APSY-NMR with proteins: practical aspects and backbone assignment

Abstract

Automated projection spectroscopy (APSY) is an NMR technique for the recording of discrete sets of projection spectra from higher-dimensional NMR experiments, with automatic identification of the multidimensional chemical shift correlations by the dedicated algorithm GAPRO. This paper presents technical details for optimizing the set-up and the analysis of APSY-NMR experiments with proteins. Since experience so far indicates that the sensitivity for signal detection may become the principal limiting factor for applications with larger proteins or more dilute samples, we performed an APSY-NMR experiment at the limit of sensitivity, and then investigated the effects of varying selected experimental parameters. To obtain the desired reference data, a 4D APSY-HNCOCA experiment with a 12-kDa protein was recorded in 13 min. Based on the analysis of this data set and on general considerations, expressions for the sensitivity of APSY-NMR experiments have been generated to guide the selection of the projection angles, the calculation of the sweep widths, and the choice of other acquisition and processing parameters. In addition, a new peak picking routine and a new validation tool for the final result of the GAPRO spectral analysis are introduced. In continuation of previous reports on the use of APSY-NMR for sequence-specific resonance assignment of proteins, we present the results of a systematic search for suitable combinations of a minimal number of four- and five-dimensional APSY-NMR experiments that can provide the input for algorithms that generate automated protein backbone assignments.

Appendix 1

Here, we describe the strategy followed for the selection of the APSY-NMR experiments to be used for sequence-specific polypeptide backbone resonance assignments in proteins (Tables 2, 4). This includes the criterion that the experiment must yield the heteronuclear correlations needed for the resonance assignments, and among the different possible experiments the one with the highest sensitivity is then selected.

APSY-NMR and polypeptide backbone resonance assignments

The process of sequence-specific resonance assignment for the polypeptide backbone includes the generation of molecular fragments from experimental scalar coupling correlations and chemical shift comparisons, and the mapping of these fragments onto the known amino acid sequence (Wüthrich 1986; Güntert et al. 2000; Moseley et al. 2001; Baran et al. 2004). In APSY-NMR experiments the magnetization transfer pathways cover short stretches of the polypeptide backbone and combine nuclei contained in these fragments into a single correlation (Fig. A1). Mutual overlap of the small molecular fragments defined by such correlations from different experiments is exploited for generating longer fragments, which can then be matched with the amino acid sequence.

When using exclusively signal acquisition on the amide proton, the amide proton and amide nitrogen-15 chemical shifts are contained in each correlation, and can therefore serve to match correlations from different experiments for the generation of longer fragments. When using experiments with dimensions below 6 (Fiorito et al. 2006; Hiller et al. 2007), additional chemical shift matching is needed. As illustrated in Fig. A1, the C^α atom is always available for this additional matching, and a second matching atom can be either HA, CB or CO. Thereby, the C^β chemical shift allows the unambiguous distinction between different amino acid types, thus increasing the reliability of sequence-specific assignments and allowing unambiguous assignments of smaller fragments (Güntert et al. 2000; Hiller et al. 2007). With the requirement that there must be two overlapping chemical shifts between adjoining correlations, groups of two or three four- and five-dimensional correlations can be devised (Fig. A1). To identify suitable pulse sequences to be used for each assignment strategy outlined in Fig. A1, we compared the relative sensitivities of different experiments that could, in principle, provide the desired correlations.

