A NMR Experiment for Simultaneous Correlations of Valine and Leucine/Isoleucine Methyls with Carbonyl Chemical Shifts in Proteins

Abstract

A methyl-detected ‘out-and-back’ NMR experiment for obtaining simultaneous correlations of methyl resonances of Valine and Isoleucine/Leucine residues with backbone carbonyl chemical shifts, SIM-HMCM(CGCBCA)CO, is described. The developed pulse-scheme serves the purpose of convenience in recording a single data set for all ILV methyl positions instead of acquiring two separate spectra selective for Valine or Leucine/Isoleucine residues. The SIM-HMCM(CGCBCA)CO experiment can be used for ILV methyl assignments in moderately sized protein systems (up to ~100 kDa) where the backbone chemical shifts of ¹³C^α, ¹³C_β and ¹³CO are known from prior NMR studies and where some losses in sensitivity can be tolerated for the sake of an overall reduction in NMR acquisition time.

Selective ¹³CH₃ labeling of Ile^δ1, Leu^δ and Val^γ (ILV) methyl positions on a deuterated background (Goto et al. 1999; Ruschak and Kay 2010; Sheppard et al. 2010; Tugarinov et al. 2006; Tugarinov and Kay 2004) in combination with Methyl-TROSY (Amero et al. 2009; Ollerenshaw et al. 2003; Tugarinov et al. 2003) have had a profound impact on NMR studies of structure and dynamics in high-molecular-weight proteins (Gelis et al. 2007; Hamel and Dahlquist 2005; Religa et al. 2010; Rosenzweig et al. 2013; Ruschak et al. 2010; Sprangers and Kay 2007; Sprangers et al. 2007; Tugarinov et al. 2004; Velyvis et al. 2007). The success of NMR studies that make use of the ILV labeling methodology is predicated upon the availability of methyl resonance assignments. While a variety of approaches have been proposed for assignment of methyl resonances in very large (>> 100-kDa) proteins and protein complexes over the past decade (John et al. 2007; Sprangers and Kay 2007; Sprangers et al. 2007; Velyvis et al. 2009; Venditti et al. 2011; Xu et al. 2009), methyl-detected ‘out-and-back’ NMR experiments that correlate the (¹H, ¹³C) chemical shifts of ILV methyl groups with the backbone and ¹³C_β carbon positions (Sheppard et al. 2009; Sheppard et al. 2009; Tugarinov et al. 2004; Tugarinov and Kay 2003), provide the most sensitive and robust methodology for unambiguous methyl assignments in medium-sized (< ~100 kDa) protein systems without recourse to site-directed mutagenesis and/or analysis of NOE data. These experiments benefit from isotope labeling strategies that ‘linearize’ the ¹³C spin-systems of Leu and Val side-chains via the use of α-keto-acid (Tugarinov et al. 2006) or aceto-lactate (Gans et al. 2010) biosynthetic precursors that have only one of their methyl positions labeled with ¹³CH₃, while the other remains at natural abundance and deuterated (¹²CD₃). The ‘linearization’ of carbon spin-systems of (branched) LV side-chains has been shown to simplify NMR spectra and provide significant sensitivity gains in methyl resonance assignment experiments (Tugarinov and Kay 2003).

