Journal of Biomolecular NMR

, Volume 47, Issue 3, pp 183–194

MQ-HNCO-TROSY for the measurement of scalar and residual dipolar couplings in larger proteins: application to a 557-residue IgFLNa16-21

Authors

  • Sampo Mäntylahti
    • NMR Laboratory, Program in Structural Biology and Biophysics, Institute of Biotechnology/NMR LaboratoryUniversity of Helsinki
  • Outi Koskela
    • Laboratory of Organic Chemistry, Department of ChemistryUniversity of Helsinki
  • Pengju Jiang
    • Biochemistry DepartmentUniversity of Oxford
    • NMR Laboratory, Program in Structural Biology and Biophysics, Institute of Biotechnology/NMR LaboratoryUniversity of Helsinki
Article

DOI: 10.1007/s10858-010-9422-z

Cite this article as:
Mäntylahti, S., Koskela, O., Jiang, P. et al. J Biomol NMR (2010) 47: 183. doi:10.1007/s10858-010-9422-z

Abstract

We describe a novel pulse sequence, MQ-HNCO-TROSY, for the measurement of scalar and residual dipolar couplings between amide proton and nitrogen in larger proteins. The experiment utilizes the whole 2TN polarization transfer delay for labeling of 15N chemical shift in a constant time manner, which efficiently doubles the attainable resolution in 15N dimension with respect to the conventional HNCO-TROSY experiment. In addition, the accordion principle is employed for measuring (J + D)NHs, and the multiplet components are selected with the generalized version of the TROSY scheme introduced by Nietlispach (J Biomol NMR 31:161–166, 2005). Therefore, cross peak overlap is diminished while the time period during which the 15N spin is susceptible to fast transverse relaxation associated with the anti-TROSY transition is minimized per attainable resolution unit. The proposed MQ-HNCO-TROSY scheme was employed for measuring RDCs in high molecular weight protein IgFLNa16-21 of 557 residues, resulting in 431 experimental RDCs. Correlations between experimental and back-calculated RDCs in individual domains gave relatively low Q-factors (0.19–0.39), indicative of sufficient accuracy that can be obtained with the proposed MQ-HNCO-TROSY experiment in high molecular weight proteins.

Keywords

Dipolar couplingsHNCONMRProteinsScalar couplingsTROSY

Introduction

Introduction of tunable alignment of proteins in a dilute liquid crystal medium gives rise to residual dipolar couplings (RDCs), which have revolutionized structural characterization of biological macromolecules (Blackledge 2005; Tjandra and Bax 1997; Bouvignies et al. 2007; Prestegard et al. 2000). RDCs are direct product between gyromagnetic ratios of two nuclei, and the magnitude of the coupling is depended on both the bond vector orientation and the internuclear distance (Annila and Permi 2004). RDCs enable not only determination of macromolecular structures with higher precision but also studies of conformational changes (Pääkkönen et al. 2000), domain orientation (Fischer et al. 1999; Tugarinov and Kay 2003), dynamics and folding (Fredriksson et al. 2004; Lakomek et al. 2006; Tolman and Ruan 2006). Effective application of these orientational restraints necessitates accurately determined residual dipolar contribution (D), which can be derived from measurements of scalar (J) splittings in water and (J + D) splittings in the presence of orienting medium. The accuracy is often compromised due to additional line broadening of NMR signals induced not only by dipole–dipole (DD) relaxation, but also by non-averaged dipolar contribution (RDCs) in aligned medium, from proximate protons. For this reason, higher accuracy is typically obtained when (J + D) splitting is measured from heteronuclear dimension instead of directly detected proton dimension. One-bond RDC between the amide proton and nitrogen (1DNH) can be measured with modest effort and high accuracy in case of small to medium sized proteins. To that end, several two-dimensional experiments have been developed, which utilize the well-dispersed 15N, 1H correlation map together with the spin-state separation of up- or downfield 15N–1H multiplet components into the subspectra, which eliminates additional spectral crowding due to increased number of cross peaks in J coupled experiments (Meissner et al. 1997; Ottiger et al. 1998; Andersson et al. 1998; Weigelt 1998; Lerche et al. 1999; Permi 2002; Würtz and Permi, 2007). However, in case of larger proteins (>30–40 kDa), the accurate measurement of JNH and (J + D)NH coupling becomes much more difficult due to increasing number of cross peaks and the drastic difference in decay rates between the so-called TROSY and anti-TROSY components of the 15N–1H multiplet (Pervushin et al. 1997; Yang and Kay 1999; Kontaxis et al. 2000). Although position of the TROSY component of the 15N–1H multiplet can be measured with high accuracy, a relative position of the 15N shift associated with 1Hβ spin state is vaguer due to broader line and lower sensitivity of the anti-TROSY component. Increased spectral overlap, together with decreased sensitivity and line broadening in high molecular weight systems, makes measurement of RDCs in these proteins very challenging. To this end Kay and co-workers have proposed HNCO-TROSY based experiment for measuring (J + D)NHs, which provides enhanced cross peak dispersion when compared to two-dimensional 15N-HSQC or 15N-TROSY based experiments (Yang et al. 1999).

In this work, we introduce a modified HNCO-based pulse sequence, which offers superior resolution in 15N dimension in comparison to traditional HNCO-TROSY experiment and is especially useful for measuring RDCs between amide proton and nitrogen in high molecular weight proteins. The proposed pulse sequence was evaluated with three proteins, 56-residue GB1, 76-residue human ubiquitin, and a 557-residue (60 kDa) filamin A, composed of six consecutive (16–21) filamin-type immunoglobulin-like (IgFLN) domains.

