Improved Sensitivity of Laser-Enhanced ¹H^α-¹³C^α-Correlation via Suppression of C^α-C’ Scalar-Coupling Evolution

Abstract

Low-concentration photochemically induced dynamic polarization (LC-photo-CIDNP) enables the spectroscopic analysis of biomolecules containing the amino acids Trp and Tyr at sub-micromolar concentration in solution. Typical LC-photo-CIDNP pulse sequences involving ¹H-¹³C correlation, however, perform well in the case of aromatic resonances but display a relatively poor signal-to-noise ratio for ¹³C^α and ¹³C^β resonances. Here, we develop a novel pulse sequence denoted as ¹³C perturbation-recovered selective-pulse photo-CINDP enhanced reverse INEPT, or ¹³C PRESPRINT, tailored to the LC-photo-CIDNP analysis of ¹H-¹³C^α pairs. Our method, which is based on full suppression of 1-bond C^α-C’ scalar-coupling evolution during the constant-time delay, results into a sensitivity improvement by a factor of 2. The enhanced performance of this pulse sequence enabled us to improve the analysis of LC-photo-CIDNP laser-power dependence at very low (200 nM) sample concentration. An improved theoretical model, developed to quantitatively describe this laser-power dependence, shows excellent agreement with our ¹³C PRESPRINT experimental data.

1. Introduction

Photochemically induced dynamic nuclear polarization (photo-CIDNP) is an established magnetic-resonance technique [1-6] that has been recently exploited in liquid-state NMR to enhance the sensitivity of biomolecules carrying aromatic functional groups [7-18]. A main advantage of photo-CIDNP is the fast (~0.2 s) in situ generation of nuclear hyperpolarization under physiologically relevant conditions.

Low-concentration photo-CIDNP [10] (LC-photo-CIDNP) is a branch of this method that requires specific photo-sensitizers that are tailored to sub-μM sample concentrations and oxygen-scavenging agents. LC-photo-CIDNP is typically performed in the presence of cryogenic probes, and it enables the rapid detection of tryptophan (Trp) and tyrosine (Tyr), either in their free state or within proteins, down to 200 – 500 nM concentration.

C^α and, to a lesser extent, H^α chemical shifts yield semi-quantitative information on protein secondary structure [19]. Hence, detecting C^α-H^α pairs at the highest possible sensitivity is particularly desirable. ¹³C PREPRINT (Fig. 1B) is a LC-photo-CIDNP pulse sequence tailored to the ultra-sensitive detection of aromatic H-C or H^α-C^α pairs in Trp (Fig. 1A) and Tyr, in isolation or within proteins [10]. This pulse sequence is particularly useful for the characterization of backbone secondary structure of proteins via H^α-C^α pairs.

Figure 1.

(A) Structure of L-tryptophan (Trp), highlighting relevant chemical shifts and scalar coupling constants. Numerical values of chemical shifts and coupling-constants were determined experimentally in 90%H₂O and 10% D₂O at room temperature via a simple 1D pulse-acquire ¹³C sequence. (B) scheme illustrating the ¹³C PREPRINT pulse sequence [10].

¹³C PREPRINT includes a constant-time period T, which is typically optimized for evolution of the one-bond C^α-C^β scalar coupling interaction (¹J_C^αC^β = 33.5 Hz, Fig. 1A). On the other hand, ¹³C^αs are also coupled to ¹³C’ with a ¹J_C^αC' close to 53.4 Hz (Fig. 1A). Hence, evolution of the one-bond C^α-carbonyl carbon (i.e., C^α-C') scalar coupling interaction during T leads to additional terms in the Hamiltonian, resulting in ¹³C^α coherence loss. The situation is further complicated by Bloch-Siegert effects [20] experienced by ¹³C’, given that the ¹³C carrier frequency is centered on ¹³C^α, in experiments aimed at detecting H^α-C^α pairs. In summary, significant signal losses occur as the result of both of the above effects.

