Tailored flavoproteins acting as light-driven spin machines pump nuclear hyperpolarization

The solid-state photo-chemically induced dynamic nuclear polarization (photo-CIDNP) effect generates non-Boltzmann nuclear spin magnetization, referred to as hyperpolarization, allowing for high gain of sensitivity in nuclear magnetic resonance (NMR). Well known to occur in photosynthetic reaction centers, the effect was also observed in a light-oxygen-voltage (LOV) domain of the blue-light receptor phototropin, in which the functional cysteine was removed to prevent photo-chemical reactions with the cofactor, a flavin mononucleotide (FMN). Upon illumination, the FMN abstracts an electron from a tryptophan to form a transient spin-correlated radical pair (SCRP) generating the photo-CIDNP effect. Here, we report on designed molecular spin-machines producing nuclear hyperpolarization upon illumination: a LOV domain of aureochrome1a from Phaeodactylum tricornutum, and a LOV domain named 4511 from Methylobacterium radiotolerans (Mr4511) which lacks an otherwise conserved tryptophan in its wild-type form. Insertion of the tryptophan at canonical and novel positions in Mr4511 yields photo-CIDNP effects observed by 15N and 1H liquid-state high-resolution NMR with a characteristic magnetic-field dependence indicating an involvement of anisotropic magnetic interactions and a slow-motion regime in the transient paramagnetic state. The heuristic biomimetic design opens new categories of experiments to analyze and apply the photo-CIDNP effect.


Results
To rationalize the key properties of molecular spin-machines that can be used to generate photo-CIDNP, we proposed a design strategy based on mutations, supported by field-dependent CIDNP studies. By using various mutations, as described below, we were able to vary the distance between the electron donor and acceptor. In this way, we affected the rate of SCRP formation and recombination, and also varied the electron-electron spin-spin interaction in the SCRP. To probe the reaction and spin dynamics in the SCRP, we used the field dependence of photo-CIDNP.
Screening LOV domains for induction of photo-CIDNP. Aiming for designed molecular spinmachines producing light-induced nuclear hyperpolarization, we have designed a series of protein mutants, which will be presented in parts (i)-(iii) ( Table 1).
(i) So far, the occurrence of the solid-state photo-CIDNP effect was limited to cysteine-lacking LOV domains of phototropin [39][40][41] ; for this reason, here we explored other potential LOV-based light-induced hyperpolarization generators. Alignment and comparison of the amino-acid sequences of Asphot1-LOV2, Scientific Reports | (2020) 10:18658 | https://doi.org/10.1038/s41598-020-75627-z www.nature.com/scientificreports/ Crphot-LOV1 and Ptaureo1a-LOV show about 50% of identity ( Fig. 1A) and the crystal structures show almost identical tertiary structures (Fig. 1B). In particular, the distance between FMN and tryptophan ( Fig. 1C), which determines the strength of the spin-spin coupling in the SCRP and is therefore central to generate photo-CIDNP, is nearly the same, being approximately 11 Å. Therefore, we used the mutant Ptaureo1a-LOV-C287S ( r FW ~ 11 Å) for the liquid state photo-CIDNP NMR experiment, in which the conserved cysteine 287 is replaced by serine. (ii) Formation of an SCRP by electron transfer to excited FMN can occur if a nearby tryptophan is present to act as the electron donor. It has been shown that the amino acid tryptophan is able to provide this function 41,54 . Therefore, in the LOV protein Mr4511, lacking the conserved tryptophan, we introduced a tryptophan at its canonical position Q112 by mutation resulting in the Mr4511-C71S-Q112W double