⁷⁷Se NMR Probes the Protein Environment of Selenomethionine

Abstract

Sulfur is critical for the correct structure and proper function of proteins. Yet, lacking a sensitive enough isotope, nuclear magnetic resonance (NMR) experiments are unable to deliver for sulfur in proteins the usual wealth of chemical, dynamic, and structural information. This limitation can be circumvented by substituting sulfur with selenium, which has similar physicochemical properties and minimal impact on protein structures but possesses an NMR compatible isotope (⁷⁷Se). Here we exploit the sensitivity of ⁷⁷Se NMR to the nucleus’ chemical milieu and use selenomethionine as a probe for its proteinaceous environment. However, such selenium NMR spectra of proteins currently resist a reliable interpretation because systematic connections between variations of system variables and changes in ⁷⁷Se NMR parameters are still lacking. To start narrowing this knowledge gap, we report here on a biological ⁷⁷Se magnetic resonance data bank based on a systematically designed library of GB1 variants in which a single selenomethionine was introduced at different locations within the protein. We recorded the resulting isotropic ⁷⁷Se chemical shifts and relaxation times for six GB1 variants by solution-state ⁷⁷Se NMR. For four of the GB1 variants we were also able to determine the chemical shift anisotropy tensor of SeM by solid-state ⁷⁷Se NMR. To enable interpretation of the NMR data, the structures of five of the GB1 variants were solved by X-ray crystallography to a resolution of 1.2 Å, allowing us to unambiguously determine the conformation of the selenomethionine. Finally, we combine our solution- and solid-state NMR data with the structural information to arrive at general insights regarding the execution and interpretation of ⁷⁷Se NMR experiments that exploit selenomethionine to probe proteins.

Graphical Abstract

INTRODUCTION

Cysteine (Cys) and methionine (Met) hold a special place in the pantheon of proteinogenic amino acids because only they contain sulfur. The sulfur in cysteine is part of its thiol group where it frequently acts either as a nucleophile for enzymatic reactions or as a critical structural element when it forms disulfide bonds with another cysteine. Frequently, cysteine’s sulfur is also involved in signaling pathways. In Met, sulfur is part of a thioether group that can stabilize protein structure via sulfur–π interactions with aromatic rings and electrostatic interaction through low-lying σ* orbitals.¹ Met’s polarity caused by the sulfur is also exploited for the recognition of ligands and protein partners. In addition, both Cys and Met are sites for posttranslational modifications that alter protein function.^2,3 Because of the importance of sulfur for both the function and structure of proteins, this nucleus and its chemical environment are central in deciphering the workings of these molecular machines. However, in the case of sulfur, nuclear magnetic resonance (NMR) spectroscopy, the tool that has made the study of hydrogen, carbon, nitrogen, and phosphorus in biological systems routine, cannot deliver the same wealth of information, for its only NMR-sensitive isotope, ³³S, is a quadrupolar nucleus with very low sensitivity.⁴

One promising approach to cut this Gordian knot is to substitute the sulfur with selenium. The two elements share many of the physicochemical properties that are important in the context of biological macromolecules, including comparable electronegativities and redox states.⁵ Fortunately, both the atomic radii of the two elements and their respective covalent bond lengths with carbon differ only by about 10%. In most cases, this translates into small, if not negligible, structural changes when sulfur is replaced by selenium. The evidence for this is plentiful because Met is routinely replaced by selenomethionine (SeM) in proteins prepared for X-ray crystallography. That way one can take advantage of selenium’s unique anomalous diffraction signature which significantly eases the process of extracting the proper protein structure from the diffraction data. In most of these cases, the substitution causes little, if any, effect on either structure or function.

Such a substitution approach would also be beneficial for NMR experiments because, in contrast to sulfur, one of selenium’s isotopes, ⁷⁷Se, is a spin 1/2 nucleus that is NMR sensitive. This isotope could thus serve as a valuable spectroscopic surrogate for sulfur.⁶ The chemical shifts of the ⁷⁷Se nucleus in biological systems span over 2000 ppm, which promises ample room to resolve the different chemical species that selenium-containing amino acids can form when introduced into proteins. For example, the selenoether of SeM alone spans more than 160 ppm. Already a useful tool to study selenoorganic and -inorganic compounds,^{7,8 77}Se NMR is also increasingly applied to biological systems.^6,9 So far it has been used to detect disulfide connectivities, enzymatic reactions, and conformational dynamics.^10–16 However, attempts to develop ⁷⁷Se NMR into a truly routine technique that can study the sulfur-dependent mechanisms in proteins have been less successful. There are two main reasons for this. First, methods for the efficient isotopic enrichment of selenoproteins were unavailable in the past. Second, there is currently no library or data bank of experimental spectra nor are there computationally derived spectra of high enough fidelity to guide analysis and interpretation of ⁷⁷Se spectra of proteins. Essentially, as the number of examples is so small, it is currently next to impossible to assign ⁷⁷Se spectra or extract information about dynamics.

Over the past few years, we and others have made good progress in overcoming the bottleneck of producing large enough quantities of proteins containing selenium and specifically the NMR sensitive isotope ⁷⁷Se. For example, we previously developed a straightforward and cost-effective method to isotopically enrich proteins by randomly incorporating ⁷⁷Se during their heterologous expression in Escherichia coli.¹⁴ This method requires no expertise in selenium chemistry and is simple to implement. More recently our group devised a method in which bacterially expressed protein fragments are chemically ligated.¹⁷ This flexible and efficient approach can place ⁷⁷Se at almost any desired place along a protein’s peptide chain. Now with the sample production problem tackled, we are finally able to make inroads on the second issue and build a systematic collection of experimental ⁷⁷Se NMR spectra of proteins. Our initial efforts presented here focus on SeM because its presence in proteins affects structure and function minimally.

To date, information about [⁷⁷Se]-SeM NMR properties in proteins is limited. In dihydrofolate reductase, the ⁷⁷Se chemical shift spread for five SeM residues was 40 ppm.¹⁸ In SeM-enriched calmodulin the ⁷⁷Se chemical shift dispersion for the nine SeM residues is approximately 60 ppm (and carbon chemical shifts of SeM methyl groups are upfield shifted as much as 11 ppm).¹⁹ This range for a given bond type is significantly larger than that reported for a given ¹³C and ¹⁵N in proteins (http://www.bmrb.wisc.edu/histogram) and is notable because it prevents spectral overlap even in large proteins and complexes. However, a more detailed analysis of the selenium spectra and recommendations for data interpretations are not available because the correlation between the exact SeM conformation and environment and its NMR parameters has not been determined. Also, data about essential NMR parameters for selenium in proteins are missing. A prominent example is the chemical shift tensor because it informs on the electronic surrounding of a nucleus.²⁰ While it has been reported for the free SeM amino acid, at this point no solid-state NMR experiments have been performed to extract this tensor for selenium in proteins.

It is therefore not well understood how the chemical environment of the selenium nucleus in a protein is related to the chemical shifts or the chemical shift tensors in potential NMR experiments. As far as computational efforts are concerned, there are but a few density functional theory (DFT) calculations of selenium NMR parameters in biological systems, and they focused primarily on selenocysteine (Sec) and its role in native selenoproteins.^11,21,22 Thus, establishing a library of experimental ⁷⁷Se chemical shifts and chemical shift tensors that are paired with high-resolution structural information on selenium in proteins will be of great aid to future research of Se NMR of proteins, whether it be experimental or numeric in nature.

As a first step, we present here a systematic study that combines high-resolution X-ray crystallography, solution-state, and—for the first time—solid-state ⁷⁷Se NMR spectroscopy to correlate the specific chemical environment of selenium with its NMR signatures for a variety of positions in a well-characterized protein.

