Specific Labeling of Threonine Methyl Groups for NMR Studies of Protein-Nucleic Acid Complexes

Abstract

Specific ¹³C labeling of Thr methyl groups has been accomplished by growth of a standard laboratory strain of Escherichia coli on [2-¹³C] glycerol in the presence of deuterated isoketovalerate, Ile, and Ala. Diversion of label from the Thr biosynthetic pathway is suppressed by including Lys, Met, and Ile in the growth media. This method complements the repertoire of methyl-labeling schemes for NMR structural and dynamic studies on proteins and is particularly useful for the study of nucleic acid binding proteins due to the high propensity of Thr residues at protein-DNA and -RNA interfaces.

NMR spectroscopy is a powerful tool for probing protein dynamics over a range of time-scales(1). The combination of progress in optimized pulse sequences, better hardware, and better isotopic labeling schemes has allowed researchers to study proteins of high molecular weight(2, 3). Historically, dynamics studies using NMR spectroscopy have focused on the protein backbone (amide groups). However, the N-H group may not reflect side-chain motions and hence the need to study side-chain dynamics. Among all the groups available to study side-chains, methyls are especially suited for the purpose as they are well-distributed through the protein structure. Furthermore, the methyl HMQC (heteronuclear multiple-quantum coherence) experiments can exploit the TROSY effect resulting from cancellation of intra-methyl dipolar relaxation interactions to give better resolved and more sensitive spectra(4). This has enabled the use of methyl groups to study the dynamics of high molecular weight proteins and their complexes.

A prerequisite for ¹³C-based methyl dynamics studies in large proteins is that the methyl carbons should be specifically ¹³C labeled (adjacent to ¹²C), and protonated in a perdeuterated background. Initially, methyl labeling utilized ¹H,¹³C-pyruvate (Pyr) as the sole source of carbon in deuterated media(5). More recent strategies involve the use of selectively labeled α-keto acids to specifically incorporate protonated ¹³C methyls in proteins(6, 7). Commercially available precursors enable incorporation of ¹³CH₃, ¹³CH₂D, or ¹³CHD₂ groups at the Ile, Leu and Val methyl sites in the protein. The CH₃ isotopomer allows characterization of ms-μs motions using relaxation dispersion experiments(8, 9). The CH₂D and the CHD₂ isotopomers are well suited for deuterium relaxation studies (10), while the CHD₂ isotopomer is ideal for carbon relaxation (11). More recently, a procedure to specifically label the Ala methyls has been published (12). This protocol involves using perdeuterated α-ketoisovalerate, succinate and isoleucine to prevent the scrambling of the ¹³C label. A more general labeling approach uses [1-¹³C]- or [2-¹³C]- glucose(13); or [1,3-¹³C]- or [2-¹³C]- glycerol(14) as the sole carbon source to express proteins which are labeled at the methyl carbons. A disadvantage of using glucose instead of glycerol as the carbon source is that the conversion of glucose to Pyr dilutes the ¹³C label by 50%.

The methyl-labeling techniques developed to-date have focused on Ala, Ile, Leu and Val side-chains, all of which lack hydrogen bonding capabilities. In contrast, Thr can participate in hydrogen bonding as well as nonpolar interactions. Due to the different properties of methyl-containing amino acids, the distribution of these residues differs considerably for protein-protein, protein-RNA, and protein-DNA interfaces (Figure 1). In the case of DNA-protein interfaces, Leu (−.69), Ile (−.25), Ala (−.16), and Val (−.17) are all under-represented, showing negative propensities (given after residue name), while Ser (0.25), Thr (.27), Phe (.35), Trp (.41), Tyr (.46), and Arg (.57) are over-represented at protein-DNA interfaces. Thr residues are found with nearly equal propensity in both the core and rim of protein-DNA interfaces. Consequently, Thr residues potentially provide better probes of DNA- and RNA-protein interfaces than any other methyl containing amino acid currently utilized for NMR studies. Even though Thr methyls have a narrow chemical shift range, their proximity to nucleic acids is expected to result in a broader dispersion of shifts.

Figure 1.