mutant. Previously, transient absorption experiments have been used to test the function of these mutants. In Mr4511, when the cysteine residue was mutated to serine (C71S) or glycine (C71Q) and no tryptophan was present, the lifetime of 3 FMN, τ T , was around 240 μs 51 . Introduction of tryptophan to the canonical position, Mr4511-C71S-Q112W, gave rise to faster quenching of 3 FMN reducing τ T to ~ 24 μs 51 , a value very close to ~ 27 μs observed in Crphot-LOV1-C57S ( r FW ~ 11 Å) 43 . Hence, the double mutant Mr4511-C71S-Q112W ( r FW ~ 11 Å) is the second candidate for the generation of the solid-state photo-CIDNP effect. (iii) Finally, we introduced the electron-donating tryptophan at non-canonical positions. The introduction of tryptophan to a new position of the protein allows to change the distance between FMN and tryptophan, their relative orientation and chemical environments and, therefore, to affect the kinetic and magnetic parameters, critical for the formation of the solid-state photo-CIDNP effect. It is difficult to fine-tune all relevant reaction and magnetic resonance parameters simultaneously, therefore, we focused on creating mutants with different r FW . Lacking a crystal structure of Mr4511, the design relied on a structural model created using SWISS-MODEL and the crystal structure of aureochrome1a-LOV (PDB: 3UE6) from a eukaryotic photosynthetic stramenopile as template 58 . The report of the modeling parameters is provided in Supplementary Information Table S1. Additionally, comparison of the amino-acid sequence (Fig. 1A) allowed us to predict the occurrence of α-helix and β-sheet secondary structures and to reconstruct the tertiary structure of Mr4511. As targets for mutation, we considered amino acids that do not interact with the FMN 50 and also have a bulky side chain similar as tryptophan. Using these ideas, we have designed the following mutants with different positions of tryptophan with various r FW values: Table 1).
Another aspect relevant for rational design of a biomimetic light-driven spin-machine for production of photo-CIDNP is the possibility to introduce isotope labels. In particular, for the measurement of the 15 N photo-CIDNP we employed 15 NH 4 Cl as the sole nitrogen source in the bacterial growth medium during protein expression and produced uniformly 15 N-labelled protein and cofactor (see "Methods"). For the 13 C photo-CIDNP experiment, different labelling strategies were previously applied, either by incorporating the 13 C-labelled FMN into a natural abundant protein moiety of the phototropin-LOV domain 40,46 or by selective 13 C-labelling of the single tryptophan of the phototropin-LOV domain 39 . This enables unambiguous assignment of hyperpolarized 13 C signals and analysis of the photo-CIDNP effect generated by electron donor and acceptor separately. A complete picture of the effect, however, involving both electron donor and acceptor is still missing. Therefore, in the present work we produced a uniformly 13 C-labelled Crphot-LOV1-C57S ( r FW ~ 11 Å), aiming to compare the photo-CIDNP effect of FMN and tryptophan under the same conditions. The hyperpolarization effect in combination with isotope labelling paves the way to field-dependent NMR measurements, providing knowledge Table 1. LOV proteins employed for photo-CIDNP NMR in this work. a The FMN-Trp distance ( r FW ) is given as edge-to-edge distance and estimated based on a structural model obtained by SWISS-MODEL 55 . b The percentage identity is compared with the amino-acid sequence of Crphot-LOV1-C57S. www.nature.com/scientificreports/ of the relationship between enhancement factor and magnetic field, which might provide the key data for future theoretical analysis of the exact photo-CIDNP mechanism.
