Aromatic ¹⁹F-¹³C TROSY: A background-free approach to probe biomolecular structure, function and dynamics

¹⁹F-¹³C DD-CSA cross-correlation in 3-¹⁹F¹³C Tyr and the TROSY effect

While R₂ in spin ½ nuclei like ¹⁹F and ¹³C is dominated by DD and CSA mechanisms, their individual contribution can vary. In the case of aromatic ¹⁹F-¹³C systems, the line width is dominated by the CSA mechanism, especially at high magnetic fields. We calculated chemical shielding tensors for ten ¹⁹F-¹³C spin pairs at various positions in aromatic amino acid side chains using DFT methods (Supplementary Note I; axiality and rhombicity parameters of the CSA tensors are collected in Supplementary Table 1) and then used the Bloch-Redfield-Wangsness relaxation theory module available in the Spinach²⁰ library to compute the R₂ relaxation rates for each ¹³C-¹⁹F pair (Supplementary Fig. 1). While there are variations in R₂, dependent on the amino acid type and ¹⁹F position, the calculations indicate that the R₂ of the TROSY component of ¹³C_F is 3 to 18-fold slower than that of the corresponding decoupled resonance. Thus, selective detection of the ¹³C_F TROSY resonance would yield sensitive, high-resolution NMR spectra for ¹⁹F-¹³C pairs in any aromatic amino acid.

CSA contribution to the R₂ scales quadratically with the magnetic field, while the DD contribution is much less sensitive to increase in magnetic field strength. This is reflected in the magnetic-field-dependent increase in R₂ calculated for decoupled resonances (e.g.: 3-¹⁹F Tyr; Fig.1, blue triangles). However, CSA relaxation that is comparable to DD relaxation often generates favourable DD-CSA cross-correlation, which makes one of the components of the ¹³C_F coupled doublet relax slower compared to the other (Fig. 1a, red squares). Which doublet component relaxes slower depends on the angles between the DD and the CSA tensor axes, and on the sign of the J-coupling. The narrower (slower relaxing) line is referred to as the transverse-relaxation-optimized “TROSY” line^{1, 2}. For the ¹³C_F resonances, our estimates indicate that the TROSY component of the ¹³C_F doublet relaxes 8.7 times slower (6.1 s⁻¹) compared to the decoupled signal (53 s⁻¹) for 3-¹⁹F Tyr incorporated in a 42 kDa protein at 600 MHz (¹H frequency; τ_c = 25 ns at 25 °C) (Fig. 1a). By comparison, the TROSY components of ¹³C_H resonances relax with a rate of 43 s⁻¹ (Fig. 1b) under identical conditions. On the other hand, the TROSY effect on the ¹⁹F_C and ¹H_C resonances is rather small (Fig. 1c and Fig. 1d). Thus, selective detection of TROSY resonances from ¹³C_F would allow detection of the ¹⁹F-¹³C pair with high sensitivity and resolution.

Experimental observation of ¹⁹F-¹³C TROSY effect from 1D ¹³C spectra

In order to harness the power of aromatic ¹⁹F-¹³C TROSY selection and to avoid splitting of ¹³C resonances by the one-bond C-C coupling (¹J_CC ~60Hz), we synthesized a selectively ¹³C-labeled and fluorinated tyrosine, 3,5-¹³C₂-3-fluoro-L-tyrosine, which has a ¹⁹F-¹³C pair at the 3^rd position in the aromatic ring (3-¹⁹F_13C Tyr), using a modified method described by Kitevski-LeBlanc et al.²¹. To demonstrate the applicability of the ¹⁹F-¹³C TROSY experiment, we expressed two model systems, a small protein, G B1 domain (GB1), and a larger protein system, maltose binding protein (MBP) in E. coli using 3-¹⁹F_13C Tyr.

GB1 (τ_c = 5 ns) which harbours three tyrosine residues, gives three resonance pairs in its 1D ¹³C spectrum, each of which are separated by ~240 Hz, the expected ¹⁹F -¹³C (¹J_CF) coupling value (Fig. 2a). The TROSY effect is clearly seen in these 1D spectra acquired at different field strengths. The TROSY resonances which are up-field are intense and narrow while the down-field anti-TROSY signals are weak and broad. ¹H decoupling was applied during acquisition to remove remote (>two-bond) ¹H couplings. Although our labelling technique removed the strong one-bond ¹³C-¹³C coupling (¹J_CC ~ 60Hz), the ¹³C-labeling at the 5^th position in the aromatic ring, originating from the stable-isotope precursor, caused a two-bond ¹³C-¹³C coupling (²J_CC ~7 Hz). The narrow TROSY lines make this small coupling clearly visible in the 1D ¹³C spectra as a testament to the resolution gain (Fig. 2a).

