¹⁷O NMR studies of yeast ubiquitin in aqueous solution and in the solid state

Abstract

We report a general method for amino acid-type specific ¹⁷O-labeling of recombinant proteins in Escherichia coli. In particular, we have prepared several [1-¹³C,¹⁷O]-labeled yeast ubiquitin (Ub) samples including Ub-[1-¹³C,¹⁷O]Gly, Ub-[1-¹³C,¹⁷O]Tyr, and Ub-[1-¹³C,¹⁷O]Phe using the auxotrophic E. coli strain DL39 GlyA λDE3 (aspC⁻ tyrB⁻ ilvE⁻ glyA⁻ λDE3). We have also produced Ub-[η-¹⁷O]Tyr, where the phenolic group of Tyr59 is ¹⁷O-labeled. We showed for the first time that ¹⁷O NMR signals from protein terminal residues and sidechains can be readily detected in aqueous solution. We also reported solid-state ¹⁷O NMR spectra for Ub-[1-¹³C,¹⁷O]Tyr and Ub-[1-¹³C,¹⁷O]Phe obtained at an ultrahigh magnetic field, 35.2 T (1.5 GHz for ¹H). This work represents a significant advance in the field of ¹⁷O NMR studies of proteins.

Graphical Abstract

Last piece of the puzzle: We report amino acid-type specific incorporation of ¹⁷O isotope into yeast ubiquitin with an auxotrophic E. Coli strain. We show that ¹⁷O NMR signals from ¹⁷O-labeled proteins can be observed both in aqueous solution and in the solid state.

Biomolecular NMR spectroscopy has become an indispensible tool in structural biology. Now, it is equally possible to determine the three-dimensional structure of biological macromolecules either in aqueous solution^[1-5] or in the solid state.^[6-9] While tremendous progress has been made over the past few decades in many aspects of biomolecular NMR, most NMR studies in this field have continued to rely on detection of “NMR friendly” spin-1/2 nuclei such as ¹H, ¹³C, ¹⁵N, and ³¹P. Although the oxygen element is also ubiquitous in organic and biological molecules, its detection is severely hampered by the fact that the quadrupolar nature of the only NMR-active oxygen isotope, ¹⁷O (I = 5/2), often causes significant line broadening in NMR signals. In addition, because the natural abundance of ¹⁷O is exceedingly low (0.037%), ¹⁷O isotopic labeling is usually a prerequisite of any ¹⁷O NMR study. In recent years, important advances have been made in the development of ¹⁷O NMR spectroscopy as a useful tool in studying organic and biological molecules in the solid state.^[10-12] In the context of biological macromolecules, ¹⁷O NMR has been applied to study [¹⁷O]ligand-protein complexes and in the solid state^[13-15] and in aqueous solution.^[16-17] Solid-state ¹⁷O NMR was also useful in probing membrane-bound proteins/peptides.^[18-21] A particularly promising approach is to study large biomolecular systems with the ¹⁷O quadrupole-central-transition (QCT) methodology.^[22-24] However, one major obstacle is the lack of a general approach to incorporate ¹⁷O isotopes into proteins. It is somewhat surprising that, although it is routine to produce selectively or uniformly ¹³C- and ¹⁵N-labeled recombinant proteins by heterologous expression in various host cell systems,^[25-30] the same approach has not been reported in the literature for ¹⁷O-labeling of proteins. In this work, we report successful amino acid-type specific ¹⁷O-labeling of recombinant yeast ubiquitin by expression with an auxotrophic Escherichia coli strain and acquisition of ¹⁷O NMR spectra for selectively ¹⁷O-labeled yeast ubiquitin samples. The main goals of the present work are threefold: (1) to demonstrate the validity of this new ¹⁷O-labeling approach, (2) to test the hypothesis that protein sidechain groups may be readily observable by ¹⁷O NMR in aqueous solution, and (3) to explore the potential of solid-state ¹⁷O NMR of proteins at an ultrahigh magnetic field, 35.2 T (1.5 GHz for ¹H).^[31]