Sensitivity of APSY-NMR experiments

Sensitivity comparisons of different APSY-NMR experiments that are detected on the same nucleus can be based on a comparison of the signal intensities at the time domain origin, $s_{m} \left( 0 \right)$ Eq. 4. Estimates of $s_{m} \left( 0 \right)$ for different experiments were obtained with the following assumptions (Sattler et al. 1999): (1) All 90° and 180° pulses are considered to be ideal pulses. (2) Magnetization losses occur only by transverse relaxation and from incomplete evolution under J-couplings. (3) The relaxation rate of a product operator is written as the sum of the relaxation rates of the individual nuclei, and relaxation of longitudinal operators is neglected. This yields, for example, that R ₂(H^N _xN_y) = R ₂(H^N) + R ₂(N) and R ₂(H^N _xN_z) = R ₂(H^N). (4) The steady-state polarization of single protons was assumed to be independent of the atom position, so that P(H^N) = P(H^α) = P(H^β). We further used standard values for the J-coupling constants and the transverse relaxation rates R ₂ (Table A1) and based the calculations on ideal pathways with transfer delays that optimize the transfer function.

As an illustration we describe the sensitivity evaluation for the 4D APSY-HACANH experiment (Fig. A2). The magnetization pathway of four INEPT steps, A–D, can be described as

$$ {\text{H}}_{\text{z}}^{{{\upalpha}}}\,\mathop{\rightarrow}\limits^{\text{A}}\,2 {\text{H}}_{\text{z}}^{{{\upalpha}}} {\text{C}}_{\text{z}}^{{{\upalpha}}}\,\mathop{\rightarrow}\limits^{\text{B}}\,2 {\text{C}}_{\text{z}}^{{{\upalpha}}} {\text{N}}_{\text{z}}\,\mathop{\rightarrow}\limits^{\text{C}}\,2 {\text{N}}_{\text{z}} {\text{H}}_{\text{z}}^{\text{N}}\,\mathop{\rightarrow}\limits^{\text{D}}\,{\text{H}}_{\text{x/y}}^{\text{N}} $$

The step A transfers H^α polarization to H^αC^α longitudinal 2-spin order, with the transfer function $\Upgamma \left( {\text{A}} \right) = \sin \left( {{{\pi}}{}^{1}J_{\text{HA,CA}} {{\delta}}} \right)\exp \left( { - {{\delta}}R_{2} \left( {\text{HA}} \right)} \right)$, where δ is the INEPT transfer delay. If the rotational diffusion of the protein is characterized by a correlation time, τ _c, of 10 ns, the optimal value for δ is 3.3 ms and the resulting optimal transfer amplitude is Γ(A) = 0.75. The step B has the function $\Upgamma \left( {\text{B}} \right) = 0.86 \cdot \sin \left( {{{\pi}}{}^{1}J_{\text{N,CA}} {{\rho}}} \right)\cos \left( {{{\pi}}{}^{2}J_{\text{N,CA}} {{\rho}}} \right)\cos \left( {{{\pi}}{}^{1}J_{\text{CA,CB}} {{\rho}}} \right)\exp \left( { - {{\rho}}R_{2} \left( {\text{CA}} \right)} \right)$, with ¹ J _CA,CB = 0 Hz for glycine residues. This element has an optimal value of ρ = 25 ms for τ _c = 10 ns, and the resulting transfer amplitude is Γ(B) = 0.20. The step C has the transfer function ${{\Upgamma}}\left( {\text{C}} \right) = \sin \left( {{{\pi}}{}^{1}J_{\text{N,CA}} {{\chi}}} \right)\cos \left( {{{\pi}}{}^{2}J_{\text{N,CA}} {{\chi}}} \right)\exp \left( { - {{\chi}}R_{2} \left( {\text{N}} \right)} \right)$, with an optimal value of χ = 27 ms, yielding a transfer amplitude of Γ(C) = 0.45. For the step D, the transfer function is $\Upgamma \left( {\text{D}} \right) = \sin \left( {{{\pi}}{}^{1}J_{\text{HN,N}} {{\xi}}} \right)\exp \left( { - {{\xi}}R_{2} \left( {\text{HN}} \right)} \right)$, with a theoretical optimum for ξ of 4.8 ms, yielding a transfer amplitude of Γ(D) = 0.79. From the four individual transfer functions, the overall sensitivity was calculated as $s_{m} \left( 0 \right) = P\left( {\text{HA}} \right) \cdot {{\Upgamma}}\left( {\text{A}} \right) \cdot {{\Upgamma}}\left( {\text{B}} \right) \cdot {{\Upgamma}}\left( {\text{C}} \right) \cdot {{\Upgamma}}\left( {\text{D}} \right)$, where P(HA) is the equilibrium polarization of the α-proton.