Comparison of the photo-CIDNP effect between phototropin and aureochrome. Figure 2A shows the 1 H photo-CIDNP effect and its field dependence observed in Crphot-LOV1-C57S ( r FW ~ 11 Å). The protein produced for the experiment initially contained all the nuclei in their natural abundance. Then the protonated buffer of the sample was exchanged to a deuterated buffer (see "Methods" section). The final protein solution may contain ~ 0.4% residual 1 H. From this sample, the effect has not been observed directly on the protons of FMN and tryptophan, however, the light-minus-dark spectra show a negative enhancement (emissive signal, i.e., opposite to the thermal polarization) of the HDO signal at 4.7 ppm, which agrees with the previous publication 39 . A closer look at the light-minus-dark spectra shows that the entire range (-2 to 10 ppm) in the proton NMR spectra exhibits hyperpolarization, in particular the aliphatic region (0 to 2.5 ppm) as well as the HDO signal. Integrating either the HDO signal or the range of 0 to 2.5 ppm or the range of -2 to 10 ppm, we obtain a field dependence with a maximum at 0.6 T, as shown in Fig. 3A. Since the position of the maximum is the same for all protons, it is plausible that the 1 H hyperpolarization has originated from the same SCRP and has been distributed over the whole protein as well as to the residual protons in the deuterated solvent. We assume that the spread of hyperpolarization under liquid-state conditions is due to cross relaxation or due to spin diffusion mediated by non-averaged proton-proton dipolar couplings 59,60 . 13 C photo-CIDNP of a uniformly 13 C-labelled Crphot-LOV1-C57S ( r FW ~ 11 Å) induced at various field strengths ranging from 0.5 to 9.4 T and detected always at 9.4 T is shown in Fig. 2B. The hyperpolarized 13 C signals are tentatively assigned according to previous studies 39,46,59,61 , and are summarized in Table S2 in Supplementary Information. Figure 3B presents the integrated areas of some selected hyperpolarized carbon signals of FMN and tryptophan against the magnetic field strength. Overall, the carbons on the tryptophan indole ring seem to be stronger polarized than those on the isoalloxazine ring of FMN. The hyperpolarized carbon nuclei of tryptophan show emissive polarization at all fields and maximal polarization at around 3 T in agreement with previous results 39 . The selected 13 C atoms of the isoalloxazine ring of FMN show different enhancement patterns. As the magnetic field increases, the signal of, e.g., FMN C-8 stays negative, while FMN C-2 changes the sign of polarization from emission to enhanced absorption. Due to the difference of labelling strategies, the comparison of the 13 C photo-CIDNP spectrum of Crphot-LOV1-C57S ( r FW ~ 11 Å) with a previously published 13 C photo-CIDNP spectrum of Asphot1-LOV2-C450A ( r FW ~ 11 Å) 40,46 is not straightforward. Despite that, we can compare the signal of Trp C-3 carbon in both spectra. In Asphot1-LOV2-C450A ( r FW ~ 11 Å) where the tryptophan nucleus is at natural abundance, strong hyperpolarization at Trp C-3 occurs. The chemical shift of C-3 of Crphot-LOV1-C57S (109.0 ppm) slightly differs from δ (C-3) = 113.5 ppm for Asphot1-LOV2-C450A which might result from the different protein environment of the tryptophan residues. In both cases, the sign of the signal of Trp C-3 is always emissive at all fields studied. Nevertheless, the field at which the maximum polarization at Trp C-3 occurs, B max = 3 T for Crphot-LOV1-C57S and B max = 7 T for Asphot1-LOV2-C450A, is well distinguished. We assume that the difference in the label positions causes the difference in field dependence although details are not yet understood. Furthermore, in a previous work by Kothe et al. 46 the photo-CIDNP effect has been measured at four different magnetic fields (5.9, 7.1, 9.4, and 11.8 T), limiting comparability of data points and localization of the maximum B max , while in the present work, we overcame this problem by using a shuttle system (10 nT < B o < 9.4 T).
Figures 2C and 3C depict the field dependence of the 15 N photo-CIDNP obtained in uniformly 15 N-labelled Crphot-LOV1-C57S ( r FW ~ 11 Å) at 11 magnetic field strengths ranging from 0.1 to 9.4 T. In Fig. 2C, the three hyperpolarized signals and their field dependencies are shown (for assignment, see Table 2). The Trp N-1 CIDNP is a low absorptive polarization at low fields, turning into strong emissive polarization at about 1 T field and having a maximum at around 5 T. In contrast, the photo-CIDNP of FMN N-5 is always absorptive with a maximum at 3 T. At increasing field, the hyperpolarization of FMN N-10 turns from enhanced absorption to emission at around 3 T, similar to the photo-CIDNP of Trp N-1 and in clear contrast to the photo-CIDNP of FMN N-5. A common feature for these three nuclei is that the high-field maximum occurs at around 4 T.
