fields from more polar solvents (Figures S20–S21, Table S6). Correlations between the MD-simulated average electric fields and experimentally observed vibrational frequencies demonstrate good linearity $(R^2\ge 0.93)$ with ~0.6 cm⁻¹/(MV/cm) slopes comparable to and within error of that determined for DHMN (Figure 4). The striking agreement in Stark tuning rates between the two molecules even when the identity of the ortho substituent is changed is strong evidence for the robustness of $\Delta {\overrightarrow{\mu }}_{\text{C}-\text{D}}$ with respect to substituent conformation. As an aside, the solvatochromic results for the nitrile-substituted model compound C–D are intriguing due to the linear correlation still holding across solvents that stimulate both positive and negative fields on the C–D. The same linear relationship on either side of zero is indicative that the first-order vibrational Stark equation is a good model for the solvatochromic shifts of the C–D.

Figure 4.

Vibrational solvatochromism of 3-cyanopyridine, singly reduced by deuterium at the reactive C4 position. Two primary bands are observed in the C–D region depending on whether the deuterium is on the same or opposite side of the nitrogen lone pair (Figures S20–S21). Here, only the 4R-stereoisomers are shown for simplicity though deuterium addition is not stereospecific. The Stark tuning rates (slopes) of the C–D for both conformations of the molecule match and are within error of the Stark tuning rate for the amide-containing molecule (Figure 2). These results demonstrate the robustness of the C–D Stark tuning rate with respect to the substituent identity and molecular conformation. Error bars shown are 95% confidence intervals for Voigt peak fits to three replicate FTIR spectra collections (Figure S21).

Characterization of Deuterated NAD(P)H Vibrational Spectra and Electrostatics.

Having calibrated the vibrational Stark tuning rate of the model compounds, we note that the 4-deuterated full cofactors NADPH and NADH can be used as reporters of electric field, where the C–D oscillator axis can approximate the hydride transfer axis in enzymes. NADPH can be deuterated at either the 4R or 4S positions using stereoselective enzymes and deuterated substrates to reduce the oxidized cofactor. Example enzymes include alcohol dehydrogenase from E. coli or Thermoanaerobium brockii for 4R-deuterated-NADPH,¹⁷ horse liver alcohol dehydrogenase for 4R-deuterated NADH, or glucose dehydrogenase from Pseudomonas sp. for 4S-deuterated NADH or NADPH.¹⁶ Deuterated cofactors [4R-²H] NADH and [4R-²H] NADPH were prepared in aqueous buffers (Figure 5; see SI section 1) for vibrational analysis. In contrast to the model compound, which could be analyzed using FTIR, the lower concentrations resulting from limited solubility of the full cofactors made Raman spectroscopy more advantageous than FTIR for differentiating the C–D signal from the solvent background. Furthermore, to ensure consistency with prior FTIR analyses, we include the Raman spectrum of DHMN in the Supporting Information together with FTIR spectra of the 4R-deuterated NADPH (Figures S22–S24). We note the similarity of the spectral shape and fitted peak frequencies within 2 cm⁻¹ of one another, with slight deviations in relative intensity from the FTIR spectra.

Figure 5.

(A) Deuterium labeled NADPH. The phosphate group in orange is distinctive of NADPH but not NADH. (B) Stokes-scattering Raman spectrum of aqueous NADPH and NADH (~30 mM, pH 8–9) with a 532 nm laser using a Horiba Labram Raman microscope and 600 gratings/mm, smoothed using a 17-point rolling average. (C) The cofactor (NADPH shown) can adopt folded conformations such as the one shown in addition to an elongated conformation. Folded conformations may stabilize particular conformations of the amide substituent in nicotinamide.

It is immediately notable that the 4R-deuterated NADH and NADPH spectra, which overlap well with a peak at 2112 cm⁻¹, demonstrate significantly less heterogeneity compared to the DHMN model compound in Figure 1 (a greater change than explainable by the Raman intensity differences). The 2112 cm⁻¹ frequency in water agrees well with previously reported measurements of the full cofactor,¹⁴ but now with a clarified understanding of nicotinamide’s conformational preferences in solution and the significant effects of electrostatics on the C–D frequency. The rest of the cofactor thus plays a stabilizing role for one of the nicotinamide conformations, likely the cis conformation most resembling conformer-c in Figure 1, due to the frequency similarity. The primary effect of the rest of the cofactor is in changing the relative populations of the various nicotinamide conformers. Compared to DHMN, there are additional contributions to the electric field on C–D from the cofactor tail, but these effects are calculated by MD simulations to be small (stabilizing by 6–8 MV/cm, Table 1 and Figure S25) relative to the local electrostatic changes wrought by the nicotinamide conformational changes within NADPH (Figure S8).

Table 1.

Calculated Fields and Observed Frequency Differences between Model Compound DHMN and Full Cofactor NADPH in Water

	Cis-DHMN	Trans-DHMN	Cis-NADPH	Trans-NADPH
Calculated Electric Field (MV/cm)	24.0	4.6	18.0	−3.1
Observed Frequency (cm−1)	2084.7 and 2116.5a	2133.8 and 2157.4b	2112.3a	Not resolved

^aTwo frequencies are reported for the two pyramidal inversions of the amide nitrogen in the cis rotamer.

^bTwo frequencies are also reported for the ring nitrogen inversions of the trans rotamer of DHMN.