Ubiquitin is a small regulatory protein (76 amino acid residues) and one of the standard proteins for NMR methodology development. In this work, we utilized an auxotrophic E. coli strain DL39 GlyA λDE3 (aspC⁻ tyrB⁻ ilvE⁻ glyA⁻ λDE3)^[32-33] to produce yeast ubiquitin (Ub) where glycine (Gly), tyrosine (Tyr), phenylalanine (Phe) residues were selectively ¹⁷O-labeled. This general approach was first tested with ¹⁵N-labeled amino acids and subsequently used to incorporate [1-¹³C,¹⁷O]-labeled Gly, Phe, Tyr as well as [η-¹⁷O]Tyr into yeast ubiquitin for ¹⁷O NMR studies. We chose this [1-¹³C,¹⁷O] double-labeling scheme for two reasons. First, the ¹³C-labeling can be used to monitor for any potential isotope dilution and scrambling. Second, [1-¹³C,¹⁷O]-labeled protein samples may be useful for further development of heteronuclear ¹³C/¹⁷O correlation spectroscopy. All experimental details on chemical synthesis of ¹⁷O-labeled amino acids, recombinant protein expression, purification, and characterization are provided in the Supporting Information.

Now let us first examine the hypothesis that, because protein sidechains often experience substantial local motion in aqueous solution, their nuclear quadrupole relaxation characteristics may be similar to those of small molecules. As a result, it may be possible to detect ¹⁷O NMR signals from protein sidechains in aqueous solution. To this end, we prepared two samples: yeast ubiquitin (Ub)-[1-¹³C,¹⁷O]Gly and Ub-[η-¹⁷O]Tyr. In yeast ubiquitin, there are six Gly residues, one of which, G76, is at the C-terminus. Thus, we may consider the COOH group of G76 as a “sidechain”. Since there is only one Tyr residue, Y59, in yeast ubiquitin, any ¹⁷O NMR signal from Ub-[η-¹⁷O]Tyr would be from the phenolic group of Y59. The ¹³C solution NMR spectrum of Ub-[1-¹³C,¹⁷O]Gly confirmed that there was no substantial isotope scrambling and that all six Gly residues in yeast ubiquitin were [1-¹³C,¹⁷O]-labeled (see Figure S6; our ¹H-¹⁵N HSQC signals and ¹³C chemical shifts are consistent with those reported in the literature).^[34-36] Figure 1 shows the ¹⁷O NMR spectra of Ub-[1-¹³C,¹⁷O]Gly and Ub-[η-¹⁷O]Tyr in aqueous solution. Clearly, relatively sharp ¹⁷O NMR signals are observed in both cases. The ¹⁷O NMR signal from Ub-[1-¹³C,¹⁷O]Gly appears at 272 ppm, typical of carboxylic acids. This signal broadened as the temperature was lowered, but the signal position and line width were independent of the applied magnetic field, which suggests that G76 is under the so-called extreme narrowing condition (i.e., ω₀τ_C << 1); see Figure S8. The ¹⁷O NMR signals from other 5 Gly residues in yeast ubiquitin were too broad to be detected in aqueous solution at room temperature because they are in the slow motion regime, ω₀τ_C > 1. It is interesting to note that while G75 and G76 may have similar order parameters from the perspective of ¹³C and ¹⁵N NMR, the additional carboxylate 180°-flipping motion^[37] must contribute to the very short correlation time of G76 from the ¹⁷O NMR point of view. When the tumbling motion of yeast ubiquitin was slowed down by dissolving the protein in glycerol/H₂O (20%/80%), ¹⁷O QCT NMR signals from G10, G35, G47, G53, and G75 residues were observed at 253 K, as shown in Figure S8. This suggests that, under this condition, the protein is in the ultraslow motion regime (i.e., ω₀τ_C >> 1).^[12] However, we will not pursue this line of investigation further in the present study. As also seen in Figure 1, the ¹⁷O NMR signal from Ub-[η-¹⁷O]Tyr appears at 87 ppm, which is also consistent with that observed for the free amino acid L-Tyr.^[38] Further, the ¹⁷O NMR signal from the phenolic group in Y59 is broader than that for the carboxyl group in G76, because the phenolic group is known to have a larger C_Q(¹⁷O) value.^[39] This is the first time that ¹⁷O NMR signals from protein sidechains have been observed. One immediate application is to utilize ¹⁷O NMR as a new probe for pK_a measurement. As seen in Figure 1(c), the pK_a values for the carboxyl group of G76 and the phenolic group in Y59 were determined to be 2.32 ± 0.04 and 10.44 ± 0.04, respectively. It is interesting to note that, for the phenolic group of Y59, the ¹⁷O chemical shifts in the two ionization states differ by 75 ppm. This is in agreement with the previous results for free amino acid L-Tyr both in aqueous solution^[38] and in the solid state.^[39] While the two pK_a values determined for yeast ubiquitin may not have any important implication for its biological functions per se, this proof-of-concept experiment provides a novel method for studying ionization state, hydrogen bonding or ion binding of oxygen-containing amino acid sidechains (Asp, Glu, Ser, Thr, Asn, Gln) of proteins in aqueous solution.