Results and discussion

Proposed pulse sequence for the measurement of (J + D)HN couplings in 15N, 13C (2H) labeled proteins is depicted in Fig. 1. The experiment mostly resembles well-described HNCO-TROSY experiments i.e. it establishes correlations between 1HN, 15N spins, and 13C′ nucleus of the preceding residue (Salzmann et al. 1998; Yang and Kay 1999). However, some of the common features of the original scheme have been modified for the purpose of measuring RDCs in larger proteins. These changes will be briefly described in the following:
https://static-content.springer.com/image/art%3A10.1007%2Fs10858-010-9422-z/MediaObjects/10858_2010_9422_Fig1_HTML.gif
Fig. 1

a MQ-HNCO-TROSY and b modified HR-TROSY HNCO (Hu et al. 2009) pulse sequences for measuring 1(J + D)NH. Narrow and wide bars correspond to rectangular 90° and 180° pulses, with phase x unless otherwise indicated, respectively. 90° (180°) pulses for 13C′ are applied with a strength of Ω/√15 (Ω/√3), where Ω is the frequency difference between the centers of the 13C′ and the aliphatic 13Cα regions. The 1H, 15N, and 13C′ carrier positions are 4.7 (H2O), 120 (center of 15N spectral region), and 175 ppm (center of 13C′ spectral region), respectively. The semi-selective decoupling for removal of 13C′–13Cα and 15N–13Cα coupling interactions during t1 and t2, respectively, can be accomplished using either the SEDUCE-1 decoupling sequence (McCoy and Mueller 1992) or three 180° 13C′ rectangular or one-lobe sinc pulses (indicated as half ellipsoids) applied off-resonance with phase modulation by Ω. Pulsed field gradients are inserted as indicated for coherence transfer pathway selection and residual water suppression. The delays employed are: Δ = 1/(4JNH); 2TN = 1/(2JNC′) = 24 − 34 ms; ε = gradient+field recovery delay. The TROSY component is recorded by omitting all 180°(1H) pulses between time points a and d with the following phase cycling: ϕ1 = x, −x; ϕ2 = x; ϕ3 = x; ϕ4 = y; ϕ5 = x; ϕ6 = x; ψ = x; ϕrec = x, −x. The pure anti-TROSY component along 15N dimension can be selected by changing the phases ϕ3 and ϕ6 to −x. If desired, the anti-TROSY component instead of TROSY component in 1H dimension can be selected by inverting the phase of ϕ4. The downscaled anti-TROSY component is obtained by recording the experiment (a) with the 180°(1H) pulses given at points A, and using the phase cycling: ϕ1 = x, −x; ϕ2 = x, ϕ3 = x; ϕ4 = − y; ϕ5 = − x; ϕ6 = − x; ψ = − x; ϕrec = x, −x. If the up-scaled anti-TROSY component is measured using the modified HR-TROSY HNCO (Hu et al. 2009) scheme (b), the phase cycling is ϕ1 = x, −x; ϕ2 = x; ϕ3 = − x; ϕ4 = y; ϕ5 = x; ϕ6 = x; ψ = x; ϕrec = x, −x. In the MQ-HNCO-TROSY scheme (a), depending on setting of 0 ≤ κ ≤ 1, the position of the downscaled anti-TROSY component varies between ωN and ωN + πJNH (inseta′), whereas in the HR-TROSY HNCO scheme (b) with κ ≥ 0, the upscaled anti-TROSY component floats between ωN – πJNH and ωN – (κ + 1)πJNH (insetb′). Hence, the measured (J+D)NHs are scaled by (1 + κ)/2 and κ/2 in the MQ-HNCO-TROSY and HR-TROSY HNCO spectra, respectively. Quadrature detection in the 13C′ dimension is obtained by States-TPPI protocol applied to ϕ1 (Marion et al. 1989). For quadrature detection in the indirect 15N dimension, the 90°(15N) with the phase ψ is inverted simultaneously with the gradient GN to obtain echo/antiecho selection. The data processing is according to the sensitivity enhanced method (Kay et al. 1992). In addition to echo/antiecho coherence transfer pathway selection, the axial peaks are shifted to the edge of the spectrum by inverting ϕ2 together with ϕrec in every second t2 increment

The experiment starts with the 1H–15N INEPT, which transfers magnetization from 1H to 15N (time point a). This is followed by the delay 2TN during which the antiphase coherence between 15N(i) and 13C′(i – 1) is established. At time point b, 90°(13C′) pulse converts magnetization into the 15N–13C′ multiple-quantum (MQ) coherence. During the ensuing t1 period, the 13C′ chemical shift is allowed to evolve, whereas 180°(15N) pulse at the midpoint of t1 period refocuses 15N chemical shift evolution. At time point c, 90°(13C′) pulse transforms magnetization back to the 15N single-quantum (SQ) coherence, which evolves under the 15N chemical shift Hamiltonian during the t2 period. After chemical shift labeling of 15N, the desired 15N SQ coherence is transferred to amide proton for detection, using the gradient selected and sensitivity enhanced TROSY block recently introduced by Nietlispach (2005), which suppresses the anti-TROSY pathway that plagued the original scheme proposed by Yang and Kay (1999). After Fourier transformation of time domain signal, this results in a familiar HNCO-TROSY spectrum with cross peaks emerging at ωC′, ωN − πJNH, and ωHN + πJNH frequencies. Hence, we dub this experiment as MQ-HNCO-TROSY.

