A ¹³C labeling strategy reveals a range of aromatic side chain motion in calmodulin

Abstract

NMR relaxation experiments often require site-specific isotopic enrichment schemes in order to allow for quantitative interpretation. Here, we describe a new labeling scheme for site-specific ¹³C-¹H enrichment of a single ortho position of aromatic amino acid side chains in an otherwise perdeuterated background by employing a combination of [4-¹³C]-erythrose and deuterated pyruvate during growth on deuterium oxide. This labeling scheme largely eliminates undesired contributions to ¹³C relaxation and greatly simplifies the fitting of relaxation data using Lipari-Szabo model-free formalism. This approach is illustrated with calcium-saturated vertebrate calmodulin and oxidized flavodoxin from Cyanobacterium anabaena. Analysis of ¹³C-relaxation aromatic groups of calcium-saturated calmodulin indicates a wide range of motion in the sub-nanosecond time regime.

Nuclear magnetic resonance (NMR) relaxation has proven to be a versatile probe of the link between fast internal protein motions and their relevance to function.^1-4 The motions of the polypeptide chain or of the amino acid side chains in proteins of significant size are most often studied using ¹⁵N-relaxation of amide nitrogen and deuterium or carbon relaxation in methyl groups.^2,3 This is generally due to restrictions arising from requirements of isotopic labeling and unfavorable relaxation properties of some sites within proteins. In other contexts, more specific and tailored enrichment schemes are often vital in order to eliminate unwanted dipolar and scalar interactions as well as to simplify data interpretation. Examples include use of [3-¹³C]-pyruvate⁵, [2-¹³C]-glycerol or [1,3-¹³C]-glycerol^6,7 or mixtures of singly ¹³C-enriched acetates⁸ as carbon precursors to generate isolated ¹³C spins. Even more selective spin enrichment schemes are sometimes required to suppress unwanted spin interactions and often employ more complex biosynthetic precursors. Prominent examples include labeling schemes targeted for optimal relaxation in methyl groups.^5,9,10 Comprehensive chemical synthesis, though relatively expensive, has also proven viable.¹¹

Here we focus on the use of NMR relaxation phenomena to characterize the fast sub-nanosecond motion of aromatic residues. Aromatic amino acid side chains have a rich structural role within proteins^12-15 and are often central to their biological function particularly in the context of molecular recognition^16,17 and catalysis.¹⁸ Thus the motional character of aromatic residues would seem to be of central importance in a range of protein structure-function issues. In this context, NMR phenomena have long been used to characterize relatively slow motions that are manifested in line broadening or population exchange.¹⁹

For classical relaxation phenomena used to probe fast sub-nanosecond motions, aromatic residues present a difficult situation. In addition to the concern about the isolation of the spin interaction of interest from extraneous contributions, aromatic ring systems suffer from extensive homo- and heteronuclear scalar (J) interactions that can also significantly complicate the quality and information content of obtained relaxation data. Previously, 1-¹³C- or 2-¹³C-glucose has been used to create isolated ¹³C sites in aromatic side chains and thereby eliminate one-bond ¹³C-¹³C interactions. However, this labeling scheme does not eliminate remote scalar or dipolar interactions with non-bonded ¹H spins.^20-22 In our hands, the 1-¹³C glucose labeling scheme, which is designed to place isolated ¹³C at the δ position, results in minor scrambling of label that confounds somewhat the subsequent analysis (see below). In addition, though anticipated to be less of an issue than for methyl ¹³C relaxation studies,^9,23 the presence of remote ¹H spins does present a complication to the analysis of aromatic ¹³C relaxation in proteins. Isolation of the ¹H-¹³C pair in an otherwise perdeuterated background would eliminate potential complications from dipolar interactions with remote ¹H spins as well as scalar couplings with other ring hydrogens. Labeling strategies based on glucose are unable to provide this labeling scheme. To largely overcome this limitation we have developed a biosynthetic strategy that takes advantage of the flow of carbon from erythrose 4-phosphate (E4P) to the biosynthesis of the Tyr, Phe and Trp through condensation with phosphoenolpyruvate²⁴ (Figure 1). Consideration of this pathway suggests that use of [4-¹³C]-erythrose as a sole source of ¹³C in conjunction with deuterated ¹²C-pyruvate will lead to creation of an isolated bonded ¹H-¹³C pair at a single delta position (C2 position) within the aromatic ring. Protein expression during growth on [4-¹³C]-erythrose and deuterated ¹²C-pyruvate and 99% D₂O allows for exchange of all other non-aromatic hydrogens while preserving that bonded to the target ¹³C. In the case of phenylalanine and tryptophan, an additional ¹H spin is predicted to remain at the ζ position (C4 position) of the benzoid ring. Fortunately, this spin is greater than 3 Å away from the sole ¹³C nucleus rendering its contribution to relaxation negligible. All other aromatic carbons remain NMR inactive ¹²C nuclei. This labeling pattern therefore largely eliminates the potential complications of extraneous intra-ring scalar or dipolar interactions (¹H or ¹³C) with the isolated ¹³C spin. The low gyromagnetic ratio of the replacement deuterons will cause them to contribute insignificantly to relaxation of the isolated ¹³C nucleus. Similarly, the presence of random ¹³C at natural abundance will also have negligible contribution to the measured relaxation of the target ¹³C nucleus.