To compare the effect of TROSY on ¹³C_F and ¹³C_H resonances at equivalent positions, we recorded a 1D ¹³C spectrum of unlabelled GB1 using natural abundance ¹³C, to negate the ¹J_CC coupling. Although the TROSY effect can be seen clearly in the sharper, more intense up-field TROSY component of the ¹³C_H resonances, the linewidths are much broader than those of ¹³C_F (Fig. 2b), highlighting the effect of the significantly slower R₂ of ¹³C_F. It should be noted that to see the TROSY effect, ¹H decoupling could not be applied during ¹³C_H detection and therefore the remote ¹H couplings contribute to the line broadening. Nevertheless, unlike the very narrow ¹³C_F lines, which allowed resolution of the 7Hz ²J_CC couplings, the broader lines of the ¹³C_H nuclei obscure the individual detection of these remote couplings. Of note, TROSY resonance of ¹³C_F is the up-field component, whereas the TROSY resonance of ¹³C_H is the down-field component due to the opposite sign of the ¹⁹F-¹³C and ¹H-¹³C ¹J-coupling (negative and positive, respectively).

The 2D ¹⁹F-¹³C TROSY experiment

Our 1D-¹³C_F data indicated the need to choose the up-field ¹³C_F component for TROSY selection in a 2D ¹³C-¹⁹F TROSY experiment. Theoretical calculations showed that the ¹⁹F_C TROSY component relaxed more slowly as compared to the decoupled resonance. To observe the TROSY effect on ¹⁹F_C and choose the correct TROSY resonance we recorded decoupled and coupled ¹³C-¹⁹F HSQCs of MBP (Supplementary Fig. 2). The coupled spectrum showed that the up-field ¹⁹F component was narrower, which, when taken together with the 1D-¹³C results, indicates that the top-right resonance is the TROSY component.

To selectively choose the TROSY resonance we designed an experiment with an ST2PT pulse scheme for TROSY selection (¹³C-¹⁹F TROSY-SE), similar to that used for the ¹⁵N-¹H TROSY-HSQC²² (Supplementary Fig. 2). In this sensitivity enhanced pulse-sequence design, the experiment starts with and ends on ¹⁹F magnetization, the nucleus with the larger gyromagnetic ratio, with the ¹³C frequency encoded in the indirect dimension. However, while the experiment worked well for GB1, the experiment was not sufficiently sensitive for the higher molecular weight protein MBP (42 kDa), as we were not able to observe all the fifteen 3-¹⁹F_13C Tyr resonances at 25 °C. We hypothesized that this was due to faster relaxation of the ¹⁹F_C nuclei. The CSA of ¹⁹F_C accelerates its relaxation significantly and is the primary reason for the poor sensitivity observed in the ¹³C-¹⁹F TROSY-SE experiment. Indeed, shortening the experiment by removing the sensitivity enhancement block had no significant effect on signal intensity, as the reduction in time cancels the gain of the sensitivity enhancement block (Supplementary Fig. 3). The contributions of CSA from ¹⁹F and ¹³C nuclei to the R₂ of the coherences present in the ¹³C-¹⁹F TROSY pulse sequence are shown in Supplementary Fig. 4. Any coherence with transverse ¹⁹F magnetization (¹⁹F_x/y,¹⁹F_x/y¹³C_z,¹⁹F_x/y¹³C _x/y) experiences ~22-fold faster CSA relaxation as compared to that of ¹³C (¹³C_x/y,¹⁹F_z¹³C_x/y), suggesting that it would be beneficial to minimize the time ¹⁹F spends in the transverse plane. We estimated that, for a 42 kDa protein, ~30% of the initial magnetization is lost during the first INEPT transfer from ¹⁹F to ¹³C due to relaxation. Consequently, we shortened the pulse sequence by employing an out-and-stay design in which the experiment starts with excitation and indirect encoding of one of the nuclei, ¹³C or ¹⁹F, and ends with the direct detection and encoding of the other. Using this approach, we designed two additional experiments a ¹³C-¹⁹F TROSY, which begins with ¹³C magnetization and ends on ¹⁹F, and a ¹⁹F-¹³C TROSY, which starts with ¹⁹F magnetization and ends on ¹³C, the more slowly relaxing nucleus (Supplementary Fig. 5). These experiments utilize the ST2PT block where the phases of the first 90° pulse in ¹³C and ¹⁹F, respectively, are chosen to observe the upfield ¹³C_F TROSY component (Supplementary Fig. 6).

Although the ¹³C -¹⁹F TROSY was more sensitive than the ¹⁹F-¹³C TROSY, the latter had much better resolution (Fig. 3a–c). Fifteen intense and isolated ¹⁹F-¹³C pairs were detected by the out-and-stay experiments, which correspond to the fifteen Tyr residues in MBP. This resolution demonstrates the advantage of the 2D experiments, as both the ¹⁹F and ¹³C 1D-spectra of MBP cannot resolve individual resonances (Fig. 3d and 3e).

The ¹³C- and ¹⁹F-detected out-and-stay designs had substantially greater sensitivity than the ¹⁹F-detected out-and-back style method. In the ¹⁹F-¹³C TROSY pulse program, we take advantage of the large polarization of ¹⁹F with its initial magnetization and the faster longitudinal relaxation of ¹⁹F_C (T₁ ~0.5s) to allow faster recycling delays than are possible in a design where we start with ¹³C_F magnetization (T₁ ~1.4s) (Supplementary Fig. 7). Furthermore, by starting on ¹⁹F, the nuclei with the narrower ¹³C_F TROSY resonances are encoded in the direct dimension, which allows access to high-resolution without sacrificing experimental time.