In comparison to the conventional HNCO-TROSY scheme, the MQ-HNCO-TROSY experiment utilizes the whole 2TN period instead of TN period for the 15N chemical shift labeling, thus enabling use of two-fold higher experimental resolution in the 15N dimension, which is absolutely critical in case of high molecular weight proteins with hundreds of residues. Multiple-quantum 15N–13C coherence transfer scheme has earlier been proposed by Wüthrich and co-workers (Salzmann et al. 1998) and by Clore and colleagues (Hu et al. 2009) but we prefer using the full sweep implementation (Madsen et al. 1993; Puttonen et al. 2006) which yields better sensitivity, resolution and precision when measuring RDCs.

In this work, we have employed a generalized version of the TROSY scheme by Nietlispach (2005) i.e. the proposed pulse sequence utilizes a spin-state selective filter that can be used to select any of the four possible 15N–1H multiplet components by changing the phase of the first 1H(ϕ3) and the last 15N(ϕ4) pulse in the TROSY filter (see legend to Fig. 1 for details). This was motivated by earlier studies by us and others to measure RDCs in directly detected 1H dimension (Lerche et al. 1999; Permi 2002; Würtz and Permi 2007). Indeed, measuring 15N–1H couplings from 1H dimension either between TROSY and anti-TROSY lines or between the TROSY and decoupled (HSQC) component might be an attractive possibility, especially on large proteins and highly perdeuterated samples. However, initial testing with large 60 kDa protein in a dilute liquid crystal medium at 800 MHz revealed that, as earlier reported by Bax and co-workers (Kontaxis et al. 2000), measuring the 15N–1H splitting in 15N dimension is less susceptible to additional line broadening due to unresolved 1H–1H dipolar couplings even on highly perdeuterated sample. We therefore decided to measure 15N–1H RDCs in the indirectly detected 15N dimension.

In larger proteins with hundreds of residues, precision of measured 15N–1H splitting is not solely hampered by the effective linewidth and signal-to-noise ratio (S/N) of the fast decaying 15N magnetization associated with the 1Hβ spin state, the so-called anti-TROSY component, but also by the level of cross peak overlap. Therefore, in case of measurement of (J + D)NHs in larger proteins, our aim has been in development of a pulse sequence, which offers a high resolution without sacrificing the sensitivity of experiment. To this end, we utilized an approach based on the accordion spectroscopy (Bodenhausen and Ernst 1981) to determine cross peak position of the upfield 15N-{1Hβ} component. The approach is similar to the implementations proposed by Yang et al. (1999) and Kontaxis et al. (2000), and has been successfully utilized for measuring small 1DNCα/2DNCα (Permi et al. 2000a; Puttonen et al. 2006) and 3JHNHα couplings (Heikkinen et al. 1999). In this method, additional variable delay κt2 is incorporated into the pulse sequence during which the coupling interaction between 1H and 15N is active. In practice, two 180°(1H) pulses (shown as unfilled rectangular bars in Fig. 1a) are given at the points indicated by A. The κ = 0 is a special case; the first 180°(1H) pulse does not have any influence on evolution of (J + D)NH Hamiltonian during the experiment. The second 180°(1H) pulse, applied together with the 180°(15N) pulse, decouples any 1H–15N scalar or dipolar coupling interaction during the t2 period. Afterwards, the TROSY element selects more slowly decaying TROSY component for 1H detection resulting in cross peaks appearing at ωC′, ωN, and ωHN + πJNH frequencies. Therefore (J + D)NH couplings are measured from the cross peak displacement between the TROSY and the F2-decoupled TROSY (κ = 0) spectra. Because cross peak positions differ in F2(15N) by πJNH, the measured splitting is only half of the true coupling, hence doubling the error in measurement.

In case of 0 < κ ≤ 1, the first 180°(1H) pulse inverts the 1H spin states and 15N–1H coupling interaction evolves during the κt2 period, whereas the second 180°(1H) pulse together with 180°(15N) pulse decouples the interaction during t2. Again, the ensuing TROSY element selects the most slowly relaxing component for 1H detection, resulting in cross peak at ωC′, ωN + κπJNH, and ωHN + πJNH frequencies. Thus, the apparent (J + D)NH is measured between the TROSY and downscaled anti-TROSY spectra, and the frequency of the anti-TROSY component is determined by the factor κ. True couplings are then obtained by multiplying the measured splitting with the factor of (1 + κ)/2. Using the scaling factor κ = 1 results in a ‘natural’ 15N–1H splitting as 15N chemical shift and (J + D)NH coupling evolutions are active for equal periods.

