NH for Rapid Assignment of 1HN–15N HSQC ... - ACS Publications

Article pubs.acs.org/ac

Reduced Dimensionality (4,3)D-HN(C)NH for Rapid Assignment of 1 N 15 H − N HSQC Peaks in Proteins: An Analytical Tool for Protein Folding, Proteomics, and Drug Discovery Programs Jithender G. Reddy† and Ramakrishna V. Hosur*,†,‡ †

Department of Chemical Sciences, Tata Institute of Fundamental Research, 1-Homi Bhabha Road, Colaba, Mumbai−400005, India UM-DAE Centre for Excellence in Basic Sciences, University of Mumbai, Kalina Campus, Santa Cruz, Mumbai−400098, India

‡

S Supporting Information *

ABSTRACT: While nuclear magnetic resonance (NMR) has had commendable success in atomic-level investigation of folded proteins, intrinsically unfolded and partially folded proteins have always posed a great challenge, because of poor chemical shift dispersions. We present here a reduceddimensionality-based NMR triple resonance pulse sequence, (4,3)D-HN(C)NH, which not only helps to disperse the peaks further by combining 15N and amide 1H chemical shifts, but also directly establishes correlations between 1HNi , 15Ni, 1HNi+1, and 15Ni+1 spins along the F1−F3 planes. The F2−F3 projection planes of this experiment provide unique identification of the check points in amide resonances. An assignment strategy derived by combining information along the F1−F3 planes and in the F2−F3 projection planes of the experiment has been presented and shown to be very useful for both intrinsically disordered/unfolded proteins and folded protein alike. The experiment and the protocol would be valuable for protein folding, proteomics, and drug discovery programs.

I

proposed;6,8−10 however, they involve a long minimal measurement time to obtain sufficient resolution. Later, several reduced-dimensionality (RD)-based 3D experiments were proposed.11,12 However, they lack discrimination of peaks arising from different amide (1H, 15N) correlations. Often, additional experiments with selective spin system identification that aid resonance assignments are required,13 which further involves several days of data acquisition and extensive analysis. In this background, here, we present a single RD (4,3)DHN(C)NH experiment that enables direct identification of sequential correlations using both amide 1H and 15N chemical shifts, simultaneously. This correlates (i, i + 1) amide resonances and enables identification of sequentially connected peaks in the 15N-HSQC spectrum. The sequence has been derived by simple modification of the HN(C)N experiment described earlier.14−17 Here, the chemical shift combination Ω(15N) ± Ω(1HN) in the RD dimension results in distinctly separated spectral regions. Moreover, the presence of peak pairs in the same spectrum facilitates a grouping of peaks belonging to a given system by visual inspection, especially to identify certain triplet patterns. Furthermore, the special features will significantly increase the utility of this experimental protocol for NMR applications, as mentioned in the earlier paragraphs. These special features are as follows: (1) 4D information with

ntrinsically unstructured proteins/unfolded proteins are known to be involved in many important cellular activities.1,2 However, these proteins have always posed great challenges for nuclear magnetic resonance (NMR) investigations, since the NMR spectra of these proteins exhibit severe overlap, because of the lack of stable conformation along the sequence. Hence, there is a very high demand to develop new or alternative NMR methods and strategies for unambiguous assignment of backbone resonances in such proteins.1 Similarly, in proteomics and drug discovery program, where it is necessary to identify binding sites of small ligands/drugs to proteins, the interaction of thousands of potential candidates with target proteins must be studied, and, here, rapid assignment of 15N-heteronuclear single quantum correlation (15N-HSQC) spectra of complexes becomes crucial.3 The currently available strategies involve recording of several three-dimensional (3D) spectra and elaborate analysis to elucidate specific movement of HSQC peaks on ligand/drug binding. This is a time-consuming process. In the above context, the pulse sequences proposed earlier for identifying direct correlations between amide (1H and 15N) chemical shifts of one residue with those of its N-terminal and/ or C-terminal neighbors can be very useful. In some of these,4−7 the sequential connectivities between neighboring residues are established using two 3D experiments: one to correlate 1HNi /15Ni with 15Ni−1/15Ni+1 and another to correlate 1 N 15 Hi / Ni with 1HNi−1/1HNi+1. Subsequently, some four-dimensional (4D) experiments that provide these correlations were © 2012 American Chemical Society