Commonly, the HMCM[CG]CBCA experiment that correlates methyl chemical shifts with the chemical shifts of ¹³C_α/¹³C_β nuclei (Tugarinov and Kay 2003) is sufficient for reliable assignments of ILV methyls in proteins with a moderate content of ILV methyl groups. The degeneracy of ¹³C_α/¹³C_β pairs of chemical shifts that inevitably occurs for some residues, even in medium-sized proteins, can be resolved by recording spectra that correlate ILV methyl resonances with carbonyl chemical shifts: one experiment selective for Ile and Leu, ILE/LEU-HMCM(CGCBCA)CO, and a separate data set for Val, VAL-HMCM(CBCA)CO (Tugarinov and Kay 2003). Here, we describe a three-dimensional methyl-detected ‘out-and-back’ NMR experiment developed for simultaneously correlating the methyl resonances of Val and Ile/Leu residues with the backbone carbonyl chemical shifts, SIM-HMCM(CGCBCA)CO. The pulse scheme provides the convenience of acquiring only a single data set for all ILV methyl positions instead of two separate spectra for Val and Ile/Leu residues. The SIM-HMCM(CGCBCA)CO experiment can be used for ILV methyl assignments in medium-sized proteins (up to ~100 kDa) where the backbone chemical shifts of ¹³C^α, ¹³C_β and ¹³CO are known from prior NMR studies and where some losses in sensitivity can be tolerated for the sake of overall reduction in NMR acquisition time. We primarily address the problem of how carbon magnetization in homonuclear ¹³C spin-systems of different topologies (as in Val versus Ile/Leu) can be manipulated to obtain the desired correlations for all ILV positions with minimal losses in sensitivity. As a corollary result, we report the assignments of ILV methyl resonances for the 70-kDa homo-dimeric C-terminal domain of E. coli Enzyme I (EIC) (Venditti et al. 2012) obtained via combined analysis of methyl-detected ‘out-and-back’ HMCM[CG]CBCA and SIM-HMCM(CGCBCA)CO data sets.

Despite the fact that Val and Ile/Leu side-chains have different topologies, careful manipulation of magnetization via application of selective pulses allows one to ‘store’ the magnetization terms of (shorter) Val side-chains to ensure the subsequent ‘read-out’ of carbonyl frequencies of Val in synchrony with that of Ile/Leu residues. In the following, we briefly describe how this can be achieved in practice.

Figure 1 shows the SIM-HMCM(CGCBCA)CO pulse-scheme that allows methyl-carbonyl correlations in Val and Ile/Leu residues to be obtained simultaneously. The pulse-scheme in Figure 1 utilizes the Methyl-TROSY principle (Tugarinov et al. 2003) where possible by keeping the methyl magnetization in a ¹H-¹³C multiple-quantum state as long as it resides on methyl groups (i.e. before the ¹³C_ϕ2/H_y pair of pulses and after the ¹³C_ϕ7/H_y pair of pulses). At time-point a of the scheme, the generated magnetization terms are of the form $4H_z^m{C}_x^{\beta }{C}_z^{\alpha }$ for Val and $4H_z^m{C}_x^{\gamma }{C}_z^{\beta }$ for Ile/Leu residues, where $C_r^i\text{and}{H}_r^m=H_r^1+H_r^2+H_r^3$ are product operators for carbon nucleus i and the three methyl proton nuclei, respectively, r = x, y, z, and the signs of the terms are arbitrary. Fortuitously, the region of Val ¹³C^α chemical shifts is well separated from the rest of the carbon chemical shifts in the terms above. Therefore, a selective 90° pulse can be chosen that excites all carbon nuclei in these terms except for ¹³C^α of Val. Application of a universal Gaussian Cascade (Q5) pulse (Emsley and Bodenhausen 1987) with phase ‘y’ centered at 33 ppm with an excitation bandwidth of ± 12 ppm (600 MHz) converts the terms above to $4H_z^m{C}_z^{\beta }{C}_z^{\alpha }$ for Val and $4H_z^m{C}_z^{\gamma }{C}_x^{\beta }$ for Ile/Leu (time point b; Figure 1). Thus, the magnetization of Val residues is effectively ‘stored’ in a triple-spin-order state that does not evolve due to ¹³C-¹³C couplings during the following 4T_C period. Subsequently, the evolution of magnetization proceeds via the following pathways:

$$4H_z^m{C}_z^{\beta }{C}_z^{\alpha }(b)\stackrel{4T_C,{90}_{C,\varphi 2}^{°}}{\to }4H_z^m{C}_x^{\beta }{C}_x^{\alpha }(c)\stackrel{4{\tau }_d,{90}_{C,y}^{°}{90}_{CO,\varphi 4}^{°}}{\to }8H_z^m{C}_z^{\beta }{C}_y^{\alpha }{C}_y^{\prime }(d)$$ (1)

