LC-Photo-CIDNP Hyperpolarization of Biomolecules Bearing a Quasi-Isolated Spin Pair: Magnetic-Field Dependence via a Rapid-Shuttling Device

Abstract

Liquid-state low-concentration photochemically induced dynamic nuclear polarization (LC-photo-CIDNP) is an emerging technology tailored to enhance the sensitivity of NMR spectroscopy via LED-mediated optical irradiation. LC-photo-CIDNP is particularly useful to detect solvent-exposed aromatic residues (Trp, Tyr), either in isolation or within polypeptides and proteins. This study investigates the magnetic-field dependence of the LC-photo-CIDNP of Trp-α-¹³C-β,β,2,4,5,6,7-d₇, a Trp isotopolog bearing a quasi-isolated ¹H^α-¹³C^α spin pair (QISP). We employed a new rapid-shuttling side-illumination field-cycling device that enables ultra-fast (90–120 ms) vertical movements of NMR samples within the bore of a superconducting magnet. Thus, LC-photo-CIDNP hyperpolarization occurs at low field, while hyperpolarized signals are detected at high field (700 MHz). Resonance lineshapes were excellent, and the effect of several fields (1.18–7.08 T range) on hyperpolarization efficiency could be readily explored. Remarkably, unprecedented LC-photo-CIDNP enhancements $\epsilon \cong \mathrm{1,200}$ were obtained at 50 MHz (1.18 T), suggesting exciting avenues to hypersensitive LED-enhanced NMR in liquids at low field.

Graphical Abstract

Nuclear magnetic resonance (NMR) is a well-known non-invasive technique to elucidate molecular structure and dynamics at atomic resolution. However, NMR spectroscopy is intrinsically very insensitive, primarily due to the small population differences between nuclear spin eigenstates, even in the presence of high applied magnetic fields.[1, 2] In order to enhance NMR sensitivity in liquids, a variety of hyperpolarization strategies have been employed to date,[2, 3] including Overhauser dynamic nuclear polarization (ODNP),[4–6] dissolution dynamic nuclear polarization (DDNP),[7–9] parahydrogen-induced polarization (PHIP)[10–16] and signal amplification by reversible exchange (SABRE).[17–34] On the other hand, the above technologies typically require expensive instrumentation, harsh sample conditions, and long hyperpolarization times, hampering broad applicability.

An alternative recently developed approach, known as low-concentration photochemically induced dynamic nuclear polarization, or LC-photo-CIDNP,[35–37] stands out for its simplicity and straightforward practical implementation, especially when LED irradiation sources are employed.[37, 38] LC-photo-CIDNP is an optically enhanced technology tailored to the analysis of very diluted NMR samples, within the low-μM to sub-μM concentration range.[35, 39] This approach has been employed in both structural [35–37, 39–43] and screening [44,45] studies, to date. The major mechanistic features of LC-photo-CIDNP are shared with the parent technology, known as photo-CIDNP, and have been described elsewhere[3, 41, 46–48] (see also Supplementary Material). Yet, LC-photo-CIDNP is unique in its ability to effectively utilize much smaller sample concentrations than conventional photo-CIDNP.

In combination with a suitable isotopic substitution leading to a ¹H^α-¹³C^α quasi-isolated spin pair (QISP), LC-photo-CIDNP enables the straightforward detection of sample concentrations as low as 20 nM, as in the case of the tryptophan isotopolog Trp-α-¹³C-β,β,2,4,5,6,7-d₇ (Figure 1a),[39] in an oxygen-depleted environment[49] and in the presence of reductive radical quenchers.[43] Computational simulations based on theory[40] showed that, in the presence of fluorescein (Figure 1a) as photo-sensitizer, the ¹³C^α polarization of Trp-α-¹³C-β,β,2,4,5,6,7-d₇ due to geminate recombination is particularly sensitive to the applied static magnetic field (B₀) and is expected to yield higher polarization at lower Larmor frequencies (50–300 MHz). These predictions are summarized in Figure 1b. However, no LC-photo-CIDNP field-dependence work has been done on spectrometers operating at <400 MHz, to date.