Propensity for selected residues found within protein-protein (P-P), protein-RNA (P-RNA), or protein-DNA (P-DNA) interfacial regions. The P-DNA interface is divided into a core (solvent inaccessible) and a rim region (partially solvent accessible). Propensity is defined as ln(f_i/f_i°) where f_i is the area of i^th residue in the interfacial region and f_i° is the area occupied by the residue on the remaining surface of the protein. Total P-DNA propensities for each residue are given in the text. Data from (15–17).

Presented here is a technique to uniquely label the methyl group of threonine residues of a protein expressed using a standard Escherichia coli T7 expression host(18) without genetic modification. The technique works in H₂O as well as ~100% D₂O media. We show that there is no ¹³C-¹³C coupling to the adjacent carbon and that ¹³C-H cross-peaks in the deuterated sample arise from the -¹³CHD₂ isotopomer.

We use [2-¹³C]-glycerol as the primary carbon source in the bacterial growth medium. The metabolism of glycerol in E. coli is well understood (see Figure S1). Glycerol is converted to Pyr in the glycolytic pathway and then enters the TCA cycle as acetyl-CoA by condensation with oxaloacetate (OAA) to form citrate. Following the 2-C atom from glycerol through the TCA cycle, the labeled carbon becomes either C1 or C4 of OAA. The OAA can then either get converted to Asp or can re-enter the TCA cycle by reacting with another molecule of acetyl-CoA. On re-entering the TCA cycle, both C1 and C4 of OAA will be lost as CO₂. Therefore, the ¹³C label on glycerol either goes to Asp and subsequently to Thr or is lost as CO₂. The side-chain carboxyl group of Asp ultimately forms the methyl group on Thr, with the methyl protons arising from the media as well as from the carbon source. In 100% D₂O containing media it has been shown that 45% of the Thr methyls are labeled as ¹³CHD₂ while 81% of the Thr C_β are deuterated(19). The high level of deuteration on the C_β position leads to favorable relaxation properties in larger proteins.

A number of biosynthetic pathways can lead to a dilution of the ¹³C label (see Figure S1). Pyruvate can be converted to Ala, Leu, Val and Ile while Asp can be converted to Met and Lys. Thr itself can be converted to Ile. Therefore, inhibiting these pathways is important to prevent dilution of the label. This was accomplished by adding deuterated Ala, α-ketoisovalerate, Ile, and protonated Met and Lys to the media before inducing protein expression (see supporting information for more detail). Since the carbon lost from the TCA cycle due to biosynthetic needs is obtained from CO₂, NaH¹³CO₃ was added to the media to enhance labeling (see below). The labeling scheme was tested on the 130 residue RNA binding domain of the transcriptional termination factor rho (rho130) (20). To estimate the ¹³C incorporation at the methyl group of Thr, Met was labeled by adding ¹³C methyl labeled Met to the media. The Thr labeling efficiency was tested using the following sources of ¹³C: a) [2-¹³C] glycerol b) [2-¹³C] glycerol and NaH¹³CO₃ (2g/L) c) [2-¹³C] glycerol and NaH¹³CO₃ (10g/L) and d) NaH¹³CO₃ (10g/L). Media containing 50% and 100% D₂O was used to test the contribution of the solvent to protonation of the Thr methyl.

The methyl region of the C-H correlated spectra of uniformly ¹³C labeled rho130 and the specifically labeled sample (only the Thr and Met methyls labeled) is shown in Figure 2 (panels A and B). The Thr labeling protocol is extremely specific and reduces the intensity of the other methyl peaks by ~95%, significantly reducing the crowding in this region of the spectrum. Figure 2 B also shows the 1-D slice along the carbon dimension of a threonine peak. The lack of any splitting shows that the carbon adjacent to the methyl ¹³C carbon is not labeled. Figure 2 C shows the spectrum of the deuterated specifically labeled sample acquired using a refocused-INEPT experiment which selects for ¹³CHD₂ signals(11). A C-H HMQC experiment showed that that ¹³CHD₂ isotopomer is the predominant protonated species (approx. 90% ¹³CHD₂ and 10% ¹³CH₂D) when the cells are grown in 100% D₂O. In the case of expression in 50% D₂O, the relative proportion of ¹³CH₃, ¹³CH₂D and ¹³CHD₂ is approx. 40:40:20, thus some fraction of the hydrogens on the Thr methyl group are supplied directly from the solvent as well as from the protonated carbon source.