for Val and,

$$4H_z^m{C}_z^{\gamma }{C}_x^{\beta }(b)\stackrel{4T_C,{90}_{C,\varphi 2}^{°}}{\to }4H_z^m{C}_z^{\beta }{C}_x^{\alpha }(c)\stackrel{4{\tau }_d,{90}_{C,y}^{°},{90}_{CO,\varphi 4}^{°}}{\to }4H_z^m{C}_z^{\alpha }{C}_y^{\prime }(d)$$ (2)

for Ile/Leu, where the letters in parentheses correspond to the time points in the pulse scheme (Figure 1) and the signs of the terms are neglected. The ¹³C^β-¹³C^γ2 couplings in Ile side-chains are refocused by application of a pair of ‘Iγ2’ pulses selective for the Ile ¹³C^γ2 methyl region between time-points b and c, while evolution due to ¹³C^β-¹³C^γ couplings in Val between time-points c and d is eliminated by application of a pair of ‘Vγ’ pulses selective for the Val ¹³C methyl region. The terms on the right-hand side of pathways (1) and (2) evolve during the t₁ period, with the chemical shifts of ¹³C^α nuclei of Val refocused by the ¹³C 180° pulse ‘α’ applied in the middle of t₁ and selective for the ¹³C^α chemical shifts of all ILV residues (Figure 1). The selectivity of this pulse also ensures that the ¹³C^α-¹³C^β J couplings in Val are refocused during the t₁ evolution period. After the terms generated at time point d of pathways (1) and (2) are labeled with ¹³CO chemical shifts, the magnetization is transferred back to methyl carbons that evolve during the time t₂, and subsequently to methyl protons for detection by reversing the pathways.

Figure 1.

The SIM-HMCM(CGCBCA)CO pulse-scheme for simultaneous methyl-carbonyl correlations in Val and Ile/Leu residues. All narrow and wide rectangular pulses are applied with flip angles of 90° and 180°, respectively. All the pulses (including shaped pulses) are applied along the x-axis unless indicated otherwise. The ¹H, ²H and ¹⁵N carrier positions were set to 4.7, 2.5 and 119 ppm, respectively. The ¹³C carrier is positioned at 18 ppm at the beginning of the scheme, switched to 33 ppm before the ¹³C pulse with phase ϕ1, and returned to 18 ppm after the ¹³C pulse with phase ϕ8. All the pulses shown with black rectangles are applied with the highest possible power except for the ~1.4-ms long water-selective rectangular pulses flanking the last 180 ¹H pulse and the water-selective ~7.0-ms ¹H E-BURP1 pulse (Geen and Freeman 1991) shown with the open arc. ¹³C WALTZ-16 decoupling (Shaka et al. 1983) during acquisition is achieved using a 2.5 kHz field, while ²H WALTZ-16 decoupling uses a 0.8 kHz field. The ¹³C shaped pulses marked with asterisks are high-power 350-µs (600 MHz) RE-BURP pulses (Geen and Freeman 1991) applied on-resonance. The ¹³C shaped pulses marked with daggers are high-power 350-µs RE-BURP pulses applied at 176 ppm by phase modulation of the carrier (Boyd and Soffe 1989; Patt 1992). The ¹³C pulses marked with ‘Iγ2’ are 4.0 ms RE-BURP pulses selective for Ile ¹³C^γ2 positions and are applied at 16 ppm by phase-modulation of the carrier. The ¹³C pulses marked with ‘Vγ’ are 3.0 ms RE-BURP pulses selective for ¹³C^γ methyls of valine residues and are applied at 21 ppm by phase modulation of the carrier. The ¹³C pulse labeled ‘α’ is a 1.8 ms RE-BURP pulse (600 MHz) selective for ¹³C^γ positions of ILV residues and applied at 60 ppm by phase modulation of the carrier. The ¹³C pulses labeled ‘Q5’ are 1.4 ms Q5 Gaussian Cascade pulses (Emsley and Bodenhausen 1987) applied on-resonance (see text for details). ¹³C and ¹³CO 90° pulses shown with open rectangles are applied at 45 and 176 ppm, respectively, by phase modulation of the carrier with a field strength of Δ/√15 where Δ is the difference (in Hz) between ¹³C_α/β (45 ppm) and ¹³CO (176 ppm) chemical shifts (Kay et al. 1990). The 180° ¹³CO pulse following the t₁ period shown with the shaded rectangle is applied at 176 ppm by phase modulation of the carrier with the same field strength as the 90° ‘open’ pulses. This ensures complete refocusing of ‘Bloch-Siegert’ shifts of both ¹³C_α and ¹³CO nuclei during t₁. Vertical arrows at the start and end of the 4τ_d periods and after the t₁ period indicate the position of Bloch-Siegert compensation pulses (Kay et al. 1990). Delays are: τ_a = 2.0 ms; T_C = 3.5 ms; τ_d = 2.6 ms; the delay ε is adjusted to accommodate the shaped Bloch-Siegert compensation pulse ‘α’. The phase-cycle is: ϕ1 = x,-x; 2 = 2(y),2(−y); ϕ3 = x; ϕ4 = 2(x),2(−x); ϕ5 = 4(x),4(−x); ϕ6 = 4(y),4(−y); ϕ7 = 8(y),8(−y); ϕ8 = x; receiver = x,-x,-x,x. Note that if the phase of one of the ¹³C_ϕ3 pulses is shifted by 90° (±y instead of x), the signs of Ile/Leu correlations will be inverted while retaining the signs of Val correlations. Quadrature detection in F₁ snd F₂ is achieved by States-TPPI (Marion et al. 1989) of ϕ4 and ϕ8, respectively. Durations and strengths of the pulsed-field gradients in units of (ms; G/cm) are: g1 = (1.0; 15), g2 = (0.3; 8), g3 = (0.45; 10).