The relative overall sensitivities calculated with this approach for a selection of APSY-NMR experiments that can provide the correlations of Fig. A1 are listed in Table 2.

Table A1 J-couplings and relaxation rates R ₂ used for the calculation of the relative sensitivities for selected APSY-NMR experiments

Fig. A1

Combinations of intraresidual and sequential correlations to be recorded with H^N-detected APSY-NMR experiments used to collect the data needed for polypeptide backbone assignment of ¹³C,¹⁵N-labeled proteins. Each colored shape contains the nuclei correlated by a 4D or 5D experiment. The yellow areas contain the nuclei for which the individual correlations overlap. Chemical shift matching for these nuclei results in the generation of larger molecular fragments from the correlations obtained by the individual experiments. The notations used for the different groups of experiments reflect these matching chemical shifts, and indicate in parentheses additional important information for obtaining sequence-specific assignments (see text)

Fig. A2

Pulse sequence used for the 4D APSY-HACANH experiment. Radio-frequency pulses were applied at 4.7 ppm for ¹H, 118.0 ppm for ¹⁵N, 173.0 ppm for ¹³C′ and 54 ppm for ¹³C^α. Narrow and wide bars represent 90° and 180° pulses, respectively. Pulses marked “A” were applied as Gaussian shapes, with 120 μs duration on a spectrometer with a ¹H frequency of 500 MHz. All other pulses on ¹³C^α and ¹³C′ had rectangular shape, with a duration of √3/(Δω(C^α,C′)·2)) and √15/(Δω(C^α,C′)·4)) for 90° and 180° pulses, respectively. Pulses on ¹H and ¹⁵N were applied with rectangular shape and high power. The last six ¹H pulses represent a 3-9-19 WATERGATE element (Sklenar et al. 1993). The grey pulse on ¹³C′ was applied to compensate for off-resonance effects of selective pulses (McCoy and Mueller 1992). Decoupling using DIPSI-2 (Shaka et al. 1988) on ¹H and WALTZ-16 (Shaka et al. 1983) on ¹⁵N is indicated by rectangles. t ₄ represents the acquisition period. On the line marked PFG, curved shapes indicate sine bell-shaped pulsed magnetic field gradients applied along the z-axis with the following durations and strengths: G₁, 800 μs, 18 G/cm; G₂, 800 μs, 26 G/cm; G₃, 800 μs, 13 G/cm; G₄, 800 μs, 26 G/cm; G₅, 800 μs, 23 G/cm. Phase cycling: ϕ₁ = {x, x, –x, –x}, ψ ₂ = {x, –x}, ϕ₂ = {x, –x, –x, x}, ψ ₁ = y, all other pulses = x. Fixed delays were δ′ = 4.75 ms and ξ′ = ξ = 5.4 ms. Initial delays were t ^a₁  = t ^c₁  = 1.8 ms, t ^a₂  = t ^c₂  = 11.5 ms, t ^a₃  = t ^c₃  = 11 ms, t ^b₁  = t ^b₂  = t ^b₃  = 0 ms. The switched constant-time/semi-constant time elements were used as described (Fiorito et al. 2006), with λ′ = (t ^a₃  + t ^b₃  − t ^c₃ )/2. Quadrature detection for the indirect dimension was achieved using the hypercomplex Fourier transformation method for projections (Brutscher et al. 1995; Kupce and Freeman 2004) with the angles ψ ₁, ψ ₂ and ψ ₃ (ψ ₁ and ψ ₂ were incremented and ψ ₃ was decremented in 90° steps)