The choice between the ¹⁹F and ¹³C detected out-and-stay experiments is dependent on the sample concentration, the rotational correlation time of the sample, the sample-dependent spectral width in the individual dimensions, the field strength and the probe design. Although we recorded both ¹⁹F- and ¹³C-detected experiments for comparison and have outlined the advantages of each above, we favour the ¹⁹F-¹³C TROSY pulse program. In this design, the maximum evolution in the indirect ¹⁹F dimension is limited to ~20 ms for sensitivity, and the more slowly relaxing ¹³C is encoded in the direct dimension for >200 ms, giving access to the high resolution unlocked by the ¹⁹F-¹³C TROSY effect.

Comparison of ¹⁹F-¹³C TROSY with ¹H-¹³C TROSY

The TROSY resonances of ¹³C_F in 3-¹⁹F Tyr are ~7 times narrower than the corresponding ¹³C_H for systems across a wide range of molecular weights (Supplementary Fig. 8). As an illustration, the linewidth of the ¹³C_F TROSY resonance of a protein with τ_c = 95 ns is narrower than the ¹³C_H TROSY resonance of a protein with τ_c = 15 ns. This is due to the more efficient cross-correlation process (or TROSY effect) in the ¹⁹F-¹³C system, where the magnitude of the CSA term in the spin Hamiltonian is comparable to the dipole-dipole interaction. Thus, selecting the TROSY component of ¹³C_F will yield sharper lines than those of ¹³C_H. To experimentally validate this, we took advantage of the presence of two ¹³C-labelled carbons in 3-¹⁹F_13C Tyr. While the carbon at position 3 is bonded to a ¹⁹F atom, the carbon at position 5 is bonded to ¹H. This labelling pattern allows the measurement of ¹H-¹³C and ¹⁹F-¹³C aromatic TROSY on the same sample. We recorded ¹³C-detected ¹⁹F-¹³C and ¹H-¹³C TROSY spectra of 3-¹⁹F_13C Tyr-labelled MBP at 25 °C and at 10 °C (Fig. 4). A 42 kDa MBP at 10 °C, behaves with the relaxation properties of a ~65 kDa protein at 25 °C. At 25 °C, in the ¹⁹F-¹³C TROSY we observe 15 sharp and well-resolved cross-peaks as expected for the 15 Tyr. By comparison, only 12 strong and 2 weaker peaks are observed in the ¹H-¹³C TROSY, and the cross-peaks are clearly broader when compared to their ¹⁹F-¹³C TROSY counterparts. At 10 °C, the advantage of the ¹⁹F-¹³C TROSY compared to the ¹H-¹³C TROSY is more pronounced. ¹⁹F-¹³C TROSY lines are moderately broadened at 10 °C, while the linewidths of the ¹H-¹³C TROSY cross-peaks are severely broadened, so that only two cross-peaks can still be recognized.

The effect of molecular weight on the relaxation of the ¹⁹F-¹³C TROSY component can be further highlighted by comparing the TROSY and the anti-TROSY resonances. (Supplementary Fig. 9). A larger molecular weight system can be mimicked by observation at lower temperatures. The linewidth and intensity of the TROSY resonances were barely affected by lowering the temperature from 25 °C to 10 °C, whereas the anti-TROSY resonances were severely broadened and diminished in intensity.

In addition to the fifteen intense peaks observed for MBP, there are some additional weaker signals observed in the ¹⁹F-¹³C TROSY spectrum (Supplementary Fig. 10). Rotation along the Cγ-Cβ bond is commonly referred to as an aromatic ring flip and has been previously observed for several aromatic amino acids (non-fluorinated). Some of these ring flips have been shown to be in slow exchange on the chemical shift time scale²³. In unlabelled Tyr residues this leads to two distinct signals of equal intensity due to the symmetry of the aromatic ring. In contrast, 3-¹⁹F_13C Tyr is no longer symmetric and the differences between ¹⁹F and ¹H, e.g. the ability of fluorine to hydrogen bond, could shift the equilibrium towards one conformation, producing the observed additional peaks of lesser intensity

To show the applicability of the ¹⁹F-¹³C TROSY to a more challenging and biologically relevant system, we prepared the 3-¹⁹F_13C labelled 180-kDa α7 single-ring of the 20S proteasome core particle (CP) from Thermoplasma acidophilum³. This particle lacks residues 97-103 (including one Tyr), to prevent formation of the α7-α7 subunit¹⁹, and thus, our final construct contained 10 Tyr residues. ¹⁹F-¹³C TROSY spectra at 45 °C (τc = 66 ns), 50 °C (τc = 60 ns)^{3, 24} and 60 °C (τc = 50 ns), show 8-10 discrete 3-¹⁹F_13C Tyr signals (Figure 5).