Figure 1.

Carbon inflow into the aromatic pathway through the condensation reaction of erythrose 4-phosphate and phosphoenolpyruvate. Green and purple highlighting indicates the carbon originating from pyruvate and erythrose, respectively. The position of the ¹³C label is indicated by ‘•’.

Vertebrate calmodulin was expressed in E. coli²⁵ to test this strategy. ¹²C,²H-pyruvate was used to suppress scrambling of ¹³C to other amino acids and to aid in perdeuteration of the aromatic ring. One-dimensional ¹³C-filtered and unfiltered ¹H spectra reveal essentially complete deuteration of the protein except at single delta-positions in Tyr and Phe residues and the ζ position (C4) of Phe (Figure 2). No protonation at the ε-position (C3, C5) of the aromatic ring is observed. ¹³C-labeling is restricted to the delta positions of Tyr and Phe. The enrichment of ¹³C-¹H pairs at the delta position was determined by comparison of the unfiltered ¹H spectrum and the ¹³C-filtered ¹H spectrum with and without ¹³C decoupling, and was found to be uniformly 67% (Supplementary Fig. S1). Analysis of the one-dimensional ¹³C spectra with and without ¹H coupling confirmed that the isolated ¹³Cδ-¹H has not been diluted with ²H from solvent (Supplementary Fig. S3). The under-labeling of ¹³C is a consequence of using a ratio of 3:2 for ¹²C,²H-pyruvate to [4-¹³C]-erythrose, which is motivated by the need to suppress scrambling of ¹³C to other amino acids or other positions in aromatic rings. This is a reasonable price to pay to maintain the fidelity of the ¹³C nucleus for relaxation studies.

Figure 2.

(a) Overlay of the one-dimensional ¹H spectrum of aromatic region from ¹³C filtered and decoupled (lower, blue) and unfiltered ¹H spectrum (upper, black) of calcium-saturated calmodulin expressed during growth on [4-¹³C]-erythrose, deuterated pyruvate and D₂O indicating a selective introduction of a ¹H-¹³C pair at a single δ-carbon of Phe and Tyr. Also evident are resonances arising from hydrogen bonded to ¹²C at the ζ position of the aromatic ring of Phe (red arrows). These resonances have narrow line widths due to the absence of scalar coupling. A more detailed analysis of the ¹H spectrum is presented in the supplementary (Fig. S1 and S2). (b) The two-dimensional ¹³C-HSQC spectrum of the aromatic region and the corresponding one-dimensional ¹H spectrum (¹³C filtered) are shown. No significant ¹³C labeling of other amino acids was observed.

Calmodulin does not contain tryptophan. To confirm that appropriate labeling of Trp is also achieved, flavodoxin C55A was similarly prepared.²⁶ Flavodoxin has four Trp, eight Tyr, eight Phe and one His residue. The 22 anticipated aromatic ¹H-¹³C correlations are seen in the ¹³C-HSQC spectrum (Supplementary Figure S4). No other significant ¹³C-labeling was observed. Thus the desired labeling pattern was observed for all four aromatic amino acid side chains in the context of a perdeuterated background.

¹³C R₁ and R_1ρ relaxation in calcium-saturated calmodulin prepared using this labeling strategy was measured at three magnetic fields (11.7, 14.0 and 17.6 T) using standard two-dimensional sampling pulse sequences.¹⁰ For comparison, similar measurements were done using calmodulin prepared with a labeling strategy based on [1-¹³C]-glucose^27,28 (Figure 3). The anisotropy of global macromolecular tumbling was characterized in the usual way²⁹ using the crystal structure of calcium-saturated calmodulin (pdb code: 3CLN) and assessing the two globular domains separately.³⁰

Figure 3.

Aromatic ¹³C R1 and R1ρ relaxation in calcium-saturated calmodulin prepared using [4-¹³C]-erythrose/deuterated pyruvate/D₂O strategy (Panels a and b) or a [1-¹³C]-glucose strategy (Panels c and d). Relaxation at the delta-2 position of F92 is shown.