Here, we present a modified version of the ¹³C PREPRINT pulse sequence, denoted as ¹³C PRESPRINT (perturbation-recovered selective-pulse photo-CIDNP enhanced reverse INEPT), that circumvents the above disadvantages. LC-photo-CIDNP signal-to-noise (S/N) is known to depend weakly on laser irradiation power [10]. The enhanced sensitivity achieved via ¹³C PRESPRINT enabled us to quantitatively analyze the power dependence of LC-photo-CIDNP at very low sample concentration (200 nM) across a wider range of laser powers than previously possible, starting at 50 mW. In addition, we developed an improved theoretical model to predict the LC-photo-CIDNP laser-power dependence. This model was shown to adequately fit experiment data, and was therefore able to account for a wider laser-power range than previously possible.

2. Results and discussion

2.1. Suppression of C^αC’ scalar-coupling evolution leads to improved sensitivity in LC-photo-CIDNP ¹H–¹³C correlation

In the conventional ¹H-¹³C heteronuclear-correlation ¹³C PREPRINT pulse sequence [10] (Fig. 1B) optimized for the detection of C^α resonances, some of the ¹³C^α coherence is lost due to two independent effects.

First, the widely different values of the ¹J_C^αC' (53.4 Hz) and ¹J_C^αC^β (33.5 Hz) scalar-coupling constants prevent the C^α-C’ scalar-coupling interaction from evolving by exactly 180° during the constant time T. The duration of T is typically optimized for 180° evolution of the C^α-C^β scalar-coupling interaction (T ~ 1/¹J_C^αC^β–2τ empirically optimized to 26.6 ms). Note that the scalar-coupling constants reported here are for the ¹³C-¹⁵N-labeled photo-CIDNP-active amino acid Trp. As a result, T is not optimized for C^α-C' coupling, resulting in significant signal losses. For illustrative purposes, the above can be readily verified (upon ignoring off-resonance effects and pulse-field gradients) by considering the experimentally detectable product operators at the end of the constant time: $2{\widehat{I}}_z{\widehat{A}}_x\cos [{\pi }^1{\mathrm{J}}_{{\mathrm{C}}^{\alpha }{\mathrm{C}}^{\prime }}(2\tau +T)]\cos [{\pi }^1{\mathrm{J}}_{{\mathrm{C}}^{\alpha }{\mathrm{C}}^{\beta }}(2\tau +T)]\cos ({\omega }_A{t}_1)+2{\widehat{I}}_z{\widehat{A}}_y\cos [{\pi }^1{\mathrm{J}}_{{\mathrm{C}}^{\alpha }{\mathrm{C}}^{\prime }}(2\tau +T)]\cos [{\pi }^1{\mathrm{J}}_{{\mathrm{C}}^{\alpha }{\mathrm{C}}^{\beta }}(2\tau +T)]\sin ({\omega }_A{t}_1)$, where the symbols ${\widehat{I}}_z$ and ${\widehat{A}}_{x,y}$ refer to the pertinent components of the nuclear-spin angular-momentum operators of H^α and C^α, respectively, ω_A is the angular precession frequency of the C^α spin in the rotating frame, and the time delays τ and T are defined as in Fig. 1B.

Second, the widely different chemical shifts of the Trp C’ (174.5 ppm) and C^α (55.2 ppm) generate Bloch-Siegert effects (ca. 119 ppm off-resonance from the carrier frequency) on the carbonyl carbon, given that the ¹³C pulses are centered in the C^α spectral region.

To quantify the extent of the above complications, we carried out computer simulations estimating relative expected signals in ¹³C PREPRINT experiments in the absence and presence of C^α-C' scalar coupling and off-resonance effects. The results are shown in Fig. 2A (see also SI for details). Interestingly, the simulated ¹³C PREPRINT signal is expected to increase slightly as the ¹³C π hard-pulse duration gets longer. This result applies to both dark (i.e., laser off) and light (i.e., laser on) conditions. The duration of this pulse is ca. 24 μs (γB₁ = 20.83 kHz) for our NMR probe, i.e., a typical value on modern NMR spectrometers.

Figure 2.