¹⁹F-¹³C TROSY effect in ribonucleic acids

The aromatic ¹H-¹³C TROSY effect has been previously harnessed to obtain narrow lines in nucleic acid NMR²⁵, and independently, ¹⁹F-substituted nucleobases have been used in the study of nucleic acids, including in cell NMR applications^{13, 26}. Therefore, we posited that the aromatic ¹⁹F-¹³C TROSY described here could also be applied to nucleic acids. We calculated R₂ to determine the TROSY effect for guanine, uracil, thymine, cytosine and adenine nucleobases, with ¹⁹F substitutions in eight different positions (Supplementary Fig. 11). For the following discussion, we will consider a fluorinated nucleotide and its non-fluorinated counterpart, to be part of a nucleic acid with a τ_c of 25 ns measured at 600 MHz.

For a nucleobase harbouring a ¹⁹F-¹³C pair, our calculations predicted that the TROSY component of ¹³C_F would, on average, relax 4 times slower than the equivalent decoupled resonance. Similar to the aromatic amino acids, the cross-correlation of ¹³C-¹⁹F pairs is more effective than that of ¹³C-¹H pairs in nucleic acids (e.g. 5-¹⁹F Uracil, Supplementary Fig. 12). This makes the R₂ of the TROSY component on average 3 times slower for ¹³C_F than that of ¹³C_H at the same position. 2-fluoroadenine represents a special case, where the TROSY component of ¹³C_F relaxes 58 times slower as compared to the decoupled resonance, yielding theoretically predicted linewidths of < 1Hz. This ¹³C_F TROSY relaxation rate is approximately 29 times slower than the TROSY component of the corresponding ¹³C_H. Intriguingly, the ¹⁹F_C R₂ rates of several nucleotide configurations are considerably slower than those of aromatic amino acids. For example, the ¹⁹F_C relaxation rate of the TROSY component in 5-fluorocytosine is only 181 Hz in comparison to the 520 Hz of 3-¹⁹F Tyr.

To experimentally observe the ¹³C-¹⁹F TROSY effect in nucleic acids, we synthesized a double stranded 16-mer 5-fluorouracil-substituted DNA. We used natural abundance ¹³C to record ¹⁹F-¹³C correlations at 5 °C (Fig. 6). An approximate rotational correlation time of 11.6 ns was calculated using HYDRONMR²⁷. The coupled ¹⁹F-¹³C HSQC spectrum shows four resonances with a sharp and intense top right component. Similar to what was previously observed for 3-¹⁹F_13C Tyr, the upfield shifted carbon components are clearly narrower than the downfield counterparts. Figure 6b and c, show selection for the TROSY and anti-TROSY components in the 5-fluoruracil 16-mer at 5 °C, respectively. The TROSY effect can also be observed for the upfield ¹⁹F_C component, confirming the prediction of an effective ¹⁹F_C TROSY for 5-fluorouracil.

Synthesis of 3,5-¹³C₂-3-fluoro-L-tyrosine (3-F Tyr)

Selectfluor® was purchased from Fluorochem (Hadfield, UK). HPLC-grade acetonitrile and methanol as well as all other reagents were purchased from Labimex Ltd (Sofia, Bulgaria) and used without further purification. 3,5-¹³C₂-L-tyrosine was purchased from Cambridge Isotope Laboratories (Tewksbury, MA). Preparative HPLC was performed on a Waters® 600 Delta system equipped with a Phenomenex® Luna C-18 column (250 × 21 mm, 5μ) and electrospray ionization mass spectrometry (ESI-MS) detection in the positive mode using a Waters® Micromass ZQ 2000. NMR spectra were recorded on a Bruker 500 MHz spectrometer equipped with an AvanceIII console.

The synthesis follows a modified procedure of Kitevski-LeBlanc et al.²¹ (Supplementary Note II). L-tyrosine (phenol-3,5-¹³C₂, 95-99%) (100 mg, 0.55 mmol) was suspended in 20 mL of acetonitrile:water (4:1) in a screw cap vial and heated to 80°C for 30 minutes in a polyethylene glycol 400 bath. The vial was removed from the bath briefly, Selectfluor® (N-Fluoro-N’-chloromethyltriethylenediaminebis(tetrafluoro-borate)) (586 mg, 2.4 equivalents) was added, and the vial was returned to the heating bath. The suspension cleared within minutes and stirring was continued for 1h at 80°C. The course of the reaction was monitored by mass spectrometry. At the end of the 1 h incubation, the ratio of tyrosine to 3-fluorotyrosine and 3,5-difluorotyrosine was 30:60:10. Longer reaction times or excess of fluorinating agent resulted in larger fractions of difluorinated product, which could not be recycled, unlike unreacted tyrosine. The reaction solution was diluted with 20 mL of 0.1 M HCl and the acetonitrile was evaporated fully under reduced pressure. The remaining aqueous solution was filtered through a 0.45 μm PTFE syringe filter and purified in 5 mL volumes by preparative HPLC with ESI-MS detection. Gradient elution over a 15-min period at a flow rate of 13 mL/min with a starting solvent ratio of 1.25:1.25:97.5 (methanol:acetonitrile:0.1% HCOOH) and an ending solvent ratio of 10:10:80 (methanol:acetonitrile:0.1% HCOOH) was used. Retention times of tyrosine, 3-fluorotyrosine and 3,5-difluorotyrosine were 9.4 min, 10.3 min and 11.8 min, respectively. 3-fluorotyrosine-containing fractions were evaporated under reduced pressure to a film, which was then triturated with diethyl ether to yield 40 mg of tan powder (36% from tyrosine). In addition, another 15 mg of unreacted tyrosine was recovered.