More specifically, we studied the immunoglobulin-binding B1 domain of streptococcal protein G (GB1). A 56 amino acid protein, GB1 possesses a remarkably compact and stable fold, similar to ubiquitin.²³ GB1’s fold also contains all major secondary structural elements, i.e., helices, β-sheets, and loops, making it a good fit for studies involving different environments. In-depth biochemical and biophysical studies on GB1 are available,^23–26 and its structure has been determined extensively by X-ray crystallography and solution- and solid-state NMR. For the work here we engineered a series of GB1 variants, each containing a single SeM located in a different protein environment, which enabled us to probe location-specific information. The high-resolution structure of five of these GB1 variants was solved by X-ray crystallography to study the relationship between a particular nuclear environment of ⁷⁷Se and the resulting NMR spectra and parameters. We were also finally able to include the full chemical shift tensors of selenium in a protein environment.

RESULTS

In order to survey SeM in different protein environments, we have designed a series of GB1 variants with a single SeM at different locations. To ensure that all GB1 variants have only one SeM at the desired position, the purified proteins were based on a “SeM-free” GB1 variant, which contains Gly in the first position and Gln at position 2 for consistency with the vast literature about GB1 structure and stability. Of the six GB1 variants, four were designed to have SeM positioned in the hydrophobic core (GB1 L5SeM, A34SeM, V39SeM, and V54SeM) and two were designed with SeM in a solvent-accessible position (GB1 I6SeM and V29SeM). Since GB1 has been extensively characterized, amino acid changes at these locations have been previously characterized; however, none of these involved SeM.^23,26–28

In the following, we first describe the molecular information about the SeM environment in GB1 variants from the crystallography. We then report and compare their solution NMR parameters, and finally summarize all the additional information provided by the solid-state NMR experiments.

Structure Determination.

To record the chemical environment surrounding the SeM, as well as SeM conformation and mobility at each site, the six GB1 variants were crystallized, and high-resolution structural information was obtained for five. Crystallization conditions were selected to match those used for the solid-state NMR samples (see later) to aid in data interpretation. All six GB1 variants yielded diffracting crystals, and the structures of five variants were obtained by X-ray crystallography at a resolution of 1.1–1.2 Å and deposited in the Protein Data Bank (PDB; Figure 1, Table 1, and Table S3 in the Supporting Information). GB1 V54SeM was the only variant that required imidazole for crystallization and whose crystals were frequently twinned. Thus, it was not possible to refine a model to the precision required for inclusion in the PDB. The refinement statistics and information about the different conformations of SeM and their local environment in the five variants are summarized in Table 1, Table 2, and Table S3. Overall, as evidenced by comparison with the wild-type GB1, the secondary and tertiary structures of all five variants were not substantially impacted by the introduction of the SeM in different locations, and structural perturbations were restricted to SeM’s immediate vicinity. The most notable effect on protein structure was observed for GB1 A34SeM. There, Trp43 rotated toward the third β-sheet to accommodate SeM’s long side chain (see also discussion below). Mutations at positions Ala34 and Val54 were reported to have such effects on GB1 structure.²³

Figure 1.

Structures of GB1 variants carrying SeM at different positions (top row), close-up of the SeM (middle row), and details of SeM’s local environment within 4.2 Å. (A) GB1 L5SeM PDB entry 6CNE; (B) GB1 I6SeM PDB entry 6CPZ; (C) GB1 V29SeM PDB entry 6C9O; (D) GB1 A34SeM PDB entry 6CHE; (E) GB1 V39SeM PDB entry 6CTE; (F) GB1 V54SeM low-resolution model. An imidazole ligand is displayed for GB1 A34SeM, and an MPD ligand is shown for GB1 I6SeM chain B. Hydrogen atoms were omitted for clarity. The selenium atom is colored orange, carbons are depicted in green, oxygens are depicted in red, and nitrogens are depicted in blue.

Table 1.

Crystallization Conditions and Key Refinement Statistics for GB1 SeM Variants^a

	L5SeM	I6SeM	V29SeM	A34SeM	V39SeM
crystallization condition	49% MPD	48% MPD	46% MPD	49% MPD	47% MPD
	20% IPA	20% IPA	20% IPA	20% IPA	20% IPA
	25 mM NaCH3CO2	25 mM NaCH3CO2	25 mM NaCH3CO2	25 mM NaCH3CO2	25 mM NaCH3CO2
	pH 4.9	pH 4.7	pH 4.5	pH 4.9	pH 4.8
protein solution	20 mg/mL	20 mg/mL	20 mg/mL	20 mg/mL	20 mg/mL
	25 mM NaCH3CO2	25 mM NaCH3CO2	25 mM NaCH3CO2	25 mM NaCH3CO2	25 mM NaCH3CO2
	pH 5.5	pH 5.5	pH 5.5	pH 5.5	pH 5.5
	no TCEP	2 mM TCEP	2 mM TCEP	2 mM TCEP	no TCEP
growth time	1 week	1 week	1 week	1 week	1 week
resolution (Å)	25.74 (1.20)	27.57 (1.12)	29.10 (1.20)	26.96 (1.10)	27.17 (1.20)
R-work	0.1930 (0.2419)	0.1453 (0.1253)	0.1616 (0.1926)	0.1746 (0.1624)	0.1629 (0.2218)
R-free	0.2160 (0.2704)	0.1615 (0.1591)	0.1869 (0.2338)	0.1853 (0.1632)	0.1793 (0.2328)
CC1/2	0.992 (0.429)	0.999 (0.996)	0.997 (0.979)	0.998 (0.99)	0.999 (0.963)
CC*	0.998 (0.775)	1 (0.999)	0.999 (0.995)	0.999 (0.997)	1 (0.99)
PDB ID	6CNE	6CPZ	6C9O	6CHE	6CTE

^aThe complete list of refinement parameters is provided in Table S3.

Table 2.

Side Chain Torsion Angles and Local Interactions of SeM Residues^a

GB1 variant	C–Se–C (A)	χ1 (A)	χ2 (A)	χ3 (A)	C–Se–C (B)	χ1 (B)	χ2 (B)	χ3 (B)	residues within 4.5 Å of Se	ligands within 4.5 Å of Se
L5SeM	100.6	169.4	176.1	52.1	99.9	176.5	−180.0	49.4	Leu7 Phe30 Ala34 Trp43 Val54
I6SeM	99.3	−62.3	179.3	69.5	100.0	−64.6	−179.3	68.1	Lys4 Glu15 (A)	MPD
	99.1	−50.4	−67.8	74.4	98.8	−49.1	−68.1	−77.3	Thr51 Thr53 (B)
V29SeMb	104.6	−63.3	174.4	−45.1	94.3	−64.3	166.1	86.5	Thr18 Glu19 Ala20 Thr25 Ala26	MPD
A34SeMc	97.9	−69.9	162.3	53.2					Leu5 Leu7 Phe30 Lys31 Val54	IMD
V39SeM	100.0	−55.4	−175.8	85.5	99.9	−53.2	−177.2	86.4	Leu7 Leu12 Ala34 Val54	MPD

^aThere are two GB1 chains in the asymmetric unit for all GB1 variants except for GB1 A34SeM. Hence “C–Se–C (A)” means bond angle C–Se–C in GB1 copy A, and “C–Se–C (B)” means bond angle C–Se–C in GB1 copy B, etc. For SeM residues, χ₁ is dihedral angle N–C_α–C_β–C_γ χ₂ is the dihedral angle C_α–C_β–C_γ–Se; χ₃ is the dihedral angle C_α–C_β–Se–C_ε.

^bStatistics for GB1 V29SeM conformation in chain A are the least accurate due to conformational mobility.

^cGB1 A34SeM has only one GB1 in the asymmetric unit.