(A) Simulated ¹³C PREPRINT signal of Trp H^α as a function of ¹³C π hard-pulse duration, assuming a 600 MHz NMR spectrometer. (B) Simulated ratio of ¹³C PREPRINT S/N of Trp H^α as a function of ¹³C π hard-pulse duration, for a 600 MHz NMR spectrometer.

As apparent in Fig. 2A and more explicitly highlighted in Fig. 2B, the predicted signal increases by ca. 2-fold when C^α-C’ scalar-coupling evolution and off-resonance effects are not present.

Encouraged by the above predictions, we proceeded to implement the following updates to the ¹³C PREPRINT pulse sequence. First, we replaced a few key ¹³C hard π pulses with corresponding semi-selective pulses (not exciting the carbonyl carbons) centered in the C^α region, to refocus the C^α-C' scalar-couplings. Semi-selective excitation was carried out with the Q3_surbop pulse, which is an improved version the Q3 pulse shape [21]. This pulse has an excellent excitation profile, as shown in Fig. 3A. When centered in the ¹³C^α region, Q3_surbop produced negligible excitation in the ¹³C’ region, which is ca. 18 kHz off-resonance from the pulse center frequency, at 600 MHz. An additional Q3_surbop band-selective ¹³C pulse centered on the ¹³C’ resonances was introduced during t₁ evolution, to decouple ¹³C’from ¹³C^α, thus preventing any peak splitting in the indirect dimension due to ¹J_CαC' coupling (see 2D spectrum in Supporting Fig. S4).

Figure 3.

(A) Excitation profile of 180° ¹³C Q3_surbop shaped pulse used in the ¹³C PRESPRINT pulse sequence. The predicted excitation profile was generated via the Bruker waveform simulator. Experimental data points were generated upon changing the center frequency of the 180° ¹³C shaped pulse of a simple reference pulse sequence (see Supporting Fig. S2). (B) ¹³C PRESPRINT pulse sequence. The inverted-bell shapes denote Q3_surbop shaped pulses (180° flip angle), which are improved versions of the Q3 selective pulse [21] developed by Bruker. The symbol t_L denotes laser irradiation time, Δ is the perturbation-recovery delay, τ = 1/(4J_CH), and T is the total evolution time in the indirect dimension (optimized to 26.6 ms for C^α). The phase cycling is ϕ_rec = x, −x, −x,x; ϕ₁ = y, −y; ϕ₂ = y, y, −x, −x, −y, −y, x, x. For convenience, a direct side-by-side comparison between the ¹³C PRESPRINT and ¹³C PREPRINT pulse sequences is provided in Supporting Fig. S3.

We denote the resulting improved pulse sequence, shown in Fig. 3B, as ¹³C-perturbation-recovered band-selective-photo-CIDNP enhanced reverse INEPT, or ¹³C PRESPRINT (note the relatively minor spelling difference relative to ¹³C PREPRINT). Conveniently, this pulse sequence achieves complete refocusing of C^α-C’ scalar coupling, and off-resonance effects are no longer present. The performance of ¹³C PRESPRINT was experimentally tested and compared with that of ¹³C PREPRINT.

Data were first collected on 2 μM uniformly ¹³C-¹⁵N-labeled Trp (32 scans, ~ 1 min. total experiment time). As shown in Fig. 4A, under light (laser on) conditions a fairly high S/N was rapidly observed for ¹³C PRESPRINT. This value is ca. 2-fold higher than the S/N achieved for ¹³C PREPRINT, in full agreement with the computer simulations. At 200 nM Trp (Fig. 4B), a similar 2-fold gain was achieved, though the S/N was lower, due to both the lower concentration and the smaller number of acquired transients. When we measured the experimental enhancement on a different cryogenic probe with a longer ¹³C π pulse (i.e., 30 instead of 24 μs, data not shown), the results were consistent with the theoretical predictions of Fig. 2.

Figure 4.