Figure 2 shows plots of the magnetization transfer function for Val methyl (green curve) and Ile/Leu methyl (red curve) (¹H,¹³C)-¹³CO correlations as a function of the delay 4τ_d in the SIM-HMCM( CGCBCA)CO pulse scheme. These plots take into account the fact that Val signals relax faster during the 4τ_d period than Ile/Leu signals as Val magnetization is present in a ¹³C-¹³C multiple-quantum state, with both ¹³C_α and ¹³C_β nuclei relaxing in the transverse plane. This is also true for Val residues with respect to the t₁ evolution period where the multiple-quantum ¹³C_α-¹³CO state relaxes. The duration of the τ_d delays should be chosen to ensure the desired compromise between the intensities of Val and Ile/Leu correlations (τ_d = 2.6 ms in this work; Figure 2), and the time t_1,max should be adjusted with faster relaxation of Val correlations in mind (t_1,max ≈ 15 ms at 600 MHz allows the observation of all Val correlations for EIC). The terms $8H_z^m{C}_x^{\beta }{C}_y^{\alpha }{C}_y^{\prime }$ in Ile/Leu residues are omitted from the end of pathway (2) above but are not filtered out before the t₁ evolution period. These terms (blue curve in Figure 2) are, however, much lower in intensity for the duration of τ_d chosen here, providing correlations are observed for only a few very intense peaks in the 3D spectra of EIC at the frequencies Ω_CO ± Ω_Cβ, where Ω_i is the offset of nucleus i from the carbon carrier. We note that the same transfer function as shown for the $8H_z^m{C}_x^{\beta }{C}_y^{\alpha }{C}_y^{\prime }$ terms of Ile/Leu in Figure 2 also applies for the correlations of Val residues in the ILE/LEU-HMCM( CGCBCA)CO data sets optimized for observation of Ile, Leu methyl-(¹H,¹³C)-¹³CO correlations. That is why the correlations of Val residues are either not observed or observed with low intensities in the ILE/LEU-HMCM(CGCBCA)CO data sets.

Figure 2.