In all five X-ray structures the quality of SeM’s electron density was sufficient to unambiguously place the side chain (Figure 2 and Figure S3) and deduce its conformation. The electron density for the terminal methyl group was observed for all SeM residues, although it was weaker for the solvent-exposed positions in GB1 I6SeM and GB1 V29SeM. Nevertheless, in all variants the terminal methyl group could be unambiguously located. Locating these terminal methyl groups is essential for establishing SeM’s conformation and notable because, even for high-resolution structures in the Dunbrack 1.0 Å database, the methyl electron density is observed for only about 50% of Met and SeM residues.²⁹ Exploiting the anomalous diffraction maps, the position of the selenium atoms was determined, including—as will be further discussed for GB1 V29SeM and GB1 A34SeM—alternative conformations. Additional assessment of the structural data quality can be gathered from the temperature factors (B factors) arising from the static disorder and dynamic disorder in the crystal. Analysis of the scaled B factors of all C_α atoms (B′ factor) of the five GB1 variants compared to the original GB1 (PDB 2QMT) showed (Figures S4 and S5) that the third loop remains the secondary structure element with the highest mobility (residues 37–41). The selenium B factor in the different SeM is <8 for GB1 L5SeM, <19 for GB1 I6SeM, <21 for GB1 V29SeM (for chain B), <8 for GB1 A34SeM, and <10 for GB1 V39SeM. These low B factors, in addition to the anomalous maps, enabled us to unambiguously determine the SeM side chain conformation for all structures except GB1 V29SeM.

Figure 2.

Electron density of each SeM side chain found in the five crystal structures of GB1 variants. (A) GB1 L5SeM; (B) GB1 I6SeM; (C) GB1 V29SeM; (D) GB1A34SeM; (E) GB1 V39SeM. For variants with two copies of the protein in the asymmetric unit, the SeM in chain A is shown on the left and that of chain B is shown on the right. A34SeM has only one GB1 in the asymmetric unit. The selenium atom is shown in orange, carbons are shown in green, oxygens are shown in red, and nitrogens are shown in blue. Hydrogens are omitted for clarity. The 2F₀–F_c electron density map is shown in blue. The anomalous diffraction map is shown in magenta.

At this high resolution, it is not uncommon to observe more than one conformation for amino acid side chains. A sole SeM conformation was observed for the structures of GB1 L5SeM and V39SeM, while two SeM conformations were observed for GB1 I6SeM. For both GB1 V29SeM and GB1 A34SeM one dominant and some lower populated SeM conformations were found. In the case of GB1 V29SeM those lower populated conformers differ primarily only in the methyl position. In contrast, for GB1 A34SeM, the additional, lower-occupancy conformations indicated by the electron density and anomalous maps deviated more substantially (Figure 2). For all GB1 variants, the SeM torsion and bond angles, summarized in Table 2, agree with both the preferences determined by calculations³⁰ and a survey of high-resolution protein structures carried out by the Regan group.²⁹ Among these angles the χ₃ dihedral angle (C_β–C_γ–Se–C_ε) is particularly useful because it informs on the geometric positioning of the two lone electron pairs of selenium. The importance of χ₃ is underscored by how much it impacts—together with the C_γ–Se–C_ε angle—DFT calculations of selenium NMR parameters.^22,31 The χ₃ dihedral angles for the 871 SeM residues contained in the Dunbrack 1.0 Å database cluster in roughly equal parts around 60 and 300°, with a minor population grouped around 180° according to the analysis by the Regan group. The SeM χ₃ dihedral angles in the high-resolution structures of our GB1 variants follow these conformational preferences, as they are bunched around 70°. Also, the C_γ–Se–C_ε angles in our structures, whose average over the 11 SeM conformations is 100°, is in line with the reported 98 ± 2° of the Regan survey.

The details of the SeM conformation in the different GB1 variants can be rationalized based on the local protein environment and the weak interactions it has with selenium. In the following we describe these specific details for all five GB1 variants.

GB1 L5SeM.

In GB1 L5SeM, the SeM is located in the hydrophobic core close to aromatic residues (Figure 1A). There is little free volume for motion that would not require the displacement of nearby atoms. As Figure 2A and the low B factors (see discussion above) show, the conformation of the SeM side chain is well resolved. Since Leu and SeM are both relatively bulky, the structural impact of the amino acid substitution is limited to the immediate vicinity of SeM (Figure 1A). The selenium atom is interacting with the indole ring of Trp43. Such favorable electrostatic interactions between selenium and aromatic rings are well documented in proteins.^1,32 The selenium is approaching Trp43 edge-on at a distance of 4.75 (A) and 4.72 Å (B) at an angle of 65.3 (A) and 66.8° (B). (The distance was measured from the selenium atom to the center of the indole’s benzene ring; the angle is between the selenium–aromatic ring center line and the normal of the aromatic ring plane; see details in Figure S6.) Concurrently it is also interacting with Phe30 at a distance of 4.77 (A) and 4.69 Å (B) at an angle of 39.8 (A) and 37.8° (B). In a survey of Met interacting with aromatic amino acids in proteins, it was reported that Met interacts with Phe and Trp at angle larger than 60° or smaller than 30°, but not in between.³³

GB1 I6SeM.

In GB1 I6SeM, the SeM is in the solvent-exposed surface of a β-strand, close to the positive charge of Lys4. The change of Ile to SeM caused little structural perturbation. In both GB1 chains in the asymmetric unit, the SeM adopts two equally populated conformations (Figures 1B and 2B). In one conformation, which is identical in both GB1 chains, the selenium is within 4.1 Å to the Thr51 and Thr53 methyl groups and the SeM methyl group moved to accommodate a shift in the position of the Lys4 amine group. In the second conformation, which is not identical in both GB1 copies, there is a hydrogen bond between the Glu15 side chain and the selenium atom (3.2 and 3.3 Å in chains A and B, respectively). Indeed, the selenium atom of SeM was previously shown to form a hydrogen bond with both nitrogen and oxygen in proteins.³⁴ In this second conformation, the selenium atom is in an identical location in both chains, but it has alternate methyl group positions. For both conformations, the SeM conformation is also evidently influenced by a nearby 2-methyl-2,4-pentanediol (MPD) ligand. The MPD interacts via hydrogen bonds not with SeM itself, but with neighboring water molecules. Nevertheless, the presence of MPD does change selenium’s immediate environment because the ligand has a lower dielectric constant than water.

GB1 V29SeM.

In GB1 V29SeM, the SeM is positioned on the α-helix and partially exposed to solvent (Figure 1C). The change of Val to SeM did not result in a significant structural change because this position can accommodate the bulkier SeM. The selenium in this protein has the widest distribution of positions among the five variants due to motion (specifically in chain A, Figure 2C), and therefore the bond angles are the least accurately determined. The χ₃ dihedral angle differs for the two chains and so does the position of MPD ligands (Table 2) nearby. In chain A, selenium’s distance from the Thr25 backbone oxygen is 4.0 Å. In chain B, the selenium is positioned 3.8 Å from the Thr25 backbone oxygen and 4.0 Å from the oxygen of the Thr18 side chain. The slight adjustment in the position of the selenium atom and the methyl group appears to maximize their van der Waals contact (Table S4) with a nearby MPD. Nonbonded electrostatic interactions between selenium and oxygen and nitrogen are well-known.¹ In both protein chains, the selenium is within 5 Å from the Lys13 amine group, contributed by a nearby protein chain in the crystal.

GB1 A34SeM.

In GB1 A34SeM, the SeM is situated in the hydrophobic core of the protein (Figure 1D). The replacement of the short Ala by the longer SeM caused a shift in the position of Trp43 in the hydrophobic core. This, in turn, led to a change in the location of the third β-sheet and may be the reason why this variant crystallized with only one protein chain in the asymmetric unit (Table 1). A solvent pocket allowed imidazole, which was present during the initial purification steps but not the crystallization setup, to bind and pack between SeM34 and Trp43 (Figures S6 and S7). There is evidence for at least three SeM side chain conformations: a predominant conformation that corresponds to the presence of an imidazole group packed against Trp43 and SeM side chains, and two low occupancy conformations that may arise from the inherent flexibility of the SeM side chain in the absence of an imidazole group and the associated stabilizing cation–π interactions. The structure was refined using the most populated conformation because the electron density of the weakly populated side chain conformations was too diffuse to assign (Figure 2D). SeM34 lies 3.9 Å from the center of the imidazole ring at an angle of 11° to the imidazole plane (en-face) and 3.9 Å to the imidazole N₃, which is its nearest neighbor (Figure S6). In our structure, the selenium is right above the nitrogen and could be stabilized by orbital overlap. Moreover, SeM is in close contact with Val54 and Leu5.