Comparison between 1D ¹³C PREPRINT and ¹³C PRESPRINT spectra of (A) 2 μM and (B) 200 nM ¹³C-¹⁵N-enriched Trp in 90% H₂O and 10% D₂O. Data for each dark/light experiment pair were collected on identical independently prepared samples, with the same experimental parameters and conditions except for pulse-sequence-dependent values. Spectra were acquired with 2,630 total points and a sweep width of 6,602 Hz. The recycle delay was 2.5 s. Spectra were zero-filled to 4,096 complex points, and an exponential window function was applied (5 Hz line-broadening). ¹³C PRESPRINT shows a 2-fold higher S/N than ¹³C PREPRINT, under light conditions.

The ¹³C PRESPRINT pulse sequence was also tested on free tyrosine (Tyr) in solution (Supporting Fig. S5). However, due to the much lower photo-CIDNP polarization of Tyr C^α, enhancement values could not be quantified.

2.2. Theoretical prediction of laser-power dependence of LC-photo-CIDNP at nanomolar sample concentration

LC-photo-CIDNP enhancements depend strongly on laser irradiation power at high sample concentration, i.e., ≥ μM. On the other hand, this dependence is known to be significantly weaker at lower sample concentrations, i.e., ≤ 500 nM [10]. Recent investigations by Cavagnero and coworkers showed that steady-state photo-CIDNP polarization at low sample concentration (typically < 1 μM), where the free-radical of the molecule of interest M^•+ undergoes complete nuclear spin relaxation before termination, can be quantitively described by a theoretical expression [10]. This sample-concentration regime typically obeys the condition $1∕{\mathrm{T}}_1^{{\mathrm{M}}^{•+}}>>k_{ter}[{\mathrm{M}}^{•+,\text{SS}}]+k_{de}[{\mathrm{M}}^{\text{SS}}]$, where ${\mathrm{T}}_1^{{\mathrm{M}}^{•+}}$ is the spin-lattice nuclear relaxation time of the molecule of interest in free-radical form, k_ter is the effective rate constant for the regeneration of M taking into account all the elementary steps leading to it starting from M^•+ [18, 22], k_de is the rate constant for degenerate electron exchange, and the symbols [M^SS] and [M^{•+, SS}] denote the steady-state concentrations of the molecule of interest in neutral and free-radical form. Under the above conditions, the steady-state polarization of the k^th nucleus, $P_k^{SS}$, is [10]

$$P_k^{M,SS}=\frac{{\mathrm{T}}_1^{\mathrm{M}}{k}_{et}{[}^{\mathrm{T}}{\mathrm{D}}^{\text{SS}}][{\mathrm{M}}^{\text{SS}}](1+\gamma ){\xi }^{\mathrm{G}}{\Phi }_{\mathrm{G}}}{[\mathrm{M}{]}_{{}_O}},$$ (1)

where γ is defined as

$$\gamma =\left(\frac{{\xi }^{\mathrm{F}}(1-{\Phi }_{\mathrm{G}})}{{\xi }^{\mathrm{G}}{\Phi }_{\mathrm{G}}}\right),$$ (2)

with ${\xi }_k^{\mathrm{F}}$ and ${\xi }_k^G$ denoting the normalized probability differences to generate a recombination product in the α and β spin states of the k^th nucleus per F-pair and geminate recombination event, respectively. The parameter Φ_G denotes the total probability of geminate recombination per geminate radical pair. The initial concentrations of the molecule of interest are [M]₀. The symbols [^TD^SS] and [D^•−,SS] denote steady-state concentrations of the dye in the triplet excited-state and radical forms, respectively. The effective rate constant for the bimolecular electron transfer between the molecule of interest M and the triplet excited-state dye ^TD is denoted as k_et. Finally, ${\mathrm{T}}_1^{\mathrm{M}}$ is the spin-lattice nuclear relaxation times of the k^th nucleus in M.