Figure 1.

¹⁷O NMR spectra obtained at 16.4 T (700 MHz for ¹H) for (a) Ub-[1-¹³C,¹⁷O]Gly and (b) Ub-[η-¹⁷O]Tyr in aqueous solution (1.2 mM protein, H₂O, 25 mM Tris buffer, 250 mM NaCl, pH 8, 25 °C). (c) Titration results. For Ub-[1-¹³C,¹⁷O]Gly at low pH values, a 50-mM phosphate buffer was used.

To explore the potential of solid-state ¹⁷O NMR in detecting ¹⁷O signals from a protein backbone, we prepared two yeast ubiquitin samples: Ub-[1-¹³C,¹⁷O]Tyr and Ub-[1-¹³C,¹⁷O]Phe. Once again, since only one Tyr residue is present in yeast ubiquitin, Ub-[1-¹³C,¹⁷O]Tyr serves as an ideal sample to avoid any ambiguity in spectral analysis. Figure 2(a) shows an overlay of the 2D ¹H-¹⁵N HSQC spectra of Ub-[¹⁵N]Tyr and Ub-[¹⁵N]Phe in aqueous solution. The observed ¹H and ¹⁵N chemical shifts are consistent with the literature values.^[34] This data confirmed that no isotope scrambling occurred. Figures 2(b) and 2(c) show the ¹³C CP/MAS NMR spectra of hydrated solid samples of Ub-[1-¹³C,¹⁷O]Tyr and Ub-[1-¹³C,¹⁷O]Phe. The observed line width of 0.4 ppm is similar to those observed for ubiquitin microcrystals.^[40-42] The ¹³C chemical shifts for Y59, F4, and F45 are also in agreement with those reported in previous solution^[35-36] and solid-state NMR studies.^{[40, 42]} Figure 2(d) shows the ¹⁷O MAS NMR spectrum of Ub-[1-¹³C,¹⁷O]Tyr obtained at 35.2 T. Since there is a single Tyr backbone carbonyl oxygen in this sample, the spectrum can be readily analysed, which yielded the following ¹⁷O NMR tensor parameters: δ_iso = 308 ppm, C_Q = 9.0 MHz, η_Q = 0.7, Δ_CS = δ₃₃ – δ_iso = −320 ppm, η_CS = (δ₂₂ – δ₁₁)/Δ_CS = 0.2. The ¹⁷O MAS NMR spectrum of Ub-[1-¹³C,¹⁷O]Phe, shown in Figure 2(e), displays signals that are broader than those seen for Ub-[1-¹³C,¹⁷O]Tyr, suggesting the presence of two oxygen sites. This is also consistent with the ¹³C and ¹⁵N NMR data mentioned earlier. Once again, spectral analysis yielded a complete set of ¹⁷O NMR tensor parameters for both sites: F45, δ_iso = 299 ppm, C_Q = 8.9 MHz, η_Q = 0.7, Δ_CS = −320 ppm, η_CS = 0.2; Y4, δ_iso = 281 ppm, C_Q = 8.2 MHz, η_Q = 0.4, Δ_CS = −310 ppm, η_CS = 0.3. This is the first time that complete ¹⁷O NMR tensor parameters have been obtained for oxygen atoms on a protein backbone. All ¹⁷O NMR tensor parameters obtained for the yeast ubiquitin samples are summarized in Table S1, together with some relevant literature data.^{[18-20, 43-47]} As seen from Table S1, the overall ¹⁷O NMR parameters observed for yeast ubiquitin samples are similar to those reported previously for peptides. The most striking observation is the sensitivity of δ_iso(¹⁷O) and C_Q(¹⁷O) toward molecular structure. For example, the crystal structures of ubiquitin suggest that both F45 and Y59 are involved in loop regions, whereas F4 is within a β-sheet.^[48-49] Thus, strong hydrogen bonding is expected around the backbone carbonyl oxygen of F4. Indeed, the values of δ_iso(¹⁷O) and C_Q(¹⁷O) found for F45 and Y59 are similar, but are significantly larger than the corresponding values observed for F4. It is interesting to point out that the difference in δ_iso(¹³C) values between F4 and F45 is approximately 1 ppm, as seen from Figure 2(c), whereas the corresponding δ_iso(¹⁷O) values differ by 18 ppm. These observations are entirely consistent with the previously known trends in the dependence of ¹⁷O NMR parameters on hydrogen bonding (i.e., strong hydrogen bonding leads to reduction in both δ_iso(¹⁷O) and C_Q(¹⁷O) values for carbonyl compounds).^[10] The sensitivity offered by the 35.2-T magnet is also impressive. The levels of ¹⁷O enrichment in the Ub-[1-¹³C,¹⁷O]Tyr and Ub-[1-¹³C,¹⁷O]Phe samples were about 30-40%. Approximately 20 mg of lyophilized protein was packed in a 3.2-mm MAS rotor. The total experimental times to record the ¹⁷O MAS spectra shown in Figures 2(d) and 2(e) were 4-5 hr. With higher ¹⁷O enrichment levels, it may be possible to apply techniques such as MQMAS^[50] or STMAS^[51] for proteins to obtain even higher spectral resolution. In principle, the solid-state ¹⁷O NMR approach is not impacted by the size of the protein under study.