GB1 V39SeM.

In GB1 V39SeM, the SeM is positioned in the relatively mobile loop 3 (Figure S5), but it is facing the hydrophobic core. There is little free space for molecular motions, and the electron density fits only one highly populated conformation (Figures 1E and 2E). Notably, SeM is surrounded by hydrophobic residues and lies within 3.7–4.2 Å from Leu7, Leu12, Ala34, and Val54 methyl groups. This suggests that it is primarily engaged in van der Waals interactions.

GB1 V54SeM.

In GB1 V54SeM, the SeM is located in the hydrophobic core. Crystallization of GB1 V54SeM was successful only in the presence of imidazole. It was not possible to collect high-quality diffraction data and refine a model to an acceptable level, most likely due to twinning. Yet models generated from more than five different crystals were all in agreement about the placement of SeM and aromatic residues. The resulting model can serve to guide the discussion about the overall protein structure (Figure S8), but not for in-depth analysis of the SeM conformation or selenium’s weak interactions. The model was also not deposited in the PDB due to its suboptimal quality. Nevertheless, it is still obvious that the mutation of Val to SeM resulted in the most significant structural changes. Trp43 was shifted in a manner similar to that observed in GB1 A34SeM. Moreover, Phe30 also changed its position. The impact on aromatic packing in the hydrophobic core was therefore substantial. This big structural change in the protein’s core was further confirmed by the ¹H–¹³C heteronuclear single-quantum coherence (HSQC) result (see below). While the selenium of SeM54 lies close to Trp43 and Phe30, it resides closer to the Gly41 amide group and is likely to form with it a hydrogen bond; however, the model quality did not allow definitive determination of the SeM conformation.

Finally, regarding ligands and crystal contacts, the only GB1 variants in which the conformation of SeM is influenced by crystal contacts are GB1 I6SeM and V29SeM. Also, in both cases MPD ligands are found nearby. Lastly, as detailed above, an imidazole is in contact with SeM in GB1 A34SeM.

Solution NMR.

With the information about the structures and conformational mobilities of the different GB1 variants at hand, the proteins were used to survey NMR properties in both solid- and solution-state samples. Since ⁷⁷Se’s natural abundance is 7.63%, it is possible to collect solution-state NMR spectra for GB1 samples with natural abundance. Samples were prepared by expressing the respective GB1 proteins in E. coli in defined growth medium supplemented with SeM.³⁵ For several samples, GB1 was enriched with ⁷⁷Se using our published method.¹⁴ In brief, [⁷⁷Se]-SeM was incorporated into proteins instead of Met by swamping sulfur pathways with [⁷⁷Se]-selenite during heterologous expression in E. coli. The presence of SeM was verified by intact protein electrospray ionization mass spectrometry, and it was found that between 75 and 85% of the proteins in our samples contained ⁷⁷SeM (the remaining is Met, see Figure S1 and Table S2). All GB1 variants used here and the wild-type GB1 were found to exist as dimers in solution (Figure S2).

The NMR spectra were subsequently recorded by solution NMR spectroscopy. These experiments determined the ⁷⁷Se isotropic chemical shifts (δ_iso) and their temperature dependence, the T₁ relaxation times, and the signal full width at half-maximum (FWHM) values at magnetic fields of 14.1, 11.74, and 9.4 T. In the case of selenium NMR, the FWHM can provide a good estimation for the T₂ relaxation time. The field dependence of both the spin–lattice relaxation time, T₁, and the spin–spin relaxation time, T₂, can be used to evaluate the presence of molecular motion as will be discussed.

Figure 3 shows the one-dimensional (1D) ⁷⁷Se solution NMR spectrum for the six different GB1 variants at a magnetic field of 11.74 T and a temperature of 293 K. As Table 3 shows, the selenium isotropic chemical shift of our six GB1 variants varied from 50 to 122 ppm. This range for SeM is similar to the 34–111 ppm range that has been previously reported for nine SeMs incorporated into calmodulin.¹⁹ Also it covers the chemical shift of free SeM, 72 ppm.³⁶ The selenium isotropic chemical shifts are not strictly correlated with a single parameter such as the solvent accessibility or the χ₃ dihedral angle. However, the X-ray structures suggest that a higher accessibility of SeM to water contributes to a smaller ⁷⁷Se chemical shift (i.e., a higher shielding). A compilation of ⁷⁷Se NMR data of small organic compounds reported the solvent effect on ⁷⁷Se chemical shifts is substantial, with ⁷⁷Se nuclei more shielded in polar solvents.⁷

Figure 3.

⁷⁷Se solution NMR spectra of the six GB1 variants acquired at 11.74 T at 293 K. From the top to bottom: GB1 L5SeM, GB1 I6SeM, GB1 V29SeM, GB1 A34SeM, GB1 V39SeM, and GB1 V54SeM.

Table 3.

Solution ⁷⁷Se NMR Characteristics of SeM in GB1 Variants^a

	δiso (ppm)					FWHM (Hz)			T1 relaxation time (ms)
GB1 variant	9.4 T	11.74 T	14.1 T	temp (K)	Δδiso/ΔT (ppm/K)	9.4 T	11.74 T	14.1 T	9.4 T	11.74 T	14.1 T
L5SeM		101.5		303			16
	101.4			298		10		21	186 ± 4
			101.6	295				22		166 ± 5	143 ± 4
		101.2		293	0.059		19
		100.0		277			36
I6SeM		55.0		303			10
	52.3		53.7	298		11		22	363 ± 4
				295						264 ± 3	225 ± 2
		53.0		293	0.190		14
		50.0		277			20
V29SeM		74.5		303			7
	75.2			298		7			250 ± 3
			75.9	295				13		189 ± 2	152 ± 1
		74.1		293	0.034		9
		73.6		277			13
A34SeM				303
				298
			76.5	295		NA		~335
		74.8		293	NA		201		NA	NA	NA
		75.3		277			138
V39SeM		82.6		303			17
	81.8			298		10			238 ± 1
			82.4	295				22		198 ± 3	146 ± 1
		82.1		293	0.066		15
		80.9		277			26
V54SeM		118.6		303			77
	120.0			298		54			134±2
			120.6	295				115		107 ± 2	113 ± 1
		119.9		293	−0.119		52
		121.7		277			91

^aHere, δ_iso (ppm) and FWHM (Hz) are reported with no line broadening applied (i.e., natural line widths). The line broadening applied to derive the FWHM of GB1 A34SeM was 10 Hz. T₁ values for each variant were obtained by manually integrating each peak over a same-sized interval at each field strength. Line broadening applied to spectra in T₁ data sets are as follows: GB1 L5SeM, 18 Hz; GB1 I6SeM, 8 Hz; GB1 V29SeM, 5 Hz; GB1 V39SeM, 10 Hz; GB1 V54SeM, 40 Hz (these values were selected as approximately half the width (Hz) of the broadest peak in the temperature dependence set).

Temperature Dependence.

The temperature dependence of the ⁷⁷Se isotropic chemical shifts and FWHMs between 277 and 303 K were recorded to calculate the change in δ_iso per kelvin (Figure 4 and Table 3). Such changes of the isotropic chemical shifts with temperature provide additional information about the mobility at selenium sites. However, as in the case of all heavy atoms, other experimental conditions such as solvent, pH, and temperature also affect selenium’s chemical shifts.^7,8 Of all variants, the isotropic chemical shift of GB1 I6SeM changed the most with temperature. In native GB1, the I6 side chain, located on the surface of the protein, stays mobile even in GB1 crsytals.³⁷ The large Δδ_iso/ΔT of the GB1 I6SeM variant indicates that SeM at that position retains that fast, unconstrained motion in solution. There is also evidence that SeM remains mobile even in crystals of this protein, as will be discussed later.

Figure 4.

⁷⁷Se solution NMR spectra of five GB1 variants acquired at 11.74 T at different temperatures.