¹H NMR (500 MHz CD₃OD:D₂O (1:1)) δ 3.10 (dd, J = 14.7, 7.5 Hz, 1H), δ 3.23 (dd, J = 14.7, 5.5 Hz, 1H), δ 4.08 (dd, J = 7.5, 5.5 Hz, 1H), δ 6.97 (d, J = 8.3, 1H), δ 6.98 (dq, J = 161.3, 8.7 Hz, 1H), δ 7.08 (ddd, J= 11.9, 5.6, 1.63 Hz). ¹³C NMR (126 MHz, CD₃OD:D₂O (1:1)) δ 118.20 (d, J = 7.2 Hz), δ 151.28 (dd, J = 245.3, 7.2 Hz). ¹⁹F NMR (471 MHz, CD₃OD:D₂O (1:1)) δ −136.5 (d, J = 245.3 Hz). LRMS calculated for C₇¹³C₂H₁₁FNO₃⁺ [M+H]⁺ 202.08, found 202.12.

Expression and purification of 3-F Tyr GB1, MBP and the single-ring a7 particle of the 20S proteasome core particle:

BL21 (DE3) E. coli were transformed with plasmids encoding the protein G B1 domain (GB1; pET9d), maltose binding protein (MBP; pMALC4X), or the single-ring α7 particle (α7, modified pET³). Cultures were grown at 37 °C in a shaker-incubator in M9 minimal medium supplemented with 2 g/L ¹³C-glucose and 1 g/L ¹⁵NH₄Cl. To achieve uniform incorporation of 3-¹⁹F¹³C tyrosine, cells were first grown to an optical density of 0.5 measured at a wavelength of 600 nm (OD₆₀₀). Then 1 g/L glyphosate was added, along with 50 mg/L 3-¹⁹F¹³C l-tyrosine, 50 mg/L l-phenylalanine, and 50 mg/L l-tryptophan for GB1 and MBP expression. α7 was expressed with 80% 3-¹⁹F¹³C l-tyrosine labelling. To this end 70 mg/L l-phenylalanine, 70 mg/L l-tryptophan, 5 mg/L l-tyrosine and 30 mg/L 3-¹⁹F¹³C l-tyrosine were added alongside 1 g/L glyphosate. Cultures were grown to an OD₆₀₀ of 0.7-0.8 at which point protein expression was induced with 1 mM isopropyl β-d-1-thiogalactopyranoside (IPTG). The cultures were incubated for an additional 16 h at 28 °C (MBP and GB1) or 30 °C (α7) to allow for protein expression. Cells were pelleted by centrifugation for 20 min at 4 °C at 3,500 × g.

GB1-expressing cell pellets were resuspended in 40 mL of GB1 lysis buffer (50 mM Tris-HCl (pH 8.0), 350 mM NaCl, 10 mM imidazole, 5 mM β-mercaptoethanol (β-ME)) per liter of original bacterial culture and lysed by sonication. Cell debris was removed from the crude lysate by centrifugation at 4 °C for 40 min at 33,000 × g. GB1 was purified from the resulting supernatant by gravity-flow affinity chromatography using 5 mL (10-mL of a 50% slurry) of Ni-NTA resin (Qiagen). After washing the resin with 40 mL of 50 mM Tris-HCl (pH 8.0), 350 mM NaCl, 40 mM imidazole, and 5 mM β-ME, the protein was eluted in an identical buffer containing 350 mM imidazole. GB1 was then further purified using size exclusion chromatography (GE Healthcare Life Sciences, Superdex 75 10/300 GL pre-packed size exclusion chromatography column) into NMR buffer (10 mM Na₂HPO₄ (pH 6.5), 50 mM NaCl, 1mM EDTA).

MBP-expressing cell pellets were resuspended in 40 mL MBP lysis buffer (25 mM Tris-HCl (pH 8.0), 150 mM NaCl, 1 mM EDTA) per liter of original bacterial culture, lysed by sonication, and clarified by centrifugation for 40 min at 4 °C at 33,000 × g. The lysate was loaded onto 5 mL of amylose resin. After washing the resin with resuspension buffer, MBP was eluted with the addition of 40 mM d(+)-maltose. MBP was further purified using size exclusion chromatography (GE Healthcare Life Sciences, Superdex 75 10/300 GL pre-packed size exclusion chromatography column) into NMR buffer (10 mM HEPES (pH 6.5), 1mM EDTA, 1mM β-cyclodextrin).