Figure 1b shows a modified, albeit conceptually analogous, implementation of the HR-TROSY HNCO pulse sequence (Hu et al. 2009). Although for recording the TROSY component both experiments are identical, they differ in case of recording the anti-TROSY component. Indeed, when measuring the anti-TROSY component, the signal modulates in the MQ-HNCO-TROSY (Fig. 1a) according to
$$ \exp \left[ { - \left( {i\omega_{\text{N}} - i\kappa \pi J_{\text{NH}} } \right)t_{2} } \right]\exp \left[ { - \left( { - \frac{{\left( {1 + {\kappa}} \right)}}{2}\,R_{{ 2 {\text{T}}}} t_{2} } \right)} \right]\exp \left[ { - \left( {\frac{{\left( {1 + {\kappa}} \right)}}{2}\,R_{{ 2 {\text{A}}}} t_{ 2} } \right)} \right] $$
(1)
and in the HR-TROSY HNCO (Fig. 1b)
$$ \exp \left[ { - \left( {i\omega_{\text{N}} +i \left( {1 + \kappa } \right)\pi J_{\text{NH}} } \right)t_{ 2} } \right]\exp \left[ { - \left( {\frac{\kappa }{2}\,R_{{ 2 {\text{T}}}} t_{ 2} } \right)} \right]\exp \left[ { - \left( {\frac{\kappa }{2}\,R_{{ 2 {\text{A}}}} t_{ 2} } \right)} \right] $$
(2)
where R2T and R2A correspond to relaxation rates of TROSY and anti-TROSY components, respectively. As we measure the anti-TROSY components with respect to the TROSY cross peak emerging at ωN − πJNH, the required scaling factors are not similar between two experiments and they need to be normalized with respect to MQ-HNCO-TROSY to allow direct comparison of effective relaxation during the pulse sequences. Replacing κ in (2) by κ + 1 translates into
$$ \exp \left[ { - \left( {i\omega_{\text{N}} - i\left( {2 + \kappa } \right)\pi J_{\text{NH}} } \right)t_{2} } \right]\exp \left[ { - \left( {\frac{{\left( {1 + \kappa } \right)}}{2}\,R_{{ 2 {\text{T}}}} t_{ 2} } \right)} \right]\exp \left[ { - \left( {\frac{{\left( {1 + \kappa } \right)}}{2}\,R_{{ 2 {\text{A}}}} t_{ 2} } \right)} \right] $$
(3)
Now, using identical values of κ in (1) and (3) result in identical but opposite offset for the scaled anti-TROSY component from the TROSY cross peak. In this case, the relevant relaxation exponents for MQ-HNCO-TROSY and HR-TROSY HNCO then become
$$ \exp \left[ { - \left( {\left( {R_{{ 2 {\text{A}}}} - R_{{ 2 {\text{T}}}} } \right)\frac{1}{2}\left( {1 + \kappa } \right)t_{ 2} } \right)} \right] $$
(4)
and
$$ \exp \left[ { - \left( {\left( {R_{{ 2 {\text{A}}}} + R_{{ 2 {\text{T}}}} } \right)\frac{1}{2}\left( {1 + \kappa } \right)t_{ 2} } \right)} \right], $$
(5)
respectively. Hence, the effective relaxation rates for the scaled anti-TROSY components in the MQ-HNCO-TROSY and HR-TROSY HNCO experiments are \( {\raise0.7ex\hbox{$1$} \!\mathord{\left/ {\vphantom {1 2}}\right.\kern-\nulldelimiterspace} \!\lower0.7ex\hbox{$2$}} \times \left( {1 + \kappa } \right) \times \left( {R_{{ 2 {\text{A}}}} - R_{{ 2 {\text{T}}}} } \right) \) and \( {\raise0.7ex\hbox{$1$} \!\mathord{\left/ {\vphantom {1 2}}\right.\kern-\nulldelimiterspace} \!\lower0.7ex\hbox{$2$}} \times \left( {1 + \kappa } \right) \times \left( {R_{{ 2 {\text{A}}}} + R_{{ 2 {\text{T}}}} } \right) \), respectively. Line narrowing can be observed by comparing relaxation exponents between the MQ-HNCO-TROSY and HR-TROSY HNCO experiments. As linewidth is a function of effective relaxation rate, we can directly observe:
$$ {\frac{{{\text{LW}}_{\text{MQ - HNCO - TROSY}} }}{{{\text{LW}}_{\text{HR - TROSY - HNCO}} }}} = { \exp }\left[ { - \left( {1 + \kappa } \right)R_{{ 2 {\text{T}}}} t_{ 2} } \right] < 1 $$
(6)
where κ ≥ 0 and LW is peak linewidth. For instance, using the factor κ = 1, i.e. no scaling of the apparent (J + D)NH splitting, results in additional line broadening in the HR-TROSY HNCO scheme by \( \exp \left[ {2R_{{ 2 {\text{T}}}} t_{ 2} } \right] \) Despite the 2–7 times larger R2A in comparison to R2T, the additional line broadening is far from being negligible even on perdeuterated 60 kDa system.
Figure 2 shows 1D traces along 15N dimension from K10–A34 residues in small 56-residue protein GB1. Overlays of the TROSY component (middle) and downscaled and upscaled anti-TROSY components of the MQ-HNCO-TROSY (right) and HR-TROSY HNCO (left) spectra, recorded with different κ values are shown for K10 (Fig. 2a) and A34 (Fig. 2b). It can be observed that the apparent frequency difference and linewidth increase, whereas the sensitivity decreases, with increasing κ. In case of κ = 0, the line broadening for the upfield (downscaled anti-TROSY line) multiplet component is counterbalanced by refocused long-range coupling interactions between 15N and 1H during the t2 in non-deuterated sample. Two- and three-bond couplings between 15N(i) and 1Hα(i), and 1Hα(i – 1) range typically from −0.5 to −2 Hz and 0.1 to −2 Hz (Permi et al. 2000b; Permi 2003; Wang and Bax 1995), whereas three-bond scalar couplings up to 5–6 Hz between 15N(i) and 1Hβ(i) have been reported (Düx et al. 1999). In case of κ = 1, the apparent linewidth is 10 and 13% larger in comparison to decoupled 15N component with κ = 0 for K10 and A34, respectively. This line broadening is due to faster decay of 15N transverse magnetization (anti-TROSY effect) during κt2/2 as well as contribution of long-range 15N–1H couplings to the effective line width. In contrast, the line width in HR-TROSY HNCO spectrum increases by 52% for K10 and 47% for A34 when using the scaling factor κ = 2 instead of κ = 1. Comparison of attainable line widths between the MQ-HNCO-TROSY (Fig. 1a) and HR-TROSY HNCO (Fig. 1b) schemes reveals significant differences. As demonstrated for A34 and K10 in Fig. 2, the apparent line width obtained with the MQ-HNCO-TROSY scheme is significantly smaller in comparison to the HR-TROSY HNCO method (Fig. 1b), where the spin-echo period follows the chemical shift labeling period. Indeed, line width is nearly 50% larger in HR-TROSY HNCO (κ = 2) spectrum in comparison to the corresponding MQ-HNCO-TROSY (κ = 1) spectrum. This is mainly governed by additional signal decay during the κt2 period, which is implemented in a constant-time manner in Fig. 1a instead of a real-time implementation shown in Fig. 1b.
https://static-content.springer.com/image/art%3A10.1007%2Fs10858-010-9422-z/MediaObjects/10858_2010_9422_Fig2_HTML.gif
Fig. 2