The trend overall was a decrease in isotropic ⁷⁷Se δ_iso at lower temperatures; GB1 V54SeM is the only variant in which the isotropic ⁷⁷Se δ_iso increased at lower temperatures. The FWHM of this variant’s signal was also narrowest at 293 K (as opposed to 303 K for GB1 L5SeM, I6SeM and V29SeM). One potential explanation is that this is due to changes in the aromatic stacking at the protein core (Figure S8) as the temperature increases.

Relaxation Times.

The T₁ relaxation times for all GB1 variants were acquired at different magnetic field strengths to assess the contribution of chemical shift anisotropy (CSA) to relaxation. At the lowest field (9.4 T) the T₁ relaxation times of the variants ranged from 134 to 364 ms, while at the highest field (14.1 T) T₁ lay between 113 and 225 ms (Table 3, Figure S9). Regardless of field strength, GB1 I6SeM exhibited the longest T₁ relaxation time and GB1 V54SeM exhibited the shortest. In GB1 I6SeM, the SeM is in the solvent-exposed interface of the β-sheets and hence has the fewest steric constrains. This location suggests that the SeM of this variant moves the fastest and has the highest degree of freedom among the six variants in solution. Indeed, its long T₁ relaxation time suggests motions faster than the Larmor frequency (i.e., faster than 9 ns). In contrast, GB1 V54SeM’s large FWHM and short T₁ relaxation time suggest conformational mobility in the microsecond range. It was not possible to record the T₁ relaxation time for GB1 A34SeM since its short T₂ relaxation time leads to a large FWHM and results in reduced sensitivity.

To assess the role of the chemical shift anisotropy mechanism in overall longitudinal relaxation T₁(total), first the measured longitudinal relaxation rate, R₁(total) = 1/T₁(total) was plotted against the square of magnetic field, B₀² (Figure S10). A linear dependence of these two parameters establishes that the contribution of the chemical shift anisotropy mechanism is dominant. Indeed, the linear graphs of Figure S10 validate this scenario for GB1 L5SeM, I6SeM, V29SeM, and V39SeM. Thus, in these cases the CSA remains the only field dependent contribution to T₁(total) and consequently T₁(CSA) can be determined, as summarized in Table S5. How much the CSA contributes overall to the longitudinal relaxation can be expressed by the ratio of relaxation rates R₁(CSA)/R₁(total). Values of this ratio from over 50% at a low magnetic field (9.4 T) to over 70% at a moderate magnetic field (14.1 T) were observed for GB1 I6SeM, V29SeM, and V39SeM, while GB1 L5SeM exhibits a smaller CSA contribution. Overall, the chemical shift anisotropy relaxation mechanism dominates the longitudinal relaxation for all but GB1 L5SeM at 11.4 T.

In proteins, the T₂ relaxation time of selenium is primarily due to the chemical shift anisotropy driven mechanism and thus it can be approximated by the inverse of the peak width:¹⁶

$$\frac{1}{T_2}\left(\mathrm{CSA}\right)\approx \frac{1}{T_2^{\ast }}=\pi {\nu }_{1/2}$$

Here, for the majority of GB1 variants, the FWHM at 14.1 T is close to 2.25 times the FWHM at 9.4 T, which is the expected ratio when T₂ relaxation predominantly is indeed driven by the chemical shift anisotropy mechanism (see Table S6). The estimation of CSA contributions to relaxation times is further discussed in the context of the experimentally measured CSA below. The variants that exhibit large FWHMs are GB1 A34SeM and V54SeM. The mutation of Ala to SeM caused a change in the hydrophobic core of GB1 A34SeM, creating a large pocket that is in contact with the solvent. In the crystal structure that pocket was frequently occupied by an imidazole that stabilized the SeM conformation. This imidazole binding pocket is exposed to solvent (Figure S7), and both the electron density and the solution NMR suggest a significant degree of SeM’s conformational freedom. We note that the large value for GB1 V54SeM may be associated with the more extensive structural changes (Figure S8). Because it is also the variant with the most limited structural information, we followed up by recording ¹H–¹³C HSQC and ¹H–¹⁵N HSQC to characterize the structural perturbation caused by introducing SeM.

Correlation Spectra.

Natural abundance two-dimensional (2D) (¹H,¹³C)-heteronuclear single-quantum coherence (HSQC) and (¹H,¹⁵N)-HSQC were recorded for all six variants (Figures S11–S13). (¹H,¹⁵N)-HSQC was also collected for the SeM-free GB1 to assist in the comparison of the structural perturbation. The spectra further support our observation that all variants are well-folded. The differences in the (¹H,¹⁵N)-HSQC spectra between the variants and the SeM-free GB1 points again to GB1 A34SeM followed by GB1 V54SeM as the variants with the most changes as discussed under Structure Determination.

SeM Protons.

The isotropic chemical shifts of the protons adjacent to the selenium were recorded by natural abundance 2D (¹H,⁷⁷Se)-heteronuclear multiple-bond correlation (HMBC) spectra at 298 K (Figure 5). The HMBC spectra were collected for all variants, except GB1 A34SeM, for which sensitivity was too low. Table S7 lists the values of the chemical shifts of ⁷⁷Se and methyl ¹H for the SeM in the different variants. The spread of the chemical shifts of the methyl protons is similar to the range (1.56–2.03 ppm) reported for calmodulin at a comparable temperature by Vogel and co-workers. With 1.4 ppm GB1 V54SeM exhibited the most upfield ¹H chemical shift. Because these upfield shifts of the methyl ¹H for GB1 V54SeM and GB1 L5SeM are caused by nearby aromatic rings,³⁸ they can potentially guide assignments of SeMs that are close to aromatic residues.

Figure 5.

Two-dimensional ¹H–⁷⁷Se HMBC spectrum of five GB1 variants. (A) GB1 L5SeM; (B) GB1 I6SeM; (C) GB1 V29SeM; (D) GB1 V39SeM; (E) GB1 V54SeM.

Solid-State NMR.

While solution NMR yields the isotropic chemical shift, solid-state NMR records the chemical shift anisotropy (CSA) tensor—a rich source of information on the electronic structure, the chemical bonding, and the protein environment.³⁹ Because the CSA can be calculated by DFT, it links experimental measurements to numerical work. It is thus an important milestone for the future development of predictions for biological selenium NMR.

Four GB1 variants were selected to record the [⁷⁷Se]-SeM CSA in the crystalline state: L5SeM, I6SeM, V29SeM, and V39SeM. To maintain a data set that was comparable in both content and fidelity across GB1 variants, two proteins were not included in the solid-state NMR experiments: GB1 A34SeM, because its low solution NMR sensitivity resulted in an incomplete data set, and V54SeM for its suboptimal structural data. To increase the sensitivity of the samples, they were isotopically enriched with ⁷⁷Se by adding [⁷⁷Se]-selenite to the growth medium as previously described.¹⁴ Solid-state NMR samples were microcrystalline GB1 prepared by batch crystallization of the variants using MPD as a precipitant under the same conditions used for the preparation of the X-ray samples. Figure 6 presents the spectra of [⁷⁷Se]-SeM acquired by proton-decoupled cross-polarization magic angle spinning (¹H–⁷⁷Se CP/MAS) solid-state NMR at different spinning speeds ranging from 3 to 8 kHz at a magnetic field of 14.1 T and a temperature of 293 K. We note that it was not possible to directly record the selenium spectra of proteins without cross-polarization due to low sensitivity. Also, selenium atoms are positioned in the crystal 18–28 Å from each other and thus homonuclear ⁷⁷Se–⁷⁷Se dipolar coupling is negligible.

Figure 6.

Solid-state ⁷⁷Se CP/MAS NMR spectra of [⁷⁷Se]-SeM incorporated in the protein GB1 L5SeM (left, panels A–C), GB1 I6SeM (middle panels D–F), GB1 V29SeM (middle, panels G–I) and GB1 V39SeM (right, panels J–L) at different spinning speeds, acquired at 14.1 T at 293 K. (A) GB1 L5SeM at 8 kHz; (B) GB1 L5SeM at 5 kHz; (C) GB1 L5SeM at 3 kHz; (D) GB1 I6SeM at 8 kHz; (E) GB1 I6SeM at 5 kHz; (F) GB1 I6SeM at 3 kHz; (G) GB1 V29SeM at 8 kHz; (H) GB1 V29SeM at 5 kHz; (I) GB1 V29SeM at 3 kHz; (J) GB1 V39SeM at 8 kHz; (K) GB1 V39SeM at 5 kHz; (L) GB1 V39SeM at 3 kHz.