Figure 2.

2D ¹H-¹³C spectra of rho130. A) HSQC of uniformly ¹³C labeled protein, B) HSQC of rho130 specifically labeled at Thr and Met methyl groups, and C) refocused-INEPT experiment on deuterated rho130 sample showing the Thr CHD₂ peaks.

The efficiency of ¹³C labeling of the Thr methyls using glycerol alone is 32% (Figure 3). If all of the ¹³C from the 2-C of glycerol was used for Thr synthesis then the γ-methyl would be labeled to a level of 50% (see Figure S1). Numerical simulations show that with 32% labeling, approximately 40% of the carbon flux in the TCA cycle is diverted to biosynthetic needs. Because this carbon is replenished by carboxylation of phosphoenolpyruvate with ¹²CO₂ to produce OAA, the observed labeling efficiency is less than 50%. Since the incorporated CO₂ becomes the C4 position in OAA, the methyl labeling efficiency can be increased by adding NaH¹³CO₃ to the media prior to induction. Figure 3 shows that an 80% increase in signal intensity, relative to that obtained with [2-¹³C] glycerol alone, was obtained on adding 10g/L of NaH¹³CO₃ to the media. Use of NaH¹³CO₃(10g/L) alone is as efficient as using [2-¹³C] glycerol.

Figure 3.

Mean normalized peak intensities when using different ¹³C sources. Individual peaks were first normalized to the corresponding peaks in a uniform ¹³C labeled sample to account for intensity differences due to polarization transfers. Each set was then normalized to the most intense Met resonance. The bars indicates the range of values observed for the 5 Thr residues.

When [2-¹³C] glycerol is used as the carbon source the carbonyls of Arg, Asp, Asn, Glu, Gln, Leu, Pro, and Thr residues will be selectively labeled. The ¹³C carbonyl can be used to filter the 2D ¹H-¹⁵N spectrum (Figure S2), showing only peaks from NH groups preceding the labeled carbonyl. The intensity of the peaks in the filtered spectrum is consistent with the expected labeling pattern reported in earlier studies (13, 14, 21) (Figure S3). Since carbonyl editing allows the identification of the preceding residue type for NH, this information can be used in the mainchain resonance assignment process (22, 23). Note that the residues that are labeled at the carbonyl position with [2-

¹³C] glycerol are complementary to the set of residues (I, L, Y, V, A, P) identified as being useful for assignment purposes using individual ¹³C carbonyl labeled amino acids (23).

In addition to selective labeling at the carbonyl position, the use of [2-¹³C] glycerol also leads to selective labeling of the C_α position (13, 14, 21). An analysis of peak intensities in the C_α filtered ¹H-¹⁵N spectra of rho130 (Figure S4) shows that this labeling pattern is preserved with the Thr labeling scheme. The simplification of the C-H spectrum facilitates C_α relaxation studies, an important contribution to the characterization of backbone dynamics.

To summarize, the Thr labeling protocol can be easily incorporated into standard protein expression protocols. The labeled methyl carbon is not coupled to the adjacent carbon, and when using 100% deuterated media, ¹³CHD₂ is the predominant species. Also, due to the use of deuterated α-ketoisovalerate, Ala, Met, and Ile, even proteins expressed in H₂O based media would have a significantly deuterated methyl background. These characteristics make this protocol ideal for relaxation studies. We note that although the labeling efficiency of the Thr methyl with respect to ¹³C exceeds 50%, the fraction of ¹³CHD₂ methyls is estimated to be ~25% when 100% D₂O is used, reducing the overall efficiency relative to other methyl-labeling schemes. However, the fraction of the ¹³CHD₂ isotopomer can be increased by raising the H₂O content of the media. The protocol also selectively enriches the labeling of the carbonyl carbons, providing useful information for resonance assignments. This protocol complements the set of methyl labeling schemes available to the NMR community and should be particularly useful in the study of protein-DNA interactions.