α7 cell pellets were disrupted by sonication, and the insoluble fraction was removed by centrifugation for 40 min at 16,000 rpm. The protein was initially purified by gravity-flow affinity chromatography using 5 ml (10-ml of a 50% slurry) of Ni-NTA resin (Qiagen). After washing the resin with 40 ml of 50 mM Tris-HCl (pH 8.0), 350 mM NaCl, 20 mM imidazole, and 5 mM β-ME, the protein was eluted in an identical buffer containing 350 mM imidazole. The elution fraction was dialyzed against a buffer containing 50 mM Tris-HCl (pH 8.0), NaCl (100 mM), and DTT (1 mM), and the His₆-tag was removed using TEV protease. The digested proteasome was further purified using size exclusion chromatography (GE Healthcare Life Sciences “Superdex 200 10/300 GL”) and eluted in 20 mM sodium phosphate buffer (pH: 6.6) with 50 mM NaCl and 0.5 mM EDTA. The α7 single-ring particle was concentrated to 143 μM, which corresponds to a monomer concentration of 1 mM.

Typically, a 1 L expression culture produced approximately 100 mg GB1 or MBP and 50 mg α7.

5-fluorouracil-substituted DNA:

The 16-mer, double-stranded DNA sample containing a single 5-fluorodeoxyuridine nucleotide was made from two individual strands obtained from IDT (Coralville, IA) as polyacrylamide-purified samples. The 5-fluorodeoxyuridine nucleotide is known to form a base pair with an adenosine nucleotide.

The DNA sequences of the individual strands are shown below.

5’ – GCT AGG /5F-dU/ CA ATA CTC G - 3’

5’ - CGA GTA TTG ACC TAG C – 3’

where /5F-dU/ is the 5-fluorodeoxyuridine nucleotide (refer to Supplementary Note III for the structure).

The samples were annealed in TE buffer (5 mM Tris-HCl (pH 8.0), 1 mM EDTA) by heating to 80 °C and cooling to 20 °C with a ramp of 1 °C/ 5 sec. The concentration of the double stranded DNA 16-mer was determined using an extinction coefficient of 265,113 L·mol⁻¹·cm^-1. The final concentration was 1.5 mM.

NMR spectroscopy

Experiments were performed on 1 mM GB1, 700 μM MBP, 1 mM α7 single-ring proteasome particle, or 1.5 mM 5-fluorouracil-substituted DNA. All NMR spectra were collected in the indicated NMR buffer at 298 K, unless otherwise noted. All the 2D NMR spectra were processed with NMRPipe⁵¹ and analysed with CcpNmr⁵². The 1D spectra were processed with Bruker TopSpin® (version 3.2).

1D ¹³C spectra on 3-¹⁹F¹³C Tyr GB1 and unlabelled GB1

The 1D ¹³C spectra of 3-¹⁹F¹³C Tyr GB1 (1 mM) were collected on 400, 500, 600, 800, and 900 MHz (¹H frequency) spectrometers. The 400 MHz spectrometer is a Varian spectrometer equipped with an INOVA console and a room temperature probe. The 500, 600, 800, and 900 MHz (¹H frequency) spectrometers are Bruker spectrometers, equipped with cryogenically cooled probes. ¹H decoupling was applied during the acquisition and recovery time periods. The 1D ¹³C spectra of unlabelled GB1 (1 mM) were collected on 500, 600, 800, and 900 MHz (¹H frequency) Bruker spectrometers, equipped with cryogenically cooled probes. ¹H decoupling was not applied on the unlabelled sample to allow for observation of the TROSY component of the natural abundant ¹³C_H resonance at the 3^rd position of Tyr.

2D ¹³C-¹⁹F correlation spectra on MBP.

The spectra of MBP were recorded on a 600 MHz (¹H frequency) Bruker spectrometer equipped with a 5mm triple resonance inverse CryoProbe (TCI, ¹H tunable to ¹⁹F) with ¹H/¹⁹F, ¹³C, and ¹⁵N frequencies. The ¹H channel was tuned to ¹⁹F. This configuration does not allow for ¹H decoupling.

For the ¹⁹F-detected out-and-back HSQC, the spectral width in the indirect ¹³C dimension was set to 5 ppm and the spectral width in the direct ¹⁹F dimension was set to 40 ppm in both decoupled and coupled experiments. 256 complex points (real + imaginary) were collected in the indirect ¹³C dimension (acquisition time = 170 ms) and 4096 complex points were collected in the direct ¹⁹F dimension (acquisition time = 180 ms). 200 and 512 scans were collected for the coupled and decoupled HSQC, respectively, with a recycling delay of 1 second. The carrier frequencies were centred at 151.5 and −134 ppm in the ¹³C and ¹⁹F dimensions, respectively.

For the out-and-back style TROSY experiments (including the sensitivity enhanced version) the spectral width in the indirect ¹³C dimension was set to 5 ppm and the spectral width in the direct dimension was set to 40 ppm. 128 complex points were collected in the indirect ¹³C dimension (acquisition time = 170 ms) and 4096 complex points were collected in the direct ¹⁹F dimension (acquisition time = 180 ms). The carrier frequencies were centred at 151.5 and −134.0 ppm in the ¹³C and ¹⁹F dimensions, respectively. 200 scans were collected with a recycling delay of 1 seconds. The regular ST2PT and the sensitivity enhanced versions of the experiments were recorded in a time-equivalent fashion with the same acquisition parameters.