Expansions of K10 (a) and A34 (b) cross peaks from GB1 along the 15N dimensions. For both residues, four spectra with the anti-TROSY components and a spectrum with the TROSY component are overlaid. The TROSY component was recorded for the reference, using the MQ-HNCO-TROSY experiment, and is shown in the middle. Two upscaled anti-TROSY spectra (left) were recorded with κ = 1 and κ = 2 using the modified HR-TROSY HNCO experiment. Two spectra on the right show the downscaled anti-TROSY components, which were recorded with scaling factors κ = 0 and κ = 1 using the MQ-HNCO-TROSY pulse sequence

In order to verify accuracy of these approaches in practice, we measured 1JNHs in 15N/13C labeled human ubiquitin (76 residues, MW 8.6 kDa) using the MQ-HNCO-TROSY experiment with two different κ values (κ = 0 and κ = 1) and compared them with the reference values obtained using the generalized 2D 15N, 1H TROSY experiment (Andersson et al. 1998; Weigelt 1998). A pair-wise comparison of 1JNHs obtained using the proposed approach to the reference values measured using the 2D TROSY experiment, gave a Pearson correlation of R = 0.94 and R = 0.95 with an RMSD of 0.30 Hz and 0.26 Hz for κ = 0 and κ = 1, respectively. Correlation plots between the proposed method and the reference experiment shown in Fig. 3 reveal no obvious systematic errors, suggesting that 1JNHs can be measured with a sufficient accuracy by the proposed methodology, and the attainable precision is mainly governed by the apparent line width and sensitivity.
https://static-content.springer.com/image/art%3A10.1007%2Fs10858-010-9422-z/MediaObjects/10858_2010_9422_Fig3_HTML.gif
Fig. 3

Correlations between scalar 1JNH couplings obtained using the reference experiment (Andersson et al. 1998; Weigelt 1998) and the MQ-HNCO-TROSY pulse sequence. Pearson’s correlations for (a) and (b) are R = 0.94 and R = 0.95 with RMSD of 0.30 and 0.26 Hz, respectively. MQ-HNCO-TROSY spectra were collected using the scaling factors κ = 0 and κ = 1 for data in (a) and (b)

The precision by which the cross peak placement can be determined is according to Bax et al. (2001) defined by:
$$ \Updelta J = {\frac{\text{LW}}{\text{SN}}} $$
(7)
where ΔJ is a crude estimation of the attainable precision (RMSD) of a coupling in Hz, LW is the attainable signal line width at the half height in Hz, and SN is the signal-to-noise ratio of the peak.
Accurate determination of couplings with the proposed MQ-HNCO-TROSY scheme depends also on the scaling factor κ, which also scales the random error. Although it seems convenient to perform measurements with κ = 1, in which case no scaling is required, smaller κ improves the overall sensitivity and decrease the apparent line width. Hence, the precision of measured couplings is a trade-off between LW/SN and the scaling of the random error. Determination of J coupling with κ = 1 leads to precision ΔJκ=1 = LW1/SN1 whereas the same coupling can be measured, for instance with κ = 0 and multiplying the result by factor of 2, thus leading to precision ΔJκ=0 = 2 × LW0/SN0. To a first approximation, precision follows coarse relation
$$ \Updelta J_{i} = {\frac{{{\text{LW}}_{i} }}{{{\text{SN}}_{i} }}} = {\frac{{\left( {\kappa_{j} + 1} \right)}}{{\left( {\kappa_{i} + 1} \right)}}} \times {\frac{{{\text{LW}}_{j} }}{{{\text{SN}}_{j} }}} = {\frac{{\left( {\kappa_{j} + 1} \right)}}{{\left( {\kappa_{i} + 1} \right)}}} \times \Updelta J_{j} $$
(8)
where i and j refer to affiliated κ values. It is noteworthy that this description does not take into account line narrowing or signal to noise enhancement with smaller κ values. Although more sophisticated correction factors can be applied for determining LW/SN with different κ values, we preferred determination of LW/SN experimentally. As shown in recent studies, methods with the downscaled couplings provide higher precision for fast decaying signal on larger proteins (Kontaxis et al. 2000; Tugarinov and Kay 2003). As noted by Bax et al. (2001), the ΔJ defines only the lower limit of precision, while several other factors, most prominently resonance overlap, hamper the accurate determination of the splitting. These conditions are often met in the case of larger proteins, where both extensive cross peak overlap and prohibitively fast relaxation of the anti-TROSY component governs the attainable accuracy and precision of measured (J + D)NHs.
We measured MQ-HNCO-TROSY experiment with different κ values on larger monomeric protein, a 60-kDa fragment (557 residues) of filamin A, and compared linewidths and resolution of obtained spectra. Structural arrangement of filamin A is modular, comprised of six adjacent filamin-type immunoglobulin-like domains 16–21 (IgFLNa16–21), with average 15N transverse relaxation rates of 30–40 s−1, that make IgFLNa16–21 a highly challenging molecular system to be studied by solution state NMR spectroscopy. Although perdeuteration was a prerequisite for the measurement of RDCs in this high molecular weight system, increased resonance overlap due to large monomeric architecture generates an additional challenge. MQ-HNCO-TROSY experiment provided adequate resolution in the 15N dimension and enabled measurement of 84% of all possible (J + D)NHs in IgFLNa16–21 Resolution enhancement in the 15N dimension, obtained using the novel scheme instead of the conventional HNCO is illustrated in Fig. 4. Clearly the separation of cross peaks (the TROSY component) in the 15N dimension is significantly better in the MQ-HNCO-TROSY spectrum (Fig. 4a) than in the conventional HNCO-TROSY spectrum (Fig. 4b).
https://static-content.springer.com/image/art%3A10.1007%2Fs10858-010-9422-z/MediaObjects/10858_2010_9422_Fig4_HTML.gif
Fig. 4