Figure 1.

LC-photo-CIDNP key molecules and predicted field dependence of nuclear-spin hyperpolarization. (a) Structure of photosensitizer and Trp isotopolog carrying a quasi-isolated spin pair (QISP). (b) Computationally predicted LC-photo-CINDP ¹³C geminate polarization of Trp isotopolog. Data in panel b were reproduced with permission from Li, S. Y. Appl. Magn. Reson. 2023, 54, 59–75.

Here, we addressed the above information gap by exploring the potential of LC-photo-CIDNP as a sensitivity-enhancement tool in NMR at low field. We employed a novel belt-driven ultra-fast field-cycling device (FCNMR Ltd), originally developed for relaxometry,[50, 51] on a 700 MHz NMR spectrometer equipped with a cryogenic probe. The field-cycling apparatus, was customized for optically enhanced NMR (Supporting Formation Figure S1) upon addition of a vertically adjustable side-illumination module that includes 10 LED chips, and is located on the shuttle rail.[52, 53] A plunger immediately above the NMR sample prevents macroscopic liquid motions during shuttling and prevents bubble formation. This unique device is equipped with an optical reflector that enables stable and homogeneous side illumination at a variety of stray fields. Optical excitation wavelengths and irradiation powers are programmable. Upon comparison with other previously developed field-cycling tools employed in the context of photo-CIDNP,[54–58] our apparatus is particularly convenient due to (a) the concurrent presence of reproducible vertical positioning of NMR sample during shuttling, (b) the efficient side-illumination setup and (c) the extremely rapid up/down shuttling time (90–120 ms), which reduces polarization losses and yields excellent NMR-readout lineshapes.

A very diluted (1 μM) aqueous sample of the QISP tryptophan Trp-α-¹³C-β,β,2,4,5,6,7-d₇ was employed for our field-dependent hyperpolarization measurements. The shuttling/illumination scheme of Figure 2a was used to determine the extent of nuclear-spin hyperpolarization at a variety of low stray fields ranging from 1.18 to 7.03 T (corresponding to 50–300 MHz). The initial sample shuttling to low field was followed by a rapid 100 ms optical irradiation (480 nm, at irradiance 91.7 mW/mm²). Samples were then rapidly shuttled back to 700 MHz (16.4 T) for detection, taking advantage of the enhanced high-field signal-to-noise ($S/N$) readout. The 1D ¹H detected ¹³C RASPRINT[37] pulse sequence was used, and 32 scans were acquired. Note that ¹³C RASPRINT is a heteronuclear correlation experiment that transfers photochemically enhanced longitudinal nuclear magnetization from ¹³C^α to ¹H^α for detection, via a reverse-INEPT scheme.[37] Representative data for illumination at 1.18 T (50 MHz) and 5.9 T (250 MHz) magnetic fields and 32 scans per experiment are shown in Figure 2b. Prior to data collection with illumination (i.e., LED-on), control experiments were also carried out on the same sample under dark conditions (i.e., LED-off), under identical conditions except that optical illumination was omitted. Several conclusions can be drawn from the data of Figure 2b.

Figure 2.

Fundamental aspects and experimental output of rapid-shuttle field-cycling LC-photo-CIDNP. (a) Timecourse of rapid shuttling in the context of ¹H-detected ¹³C LC-photo-CIDNP (pulse sequence: ¹³C RASPRINT). (b) Representative 1D ¹³C RASPRINT spectra of 1 ^μM Trp isotopolog (32 scans, H^α region), with illumination fields $B_{ill}=50\mathrm{M}\mathrm{H}\mathrm{z}$ (1.18 T) and 250 MHz (5.86 T), under light (blue, LED-on) conditions. (c) Reference dark (black, LED-off) spectrum of 1 ^μM Trp isotopolog (32 scans, H^α region) for data collected on the same spectrometer at the same detection field (700 MHz) with no

First, excellent lineshapes and $S/N$ are obtained at both field values (1.18 and 5.9 T) under LED-on conditions, with the fast-shuttling device introduced here. Second, not surprisingly, in our control experiment under dark conditions no signal is detectable from a low-concentration (1 mM) sample with only 32 scans. This result highlights the sensitivity advantage of LC-photo-CIDNP in the context of field cycling.