Lipari-Szabo model-free squared generalized order parameters (O²) and effective correlation times (τ_e) were determined using a grid search approach³⁵ and employed an effective bond length of 1.09 Å and residue-type specific chemical shift anisotropy tensors³² with axially symmetric and anisotropic³⁶ CSA values for Phe and Tyr, respectively. The analysis was carried out with an updated version (Relxn2A) of our in-house software.^23,35 Standard statistical F-tests were used to determine if R_ex terms were justified. None were found. The contributions from ¹³C-²H remote dipolar coupling and ¹³C-¹³C dipolar interaction due to natural abundance ¹³C amounted to less than 0.05% of ¹³C-¹H direct bond dipolar interaction. Relaxation data obtained from the [4-¹³C]-erythrose labeling scheme sample fit well to the simple model-free spectra density (Table 1). Obtained squared generalized order parameters ranged from 0.47 to 0.96 indicating a rich spectrum of aromatic ring motion within calcium-saturated calmodulin on the sub-nanosecond time scale.

Table 1.

Lipari-Szabo model-free parameters for aromatic ring motion in calcium-saturated calmodulin^a

Probe	O2	τe (ps)	Probe	O2	τe (ps)
F12b	0.49 ± 0.03	236 ± 21	F89c	0.94 ± 0.02	900 ± 39
F16b,d	0.95 ± 0.01	127 ± 12	F92c	0.72 ± 0.03	624 ± 23
F19b	0.47 ± 0.01	232 ± 19	Y99c	0.79 ± 0.02	240 ± 21
F65b	0.70 ± 0.04	176 ± 23	Y138c	0.89 ± 0.02	604 ± 36
F68b	0.96 ± 0.01	292 ± 17	F141c	0.95 ± 0.01	101 ± 11

^aPrepared using the [4-¹³C]-erythrose labeling scheme. Model-free squared generalized order parameters (O²) and effective correlation times (τ_e) determined using the simple model-free spectral density,³¹ an effective C-H bond length of 1.09 Å, and the axially symmetric and fully anisoanisotropic chemical shift anisotropy tensors for the δ-¹³C of Phe and Tyr, respectively, as determined by Ye et al.³² Macromolecular tumbling was characterized using ¹⁵N-relaxation with an effective N-H bond length³³ of 1.04 Å and a simple uniform ¹⁵N chemical shift anisotropy tensor breadth³⁴ of −170 p.p.m. The N- and C-terminal domains were treated separately. The precision of the squared generalized order parameters(O²) and effective correlation times (τ_e) were estimated by Monte Carlo sampling.

^bPart of the N-terminal domain with an effective macromolecular tumbling time of 8.96 ± 0.14 ns.

^cPart of the C-terminal domain with an effective macromolecular tumbling time of 8.05 ± 0.10 ns.

^dDue to partial spectral overlap, only data obtained at 17.6 T was fitted.

In contrast, though the primary R₁ and R_1ρ relaxation time profiles derived from the calcium-saturated calmodulin obtained from the [1-¹³C]-glucose labeling scheme fitted reasonably to a single exponential decays (Figure 3), the obtained relaxation rates largely gave relatively poor fits to the L-S model-free interpretation (5-10% versus <1% residual error) (Supplementary Table S3). Inclusion of remote ¹H spin dipolar interaction as well as dipolar interaction with ¹³Cε failed to recover the excellent statistics of the relaxation data derived from the more optimal [4-¹³C]-erythrose labeling scheme (Supplementary Fig. S5, Table. S3). It seems likely that unaccounted dipolar relaxation, intra-ring 1-bond ¹³C-¹³C J-coupling, intra-ring 2-bond ¹³Cδ-¹Hε J-coupling effects contaminate the measurement and interpretation of ¹³C-relaxation in structured proteins in a protonated background. This is consistent with the presence of additional peaks in the ¹³C-¹H HSQC spectrum of calmodulin derived from the [1-¹³C]-glucose labeling scheme (Supplementary Fig. S6). These additional peaks arise from partial ¹³C labeling (~8 - 15%) at the Cε of the aromatic aromatic ring and are a consequence of the scrambling of ¹³C label when using glucose as the carbon precursor. These considerations provide a plausible scenario where the relaxation profiles still fit reasonably to single exponential decays but fail to be fit reasonably by the L-S model-free formalism. Thus the [4-¹³C]-erythrose labeling strategy described here would seem to be highly advantageous in providing high fidelity relaxation for the study of fast aromatic ring dynamics.

In summary, we have demonstrated a strategy to produce perdeuterated proteins with isolated ¹H-¹³C pairs in the aromatic ring systems. It is shown that the aromatic side chains of calmodulin have a wide range of motion on the sub-nanosecond time scale, the observation of which has thus far been restricted to detection at natural abundance using highly concentrated and relatively small proteins.³⁷ The ability to investigate the fast dynamics of aromatic side chain will provide a highly complementary perspective to that accessed by ¹³C or ²H relaxation in methyl groups.