Two different types of out-and-stay experiments were recorded with the following parameters:

1) ¹⁹F-detected, ¹³C-¹⁹F out-and-stay TROSY: Here the spectral width in the indirect ¹³C dimension was set to 20 ppm and the spectral width in the direct dimension was set to 40 ppm. 128 complex points were collected in the indirect ¹³C dimension (acquisition time = 21 ms) and 4096 complex points were collected in the direct ¹⁹F dimension (acquisition time = 90 ms). 80 scans were collected with a recycling delay of 2 seconds. The carrier frequency was centred at 151.5 ppm in the ¹³C dimension and −135.0 ppm in the ¹⁹F dimension.

2) ¹³C-detected, ¹⁹F-¹³C out-and-stay TROSY: Here the spectral width in the indirect ¹⁹F dimension was set to 10.6 ppm and the spectral width in the direct ¹³C dimension was set to 39 ppm. 116 complex points were collected in the indirect ¹⁹F dimension (acquisition time = 10 ms) and 4096 complex points were collected in the direct ¹³C dimension (acquisition time = 348 ms). 80 scans were collected with a recycling delay of 2 seconds. The carrier frequency was centred at 151.5 ppm in the ¹³C dimension and −135 ppm in the ¹⁹F dimension.

To mimic a higher molecular weight system, we recoded the ¹⁹F-¹³C out-and-stay TROSY experiment on MBP at 10 °C. Here the spectral width in the indirect ¹⁹F dimension was set to 6.9 ppm and the spectral width in the direct ¹³C dimension was set to 39 ppm. 32 complex points were collected in the indirect ¹⁹F dimension (acquisition time = 4 ms) and 1024 complex points were collected in the direct ¹³C dimension (acquisition time = 87 ms). 400 scans were collected with a recycling delay of 1 seconds. The carrier frequency was centred at 151.5 ppm in the ¹³C dimension and −136.25 ppm in the ¹⁹F dimension. Spectra generated at 25 °C were truncated when compared to spectra recorded at 10 °C.

2D ¹³C-detected- ¹H-¹³C TROSY on MBP.

Spectra were recorded on 600 MHz (¹H frequency) Bruker spectrometers equipped with 5mm CPTCI and CPPTCI Prodigy Z-GRD Cryoprobes at 25 °C and 10 °C, respectively.

At 25 °C the spectral width in the indirect ¹H dimension was set to 2 ppm and the spectral width in the direct ¹³C dimension was set to 52 ppm. 72 complex points were collected in the indirect ¹H dimension (acquisition time = 30 ms) and 2048 complex points were collected in the direct ¹³C dimension (acquisition time = 131 ms). 160 scans were collected with a recycling delay of 1 second. The carrier frequency was centred at 125 ppm in the ¹³C dimension and 7 ppm in the ¹H dimension.

At 10 °C the spectral width in the indirect ¹H dimension was set to 2 ppm and the spectral width in the direct ¹³C dimension was set to 52 ppm. 128 complex points were collected in the indirect ¹H dimension (acquisition time = 53 ms) and 2048 complex points were collected in the direct ¹³C dimension (acquisition time = 131 ms). 400 scans were collected with a recycling delay of 1 second. The carrier frequency was centred at 125 ppm in the ¹³C dimension and 7 ppm in the ¹H dimension.

2D ¹⁹F-¹³C TROSY on the proteasome single-ring α7 particle.

Spectra were recorded on a 600 MHz (¹H frequency) Bruker spectrometer equipped with a 5mm inverse triple resonance CryoProbe (TCI, ¹H tunable to ¹⁹F) with ¹H/¹⁹F, ¹³C and ¹⁵N frequencies. The ¹H channel was tuned to ¹⁹F.

All experiments recorded were ¹³C-detected, ¹⁹F-¹³C out-and-stay TROSYs. The spectral width in the indirect ¹⁹F dimension was set to 8 ppm and the spectral width in the direct ¹³C dimension was set to 97 ppm. 28 complex points were collected in the indirect ¹⁹F dimension (acquisition time = 3 ms) and 2048 complex points were collected in the direct ¹³C dimension (acquisition time = 70 ms). 1024 scans were collected with a recycling delay of 1 second. The carrier frequency was centred at 151.5 ppm in the ¹³C dimension and −136.5 ppm in the ¹⁹F dimension. The total experiment time was ~8 hours.

2D ¹³C-¹⁹F correlation spectra on GB1 with ¹H decoupling.