Under steady-state conditions and nanomolar sample concentration, it can be shown (see Supporting Information for details) that

$$[{\mathrm{M}}^{\text{SS}}]\approx \frac{(k_{ter}∕k_{et})([\mathrm{M}{]}_{{}_o}{)}^2}{{[}^{\mathrm{T}}{\mathrm{D}}^{\text{SS}}]}-2\frac{(k_{ter}∕k_{et}{)}^2([\mathrm{M}{]}_{{}_o}{)}^3}{{[}^{\mathrm{T}}{\mathrm{D}}^{\text{SS}}{]}^2}.$$ (3)

Upon substitution of relation (3) into (1), the steady-state polarization can be expressed as

$$P_k^{M,SS}\approx {\mathrm{T}}_1^{\mathrm{M}}{k}_{ter}(1+\gamma ){\xi }^{\mathrm{G}}{\Phi }_{\mathrm{G}}[\mathrm{M}{]}_0\left(1-2\frac{[\mathrm{M}{]}_0(k_{ter}∕k_{et})}{{[}^{\mathrm{T}}{\mathrm{D}}^{\text{SS}}]}\right).$$ (4)

Given equation (4) and the fact that [^TD^SS] is directly proportional to laser irradiation power P_light (see Supporting Information), and given that the signal S under light conditions is directly proportional to the extent of steady-state polarization [23], the laser power dependence of LC-photo-CIDNP can be expressed as

$$S=a-b•P_{light}^{-1},$$ (5)

where a and b are adjustable parameters.

2.3. Laser power dependence of LC-photo-CIDNP at nanomolar Trp concentration

We previously found that LC-photo-CIDNP enhancements depend more steeply on laser-irradiation power at >μM sample concentration than at lower concentration. [10] Although we were previously able to detect 200 nM Trp via LC-photo-CIDNP, the poor S/N prevented us from quantitatively investigating the laser power dependence of LC-photo-CIDNP within the low-concentration regime (≤ 500 nM) at very low laser power (ca. 50 mW). This dependence is important to experimentally validate photo-CIDNP theoretical models at low laser power and sample concentration.

The improved sensitivity of ¹³C PRESPRINT over ¹³C PREPRINT is of general significance. Next, we provide a specific example of a useful application. Fig. 5A shows how the ¹³C PRESPRINT improved sensitivity enables investigating the LC-photo-CIDNP laser-power dependence. Data at very low sample concentration (i.e. 200 nM Trp) were collected over a wide range of laser powers, starting at 0.05 W. As illustrated in Fig. 5B, the experimental laser-power dependence shows good agreement with the theoretical model developed here, based on equation (5).

Figure 5.

(A) Representative 1D ¹³C PRESPRINT light spectra illustrating the LC-photo-CIDNP laser-power dependence of 200 nM ¹³C-¹⁵N Trp. (B) Plot illustrating the ¹³C PRESPRINT (light conditions) H^α-resonance area as a function of laser irradiation power. Areas were normalized relative to the average area at 1.5 W laser power. The solid line denotes the curve fit based on the theoretical model developed in this work (Eq. 5, with a and b regarded as adjustable parameters). (C) Cumulative ¹³C PRESPRINT S/N as a function of number of scans at 0.5 and 1.5 W laser power. A total of 2,402 points, a 6,010 Hz sweepwidth and a 2.5 s recycle delay were used in each experiment. Spectra were zero-filled to 4,096 complex points, and an exponential window function was applied (10 Hz line-broadening). Data in panels B and C are reported as average ± S.E. for n=3.

Finally, we compared the cumulative S/N as a function of number of transients, at 0.5 and 1.5 W, to probe whether the extent of photodegradation varies at widely different laser power levels (Fig. 5C) and at very low sample concentration (200 nM). We did not find significant differences between the two power settings. We conclude that collecting LC-photo-CIDNP data at low laser power at nM sample concentration does not lead to any increases in the extent of photodegradation. In summary, the data in Fig. 5 are important because they highlight that LC-photo-CIDNP can be performed with the ¹³C PRESPRINT pulse sequence at low laser power (> 0.2 W), and yields excellent NMR sensitivity with no photodegradation penalties relative to higher-laser-power conditions.