Plots comparing the magnetization transfer functions for Valine methyl (¹H,¹³C)-¹³CO correlations (green curve; [sin(4πτ_d¹J_Cα-CO)exp(−4R_Cατ_d)exp(−4R_Cβτ_d)]², where ¹J_Cα-CO is the one bond ¹³C_α-¹³CO J coupling (55 Hz), and R_Cα and R_Cβ are the relaxation rates of ¹³C_α and ¹³C_β nuclei, respectively, both assumed to be equal to 30 s⁻¹ for deuterated EIC), and Ile/Leu methyl (¹H,¹³C)-¹³CO correlations (red and blue curves corresponding to [sin(4πτ_d¹J_C1-CO)sin(4πτ_d¹J_Cα-Cβ)exp(™4R_Cατ_d)]² and [sin(4πτ_d¹J_Cα-CO)cos(4πτ_d¹J_Cα-Cβ)exp(−4R_Cατ_d)]², respectively, where ¹J_Cα-Cβ is the one bond ¹³C_α-¹³C_β J coupling (35 Hz) and the rest of the notation defined above) as a 18 function of the delay 4τ_d (ms) in the SIM-HMCM(CGCBCA)CO experiment of Figure 1. Methyl (¹H,¹³C)-¹³CO correlations arise from the magnetization terms indicated on the plot created at time point d and evolving during the t₁ evolution period of the scheme in Figure 1.

Analysis of magnetization flow for the pulse-scheme in Figure 1 shows that shifting the phase of one of the ¹³C 180° pulses with phase ϕ3 (¹³C_ϕ3; in the time interval between points b and c in Figure 1) by 90° inverts the sign of the correlations belonging to Ile/Leu residues while preserving the sign of Val correlations. Although this may seem as an especially attractive feature of the SIM-HMCM( CGCBCA)CO experiment, in practice, Val residues are easily distinguished from Ile/Leu by the relative signs of the peaks obtained in the 3D HMCM[CG]CBCA data sets (Tugarinov and Kay 2003). Nevertheless, in the regions of the 3D HMCM[CG]CBCA spectra where Val correlations may overlap with Leu peaks (an unlikely event due to significantly different ¹³C_α and ¹³C_β chemical shifts for Val and Leu), it may prove useful to be able to differentiate between the two types of residues from the SIM-HMCM(CGCBCA)CO data.

The SIM-HMCM(CGCBCA)CO experiment was tested on the C-terminal domain of Enzyme I (EIC) from E. coli, a symmetric 70-kDa homodimer. Enzyme I (PDB access code 2hwg (Teplyakov et al. 2006)) is the first enzyme in the phosphotransferase system (PTS), a signal transduction pathway that couples phosphoryl transfer to active sugar transport across the cell membrane (Meadow et al. 1990). Auto-phosphorylation of Enzyme I by phosphoenolpyruvate (PEP) in a Mg²⁺-dependent manner is followed by the transfer of the phosphoryl group to the histidine phosphocarrier protein HPr (Weigel et al. 1982; Weigel et al. 1982). Recently, robust protocols for expression and purification of EIC, as well as the backbone assignments and conformational transitions occurring in EIC upon PEP binding have been reported by our laboratory (Venditti and Clore 2012; Venditti et al. 2012). NMR analysis and earlier crystallographic data show that the isolated C-terminal domain of enzyme I adopts a stable tertiary fold and exists in a monomer/dimer equilibrium that is modulated by ligand binding (Chauvin et al. 1996; Chauvin et al. 1994; Patel et al. 2006; Seok et al. 1996; Seok et al. 1998; Venditti and Clore 2012)). The symmetric homo-dimer of EIC is shown in Figure 3a. Under the sample conditions used in this study (0.5 mM protein and 4 mM Mg²⁺ concentration), ~95% of the protein is present in solution in the dimeric state (Venditti and Clore 2012). The Ile and Leu/Val regions of the high-resolution ¹H-¹³C methyl constant time (CT)-HMQC spectrum (Bax et al. 1983; Mueller 1979; Tugarinov et al. 2003) of [U-²H,¹⁵N,¹³C; Ile^τ1-{¹³CH₃}; Leu,Val-{¹³CH₃/¹²CD₃}]-labeled EIC are shown in Figures 3b–c. Among 27 Ile τ1, 56 Leu τ and 36 Val γ methyl groups present in the EIC monomer, all Ile, 52 Leu and 26 Val methyls have been unambiguously assigned via combined analysis of 3D HMCM[CG]CBCA and 3D SIM-HMCM(CGCBCA)CO data sets. The backbone chemical shifts of the remaining 5 Val methyls (Val²⁷⁸, Val²⁸¹, Val⁴²⁸, Val⁴³⁰ and Val⁴⁴⁶) were not assigned in the previous NMR study of EIC (Venditti and Clore 2012). A pair of resonances with outstanding ¹³C and/or ¹H chemical shifts from this subset of methyls (Val⁴³⁰ and Val⁴⁴⁶) could be assigned tentatively based on the magnitude of the ring-current effects (Haigh and Mallion 1979; Koradi et al. 1996; Kuszewski et al. 1995) calculated from the x-ray coordinates (Teplyakov et al. 2006).