Figure 2.

(a) Overlay of the ¹H-¹⁵N 2D HSQC NMR spectra obtained at 16.4 T (700 MHz for ¹H) for Ub-[¹⁵N]Phe and Ub-[¹⁵N]Tyr in aqueous solution (0.8 mM protein, 95%H₂O/5%D₂O, 25 mM Tris buffer, 300 mM NaCl, pH 8, 25 °C). Solid-state ¹³C CP/MAS spectra obtained at 19.6 T (833 MHz for ¹H) for (b) rehydrated Ub-[1-¹³C,¹⁷O]Tyr and (c) Ub-[1-¹³C,¹⁷O]Phe. Solid-state ¹⁷O 18-kHz MAS NMR spectra obtained at 35.2 T (1.5 GHz for ¹H) for (d) Ub-[1-¹³C,¹⁷O]Tyr (665600 transients, recycle delay 20 ms) and (e) Ub-[1-¹³C,¹⁷O]Phe (313856 transients, recycle delay 20 ms). Home-built 3.2-mm MAS probes were used at 19.6 and 35.2 T.

In summary, we have demonstrated that it is feasible to achieve amino acid-type specific ¹⁷O-labeling of proteins via recombinant expression in an auxotrophic E. coli strain. This approach allows incorporation of ¹⁷O isotopes into both the protein backbone and sidechains. We have shown that since protein sidechains often exhibit significant local motion they behave like small molecules from the perspective of nuclear quadrupole relaxation, thus making the detection of their ¹⁷O NMR signals straightforward. Such an approach offers a new method for probing protein sidechain structure, interactions and dynamics. For example, it would be interesting to apply this ¹⁷O NMR approach to study ion binding to protein sidechains or hydrogen bonding. We have also explored the use of an ultrahigh magnetic field, 35.2 T, to record high-quality solid-state ¹⁷O NMR spectra for proteins. The information offered by the complete ¹⁷O NMR tensor characterization will be complementary to that obtainable from more conventional NMR probes, such as ¹H, ¹³C and ¹⁵N. The results reported in this study are encouraging. One can envision that the sensitivity and resolution in solid-state ¹⁷O NMR for proteins can be further improved by ¹³C/¹⁷O heteronuclear correlation spectroscopy perhaps in combination with nuclear dynamic polarization. It is highly desirable to make ¹⁷O accessible by NMR in studies of biological macromolecules in aqueous solution and in the solid state, so that one would be able to utilize all magnetic nuclei available in proteins, nucleic acids, and carbohydrates (¹H, ²H, ¹³C, ¹⁴N, ¹⁵N, ¹⁷O, ³¹P) to gain information about molecular structure, chemical bonding, and dynamics. The present work represents a key step towards this ultimate goal. Research to further expand the current work is under way in our laboratories.