The GB1 spectra were recorded on a 500 MHz (¹H frequency) Bruker spectrometer equipped with an Avance III HD console and 5mm triple resonance observe probe (TXO) tuned to ¹⁹F, ¹³C and ¹H frequencies. The spectral width in the indirect ¹³C dimension was set to 5 ppm and the spectral width in the direct dimension was set to 20 ppm. 150 points were collected in the indirect ¹³C dimension (acquisition time = 238 ms) and 2048 points were collected in the direct ¹⁹F dimension (acquisition time = 218 ms). 48 scans were collected with a recycling delay of 1 second. The carrier frequency was centred at 151.5 ppm in the ¹³C dimension and −134.0 ppm in the ¹⁹F dimension. The carrier frequency in the ¹H channel was set to 8.0 ppm. During ¹⁹F acquisition in the direct dimension, ¹H decoupling was applied using a WALTZ-16 scheme⁵³.

2D ¹³C-¹⁹F correlation spectra on 5-fluorouracil-substituted DNA.

The spectra of the DNA 16-mer containing a 5-fluorodeoxyuridine nucleotide were recorded on a 600 MHz (¹H frequency) Bruker spectrometer equipped with a 5 mm inverse triple resonance CryoProbe (TCI, H tunable to F) with ¹H/¹⁹F, ¹³C and ¹⁵N frequencies. The spectra were recorded at 5 °C to mimic a larger molecular weight system. Here we recorded a ¹⁹F-detected, out-and-back style, coupled HSQC and a ¹⁹F-detected, out-and-stay ¹³C-¹⁹F TROSY spectrum. The spectral width in the indirect ¹³C dimension was set to 3 ppm and the spectral width in the direct ¹⁹F dimension was set to 40 ppm. 55 points were collected in the indirect ¹³C dimension (acquisition time = 121 ms) and 1024 points were collected in the direct ¹⁹F dimension (acquisition time = 45 ms). 256 scans were collected with a recycling delay of 2 seconds. The carrier frequency was centred at 139.0 ppm in the ¹³C dimension and −165.5 ppm in the ¹⁹F dimension.

Calculating the correlation time of 16-mer DNA

The crystal structure of a 16-mer DNA duplex (pdb id: 420D) was used as a proxy for a 16-mer oligo nucleotide to estimate rotational correlation times for 5-fluorouracil-substituted DNA at 5 °C. All water molecules were stripped from the 16-mer DNA duplex and hydrodynamic calculations were performed with HYDRONMR²⁷. The recommended value of 3.2 Å was chosen for the atomic element radius. The parameter NSIG (number of values of the radius of the minibead) was set to 6 and SIGMIN and SIGMAX were defined as 1.0 and 2.0, respectively. The temperature was set to 278.15 K and solvent viscosity was set to 1.5215 mPa·s.

T₁ Measurements on ¹³C_F and ¹⁹F_C

The longitudinal relaxation times (T₁) for ¹³C_F and ¹⁹F_C were estimated using a 1D inversion recovery NMR experiment. The experiments were recorded using GB1 to observe the isolated ¹⁹F_C and ¹³C_F resonances. The T₁ for ¹⁹F_C was measured on a Bruker Avance HD NMR spectrometer operating at 500 MHz (¹H frequency) and equipped with a room temperature probe. 8 experiments were recorded with variable inversion recovery delays of 1, 50, 100, 500, 800, 1000, 1500, and 2000 ms. 32 scans were averaged with a recycle delay of 10 s. The carrier frequency was set to −135 ppm.

The T₁ for ¹³C_F was measured on a Bruker Avance NMR spectrometer operating at 500 MHz (¹H frequency) and equipped with a cryogenically cooled probe. 24 experiments were recorded with variable inversion recovery delays of 1, 40, 100, 200, 200, 500, 800, 1000, 1250, 1500, 1750, 2000, 2250, 2500, 2750, 3000, 3250, 3500, 3750, 4000, 4250, 4500, 4750, and 5000 ms. 104 scans were collected with a recycle delay of 10 s. The carrier frequency was set to 150 ppm.

The signal intensity was measured using Bruker TopSpin® (version 3.2). The T₁ value was estimated by fitting an exponential curve described by

where M₀ corresponds to the intrinsic intensity of the ¹⁹F_C and ¹³C_F resonances without relaxation.

Simulation of the free induction decay

We simulated free induction decays (FID) for the TROSY components of aromatic ¹³C_F and ¹³C_H spin-pairs, based on transverse relaxation rates calculated for 3-¹⁹F_13C Tyr and 3-¹H_13C Tyr. To compare these to ¹H-¹⁵N spin pairs, we also calculated the FID for the TROSY component of an average ¹H-¹⁵N spin pair. The FIDs were simulated using the following expression:

Where I₀ is the initial magnetization, t is the evolution time, R₂ is the relaxation rate and Ω is the chemical shift.

The parameters I₀, and Ω were set to 0.005 and 200 Hz respectively. The signal was evolved for 1 s and sampled at 2048 points, captured by variable t. The time domain signal was Fourier transformed and the real part is displayed. The TROSY component linewidths for a macromolecule with τ_c = 25 ns are 1.9, 2.3, and 13.7 Hz, for ¹³C_F, ¹⁵N_H, and ¹³C_H, respectively. The TROSY component linewidths for a macromolecule with τ_c = 65 ns are 5.0, 6.1, and 35.7 Hz, for ¹³C_F, ¹⁵N_H, and ¹³C_H, respectively.