The environment of the nicotinamide ring of the full cofactor in water is subtly different compared to the model compound in water, leading to differences in observed populations and slight frequency shifts. Unlike DHMN which experiences symmetric solvation environments on either face, NAD(P)H cofactors in solution have been demonstrated by both simulation⁵⁰ and NMR studies^51,52 to favor a folded conformation (Figure 5C), with the nicotinamide pi-stacking on top of the adenine ring and a possible preference of the pro-4S hydrogen to be associated with the adenine ring, a preference which may be stronger for NADH than NADPH.⁵² Our MD simulations using the Amber force field parameters of Ryde⁵³ in TIP3P water with sodium counterions likewise demonstrate a preference for a folded NADPH conformation (Figure S26). An asymmetric solvation environment about nicotinamide caused by the cofactor tail can explain the difference in population between the two C–D peaks of the cis NADPH distinguished by pyramidal inversion of the NH₂. While a very clear peak at 2112 cm⁻¹ is observed similar to peak-c of DHMN in Figure 1, there is only a small shoulder near 2090 cm⁻¹ that could correspond to conformer-d of Figure 1 for DHMN, indicating its much lower population in the full cofactor compared to conformer-c. Particularly, if the pro-4S-face primarily interacts with adenine,⁵² a conformation in which the hydrogens of the nitrogen NH₂ are primarily facing the adenine would generate a state similar to conformer-c, leading to a similar frequency, albeit shifted by slight environmental differences in field.

Calculating the combined field from both the solvent and the NADPH “tail” (everything below the DHMN motif) on the corresponding C–D bond yields a field result that should be more directly comparable to the solvent field on DHMN’s C–D bond (Table 1). There are slight differences in the two: in both the cis and the trans rotamers, the C–D typically experiences a more stabilizing field by 6–8 MV/cm in the full cofactor through the combined effects of the additional charged groups and effects of the folded conformation on the nicotinamide solvation environment (Table 1). Using 0.57 cm⁻¹/(MV/cm) as the Stark tuning rate, we would expect a red shift of the C–D stretch frequency in the cis rotamer by 3.4 cm⁻¹. In fact, the observed frequency of 2112.3 cm⁻¹ for the C–D in NADPH corresponds to a 4.2 cm⁻¹ red shift from the peak-c conformer, extremely close to the expectation. We note also that slight variations in frequency may result from a different ionic strength required for solvating the purified NADPH compared to the DHMN model compound, though these effects are generally small⁵⁴ and to some extent recapitulated in the MD simulations using 100 mM NaCl in addition to NADPH tetrasodium.

Finally, we note is that it is difficult to resolve the more blue-shifted trans populations within the empirical vibrational spectrum, as these peaks are too small to accurately resolve. The results indicate that in solution, a trans amide rotamer of nicotinamide is even further destabilized in the full cofactor compared to DHMN. The result is striking given that enzyme environments typically cause the exact opposite result: a strong preference for binding the trans form of nicotinamide dinucleotide, even though both the oxidized and reduced forms of the nicotinamide energetically favor the cis form in aqueous and gas phases.²⁴ These observations strongly suggest a mechanistic role of a trans amide in NAD(P)H-coupled hydride transfer reactions, an aspect of potential significance that has been rarely discussed in the literature. Explanations could involve prepolarization of the substrate by the carbonyl or a role of intramolecular electric fields originating from the carbonyl in destabilizing the reactive carbon-hydride bond to enhance reactivity. Further exploration is yet required on this intriguing question. In the case of using deuterated-NADPH to measure electrostatics on the hydride-transfer reaction coordinate, as in oxidoreductases, the fact that NAD(P)H cofactors preferentially bind in a single trans conformation decreases much of the complexity of the peak heterogeneity (Figures S27–S28). Furthermore, narrowing of the peak that occurs in enzyme-bound contexts due to a decrease in inhomogeneous line-broadening^8,37 makes the signal much easier to visualize in FTIR spectroscopy. Raman scattering also provides a suitable alternative to pick out the C–D peak from the background for a variety of protein:NAD(P)H systems.

In summary, we have shown that the incorporation of a C–D bond into the reactive position of NAD(P)H cofactors and analogues presents an effective strategy for monitoring the electrostatics experienced by the reactive site of nicotinamide. The C–D bond serves as a linear vibrational Stark probe, mapping electric fields from the environment (e.g., solvent or protein) or through-space intramolecular electrostatics to C–D vibrational frequency shifts. Vibrational solvatochromism and in silico Stark calculations allow us to calibrate this Stark tuning rate to 0.57 cm⁻¹/(MV/cm). Meanwhile, conformational dynamics are also monitored via frequency shifts of the Stark probe, with the most striking being cis–trans rotation of the amide substituent and pyramidal inversions of the two nitrogens in nicotinamides. These small-scale conformational changes perturb the local electrostatics of the C–D bond in significant and measurable ways, such that they might be expected to play roles in NAD(P)H-coupled oxidation–reduction reactions in enzymes. Further study is warranted on the significance of the enzyme environment exerting a strong bias to binding the higher-energy trans-conformer of the nicotinamide. Use of deuterated NAD(P)H as a vibrational Stark probe for local electrostatics opens new avenues for experimentally probing electric fields along the hydride transfer axis in the multitude of oxidoreductases that couple to these prolific cofactors.