For comparison of the photo-CIDNP field dependencies for the three different NMR active nuclear isotopes, we have chosen the most enhanced signal or spectral region for each nucleus, 1 H, 13 57 , and the simulated structure of Mr4511 (yellow) without FMN. The simulation is performed with SWISS-MODEL based on the crystal structure of aureochrome1a-LOV from Vaucheria frigida (PDB entry: 3UE6) 58 . The information about structural modeling is listed in Supplementary Information Table S1. Five mutants of cysteine-lacking Mr4511 were generated, one with tryptophan placed at the canonical position, Q112W, r FW ~ 11 Å, the other four at non-canonical positions F130W, Y116W, Y129W and K57W with increasing r FW .  Table S2 and 15 N chemical shifts in Table 2. The lower trace shows stacked spectra of the CIDNP effects of the corresponding nuclei measured at five different magnetic fields.  (Fig. 4B); the corresponding 1D spectra are shown in Figure S1. Hence, a high similarity in the field dependence is observed for the two proteins. All three types of nuclei in both samples exhibit emissive NMR signals at the maxima; the positions of maxima depend on the nuclear gyromagnetic ratios (the maximum is positioned at a lower field for a nucleus with a higher gyromagnetic ratio). In both cases, the 1 H hyperpolarization reaches its emissive maximum at a field around 0.6 T, while the maxima of 13 C and 15 N photo-CIDNP appear to be at higher fields. Although the primary sequence of the two proteins does not have a high identity, their tertiary structures (Fig. 1B) and, as shown here, their functional mechanisms are almost identical. Therefore, the production of the photo-CIDNP does not depend crucially on the primary sequence of individual amino acids, whereas the distance between the two redox partners forming the SCRP and their mutual orientation plays of a decisive role.
Tryptophan and non-tryptophan derived photo-CIDNP effect generated in Mr4511 mutants. To generate a photo-CIDNP effect in Mr4511, the conserved cysteine was first replaced with the inactive serine resulting in the mutant Mr4511-C71S. Furthermore, tryptophan was introduced to different locations of the Mr4511-C71S protein, generating five additional mutants ( Table 1). The magnetic field dependence of the photo-CIDNP effect under liquid-state conditions obtained by 15 N and 1 H NMR is summarized in Fig. 5, plotted in the same way as in Fig. 3. The corresponding 15 N NMR spectra are shown in Figure S2 and the 15 N chemical shifts are provided in Table 2.   Figure S3 and S4). Time-resolved optical absorption analysis on F130W suggests that the triplet state 3 FMN is not formed to an observable extent in this protein, most likely, because of the ultrafast electron transfer between FMN and tryptophan (Aba Losi, personal communication). Accordingly, for the construction of light-induced artificial spin-machines pumping nuclear hyperpolarization, the information of a minimum distance of the redox partners of the SCRP is highly relevant.
Except for Mr4511-C71S-F130W ( r FW ~ 6 Å), all other Mr4511-C71S mutants generated both, the 15 N and 1 H photo-CIDNP effect under liquid-state conditions. As shown in Fig. 5B2, Mr4511-C71S-Y116W ( r FW ~ 9 Å) shows hyperpolarization for the nitrogen on FMN and tryptophan, therefore it is referred to as "tryptophanderived photo-CIDNP". This protein shows a maximal 1 H hyperpolarization at 2.4 T (Fig. 5B1), which is a higher field than observed for any other mutant (see below).
The r FW value of ~ 11 Å in the phototropin and aureochrome LOV domains was known to generate a photo-CIDNP effect 40,41 . Here we compare the two cases when the tryptophan is at the canonical position, Mr4511-C71S-Q112W ( r FW ~ 11 Å) (Fig. 5C1,C2) and the non-canonical position but with nearly the same r FW distance, Mr4511-C71S-Y129W ( r FW ~ 11 Å) (Fig. 5D1, D2). Although Mr4511-C71S-Q112W and Mr4511-C71S-Y129W exhibit a similar field maximum for the 1 H photo-CIDNP effect, their field-dependent 15 N photo-CIDNP effects are significantly different. This means that the efficiency of photo-CIDNP formation does not only depend on the spatial distance. Different orientations and different local mobility of the residues might be considered as the origin of this difference. However, the phase of hyperpolarized Trp N-1 signal is allways negative for mutants Y116W, Y129W and Q112W.