Figure 3.

(a) Ribbon diagram representation of the structure of the biological assembly of the C-terminal domain of Enzyme I (EIC dimer; pdb code 2hwg (Teplyakov et al. 2006)). The molecular dyad axis of the homodimer is shown with a dashed vertical line. (b–c) Regions of the ¹H-¹³C methyl CT-HMQC correlation map of [U-²H,¹⁵N,¹³C; Ile^δ1-{¹³CH₃}; Leu,Val-{¹³CH₃/¹²CD₃}]-labeled EIC (90% H₂O/ 10% D₂O; 600 MHz; 37 °C) featuring the correlations of (b) Ile^δ1 and (c) Val^γ/Leu^δ methyl groups. The high-resolution CT-HMQC correlation spectrum was acquired with the constant time period adjusted to 2/¹J_CC = 56 ms. Selected residues are labeled with methyl peak residue types and numbers. The methyl-containing residues are numbered in the context of full-length EI. The assignments of Val⁴³⁰ and Val⁴⁴⁶ are tentative. (d–f) Superposition of a selected region of the 2D ¹H-¹³CO planes plotted from the ¹H-¹³C chemical shift area highlighted in the Leu-Val region of the methyl ¹H-¹³C correlation map in panel (c) from (d) 3D VAL-HMCM( CBCA)CO, (e) 3D ILE/LEU-HMCM(CGCBCA)CO and (f) 3D SIM-HMCM( CGCBCA)CO data sets. The sample comprised 0.5 mM EIC dissolved in 20 mM Tris buffer, pH = 7.4, 100 mM NaCl, 2 mM DTT, 4 mM MgCl₂, and 1 mM EDTA (Venditti and Clore 2012). All experiments were performed on a Bruker Avance III 600 MHz spectrometer equipped with a cryogenic probe operating at 37 °C. Typically, the 3D data sets were recorded with 18*, 40* and 512* complex points in the ¹³CO, ¹³C and ¹H dimensions, respectively, corresponding to respective acquisition times of 15, 14, and 64 ms. The recovery delay of 1.4 s and 32 scans per fid led to total acquisition times of ~40 hrs per 3D experiment. NMR spectra were processed with NMRPipe/NMRDraw software (Delaglio et al. 1995) and analyzed with NMRView (Johnson and Blevins 1994) using a tcl/tk interface written in-house.