Representative expansion of two-dimensional 15N–1H MQ-HNCO-TROSY spectrum from IgFLNa16–21. Spectrum in the panel (a), showing the TROSY components, was recorded using 118 complex points in 15N dimension (t2,max~47.2 ms) and the spectrum in the panel (b)—with 59 complex points in t2 (t2,max~23.6 ms)

Fast relaxation of the anti-TROSY component establishes an additional problem in large proteins such as IgFLNa16–21. Figure 5a shows expansions of 1D 15N traces taken through K1824, Y1862 and H2061 cross peaks of 15N, 13C, 2H labeled IgFLNa16–21. Again shown are the TROSY spectrum (in the middle) overlaid with the anti-TROSY spectra recorded using the MQ-HNCO-TROSY experiment with κ = 0 and κ = 1 (right) and HR-TROSY HNCO experiment with κ = 1 and κ = 2 (left). It is clearly evident from Fig. 5a that linewidth of the upfield 15N-{1H} component increases rapidly with increasing κ. For instance, in case of MQ-HNCO-TROSY experiment with κ = 0, the non-apodized linewidth and S/N ratio for the downscaled anti-TROSY component of K1824 are 14.2 Hz and 22.2, whereas they are 16.7 Hz and 17.4 with κ = 1, respectively. However, in the HR-TROSY HNCO spectrum, the corresponding line widths for the upscaled anti-TROSY component are 16.9 Hz (κ = 1) and 23.8 Hz (κ = 2). A similar trend is observed also in the cases of Y1862 and H2061. Although line broadening is less dramatic when compared to non-deuterated GB1 (Fig. 2) due to greater R2A/R2T ratio in the perdeuterated sample, this indicates that contribution of slowly decaying (R2T) magnetization to the overall linewidth of broad multiplet component is not negligible even on large perdeuterated protein. Attainable linewidths for 25 well-resolved 15N–1H cross peaks in 15N, 13C, 2H labeled IgFLNa16–21 are presented in Fig. 5b, showing that linewidths acquired with HR-TROSY HNCO are larger than those recorded with MQ-HNCO-TROSY experiment.
https://static-content.springer.com/image/art%3A10.1007%2Fs10858-010-9422-z/MediaObjects/10858_2010_9422_Fig5_HTML.gif
Fig. 5

a Representative expansions along the 15N dimension from the two-dimensional 15N–1H MQ-HNCO-TROSY spectrum of 2H, 13C, 15N labeled IgFLNa16–21. Excerpts are taken through the cross peaks of residues K1824, Y1862, and H2061. TROSY and downscaled anti-TROSY (κ = 0 and κ = 1) components, recorded with the MQ-HNCO-TROSY scheme, and upscaled anti-TROSY (κ = 1 and κ = 2) components, recorded with the HR-TROSY HNCO pulse scheme, are shown overlaid. b Linewidths for 25 well-resolved 15N–1H cross peaks in MQ-HNCO-TROSY and HR-TROSY HNCO spectra. Blue and greenbars show apparent line widths obtained using the MQ-HNCO-TROSY experiment with κ = 0 and κ = 1, respectively. The bars corresponding to values obtained with the HR-TROSY HNCO experiment with κ = 1 and κ = 2 are highlighted with red and orange colors, respectively

It can be anticipated that performance of the proposed MQ-HNCO-TROSY experiment sustains superiority to HR-TROSY HNCO despite the fact that the ratio R2A/R2T increases as polarizing magnetic field, B0, approaches the optimal 900–1,000 MHz field strength, where the 15N chemical shift anisotropy (CSA) and 15N–1H dipolar coupling (DD) relaxation mechanisms cancel each other (Pervushin et al. 1997). Several 15N–1H spin moieties are influenced by additional relaxation mechanisms violating optimal TROSY conditions e.g. chemical exchange contribution or a slight variation in the magnitude or orientation of the 15N CSA tensor between residues in α-helix and β-sheet. Therefore, contribution of R2T during κt2 period to the overall linewidth is non-zero.

