The aminopeptidase TET2 from Pyrococcus horikoshii (UniProt entry O59196) was produced via overexpression of the pET41c-PhTET2 plasmid in Escherichia coli BL21(DE3) RIL cells. The plasmid is available from AddGene (https://www.addgene.org/182428/, last access: 15 February 2026). For the resonance assignment, tryptophans were mutated into phenylalanines in different combinations either as single mutant (mutant 1: W106F) or as triple mutants (mutant 2 (only W106): W136F, W164F, W276F; mutant 3 (only W136): W106F, W164F, W276F; mutant 4 (only W164): W106F, W136F, W276F; mutant 5 (only W276): W106F, W136F, W164F). The 19F labelling was achieved either with 5-fluoroindole (Sigma-Aldrich, catalogue number F9108) in protonated medium (mutants) or with [5-13C,3,4,6-2H3]-5-fluoroanthranilic acid (5FC-anthranilic acid) in deuterated medium (wild type).

The 5-fluoroanthranilic acid isotopologue was synthesised in-house by adapting the synthetic route reported by to the present labelling scheme; details of the synthesis will be published elsewhere.

The plasmid (kanamycin resistance) was transformed into competent BL21(DE3) RIL cells (chloramphenicol resistance) via heat shock. Unless otherwise mentioned, all cultures contained kanamycin and chloramphenicol, and shaking was performed at 200 rpm and 37 °C.

5-fluoroindole labelling was achieved as follows. After precultures in LB medium and minimal M9 medium, the main culture was inoculated to an optical density at 600 nm (OD₆₀₀) of 0.2 and shaken until it reached 0.6–0.7. At this point, 1 gL-1 glyphosate (abcr, Karlsruhe, Germany; catalogue number AB505195) was added, and the culture was grown for 15 min before 100 mgL-1 5-fluoroindole, 50 mgL-1 L-tyrosine (Sigma T3754), and 50 mgL-1 L-phenylalanine (Sigma 78019) were added. The culture was grown for 45 min, and expression was induced with 1 mM isopropyl-β-D-thiogalactopyranosid (IPTG). Cells were harvested at 6500 rcf for 15 min after 4 h of shaking.

Labelling with 5FC-anthranilic acid was achieved as follows. Cells were adjusted to deuterated M9 medium by growth in consecutive precultures of LB medium and M9 medium prepared with 100 % H2O, 50 % H2O/50 % D2O, and 100 % D2O. The final preculture and the main culture were prepared with 15NH4Cl and D-2H7-glucose. The main culture was inoculated to an OD₆₀₀ of 0.2 and shaken until it reached 0.6–0.7. At this point, 1 gL-1 glyphosate, 50 mgL-1 L-tyrosine, 50 mgL-1 L-phenylalanine, and 15 mgL-1 5FC-anthranilic acid were added, and the culture was shaken for 40 min. The temperature was then reduced to 28 °C for 15 min before induction with 1 mM IPTG. Cells were grown overnight at 28 °C and then harvested at 6500 rcf for 15 min.

The cell pellet was re-suspended in lysis buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 0.05 mgmL-1 DNase, 2 mM MgCl2, 0.025 mgmL-1 RNase, and 0.5 tablets cOmplete EDTA-free protease inhibitor), kept on ice for 30 min, and sonicated for 2 min. A heat shock was performed at 80 °C for 15 min. After the addition of 10 mL buffer A (20 mM Tris-HCl pH 7.5, 100 mM NaCl), the cell debris was collected by centrifugation at 46 000 rcf for 40 min at 4 °C. The supernatant was washed with buffer A using an Amicon ultra centrifugal filter with a molecular weight cutoff of 100 kDa before loading onto a RESOURCE Q column (Cytiva) and eluted with a gradient over 10 column volumes from buffer A to buffer B (20 mM Tris-HCl pH 7.5, 1 M NaCl). The fractions containing the protein were concentrated, loaded onto a HiLoad 10/300 Superdex 200 pg column (Cytiva), and eluted in buffer A.

Samples for solid-state MAS NMR were prepared by batch crystallisation of TET2 with 2-methyl-2,4-pentanediol (MPD; Sigma 68340) . A solution of 10 mgmL-1 protein and the paramagnetic compound (where applicable; Gd(DTPA): Sigma 381667, Gd(DTPA-BMA): GE Healthcare Omniscan (contains 4.2 % NaCa(DTPA-BMA)) in buffer A was mixed with MPD in a v/v ratio of 1:1. The concentration given for the paramagnetic compound in the following is related to the final concentration in the sample, including MPD. The microcrystals were filled into a 1.3 mM MAS rotor (Bruker) by ultracentrifugation overnight at 4 °C and 68 000 rcf.