The tryptophan residue in Mr4511-C71S-K57W ( r FW ~ 17 Å) is the most remote electron donor from FMN in the studied set of mutants and, thus, the reaction rate constant of electron transfer in this protein mutant is expected to be the smallest in this series 62 . The protein shows a 15 N hyperpolarization solely on the FMN N-5, while the FMN N-10 and Trp N-1 signals do not exhibit any enhancement in a wide range of magnetic fields (Fig. 5E). The same photo-CIDNP experiments were also performed on Mr4511-C71S in which no tryptophan was present leading to very similar results (Fig. 5F) with somewhat weaker single emissive FMN N5 signal (Figure S2 graphs D and E). This surprising result clearly indicates that the 15 N photo-CIDNP effect reported here is not derived solely from the involvement of the added tryptophan residue. It is well-known that tyrosine can also act as an electron donor in biological systems 63,64 . According to the amino-acid sequence and our structural model of Mr4511, there are four tyrosine residues located in proximity to FMN in the range of 9 to 12 Å. In line with this speculation is the fact that the 15 N photo-CIDNP effect is not observable for tyrosine, since tyrosine does not contain 15 N in the side chain. Therefore, the light-driven molecular spin-machines can probably also rely on SCRPs containing a tyrosine radical.

Discussion
Here, we show that a photo-CIDNP effect originating from the SCRP of FMN and tryptophan can be produced in artificially designed flavoproteins. We employ a systematic mutation strategy to vary reaction and magnetic parameters of the paramagnetic centers generated by light. It appears that the r FW distance of ~ 6 Å between the FMN and tryptophan is too short to provide conditions suitable for photo-CIDNP formation, whereas in the range of ~ 9 Å to ~ 11 Å, the effect has been observed ( Fig. 5; Table 2). The data on the field dependence of the photo-CIDNP effect generated by the designed LOV domains show complex dependencies, which are not expected for the liquid-state photo-CIDNP effect. In particular, the differences in field dependence obtained for LOV domains having the same distance of donor and acceptor suggest that anisotropic spin interactions come into play as they are expected for solids. In addition to the field dependence, a distance dependence has been documented. Apparently, further parameters are involved, presumably the relative orientation of donor and acceptor as well as their local dynamics. Both, anisotropy and relaxation effects require further studies. Furthermore, the effect of different label patterns requires a future study.
The presence of solid-state mechanisms in LOV domains in liquids implies that the transient SCRP occurs in a slow-motion regime, during which the anisotropic electron-nuclear interactions are conserved for the build-up of hyperpolarization. In contrast, on NMR time scale, all the anisotropic nuclear interactions, i.e., nuclear dipolar coupling and chemical shift anisotropy as present in solids are averaged out and thus the hyperpolarized signals in the NMR spectra exhibit no obvious anisotropic features. Such phenomenon was previously observed for a Regarding the 1 H photo-CIDNP, the region of chemical shifts ranging from − 2 to 10 ppm of the spectrum is integrated and plotted against the magnetic field (A1-F1). The three 15 N photo-CIDNP hyperpolarized signals correspond to FMN N-5 (green), FMN N-10 (yellow) and Trp N-1 (blue); the signals are plotted against the magnetic field strength (A2-F2) in the same range and scale as for the 1 H signal. For straightforward comparison, the signal strength at the maximal hyperpolarization magnitude of each graph is normalized to unity. The positive and negative signs of the y-axis indicate absorptive and emissive hyperpolarization, respectively. Since the mutant F130W shows no detectable photo-CIDNP effect, as presented in the Supplementary Information, Figures S3 and S4, the data points shown in (A1) and (A2) were measured the over the same spectra regions with the same noise level as in the corresponding spectra for Y129W shown in (D1) and (D2).