Further inspection of attainable linewidths reveals that in case of larger proteins, the most advantageous strategy to determine 1(D + J)NHs is to measure the corresponding splitting symmetrically around the TROSY component as reported by Tolman and co-workers (Abrogast et al. 2010). To that end, schemes in Fig. 1a and b are combined to measure downscaled and up-scaled anti-TROSY components with scaling factors κ = 0 and κ = 1, respectively. This results in two cross peaks emerging at ωC′, ωN, ωHN + πJNH and ωC′, ωN −  2πJNH, ωHN + πJNH frequencies. Therefore, 1(D + J)NHs can be obtained between two anti-TROSY cross peaks offset from the TROSY transition by ±πJNH, which provides optimal sensitivity and resolution in case of larger proteins (Abrogast et al. 2010).

Using the scaling factor κ = 1 instead of κ = 0 for measuring the anti-TROSY component provided higher accuracy in case of K1824 in IgFLNa16–21, with ΔJs of 0.96 and 1.28 Hz, respectively. However, we decided to use the scaling factor κ = 0, resulting in the apparent (J + D)NH/2 splitting, as transverse relaxation of an anti-TROSY component for several residues is significantly faster than we found for K1824. Hence, we employed the proposed MQ-HNCO-TROSY experiment at 800 MHz for measuring 15N–1H RDCs in IgFLNa16–21 from the cross peak displacement between the TROSY and downscaled anti-TROSY components, recorded with the scheme in Fig. 1a using the scaling factor κ = 0. RDCs were extracted from the difference between 15N–1H splittings measured in water and in a diluted liquid crystal medium composed of filamentous phage Pf1 particles (Hansen et al. 1998) as demonstrated for K1824, T1984 and G2008 in Fig. 6a and b. In total, we measured 431 RDCs, spanning from −60 to 50 Hz. Degree of alignment varied quite substantially between individual domains of IgFLNa16–21. Dmax of ~20–30 Hz was found for domains 16–19, whereas couplings as large as −60 Hz were measured for domains 20–21. We compared the experimentally determined RDCs from IgFLNa16–21 to those predicted from the existing NMR (IgFLNa16–17 and IgFLNa18–19) and crystal (IgFLNa19–21) structures (Heikkinen et al. 2009; Lad et al. 2007), using singular value decomposition algorithm (Losonczi et al. 1999) implemented into the software package PALES (Zweckstetter and Bax 2000). A pair-wise comparison of the calculated RDCs, based on the NMR structure of IgFLNa16–17 (PDB ID: 2K7P), to corresponding experimental couplings from domains 16 to 17 in IgFLNa16–21 yield Q-factor 0.34 (Fig. 7a), indicating good fit between experimental RDCs and NMR coordinates based on NOE restraints (Heikkinen et al. 2009). In contrast to the domain pair 16–17, the fits between predicted and experimental RDCs for domain pair 18–19 (PDB ID: 2K7Q) as well as for the crystal structure of IgFLNa19–21 (PDB ID: 2J3S) gave significantly worse correlations with Q-factors of 0.45, and 0.61, respectively (Fig. 7b, c). We take this as an indication of significant differences in domain orientations between isolated domain pairs rather than insufficient quality of experimental RDC data because the dynamic range of RDCs measured from IgFLNa16–21 is several-fold larger than contribution of random error, which we estimate to be ±(2–3) Hz for the measured RDC. Experimental RDCs measured from IgFLNa16–21 clearly correlate better with the back-calculated values from the individual domains of IgFLNa16–17, IgFLNa18–19 and IgFLNa19–21 structures, resulting in Q-factors of 0.33, 0.31, 0.38, 0.33, 0.24, and 0.19, respectively (Fig. 7d–f). In these cases, a total of 321 RDCs were utilized for fittings. RDCs of the residues residing at domain boundaries or showing elevated internal mobility, or which were missing in the crystal structure (residues 2,163–2,170 and 2,191–2,197) were not included to the fits. We reckon that the poor fit between experimental and back-calculated RDCs originate either from the difference between isolated domains (IgFLNa18–19 and IgFLNa19–21) and larger molecular construct (IgFLNa16–21) or the difference between solution and crystal lattices of IgFLNa19–21.
https://static-content.springer.com/image/art%3A10.1007%2Fs10858-010-9422-z/MediaObjects/10858_2010_9422_Fig6_HTML.gif
Fig. 6

Two-dimensional expansions of K1824, T1984 and G2008 cross peaks in water (a) and in phage (b). TROSY components are shown with black contour, downscaled anti-TROSY (κ = 0) with blue contours and downscaled anti-TROSY (κ = 1) with red contours. The measured splittings in water between the TROSY and anti-TROSY (κ = 0), and TROSY and anti-TROSY (κ = 1) are highlighted in the panel (a) The corresponding splittings (in phage) between the TROSY and anti-TROSY (κ = 0) components are shown in the panel (b)

https://static-content.springer.com/image/art%3A10.1007%2Fs10858-010-9422-z/MediaObjects/10858_2010_9422_Fig7_HTML.gif
Fig. 7

Correlations between the experimental 1DNHs and the predicted couplings calculated from a IgFLNa16–17 NMR structure, b IgFLNa18–19 NMR structure, and c IgFLNa19–21 crystal structure. d A pair-wise comparison of experimental and the predicted 1DNH values of the domain 17 from IgFLNa16–17, e the domain 19 from IgFLNa18–19, and the domain 21 from IgFLNa19–21. The fits were obtained by using the singular value decomposition method (Losonczi et al. 1999) integrated into the program PALES (Zweckstetter and Bax 2000)