Received: September 13, 2012 Accepted: November 5, 2012 Published: November 5, 2012 10404

dx.doi.org/10.1021/ac302656k | Anal. Chem. 2012, 84, 10404−10410

Analytical Chemistry

Article

Figure 1. Pulse sequences for (4,3)D-HN(C)NH experiment. Narrow (hollow) and wide (filled black) rectangular bars represent nonselective 90° and 180° pulses, respectively. Unless indicated, the pulses are applied with phase x. Waltz-16 decoupling sequence31,32 with a field strength of 6.3 kHz is applied for proton decoupling, and 15N decoupling, using the GARP-1 sequence33 with the field strength of 0.9 kHz, is applied during acquisition. The strength of the 13Cα pulses (standard Gaussian cascade Q3 (180°) and Q5 (90°) pulses)34 is adjusted so that they cause minimal excitation of carbonyl carbons. The 180° 13C′-shaped pulse (width 200 μs) had a standard Gaussian cascade Q3 pulse profile with minimal excitation of 13Cα. The delays were set to X = Z = λ = 2.7 ms, Y = 3 μs, κ = 5.4 ms, δ = 2.7 ms, τC = 4.5 ms, τCN = 12.5 ms, TN = 14 ms, τN = 15 ms, and Δ = τCN − τC. The coevolution of two chemical shifts (amide 1H and 15N spins) are achieved by simultaneously incrementing the corresponding delays tH1 and tN1 , respectively. The chemical shift evolution in the 1H channel is achieved in a semiconstant manner by varying X, Y, and Z as follows: ΔX = ΔtH1 /2; ΔY = ΔtH1 /2 − λ/Ni, and ΔZ = −λ/Ni, where Ni is the number of complex points in t1. The values for the individual periods containing tN1 were A = tN1 /2, B = TN, and C = TN − tN1 /2. The values for the individual periods containing t2 were D = τN − t2/2, E = τN, and F = t2/2. The phase cycling for the experiment is Φ1 = 2(x), 2(−x); Φ2 = Φ3 = x, −x; Φ4 = Φ5 = x; Φ6 = 4(x), 4(−x); and Φreceiver = 2(x), 4(−x), 2(x). In this experiment, frequency discrimination in t1 and t2 has been achieved using States-TPPI phase cycling35 of Φ1 and Φ5, respectively, along with the receiver phase. The gradient (sine-bell-shaped; 1 ms) levels are as follows: G1 = 30%, G2 = 30%, G3 = 30%, G4 = 30%, G5 = 50%, and G6 = 80% of the maximum strength (53 G/cm) in the z-direction. The recovery time after each gradient pulse was 160 μs. Before detection, the WATERGATE sequence36 was employed for better water suppression.

time with a recycling delay of 0.9 s was ∼20 h. For the (4,3)DHN(C)NH experiment, the spectral width(offset) in ppm along the 1H, 15N, and 15N ± Δ1H dimensions, respectively, were 9 (4.7), 26 (118), and 52 (118) for unfolded Human SUMO protein and 11 (4.7), 30 (116), and 150 (116) for MCrystallin (Holo form). F2−F3 (HSQC) projection planes of (4,3)D-HN(C)NH for identification of glycine (G-HSQC), alanine (A/G-HSQC), and serine/threonine (S/T-HSQC) residues were recorded with the same spectral widths and offsets as mentioned above, and data were acquired with 16 scans per FID, using 1024 and 64 complex points along F3 (1HN) and F2 (15N) dimensions, respectively. The F1−F3 projection plane of (4,3)D-HN(C)NH for identification of sequential correlations for M-Crystallin (Holo form) was recorded with the same spectral widths and offsets as mentioned above, and data were acquired with 16 scans per FID, using 1024 and 256 complex points along F3 (1HN)- and F2 (15N ± Δ1H)-dimensions, respectively. Acquisition times for recording F2−F3 and F1−F3 projection planes were ∼20 min and 80 min (1 h, 20 min), respectively. Data processing was carried out using Topspin software (Bruker, http://www. bruker.com/), and the resulting NMR spectra were analyzed using the Computer Aided Resonance Assignment (CARA) software18 (this is a software program used for NMR data analysis).