MAS NMR experiments were performed on a Bruker Avance Neo spectrometer operating at 14.09 T (600 MHz 1H Larmor frequency). A triple-resonance HFX probe head from PhoenixNMR equipped with a 1.3 mm MAS stator from Bruker was used, with the X channel tuned to 13C. Temperature calibration was done with an external 2H4-methanol sample , and chemical shift referencing was done indirectly via the 1H signal of 2,2-dimethyl-2-silapentane-5-sulfonate sodium salt (DSS). All experiments were performed at a MAS frequency of 55.555 kHz and a sample temperature of approximately 309 K. Spectra were processed with Bruker Topspin software (versions 4.1.4 and 4.5.0).

Pulse sequence diagrams can be found in Fig. . All experiments were performed with 19F detection (10 ms acquisition time), and 1H and 13C decoupling unless stated otherwise. Composite pulse decoupling during acquisition was typically achieved with 10 kHz swfTPPM on 1H and 10 kHz WALTZ-16 on 13C. The recycle delay was set to ≈1.3⋅T1 of 19F, depending on the concentration of the paramagnetic compound, except for the 19F saturation recovery experiment, in which the recycle delay was set to 1.2 s. The pre-saturation block in the saturation recovery experiment was repeated n=50 times with a delay Δ=4.5ms. Magnetisation transfer in 13C relaxation experiments was achieved via dipolar 19F-13C cross-polarisation (CP) steps. Typical CP spin-lock radio-frequency field strengths were 40 kHz on 13C and 90 kHz on 19F, with a linear ramp of 90 %–100 % and a transfer time of 400 µs. 19F and 13C relaxation experiments were recorded as pseudo two-dimensional spectra with one 19F frequency dimension and one pseudo dimension in which a relaxation delay τ was incremented.

Spectra were processed as pseudo two-dimensional spectra in Topspin and converted to UCSF format with the bruk2ucsf program provided in Sparky . Python scripts for analysis of relaxation rate constants were written in-house, utilising the Nmrglue package . Spectra were split into individual 1D slices, and peaks were fitted with the routine implemented in Nmrglue. The intensities were then fitted to a mono-exponential function. Errors were determined by Monte Carlo analysis (500 iterations) using one standard deviation of the spectral noise.

TET2 is a dodecameric aminopeptidase from P. horikoshii that has been studied extensively by MAS NMR previously . Each of the 12 identical subunits contains four tryptophan residues: W106, W136, W164, and W276 (Fig. a). To achieve fluorine labelling of the aromatic ring, we expressed TET2 in deuterated M9 medium and added 5-fluoroanthranilic acid, which is metabolised by the bacteria into 5-fluoro-L-tryptophan (Fig. b). This precursor has a 19F-13C spin pair at position 5 in the aromatic ring and is deuterated at positions 3, 4, and 6. We will refer to the labelled protein as 5FC-W-TET in the remainder of this discussion. As expected, the 19F MAS NMR spectrum of 5FC-W-TET shows four individual peaks (Fig. c, upper spectrum). For the assignment of the signals, we prepared five Trp to Phe mutants in non-deuterated medium using the commercially available precursor 5-fluoroindole (see the Methods section for details). The respective spectra show either one (triple mutants) or three (single mutant) signals, which allowed us to assign the four signals (Fig. c, bottom spectra). Note that the spectrum of the deuterated wild type is shifted due to an isotope shift .

In recent years, several studies have been published using biomolecular 19F MAS NMR to study structural aspects of proteins . The 19F T1 relaxation time was often reported to be several seconds long, leading to a long τr.d.opt to obtain an optimal SNR. We measured 19F T1 of 5FC-W-TET with a saturation recovery experiment (Fig. d), and obtained values between 1.67±0.08s (W164) and 6.0±0.4s (W136). Considering the highest value, this would correspond to a recycle delay of τr.d.opt=7.56s.

To reduce the 19F T1, we used paramagnetic doping with two different Gd3+ chelates. We prepared samples with six concentrations of Gd(DTPA-BMA) (0, 2, 4, 6, 8, and 16 mM, Fig. e), and measured 19F T1 and T2, and 13C T1 and T2 for each residue (Figs. S1, S2, S3, S4, S6 in the Supplement). For Gd(DTPA) (Fig. f), we prepared samples with 2 and 8 mM of the compound and measured 19F T1 and T2 (Fig. S5, S6). The bulk relaxation rates R=T-1 (average over all residues) are shown in Fig. a–d.

At 8 mM Gd(DTPA), the increase in R2 led to a broadening of the whole spectrum that made the individual peaks indistinguishable (Fig. S7). Additionally, Gd(DTPA) changed the crystallisation behaviour of TET2, possibly due to binding to the protein surface (Fig. S8) . We therefore refrained from preparing samples with other concentrations or measuring 13C relaxation rate constants.