Third, our results reveal, for the first time, the remarkable expected effectiveness of LC-photo-CIDNP for the ultrasensitive detection of aromatic molecules at low field, even in the absence of field cycling. To illustrate this point, note that, here, we take advantage of high-field (16.5 T, 700 MHz) detection, which further increases $S/N$. On the other hand, the LC-photo-CIDNP field-cycling data also provide an estimate of the expected output in the absence of high-field detection. Consider the expression,

$$\frac{{\left(\frac{S}{N}\right)}_{+FC}}{{\left(\frac{S}{N}\right)}_{-FC}}=\sqrt{\raisebox{1ex}{$B_{det}$}\!\left/ \!\raisebox{-1ex}{$B_{ill}$}\right.},$$ (1)

where $B_{det}$ and $B_{ill}$ denote the detection and illumination fields, respectively. Equation 1 takes into account the receiver-coil and noise-level dependence of $S/N$ under field-cycling $({\left(\frac{S}{N}\right)}_{+FC})$ and static-$B_0$ non-field-cycling $({\left(\frac{S}{N}\right)}_{-FC})$ conditions,[2, 59, 60] assuming equivalent-optical illumination parameters and identical non-field-cycling hardware. Equation 1 can be used to appraise the expected signal-to-noise of LC-photo-CIDNP performed in the absence of field cycling $\left((S/N{)}_{-\mathrm{F}\mathrm{C}}\right)$, based on +FC experimental results. According to relation 1, the LC-photo-CIDNP representative spectra of Figure 2b, which feature $B_{ill}=50$ and 250 MHz, are expected to yield 3.74- and 1.67-fold lower $S/N$, respectively, if performed at static-field. Note that these values are underestimates, given that LED-irradiation times can be easily increased to values more closely approximating steady-state conditions (e.g., ${\mathrm{T}}_{\mathrm{L}\mathrm{F}}\ge 200\mathrm{m}\mathrm{s}$ instead of the 100 ms employed here), in experiments performed at low static field, likely resulting in further sensitivity increases. In summary, LC-photo-CIDNP has the ability to increase the sensitivity of NMR spectroscopy of aromatic molecules at low applied field. Aromatic groups are of key importance in biomolecular structure and recognition, pharmaceutical and environmental sciences[61–63] and other fields. Moreover, in the presence of reductive radical quenchers LC-photo-CIDNP can be performed for thousands of scans with only a moderate extent of photodamage.[43] Hence, significantly higher $S/N$ values can be achieved at 1 mM sample concentrations, in the presence of prolonged signal averaging. Further, note that the $S/N$ of non-field-cycling NMR experiments is expected to drop by a factor of 11.2, when the applied field ($B_0=B_{det}$) is lowered from 250 to 50 MHz, based on the conventional definition of NMR $S/N$.[2, 59, 60] On the other hand, the observed $S/N$ losses under light conditions amount to only 2.2-fold. Importantly, the latter effect is primarily LC-photo-CIDNP-related and independent of field cycling. Yet, a prime advantage of the field cycling employed here is that it provides facile access to a wide range of illumination fields.