3. Conclusions

A ca. 100% improvement in S/N over ¹³C PREPRINT has been achieved with our newly-developed ¹³C PRESPRINT pulse sequence. This experimental advance enabled us to quantitatively assess the weak power dependence of photo-CIDNP at low sample concentration. The theoretical model that we developed to describe this laser-power dependence is fully consistent with our experimental data. In summary this study highlights the exciting prospect to perform LC-photo-CIDNP at nM concentration with the ¹³C PRESPRINT pulse sequence employing low-power inexpensive and readily available light sources.

4. Experimental methods

4.1. Materials

¹³C-¹⁵N-Trp and the fluorescein photosensitizer dye were purchased from Cambridge Isotopes (Tewksbury, MA) and Sigma-Aldrich (St. Louis, MO), respectively. The oxygen-scavenging enzymes glucose oxidase (GO, from Aspergillus niger, catalog number G7141, EC 1.1.3.4) and catalase (CAT, from bovine liver, catalog number C40, EC 1.11.1.6) were purchased from Sigma-Aldrich (St. Louis, MO). Enzymes were dissolved in 10 mM potassium phosphate (pH 7.0), divided into single-use aliquots, flash-frozen in liquid nitrogen and stored at −80 °C.

4.2. NMR sample preparation

All NMR samples contained 10% D₂O, 10 mM potassium phosphate buffer (pH 7.2), 0.15 μM GO, 0.10 μM CAT, and 2.5 μM fluorescein. In addition, D-glucose (2.5 mM) was added to the NMR samples c.a. 10 min. prior to data collection.

4.3. Photo-CIDNP NMR experiments

All photo-CIDNP experiments were carried out with a continuous-wave (CW) argon-ion laser (INNOVA SABRE DBW 24/7 Coherent Inc., Santa Clara, CA) operating in single-line mode (488 nm). The laser beam was directed inside the NMR spectrometer via a convex lens (LB4330, Thorlabs, Newton, NJ), a fiber-coupler (F-91-C1-T, Newport Corporation) and a fused silica optical fiber with 0.6 mm core diameter (FDP600660710, Molex, Lisie, IL). The optical fiber was guided into a glass coaxial insert (WGS-5BL, Wilmad-Labglass, Buena, NJ), which was inserted into the NMR tube to ensure that laser beam was centered within the NMR tube. Laser power was assessed at the tip of the coaxial insert via a power meter.

All NMR spectra were collected at 25 °C with an Avance III HD 600 MHz spectrometer (Bruker Biospin Corp., Billerica, MA). A ¹H{¹⁹F/¹³C/¹⁵N} triple-resonance (TCI) cryogenic probe fitted with a z-gradient was used for all experiments unless otherwise stated (12 μs ¹³C π/2 pulse duration). In some experiments (see section 2.1), a ¹H{¹³C/¹⁵N} triple-resonance (TXI) cryogenic probe (15 μs ¹³C π/2 pulse duration) fitted with a z gradient was employed to measure ¹³C PRESPRINT enhancements for a different ¹³C π hard-pulse duration. This TXI probe had the transmitter/receiver coils and the ¹H and ²H preamplifiers (but not the ¹³C and ¹⁵N pre-amplifiers) cooled to cryogenic temperature. This probe displayed similar sensitivity to the TCI probe for ¹H-detected experiments carried out in this work.

The carrier frequency was centered on the solvent resonance (4.70 ppm) in the ¹H channel, and at 55 ppm (C^α region) in the ¹³C channel. All spectra were referenced with external 4,4-dimethyl-4-silapentane-1-sulfonic acid (DSS). Data were processed with the MestRe Nova (version 11.0.1-20560, Mestrelab Research, Spain). Spectral widths, total number of points and window functions are as listed in individual figure legends.

4.4. Computer simulations

Computer simulations in Fig. 2 were carried out with Mathematica (version 11.2.0.0, Wolfram Research, Champaign, IL) using procedures described in the Supporting Information. The simulation input included definitions of spin operators, scalar couplings, chemical shifts, pulse strengths, as well as rotation and density matrices. The output of the simulations (see Fig.2A) were numerical values proportional to the expected NMR signal, defined as in equation S4 of the Supporting Information.