We find that both compounds lead to an increase in R1 as well as R2. To quantify the effect of the two compounds, we determined the longitudinal and transverse PREs (Γ1 and Γ2), which are given by the slope of a linear fit of the respective relaxation rate constants as a function of the concentration of paramagnetic dopant. We performed fits of the bulk and per-residue relaxation rate constants to obtain the PREs for Gd(DTPA-BMA) (19F and 13C; Fig. e–f) and for Gd(DTPA) (only 19F; Fig. g–h).

As expected, both Γ1 and Γ2 are smaller for 13C than for 19F. This is due to the smaller gyromagnetic ratio of the carbon nucleus (Eq. ).

Interestingly, the measured 19F PREs for Gd(DTPA) are much higher than for Gd(DTPA-BMA). The bulk effect of Gd(DTPA) on R1 (Γ1=0.743±0.015s-1mM-1) is roughly seven times stronger than the effect of Gd(DTPA-BMA) (Γ1=0.101±0.028s-1mM-1). For Γ2, the difference is even bigger: The bulk value for Gd(DTPA) (Γ2=556.2±0.7s-1mM-1) is over 14 times higher than the value for Gd(DTPA-BMA) (Γ2=38.2±0.6s-1mM-1). The differences in relaxation behaviour are likely the result of the specific physicochemical properties of the two compounds (e.g. the slower water exchange rate of Gd(DTPA-BMA)) .

We find that in our case, paramagnetic doping with 8 mM Gd(DTPA-BMA) is the best compromise between a significant reduction of 19F T1 without heavily compromising T2 and therefore spectral resolution. The decrease in bulk T1 from 3.2±1.3s at 0 mM to 0.92±0.25s at 8 mM Gd(DTPA-BMA) translates to a reduction of τr.d.opt from 4.0±1.6 to 1.2±0.4s. The more than 3-fold shorter recycle delay significantly reduces the measurement times of 19F MAS NMR spectra, or, in other words, increases the SNR per unit time. Although the same effect could be achieved with lower concentrations of Gd(DTPA), we prefer the use of Gd(DTPA-BMA) due to a potential interaction of Gd(DTPA) with the protein in our case (see above). Figure  shows a comparison of the 19F spectra of 5FC-W-TET with and without dopant (13C spectra are shown in Fig. S9).

The observed PREs for the individual tryptophans differ (Fig. e–h), which can be expected as they are impacted by a multitude of factors, such as the solvent accessibility of the residue, its dynamics, or the density of surrounding protons . The influences of these factors can conversely even be used to characterise the surfaces and interaction interfaces of proteins with so-called solvent PREs .

To rationalise the observed PRE data for the four Trp sites, we compared them to structural parameters. We reasoned that the relaxation properties may be impacted by the 1H spins surrounding each of the 19F Trp sites and calculated the root-sum-square dipolar coupling drss (Fig. S10a). This parameter approximates the effective dipolar coupling network and can serve as an indicator for spin diffusion, which influences the propagation of the PRE effect throughout the protein. It was calculated as the square root of the sum of squared dipolar couplings between a fluorine and all back-exchangeable protons . Moreover, we calculated the solvent accessible surface area (SASA) of each tryptophan (Fig. S10b), which is an approximate measure of the shortest distance between the paramagnetic compound and a given 19F atom. In our case, we did not find a direct correlation of these parameters with the observed PREs (Fig. S10c–d).

Interestingly, the PRE patterns (relative strength of the measured PREs for the four residues) are different for different relaxation rate constants (Γ1 and Γ2), different compounds, and different nuclei (19F and 13C) (Fig. e–h). In addition to the factors mentioned above, other influencing parameters that could explain these different patterns include a lower spin diffusion efficiency for 13C, specific binding of the Gd3+ complex to the protein, and properties of the compound, such as the rotational correlation time τc or the water exchange rate.

Differences between the patterns observed for Gd(DTPA-BMA) and Gd(DTPA) could be due to the specific binding of one of the complexes to the protein. A decrease in the distance r between the compound and residues close to the binding site would increase the observed PRE effects as both Γ1 and Γ2 are proportional to r-6 (Eq. ). The binding would also decrease the τc of the compound, which could, in combination with the dynamics of a specific tryptophan, lead to dampening or acceleration of the PRE.

The most striking observation is that the 19F Γ1 of W164 is very high relative to the other residues compared to 19F Γ2 for both Gd3+ complexes. Such a difference between the longitudinal and transverse PRE can also be a result of a local reduction of τc due to binding. As Γ1 and Γ2 sample the spectral density at different frequencies, their reaction to changes of τc is not the same (Eq. ) . The complex interplay of diverse parameters makes it difficult to understand these patterns in detail.