More quantitative assessments of the above concepts can be gained via the enhancement factor and percent polarization $\left(P_{\%}\right)$ parameters, as detailed next. The LC-photo-CIDNP enhancement factor $\epsilon$ is defined as

$$\epsilon =\frac{A_{light,+FC}{Conc}_{dark,theo,+FC}}{A_{dark,theo,+FC}{Conc}_{light,+FC}},$$ (2)

where $\mathrm{A}$ and Conc denote resonance areas and sample concentrations, respectively. The subscripts “$\mathrm{l}\mathrm{i}\mathrm{g}\mathrm{h}\mathrm{t},+\mathrm{F}\mathrm{C}$” and “$\mathrm{d}\mathrm{a}\mathrm{r}\mathrm{k},\mathrm{t}\mathrm{h}\mathrm{e}\mathrm{o},+\mathrm{F}\mathrm{C}$” refer to light and theoretical-dark conditions within field-cycling experiments. The ${\mathrm{A}}_{\mathrm{dark},\text{theo},+\mathrm{F}\mathrm{C}}$ resonance areas had to be estimated indirectly, given that quantitatively appropriate dark conditions could not be experimentally achieved. This is due to the long time (~ 3–5*T₁) required for our small-molecule Trp isotopolog to reach thermal equilibrium at low field under dark conditions. This time leads to poor lineshapes, likely due to a combination of field inhomogeneities, bulk-susceptibility effects and temperature gradients. Despite these practical challenges, dark-conditions areas could be estimated theoretically based on experiments done at static field (i.e., under −FC conditions), according to

$$A_{dark,theo,+FC}=A_{dark,B_{det/dark,-FC}}\frac{B_{ill,dark,-FC}}{B_{det}}$$ (3)

where $A_{dark,B_{det/dark,-FC}}$ denotes the area of an experimental sample collected in the absence of field cycling under dark conditions, at $B_{det}$ (details in Supplementary Material).

The resulting experimentally determined enhancement factors for the ¹H^α-detected ¹³C^α LC-photo-CIDNP of Trp-α-¹³C-β,β,2,4,5,6,7-d₇ are displayed in Figure 3a and Supporting Table S1. A remarkable trend emerges. Namely, ε steadily increases as the stray field ($B_{ill}$) gets lower, ranging from $\epsilon \cong 300$ at 300 MHz until an unprecedented $\epsilon \cong \mathrm{1,200}$ is achieved at the lowest tested illumination field of 50 MHz. Note that, even if the experimental $S/N$ of resonances at low field are effectively somewhat less intense than those at higher field (e.g., see Figure 2b), the decrease in $S/N$ is much less than expected, due to the extremely strong ¹³C LC-photo-CIDNP at the low 50 MHz field.

Figure 3.

Dependence of LC-photo-CIDNP field-cycling experiments on illumination field ($B_{ill}$). (a) enhancement factors $\epsilon$ and (b) percent polarizations $P_{\%}$ of Trp isotopolog. Values are either uncorrected (white bars) or corrected (gray bars) for small polarization losses due to ¹³C $T_1$ relaxation during shuttling to the detection field $\left(B_{det}\right)$. Data are displayed as avg ± SE (n = 2). (c) T₁-relaxation-corrected $P_{\%}$ overlayed with predicted percent polarization (red dashed line) based on geminate recombination only (details in Supplementary Material). $T_1$ measurements were performed with n = 2–3.

In addition, given the estimated theoretical equilibrium thermal polarization $P_{th,eq,B_{ill}}$

$$P_{th,eq,B_{ill}}=tanh\left(\frac{\mathrm{\hslash }\gamma B_{ill}}{2k_{B}T}\right),$$ (4)

we also determined the percent polarization $P_{\%}$ from

$$P_{\mathrm{\%},}=\epsilon \mathrm{*}{P}_{th,eq,B_{ill}}\mathrm{*}100.$$ (5)

The resulting values of $\epsilon$ and $P_{\%}$ are shown in Figure 3a,b and Supporting Table S1. We also determined corrected values of $\epsilon$ and $P_{\%}$ that take into account ¹³C T₁ relaxation during shuttling to high field (90–120 ms) and during high-field recovery (50 ms, Figure 3a,b). Relaxation effects were accounted for upon performing non-optically assisted non-field-cycling ¹³C T₁ measurements at selected experimentally accessible fields (see Supplementary Material for details). Remarkably, the T₁-corrected $\epsilon$ and $P_{\%}$ values at all illumination fields are only marginally larger than the uncorrected values, thanks to the fast shuttling and short recovery time of the field-cycling device employed here.