Figures 3d–f show the superposition of a selected region of the 2D ¹H-¹³CO planes from the 3D VAL-HMCM(CBCA)CO (Figure 3d), ILE/LEU-HMCM(CGCBCA)CO (Figure 3e) and SIM-HMCM(CGCBCA)CO (Figure 3f) spectra corresponding to the area of ¹H-¹³C chemical shifts highlighted in the Leu-Val region of the methyl ¹H-¹³C correlation map in Figure 3c. Clearly, the cross-peaks in the Val-selective VAL-HMCM(CBCA)CO (Figure 3d) and Ile/Leu-selective ILE/LEU-HMCM(CGCBCA)CO (Figure 3e) experiments represent only subsets of the correlations obtained in the SIM-HMCM(CGCBCA)CO data set. The convenience of recording a single data set comes at a relatively low price: we estimate that the sensitivity of the SIM-HMCM( CGCBCA)CO data set is 25% lower on average for Ile/Leu correlations and 50% lower for Val correlations compared to the separate (Ile/Leu- and Val-selective) experiments recorded for the same measurement time. These losses in signal-to-noise per unit time in the simultaneous data set are associated with (i) faster relaxation of the multiple-quantum magnetization terms $8H_z^m{C}_x^{\beta }{C}_{tr}^{\alpha }{C}_{tr}^{\prime }$ present during the t₁ evolution period for Val residues, (ii) the non-idealities as well as relaxation and phase evolution during the long Q5 Gaussian Cascade pulses (Emsley and Bodenhausen 1987) for all residues, and (iii) the need to adjust the length of the delay 4τ_d to achieve the desired compromise in sensitivity of Val versus Ile/Leu correlations. We note that even the loss of about half of the signal for Val residues in the SIM-HMCM(CGCBCA)CO data set is tolerable as the Val-selective VAL-HMCM(CBCA)CO spectra are usually much more sensitive than their Ile,Leu-selective counterparts (4.6-fold more sensitive in the 82-kDa Malate Synthase G at the same temperature (Tugarinov and Kay 2003)) and can therefore be recorded with shorter total acquisition times.

Although a single HMCM[CG]CBCA dataset is usually sufficient for assignments of the vast majority of methyl sites, and correlations of methyl resonances with backbone carbonyls are obtained primarily for verification purposes, even in such medium-sized systems as the EIC dimer, some cases can be identified when the correlations with carbonyl chemical shifts are absolutely indispensable for unambiguous assignments. For example, Val⁴⁸¹ and Val⁵⁶⁵ of the EIC dimer have practically identical ¹³C^α and ¹³C^β chemical shifts (31.2 and 66.4 ppm, respectively). Figures 4a–c show selected regions of the 2D ¹H-¹³C planes plotted from the 3D HMCM[CG]CBCA spectrum of the [U-²H,¹⁵N,¹³C; Ile^δ1-{¹³CH₃}; Leu,Val-{¹³CH₃/¹²CD₃}]-EIC at the methyl ¹³C chemical shifts of these two residues. The peaks of one of the two methyl sites of these two valines overlap in the ¹H-¹³C CT-HMQC map (Figure 4c). The correlations provided by the SIM-HMCM(CGCBCA)CO data set is the sole means of distinguishing between the two methyl positions, as the chemical shifts of backbone ¹³CO nuclei of this pair of valines happen to be different by 1 ppm, as can be seen in the selected region of the 2D ¹H-¹³CO planes from the SIM-HMCM(CGCBCA)CO data set plotted at the methyl ¹³C chemical shifts of Val⁴⁸¹ and Val⁵⁶⁵ shown in Figures 4d–f.

Figure 4.

(a–c) Selected regions of the 2D ¹H-¹³C planes drawn from the 3D HMCM(CG)CBCA spectrum of [U-²H,¹⁵N,¹³C; Ile^δ1-{¹³CH³}; Leu,Val-{¹³CH₃/¹²CD₃}]-EIC (90% H₂O/ 10% D₂O; 600 MHz; 37 °C) at the methyl ¹³C chemical shifts of (a) one of the two Val⁴⁸¹ methyls of EIC, (b) one of the two Val⁵⁶⁵ methyls of EIC, and (c) the position where the methyl peaks of Val⁴⁸¹ and Val⁵⁶⁵ overlap. The insets in (a–c) show the regions of the 2D ¹H-¹³C CT-HMQC correlation map of EIC centered at the methyl positions in question. Negative contours are shown in red. ¹³C chemical shift positions of ¹H-¹³C_α and ¹H-¹³C_β correlations are indicated on the right side of each plot. (d–f) Selected regions of the 2D ¹H-¹³CO planes from the SIM-HMCM(CGCBCA)CO data set plotted at the methyl ¹³C chemical shifts of (d) one of Val⁴⁸¹ methyls, (e) one of Val⁵⁶⁵ methyls, and (f) the region where the methyl peak of Val⁴⁸¹ overlaps with that of Val⁵⁶⁵.