Solid-state NMR presents a different environment from that of GB1 in solution NMR discussed under Solution NMR. First, the solvent includes MPD and isopropyl alcohol (IPA), and thus the solvent organization is substantially different between solution and crystalline conditions. Indeed, the solvent in the two states is different: solution samples employed phosphate buffer poised at pH 5.5, while the crystals were formed in sodium acetate (pH 5.5), MPD, and isopropyl alcohol. In several crystals, it was possible to refine MPD molecules in close vicinity to SeM (Table 2). Lastly, surface-exposed SeM may lose degrees of freedom when the temperature is lowered due to changes in the solvent structure. The change in the environment between solution- and solid-state NMR is reflected in the observed difference of the isotropic ⁷⁷Se chemical shifts by as much as 28 ppm (Tables 3 and 4). In addition, more than one selenium environment was observed at 293 K in all the GB1 variants’ spectra (Figures 6–8). This can be attributed to the slightly different SeM conformation and environment in the two GB1 proteins in the asymmetric unit (Table 2).

Table 4.

Chemical Shift Tensors of SeM in GB1 Variants Measured by Solid-State NMR^a

T (K)	MAS (kHz)	δiso (ppm)	δCSA (ppm)	ηCSA	δ11 (ppm)	δ22 (ppm)	δ33 (ppm)	FWHMb (Hz)	Int1:Int 2c	Δδiso/ΔT (ppm/K)
77Se Selenomethionine
299	3 (site 1)	117.4	−322	0.51	363	198	−203	217	1.0:0.8	0.04 (site 1)
299	3 (site 2)	78.9	−300	0.36	285	178	−219	323
292	3 (site 1)	117.4	−321	0.58	371	184	−203	148	1.0:0.7
292	3 (site 2)	76.0	−304	0.41	291	165	−228	207
267	3 (site 1)	117.0	−335	0.53	374	195	−218	154	1.0:1.1	0.24 (site 2)
267	3 (site 2)	70.5	−317	0.36	286	172	−247	189
248	3 (site 1)	115.2	−336	0.55	375	191	−220	178	1.0:1.0
248	3 (site 2)	66.3	−334	0.42	304	163	−268	193
77Se GB1 L5SeM
293	5 (site 1)	111.3	−256	0.31	279	199	−145	93d	1.0:1.0	−0.09 (site 1)
293	5 (site 2)	110.1	−259	0.35	284	195	−149	118d
279	5 (site 1)	111.8	−251	0.40	287	187	−139	82d	1.0:1.1
279	5 (site 2)	110.9	−260	0.33	283	199	−149	84d
270	5	112.2	−268	0.37	295	197	−155	118	NA
260	5	112.8	−273	0.38	301	198	−161	239	NA
249	5	114.7	−283	0.38	310	203	−168	867	NA
77Se GB1 I6SeM
293	5 (site 1)e	79.1	−149	0.00	154	154	−70	357	~1.0:0e	0.17 (site 1)
279	5 (site 1)	76.9	−155	0.10	161	148	−78	847	1.0:0.2
279	5 (site 2)f	90.8	−97	0.10	147	132	−6	1036
270	5 (site 1)	75.8	−161	0.35	184	128	−85	1094	1.0:0.3
270	5 (site 2)f	88.2	−114	0.40	167	123	−25	759
77Se GB1 V29SeM
293	5 (site 1)	100.0	−275	0.32	281	194	−175	72d	1.0:1.0	−0.12 (site 1)
293	5 (site 2)	98.1	−273	0.34	281	188	−175	74d
279	5	100.3	−282	0.34	289	194	−182	310	NA
270	5	100.8	−283	0.33	288	196	−182	438	NA
260	5	101.5	−284	0.28	283	204	−183	764	NA
249	5f	104.6	−318	0.33	316	211	−213	1881	NA
77Se GB1 V39SeM
293	5 (site 1)	91.8	−287	0.58	318	153	−195	154d	1.0:0.3	0.01 (site 1)
293	5 (site 2)	89.5	−274	0.51	297	157	−185	168d
279	5 (site 1)	91.1	−291	0.56	318	155	−200	163d	1.0:0.2
279	5 (site 2)	88.7	−266	0.60	302	142	−177	185d
270	5	90.4	−294	0.57	321	154	−204	191	NA
260	5	90.3	−296	0.58	324	153	−206	263	NA
249	5	90.0	−309	0.59	336	154	−219	617	NA

^aThe error for the chemical shift tensor parameters extracted by Topspin-SOLA is estimated to be ±3% based on signal-to-noise.

^bThe FWHM is reported as natural line width (no line broadening parameter was applied).

^c“Int1:Int2” means (peak intensity of site 1):(peak intensity of site 2).

^dThe overlapping peaks were deconvolved using Topspin-SOLA.

^eFor GB1 I6SeM, 293 K, MAS of 5 kHz, site 2 was not fitted due to low signal-to-noise.

^fFor GB1 I6SeM, 279 K, site 2, GB1 I6SeM, 270 K, site 2, and GB1 V29SeM, 249 K, the fitting is the least accurate because of the low signal-to-noise of these two resonances.

Figure 8.

Superimposition of ⁷⁷Se CP/MAS NMR spectra at 293 (blue), 279 (red), 270 (green), 260 (purple), and 249 K (yellow) for (A) GB1 L5SeM, (B) GB1 I6SeM, (C) GB1 V29SeM, and (D) GB1 V39SeM. The strongest spinning sideband is shown for all but the GB1 I6SeM spectrum, which has comparatively low sensitivity, and thus several spinning sidebands are shown to demonstrate the temperature dependence. All spectra are acquired at a field strength of 14.1 T NMR and a spinning speed of 5 kHz.

The large chemical shift tensor of selenium allows for straightforward extraction of the tensor elements by analysis of ¹H–⁷⁷Se CP/MAS solid-state NMR⁴⁰ at spinning rates that are slow compared to the chemical shift anisotropies being measured. The [⁷⁷Se]-SeM CSA tensor principal elements, reduced anisotropy, and asymmetry (using the Haeberlen convention as defined in the Supporting Information) were extracted using spectra acquired at a spinning speed of 5 kHz (Table 4 and Table S8). To determine motional averaging of the chemical shift tensor, we carried out variable temperature NMR experiments, which included 249, 260, 270, 279, and 293 K (Figures 7 and 8). (Due to limited sensitivity, it was not feasible to acquire spectra with sufficient signal-to-noise ratio to analyze at 249 and 260 K for some of the variants.) The tensors and the δ_iso of the four variants between 249 and 293 K are listed in Table 4.

Figure 7.

Solid-state ⁷⁷Se CP/MAS NMR spectra of GB1 variants recorded at different temperatures. All spectra were acquired at 14.1 T and a spinning speed of 5 kHz. Each column corresponds to one GB1 variant, in the order of (from left to right): GB1 L5SeM, GB1 I6SeM, GB1 V29SeM, and GB1 V39SeM. From top to the bottom: ⁷⁷Se CP/MAS NMR spectra acquired at 249, 260, 270, 279, and 293 K. Last row: superimposition of spectra acquired at 293 K (blue) and a lower temperature (red).

GB1 L5SeM.

In GB1 L5SeM, there are two equally populated resonances, with a chemical shift difference of 1.2 ppm (Table 4). This may be attributed to a variation in the position of the SeM-proximal residues Thr16 and Thr25 in the two proteins in the asymmetric unit. Specifically, Thr16 has several occupied positions. This generates different environments around SeM5 in the two protein copies, which are potentially reflected in this chemical shift difference. For this variant the isotropic chemical shifts in the crystalline state are downfield from their solution value by 11 ppm (Tables 3 and 4). The reduced CSA, denoted by δ_CSA, a measure of the separation of δ_zz from the isotropic chemical shift, is about 80% that of the free amino acid SeM.