Direct comparisons between the theoretical geminate-polarization predictions of Figure 1b and the corrected experimentally derived $P_{\%}$ values of Figure 3b cannot be made for a variety of reasons. First, the theoretical predictions only consider geminate polarization (i.e., they neglect f-pair polarization).[41, 47] Second, the theoretical predictions assume that 100% of the molecules of interest form radical pairs, which is unrealistic. Third, illumination time was reduced to 100 ms in this work (relative to the customary 200 ms),[37, 39, 40, 43] to optimize lineshapes, resulting in potential pre-steady-state conditions that may lower $\epsilon$ and $P_{\%}$. Despite the above challenges, the overall trends of experiments and theory could be compared via a semiempirical fit (see details in Supplementary Material). The results, shown in Figure 3c, display excellent overall qualitative agreement. The origin of the single outlier data point at $B_{ill}$ 150 MHz is presently unknown.

To place the results of this work within a broader context, it is worth noticing that this study features the largest enhancements reported to date in liquids at low sample concentration (1μM) by LC-photo-CIDNP. While comparable enhancement factors $\epsilon$ were previously attained in solution by dissolution DNP (dDNP)[64] or parahydrogen-induced polarization (PHIP),[65] our $\epsilon >\mathrm{1,200}$ is particularly high for 𝜀 the direct and in situ analysis of biologically relevant molecules, i.e., amino acids in solution. Indeed, unlike dDNP and PHIP, the field-cycling/LC-photo-CIDNP approach presented here has no requirements for sample dissolution, transfer between devices, or extrinsic catalysts. Moreover, LC-photo-CIDNP hyperpolarization times are much faster (ca. 100 ms) than the corresponding times required by dDNP and PHIP.

In conclusion, this work shows that combining LC-photo-CIDNP with a QISP Trp isotopolog and ultrafast field-cycling leads to excellent liquid-state NMR lineshapes and extremely high enhancement factors ($\epsilon >\mathrm{1,200}$) at low $B_{ill}$ (50 MHz). Importantly, our data also show that low-field NMR of aromatic molecules can be rendered substantially more sensitive even in the absence of field cycling. Finally, we anticipate that the field-cycling/LC-photo-CIDNP technology presented here is amenable to extensions to more complex biomolecules in solution. In support of this idea, photo-CIDNP and LC-photo-CIDNP were already shown to be applicable not only to amino acids but also to polypeptides and proteins [3, 36, 37, 39, 43, 66], even in the absence of any selectively deuterated isotopologs. For instance, the 13-residue s³² peptide prepared in the presence of commercially available ¹³C-,¹⁵N-Trp was rapidly detected by 1D and 2D photo-CIDNP [49, 66], with data acquired at 1 mM concentration with no oxygen scavengers [66]. The same peptide was detected at 5 mM concentration in the presence of O₂-depleting enzymes [67]. In addition, 1D and 2D photo-CIDNP hyperpolarization experiments were carried out on the uniformly ¹³C-,¹⁵N-labeled drkN SH3 protein (i.e., the N-terminal SH3 domain of the drk adaptor protein from Drosophila) since 2011 [66], at concentrations spanning from 300 mM to 500 nM [36, 37, 43, 66, 67].

Highlights.

• Optically enhanced NMR and ultra-rapid field cycling generate high hyperpolarization in liquids

• Up/down shuttle occurs within 90–120 ms, yields Lorentzian lineshapes and minimal T1 distortions

• Our approach is most effective fort biomolecules carrying a quasi-isolated spin pair (QISP)

• 1,200-fold S/N enhancement is achieved at 1 mM) sample concentration at 1.18 T illumination field