Methyl assignments based on HMCM[CG]CBCA data sets are complicated by the fact that ¹³C^δ (and to a smaller extent ¹³C^α) chemical shifts of ILV residues in ¹³CH₃-labeled protein samples are different from those of fully deuterated proteins by three times the value of the multiple-bond deuterium isotope shifts (Gardner and Kay 1998; Rosen et al. 1996; Venters et al. 1996). Usually, however, ¹³C^β and ¹³C^α chemical shifts are available from the assignment experiments acquired on fully deuterated protein samples as was the case in the previous study of EIC (Venditti and Clore 2012). For example, deuterium isotope shifts of ¹³C^β nuclei arising from the substitution of three methyl deuterons for protons, nΔC^β (D_m) where n is the number of bonds between the carbon β and the site of the substitution (n = 2 for Val and n = 3 for Ile/Leu), are measured to be 250 ± 50 ppb for Val and 80 ± 30 ppb for Ile/Leu residues in EIC from a comparison of shifts in the 3D TROSY-HNCACB spectra acquired on the ¹³CH₃-labeled and fully deuterated protein samples. The variability of these isotope effects may compromise the exact ‘matching’ of ¹³C^β and ¹³C^α frequencies in the HMCM[CG]CBCA and HNCACB data sets if recorded on different samples. Moreover, these deuterium isotope effects on ¹³C^β nuclei may be somewhat offset by three-bond isotope shifts resulting from backbone amide ¹H_N-to-D_N substitutions, ³ΔC^β (ND) (~37 ppb on average ranging from 2 to 65 ppb in model proteins (Zhang and Tugarinov 2013)), when the HMCM[CG]CBCA spectra are recorded in D₂O. The assignments based on correlations with carbonyl shifts, however, do not suffer from these isotope effects as (i) carbonyl carbons are removed by 4 (in Val) and 5 (in Ile/Leu) bonds from the sites of the methyl ¹H-to-D replacement, and (ii) generally, deuterium isotope effects of carbonyl nuclei tend to be much smaller (Garrett et al. 1997; Sun and Tugarinov 2012; Zhang and Tugarinov 2013). In this context, we note that ¹³CO chemical shifts in the SIM-HMCM(CGCBCA)CO experiments recorded in H₂O and D₂O solvents are different by the sum of two- and three-bond deuterium isotope shifts arising from the ¹H-to-D substitutions at backbone amide positions (²ΔCO(N_i+1D) + ³ΔCO(N_iD); equal on average to 102 ± 25 ppb in EIC in good agreement with the measurements reported earlier in model protein systems (Feeney et al. 1974; Kainosho et al. 1987; LiWang and Bax 1996; Tuchsen and Hansen 1991; Zhang and Tugarinov 2013)).

In summary, we describe a methyl-detected ‘out-and-back’ NMR experiment for obtaining simultaneous correlations of methyl resonances of Val and Ile/Leu residues with backbone carbonyl chemical shifts. The developed pulse-scheme serves the purpose of convenience in recording a single data set for all ILV methyl positions instead of acquiring two separate spectra for Val and Ile/Leu correlations. The SIM-HMCM(CGCBCA)CO experiment can be used for ILV methyl assignments in moderately sized protein systems (up to ~100 kDa) where the backbone chemical shifts of ¹³C^α, ¹³C_β and ¹³CO are known from prior NMR studies and where some losses in sensitivity can be tolerated for the sake of reduction in net NMR acquisition time. We note that though Val and Leu side-chains have been non-stereospecifically labeled in EIC via the use of [¹³CH₃/¹²CD₃]-labeled α-keto-isovalerate (Tugarinov et al. 2006; Tugarinov and Kay 2004), the developed methodology is equally applicable to stereospecifically [¹³CH₃/¹²CD₃]-labeled Val and Leu isopropyl moieties, i.e. with ¹³CH₃ labels restricted to either pro-R or pro-S methyl positions as obtained via the use of appropriately labeled aceto-lactate precursors (Gans et al. 2010).