Scientific Reports
| (2020) 10:18658 | https://doi.org/10.1038/s41598-020-75627-z www.nature.com/scientificreports/ photosynthetic RC protein-membrane complex corresponding to ~ 1 MDa, measured by 13 C liquid-state NMR 28 . For LOV proteins, having the molecular weight of less than 20 kDa, the occurrence of anisotropic mechanisms in liquids likely relies on the formation of dimers 53,57,65 or higher multimers in solution. The slow tumbling rate may lead to the presence of residual proton-proton couplings which allow for the 1 H hyperpolarization transfer from the center of the photo-reaction into the environment ( Fig. 2A). So far, flavoproteins and photosynthetic RCs are the only reported electron transfer protein systems that show solid-state photo-CIDNP effect. Even despite the different cofactor arrangements and spin-dynamics, they might share the same mechanisms. Consequently, similar features of CIDNP are expected regarding the sign change of nuclear spin hyperpolarization and the similar field at which the maximum polarization occurs 22,23 . LCs and LACs analysis suggested that a solid-state photo-CIDNP effect is not only field-dependent, but also strongly orientation-dependent because of the anisotropic interactions governing in spin dynamics of the SCRP in solids 36,37 . To the present experiments conducted under liquid-state conditions, the same theory will be applied to understand the sign change that occurred in the LOV proteins as shown in Fig. 3B,C as well as Fig. 5B2-D2.
Summarizing these considerations, we can propose the following interpretation of the experimental findings. By increasing r FW , we decrease two parameters: The SCRP recombination rate and the electron-electron spin-spin coupling, J SCRP , within the SCRP. When the r FW distance is too short, photo-CIDNP formation is suppressed, most probably, due to two reasons: The first reason is that J SCRP is too large, introducing an energy gap between the singlet and triplet SCRP spin states. This gap cannot be overcome by the relatively small hyperfine couplings, and singlet-triplet interconversion in the SCRP is thus suppressed. The second reason is that the spin-evolution of the SCRP requires sufficient time for photo-CIDNP formation: fast SCRP recombination interrupts this process and thus no photo-CIDNP is formed.
As r FW increases further, we enter the regime in which the SCRP lifetime is sufficiently long and J SCRP is sizeable, but not too large to suppress singlet-triplet mixing, giving rise to photo-CIDNP formation. In this situation, the TSM scenario comes into play and the maximum position, B max , in the photo-CIDNP field dependence is given 37,39 by the matching condition |γ N |B max ≈ J SCRP , with γ N being the nuclear gyromagnetic ratio; the sign of polarization of the three different kinds of nuclei ( 1 H, 13 C and 15 N) is also consistent with previous theoretical considerations 37,39 . Hence, the B max field is different for different nuclei, which is consistent with the experimental data.
When r FW increases further, J SCRP is decreased and other photo-CIDNP mechanisms 37 come into play. In this situation, polarization formation is no longer sensitive to the γ N value, i.e., to the nuclear spin isotopes, so that different kinds of nuclei exhibit a similar behavior. The sign changes of photo-CIDNP can be rationalized in terms of changing dominance of enhancement mechanisms, as it happens in RCs 37 .
The design strategy also leads to the discovery of a new-type of photo-CIDNP effect generated by Mr4511-C71S in which no tryptophan is present. The same effect (Fig. 5E,F) also occurs in Mr4511-C71S-K57W ( r FW ~ 17 Å). Based on the present results, we are unable to unravel the origin of the new-type photo-CIDNP effect. Recent research on a designed cysteine-lacking Asphot1-LOV2 domain indicated that, without presence of the tryptophan, the FMN was reduced, however to less extend compared to the case when the tryptophan was present. Kinetic data suggested that one of the tyrosines in the LOV domain acts as counter radical 66 . Therefore, we proposed that tyrosine might act as electron donor in the absence of tryptophan also in our case.
With this, the present work significantly extends the class of light-driven molecular spin machines, which pump nuclear spin-hyperpolarization. The LOV systems are particularly suitable for such biomimetic design, while photosynthetic RCs due to their structural complexity allow for limited manipulations only. The biomimetic design that affects the parameters of the photo-CIDNP effect provides new categories of experiments to analyze the conditions for its occurrence.