GB1 I6SeM.

Initially, GB1 I6SeM exhibited at least four resonances (Figure S14 and Table S9), but upon slow cooling of the sample and heating back to 293 K, the crystals annealed and only one intense and at least one weak powder pattern were present (Figure 6). In GB1 I6SeM, the SeM is located in the interface between proteins in the crystal, and four SeM conformations are distinctly visible in the electron density map (see discussion above about GB1 I6SeM’s X-ray structure). Following this annealing step, at 279 K the difference in isotopic chemical shift between the two GB1 I6SeM powder patterns is 14 ppm. This is a notable difference for simply an alternate conformation and reinforces the observations regarding selenium’s sensitivity. The line widths of resonances at 293 K, reported in Table 4, are the highest for GB1 I6SeM among the four variants. The isotropic chemical shifts in the crystalline state are downfield from their solution value by 28 ppm (Tables 3 and 4). This shift may be attributed to a difference in conformation due to crystal contacts, the presence of MPD and IPA, the presence of residues from a nearby protein chain, or GB1’s reported tendency to form polymorphs.⁴¹

GB1 V29SeM.

For GB1 V29SeM, more than four resonances were observed at 293 K, but only two were significantly populated, with a difference of 1.9 ppm in isotropic chemical shifts (Figures 6–8 and Table 4). As in the case of GB1 L5SeM, the two main resonances were again similarly populated, suggesting a relation to the two proteins in the asymmetric unit. The resonances with lower populations may be explained by crystal heterogeneity or alternative conformations. The line widths of each resonance site at 293 K, reported in Table 4, are narrowest for GB1 V29SeM among the four variants. The isotropic chemical shifts in the crystalline state are downfield from their solution value by 28 ppm because the SeM is—as in GB1 I6SeM—solvent exposed (Tables 3 and 4).

GB1 V39SeM.

In GB1 V39SeM, SeM is exclusively in the hydrophobic core and the weak interactions are primarily van der Waals. Like other mutants, GB1 V39SeM exhibited two resonances, 2.3 ppm apart, with an intensity ratio of 3:1 (Figures 6–8 and Table 4). The isotropic chemical shifts in the crystalline state are downfield from their solution value by 11 ppm, similar to GB1 L5SeM (Tables 3 and 4).

In terms of the temperature dependence of GB1 variants’ ¹H–⁷⁷Se CP/MAS solid-state NMR (Table 4, Figure 8, Figure S15), we note that overall variants with a surface-exposed SeM such as GB1 I6SeM and V29SeM had the largest increase in their ⁷⁷Se peak widths as the temperature was lowered. The change for GB1 L5SeM and V39SeM was appreciably smaller because of their SeM position in the hydrophobic core. Particularly, GB1 V39SeM showed the smallest change in line width and isotropic chemical shift with changing temperature. This may be caused by the restriction of motion in that protein core site. For both GB1 L5SeM and V29SeM, higher temperatures unmask previously overlapping resonances that, based on their similar intensities, may be selenium sites in the two GB1 copies in the crystallographic asymmetric unit. Overall, the line broadening behavior of the variants agrees with the frequently described broadening of the ¹³C line width as the solution freezes.⁴² This broadening is often attributed to the freezing of free water and restriction of available space for molecular motions. For selenium in SeM, slowing down of the methyl group rotation is likely to be a contributing factor as well.⁴³ Since our crystals contain the cryoprotectants MPD and IPA, we estimate that the GB1 solid-state NMR samples freeze close to 260 K.

Lower temperature spectra showed significant changes for GB1 I6SeM aside from line broadening. At 270 K, the weaker powder pattern is now more pronounced (Figures 7 and 8). There are two potential explanations: either the conformation represented by the weaker pattern was undergoing chemical exchange at 293 K, which was slowed down at lower temperature, allowing its detection at that temperature, or the change in temperature and structure of the solvent led to a population shift into that conformation once the water froze.

When comparing the temperature dependence of the CSA tensor of the four variants (Figure S15), we see that changes in the individual shielding elements are similar for GB1 L5SeM and GB1 V29SeM but different for GB1 V39SeM. The principal elements of the ⁷⁷Se chemical shift tensor in GB1 V39SeM were only mildly dependent on the measurement’s temperature between 260 and 293 K but do show an increase in δ₁₁ and a decrease in δ₃₃ at 249 K. In contrast, GB1 I6SeM showed a more extensive change in all tensor elements in a temperature range from 293 to 270 K. Unfortunately, the limited sensitivity of GB1 I6SeM at lower temperatures restricts similar analysis of this surface-exposed SeM. We mention in passing that GB1 I6SeM exhibited the largest temperature coefficient among all GB1 variants. This is the only variant with a relatively large solvent-exposed surface in which SeM is near a fully charged residue (Lys4).

Free SeM.

In addition to the four GB1 variants, we have also examined the CSA tensors of the free amino acid SeM in powder form over the same temperature range to probe which environmental factors are most pertinent (Table 4 and Table S8). Previous studies of crystalline SeM by both X-ray diffraction and solid-state NMR reported two molecules in the crystal with distinct conformations.⁴⁴ The change in conformation and environment of the SeM in the two locations was sufficient to introduce a difference of 50 ppm in the δ_iso and 28 ppm in δ_CSA (~8%) between the two selenium sites at 183 K.^44,45 In molecule A, with χ₃ = 61.2°, the environment is primarily aliphatic with the selenium 3.8 Å from methylene groups, 3.7 Å from a methyl group, and 4.6 Å from the selenium of nearby molecules in the crystal. In contrast, the selenium environment in molecule B is near a charged group, the dihedral angle χ₃ = −76.8°, and the selenium is 4.0 Å from a methylene, 3.9 Å from a methyl, 4.1 Å from its own amino group, 3.6 Å from a nearby carbonyl, and 4.6 Å from a nearby selenium. The NMR parameters collected here agree with the previous reports but were collected here over a temperature range more relevant for our studies of the GB1 variants. We note that site 1 has the largest absolute value of δ_iso and δ_CSA. Site 2, which is logically molecule B, has the largest chemical shift change per degree, and the temperature coefficient Δδ_iso/ΔT = 0.24 ppm/K (Table 4) is higher than those observed for GB1 variants. Temperature-induced shifts in δ_iso are typically positive (δ_iso increases with higher temperatures)⁷ and reflect changes in the site such as conformation, vibrations, and molecular interactions.⁴⁶ The higher temperature coefficient at site 2 may be due to the more polar environment which is rich in electrostatic interactions or to a higher mobility. Similarly, the temperature coefficients of GB1 variants are well within the reported range for small molecules (Table 4).

Correlation between Measured CSA and Solution-State Relaxation Data.

An experimentally determined CSA enables a comparison to the CSA values inferred from solution-state measurements. To that end, first the contribution of the CSA mechanism to the T₁ and T₂ relaxation times, T₁(CAS)_est and T₂(CSA)_est, were extracted from the solution-phase spectra.⁴⁷ T₁(CAS)_est was estimated from T₁’s magnetic field dependence (Table 3) according to the common procedures described in the Supporting Information alongside Table S5, Figures S10, and Figure S11. The estimate T₂(CSA)_est was obtained from the line width of the resonance (Table S10). From the ratio of these two CSA contributions a correlation time, τ_c, can be calculated that should characterize selenium’s motion. Yet, this only holds true if a single correlation time is sufficient to capture the different aspects of the nucleus’s motion. The measured CSA offers an alternative route to calculate the CSA contribution to T₂ from τ_c (denoted by T₂(CSA)_calc; see the relevant equation in Table S10). The comparison of T₂(CSA)_calc with the T₂(CSA)_est extracted from solution-state data further illuminates the mobility characteristics of selenium at the different locations in the GB1 variants (Table S10).