Conclusions

We have introduced a modified HNCO-TROSY pulse sequence for measuring scalar and residual dipolar couplings in large proteins. The proposed MQ-HNCO-TROSY experiment suits well for measuring RDCs in high molecular weight systems with large number of cross peaks. This advantage can be attributed to two main reasons: the experiment provides a significantly higher resolution in 15N dimension with respect to the traditional HNCO-TROSY, thus diminishing cross peak overlap, and minimizes the time period during which the 15N spin is susceptible to fast transverse relaxation associated with the anti-TROSY transition. In the proposed scheme, the whole 2TN period is used for labeling of 15N chemical shift in a constant time manner, which efficiently doubles the attainable resolution in 15N dimension with respect to the conventional HNCO-TROSY experiment. We employed the MQ-HNCO-TROSY scheme with κ = 0 for measuring RDCs in high molecular weight protein IgFLNa16–21 with 557 residues, resulting in 431 RDCs. Correlations between experimental and back-calculated RDCs in individual domains gave relatively low value Q factors, indicative of sufficient accuracy that can be obtained with the proposed MQ-HNCO-TROSY experiment in high molecular weight proteins. We believe that the proposed MQ-HNCO-TROSY experiment will be very useful for measuring 1DNHs in larger proteins.

Experimental

The MQ-HNCO-TROSY pulse sequence for measuring scalar couplings was tested on three proteins: 1.9 mM uniformly 15N/13C labeled human ubiquitin (8.6 kDa, 76 residues) in10 mM Na-PO4, pH 5.8, supplemented with 5% (v/v) D2O. 1.5 mM uniformly 15N, 13C labeled immunoglobulin-binding domain B1 of streptococcal protein G (GB1, 6.5 kDa, 56 residues) in 20 mM Na-PO4, pH 5.5, supplemented with 7.5% (v/v) D2O, and 0.5 mM uniformly 15N, 13C, 2H labeled IgFLNa16–21 (60 kDa, 557 residues) in 50 mM Na-PO4, pH 6.8, 100 mM NaCl, 1 mM DTT and 2 mM NaN, supplemented with 7% (v/v) D2O. In order to measure residual dipolar couplings, 15N, 13C, 2H labeled IgFLNa16–21 was dissolved in a dilute liquid crystal medium, composed of filamentous phage particles, Pf1. The phage concentration was 15 mg/ml and the final protein concentration was 0.4 mM.

Spectra for ubiquitin and GB1 were acquired on a Varian Unity INOVA 800 and 600 MHz NMR spectrometers, respectively. Spectra for IgFLNa16-21 were measured on a Varian Unity INOVA 600 and 800 MHz NMR spectrometers. Spectrometers were equipped with a 1H/13C/15N triple-resonance probeheads and actively shielded Z-axis gradient systems. Spectra were processed using the standard VNMRJ 2.1 revision B software package (Varian associates, 2006), and analyzed with VNMRJ2.1 and Sparky 3.110 (Goddard and Kneller, 2004). All spectra for ubiquitin and GB1 were measured at 25°C and at 30°C for IgFLNa16-21.

In case of GB1, the two-dimensional MQ-HNCO-TROSY (HR-TROSY HNCO) spectra with values of κ = 0 and κ = 1 (κ = 1 and κ = 2) were recorded using 2 transients per FID with 128 and 512 complex points, corresponding to acquisition times of 62 and 51.2 ms in 15N and 1H dimensions, respectively. Experimental time per each spectrum was 16 min. For each spectrum, prior to zero-filling to 4,096 × 4,096 data matrix and Fourier transform, identical shifted squared sine-bell weighting function were applied to both dimensions.

In case of ubiquitin, the two-dimensional MQ-HNCO-TROSY spectra, recorded with the pulse sequence shown in Fig. 1a using the κ = 0 and κ = 1 values, and the generalized 2D 15N, 1H TROSY experiment (Andersson et al. 1998; Weigelt 1998) were employed. Spectra were acquired with 2 transients per FID with 128 and 853 complex points in 15N and 1H dimensions, respectively. These correspond to acquisition times of 62 and 85.3 ms in t1 and t2, respectively. For each spectrum, prior to zero-filling to 4,096 × 4,096 data matrix and Fourier transform, identical shifted squared sine-bell weighting functions were applied to both dimensions.

For comparison of attainable line widths in IgFLNa16-21 at 600 MHz, two-dimensional 15N, 1H correlation spectra with κ = 0 (Fig. 1a), κ = 1 (Fig. 1a, b) and κ = 2 (Fig. 1b) were acquired using 48 transients per FID with 120 and 853 complex points, corresponding to acquisition times of 60.0 and 85.3 ms in 15N and 1H dimensions, respectively. Total experimental time was 5 h for each spectrum. For each spectrum, prior to zero-filling to 4,096 × 4,096 data matrix and Fourier transform, identical shifted squared sine-bell weighting function were applied to both dimensions.

Three-dimensional MQ-HNCO-TROSY spectrum (in water/phage) was acquired using 2 transients per FID with 32, 128 and 853 complex points, corresponding to acquisition times of 18.2, 49.2, and 85.3 ms in t1(13C′), t2(15N), and t3(1H), respectively. Interscan delays were 1.55/2.4 s for the sample in water/phage, respectively. Shifted squared sine-bell weighting functions were applied to all three dimensions before zero-filling to 256 × 1,024 × 2,048 data matrix and Fourier transform. TROSY and downscaled anti-TROSY components were measured in an interleaved manner and the scaling factor κ = 0 was used. Total acquisition time was 79 and 118 h for spectrum recorded in water and phage, respectively.

Copyright information

© Springer Science+Business Media B.V. 2010