Methods
Protein preparation. The plasmid (i) encoding the LOV-C287S mutant of aureochrome1a from P. tricornutum comprising the flanking Jα and A'α helices (238-378) was received from Peter Kroth (University of Konstanz) 53 . The plasmid (ii) that encodes Mr4511 from M. radiotolerans (1-164) was generated in our own group by genome cloning 51 Table S3. All genetic manipulations were according to standard protocols. Plasmid (iii) encodes the LOV1-C57S mutant of phototropin from C. reinhardtii (16-133) carrying a 15 × His-tag at the N-terminus 67 . Further information about all the mutants employed in this work is summarized in Table 1. The protocol of heterologous overexpression and isotope-labelling of all the mutants in Escherichia coli has been reported elsewhere 39 13 C 6 ] glucose as the sole source in the growth medium yields a uniformly 15 N or 13 C labelled protein, while supplementation of 13 C or 15 N isotope-enriched indole as precursor to the normal medium results in a selective labelling of the tryptophan side chain. The 15 N and 13 C labelled proteins were used for corresponding 15 N and 13 C NMR measurements. For 1 H NMR measurement, the employed proteins are in their natural abundance, and they were washed with deuterated phosphate buffer (300 mM NaCl, 50 mM KsPO 4 in D 2 O, pH 8.0) to a final solution containing approximately 0.4% residual protons. The final concentration of the flavoproteins were controlled at about 16 μM ( ε 450nm = 12,500 M −1 cm −1 ) before photo-CIDNP measurement.

Scientific Reports
| (2020) 10:18658 | https://doi.org/10.1038/s41598-020-75627-z www.nature.com/scientificreports/ Photo-CIDNP solution-state NMR. The field-dependent 15 N-, 13 C-, and 1 H photo-CIDNP experiments of the LOV proteins were carried out on an NMR spectrometer operating at a magnetic field of 9.4 T ( 1 H frequency of 400 MHz) (Bruker Avance III HD) equipped with a home-built field-cycling device 47 . It transfers the sample to variable magnetic fields within the range 10 nT < B o < 9.4 T at which the sample is illuminated and returns it for the NMR detection at 9.4 T. For the 13 C and 15 N photo-CIDNP experiments, pulse-acquire with WALTZ-16 proton decoupling was employed. The pulse sequence of the 1 H photo-CIDNP experiments starts with a pre-saturation composite pulse sequence 68 at 9.4 T, followed by the sample shuttle cycle that includes the sample transfer to the chosen magnetic field for illumination by LED (called "light") or the same cycle without illumination (called "dark") during the fixed time and the return to 9.4 T, and it ends with the detection sequence. For all photo-CIDNP experiments, the samples were measured in dark and light, respectively, with the same number of scans. The illumination source was a 400-nm 2-W LED (Chanzon, China) and the illumination time was optimized to 0.5 s. A fresh aliquot of the sample stock was taken for a measurement at each magnetic field to compensate the effect of photo-bleaching. The temperature was 289 K for all samples with the exception that Ptaureo1a-LOV-C287S was measured at 277 K. The line-broadening for 15 N and 13 C NMR spectra were set to 30 Hz and for 1 H spectra was set to 1 Hz. The 15 N and 13 C NMR spectra were phased to the external standard, a mixture of 0.1 M 15 N labelled urea and 0.1 M 13 C labelled methanol in DMSO. The chemical shifts of 15 N NMR spectra were relative to liquid ammonia and referenced externally to urea 15 N at 76.3 ppm 69 . To present the field-dependence of the photo-CIDNP effect, the selected hyperpolarized signal was integrated and plotted against the field at which the sample was illuminated. For straightforward comparison, the signal strength at the hyperpolarization maximum is set to unity and the other signals are normalized to this value. The positive and negative signs of the y-axis indicate absorptive and emissive hyperpolarization, respectively. The error bars of the 1 H photo-CIDNP data represent the standard deviation of the mean value obtained from three measurements; the uncertainty of the 13 C and 15 N photo-CIDNP data represents the noise level relative to the corresponding hyperpolarized signal area. The spectra shown in Fig. 2 and Supplementary Information Figure S1, S2, and S4 were created with OriginPro Version 2017.