For the GB1 L5SeM, V29SeM, and V39SeM variants the calculated T₂(CSA)_calc using this rotational correlation time matched the estimated T₂(CSA)_est well. However, for more flexible SeMs, such as the one in GB1 I6SeM, such an agreement could not be established. This surprises little, because for GB1 I6SeM the line shape was influenced by chemical exchange resulting in a poor estimation of T₂(CSA)_est from the line width. Similarly, when the estimated T₂(CSA)_est was used to calculate the reduced chemical shift anisotropy (Table S11), the agreement with the measured δ_CSA was reasonable for the GB1 L5SeM, V29SeM, and V39SeM variants but much less so for GB1 I6SeM. This demonstrates that these kinds of calculations can be used to identify unusual behavior or dynamics of SeM at specific sites.

DISCUSSION

Our main objective for this survey of the NMR properties of SeM in different environments within GB1 was to lay a foundation and create reference points for the interpretation of ⁷⁷Se NMR spectra of biological samples. Along the way, we gained insights into several aspects of protein ⁷⁷Se NMR. Some arise from the first-of-a-kind selenium NMR measurements that are part of this work, others concern the behavior of selenium in proteins and its NMR characteristics, and still others relate to technical aspects of these experiments. In the following we summarize what was learned in these different areas and how it points naturally toward the next steps and new directions for ⁷⁷Se NMR of proteins.

While common practice for other NMR nuclei, this is the first report of the chemical shift tensors of selenium in a protein. This was only possible because our previous work finally enabled us to efficiently prepare large amounts of proteins containing SeM and specifically [⁷⁷Se]-SeM. Consequently, solid-state NMR samples could be produced that had enough sensitivity to measure the CSA and extract the tensor elements. We took advantage of this ability and created a library of SeM containing GB1 variants not only to measure the environment dependent selenium tensors, but also to determine the high-resolution structure of the very same proteins with X-ray crystallography.

The purpose is not to simply showcase new abilities, but to provide, together with the NMR parameters, as much structural information on the selenium environment as possible. This is necessary to aid developments of density functional theory calculations that can investigate the origin of selenium NMR parameters in biological systems. So far the advancement of theoretical calculations of selenium’s magnetic shielding tensors could only be based on measured chemical shift tensors of seleno compounds.⁶ Now as we provide here chemical shift tensors for selenium in proteins, it will be exciting to see how previous numerical efforts on small molecules carry over to biological macromolecules. Only isotropic chemical shift calculations for selenocysteine in biological systems have been reported.¹¹ The quantitative prediction of the information-rich full chemical shift tensors for selenium in a biological system has yet to be achieved. We hope our NMR and structural data can provide a test bed not only to develop suitable DFT calculations, but also to work out force fields for SeM in molecular dynamics simulations and to help in the selection of promising strategies for emerging computational approaches.

From a technical perspective, we confirmed that concentrated protein samples with a natural abundancy of ⁷⁷Se are sufficient for data collection in solution ⁷⁷Se NMR experiments. Also, HMBC experiments can readily employ natural abundance protein samples at moderate magnetic fields. We found that sample stability was not an issue for GB1 variants, although SeM and Met oxidation is known to be protein and position specific.³ For ⁷⁷Se NMR, however, this seems to be not a concern since SeM oxidation is easy to detect by the chemical shift difference. Furthermore, SeM oxidation was not observed in our GB1 samples even when stored over 1 month at 277 K.

Beyond these utilitarian aspects, our measurements already provide a wealth of information on SeM at different sites, thus deepening our understanding of SeM in proteins and aiding prediction of its behavior. Case in point are the observed correlation between SeM’s location in the protein, the associated mobility, and the observed ⁷⁷Se NMR characteristics. From the structural analysis and comparison to dynamics of the side chain in wild type GB1, it is clear that in GB1 variants where the packing of the protein core stayed intact, such as L5SeM and V39SeM, the side chain of SeM remained rigid, and narrow NMR resonances resulted. In contrast, the side chain gains mobility in the looser core packing of GB1 A34SeM and V54SeM, where the mutations changed the aromatic stacking. This is mirrored by the observed line broadening and shortest T₁ relaxation times for these variants. Finally, if there is little constraint to rapid movement of the side chain, as exemplified by the variants where SeM is fully (I6SeM) or partially (V29SeM) solvent exposed, again narrow resonances are observed, only this time caused by motional averaging. However, T₁ relaxation times are the longest for these positions, because the fast motions contribute little to the longitudinal relaxation. Yet, even on the surface of proteins, SeM seeks molecular interactions with other residues, and its weak nonbonding interactions do result in conformational preferences as exemplified by the structure of the GB1 I6SeM variant. These observations invite the generalization that both SeM within a tight protein core and SeM on a protein surface should be straightforward to detect due to the narrow lines. Nevertheless, in cases of partial constraint of SeM such as in binding pockets or interfaces of proteins and domains, broader and thus harder to detect lines can be expected.

Because wild type GB1 has been extensively characterized by both solid- and solution-state NMR, it is possible to compare the evidence for SeM’s mobility to that of the original residue at this site.^37,48,49 These previous studies of crystalline GB1 samples reported low mobilities and thus rigid side chains for L5, V29, and V54, due to their tight and constraint packing in the core. A higher mobility was exhibited by V39, and for I6 a very dynamic side chain, appropriate for its surface exposure, was established. In contrast, the motion of alanine’s methyl group was mostly unaffected by the local packing density.³⁷ Here, we have shown that when SeM replaces a small amino acid at the core, the resulting modifications of the weak interactions at the site lead to an increased mobility. The substitution of a small amino acid at a more flexible location, such as V39, proved however much less disruptive and had a reduced impact on side chain mobility. If SeM replaced a structurally similar amino acid such as Leu or Ile, dynamics changed, as one would expect, little, regardless if the site is located at the core (L5) or on GB1’s surface (I6). Overall, these effects on the local dynamics at the substitution site are mirrored and supported by the B factors, electron densities at the selenium site, and the measured relaxation times. We also find that for variants in which the GB1 structure was mostly undisturbed, and SeM was not highly mobile, both T₁ and T₂ relaxation times were dominated by CSA-driven relaxation. The exception to this was GB1 L5SeM, whose proximity to aromatic rings maybe important in this regard. Generally, when designing selenium NMR experiments, data interpretation can be simplified if either native Met residues or similar amino acids (Ile and Leu) are substituted by SeM. Otherwise, the changes to the protein structure may need to be characterized.

We also observe that selenium’s sensitivity to its chemical environment within the protein is so pronounced that alternative SeM conformations within the same sample could be differentiated. For example, more than four distinct resonances were observed when GB1 I6SeM crystals were freshly packed, with a chemical shift difference of 15.2 ppm at 299 K. Even in the cases where one crystallographic asymmetry unit contained two GB1 proteins, the small difference between the selenium environment in the two copies led to detectable variations in the NMR signal.

The above insights and observations suggest several avenues moving forward. For one, it seems, once equipped with a protein structure and knowledge of the selenium positions, the assignment of multiple SeM residues could be already successful. If, however, more information regarding the immediate environment around selenium sites is needed, performing NMR experiments utilizing dipolar and J-couplings is possible.

CSA tensors, particularly those of electron-rich nuclei, such as fluorine, metals, and here selenium, are sensitive not only to the chemical environment but also to molecular motions. Thus, selenium’s large chemical shielding tensor could be used to detect motions on the microsecond time scale which is usually probed by deuterium. Furthermore, ⁷⁷Se introduces a large chemical shift in directly bonded carbon atoms allowing their immediate identification.¹⁹ In addition, selenium’s sensitivity to both the chemical environment and SeM’s mobility could turn ⁷⁷Se NMR into a valuable technique to study the dynamics and structural changes at protein interfaces and domain junctions. When studying large proteins, protein complexes, and protein interactions, such capabilities would prove especially useful. The ability to report on minute changes in the environment and weak interactions could also be beneficial in other areas, for example, ligand binding events or even interactions with intrinsically disordered proteins. With the progress made on the foundational work, it seems there is very little that stands in the way of ⁷⁷Se NMR becoming an important tool to study the molecular machinery of life.