good spectral resolution within 3D data acquisition time (removes ambiguity); (2) easy self-and sequential peak discrimination, as they exhibit different phases; (3) ready amino-acid-type identification in 15N-HSQC spectrum produced by tuning the same experiment; and (4) efficient assignment strategy (everything can be easily mapped onto the HSQC spectrum). It also enables protein folding studies, starting from the highly denatured state of a protein. The performance of the experiments and the application of the protocol have been demonstrated here using an unfolded 110 residue Human SUMO protein, and folded M-Crystallin (Holo form) protein.

■

EXPERIMENTAL SECTION The (4,3)D-HN(C)NH pulse sequence (Figure 1) was developed using a 15N/13C-labeled GFL (Gly−Phe−Leu) peptide standard sample and tested using a 1.0 mM sample of uniformly 15N/13C-labeled Human Ubiquitin protein dissolved in a pH 6.4 phosphate buffer in 90% H2O and 10% D2O (both samples are from Cambridge Isotope Laboratories, Inc.). Application of the experiment was demonstrated on uniformly 15N/13C-labeled 0.9 mM unfolded His-tagged Human SUMO protein unfolded using 6 M urea in a pH 6.4 phosphate buffer in 90% H2O and 10% D2O and a folded 1.0 mM M-Crystallin (Holo form) protein dissolved in TrisHCL buffer (25 mM TrisHCL, 100 mM NaCl at pH 7.4) in 90% H2O and 10% D2O. All data were acquired at 25 °C using a cryoprobe on a Bruker Avance III spectrometer with a 1H operating frequency of 800 MHz. The (4,3)D-HN(C)NH data were acquired using 1024 complex points in direct dimension, with 8 scans per FID. The number of increments in the indirect 15 N ± Δ1H and 15N dimensions were 128 and 64 complex points, respectively. The delays 2TN, 2τC, 2τCN, and 2τN were set to 28, 9, 25, and 30 ms, respectively. The total experimental

■

RESULTS AND DISCUSSION The (4,3)D-HN(C)NH pulse sequence for directly identifying sequentially connected HSQC peaks (HiNi → Hi+1Ni+1) is shown in Figure 1. This has been derived from the basic 3D HN(C)N experiment16 by involving the reduced dimensionality (RD) coevolution approach,19,20 during evolution period t1. The magnetization transfer pathway in (4,3)D-HN(C)NH (Figure 2A) is identical to that in HN(C)N,16 except that 10405

dx.doi.org/10.1021/ac302656k | Anal. Chem. 2012, 84, 10404−10410

Analytical Chemistry

Article

Figure 2. (A) Schematic diagram showing the magnetization transfer pathway and frequency labeling in (4,3)D-HN(C)NH pulse sequence. 2TN, 2τC, 2τCN, and 2τN are the delays during which the transfers indicated by the arrows take place in the pulse sequence. (B) Schematic three-dimensional spectrum of (4,3)D-HN(C)NH experiment and the correlations observed in the F1−F3 (right side) and F2− F3 (left side) planes, respectively, at the 15N and 15N ± Δ1HN chemical shifts of residue i. Squares and circles represent the self-and sequential peaks, respectively. Red and black represent positive and negative phase of peaks, respectively.

Figure 3. (A) Schematic peak patterns in the F1−F3 strips of the (4,3)D-HN(C)NH spectrum and the corresponding triplet of residues. Red and black colors represent positive and negative signs, respectively. Squares and circles represent the self-and sequential peaks, respectively. X, Y, and Z are any residues other than glycine and proline. G/G′ and P represent glycine and proline, respectively. The patterns involving G provide identification of spin systems following glycines in the sequence, which help during the sequential assignment walk through the spectrum (middle panels). The patterns involving P identify break points during a sequential walk (last panel). (B) Projection of same strips along the F1-dimension, shown as F2 (1HN)− F3 (15N) plane (HSQC-type plane). (C) Strategy of a sequential walk through the 15N-HSQC spectrum using the (4,3)D-HN(C)NH spectrum. The assignment strategy here has been described using the polypeptide sequence “−XYZ−”. For each spin system “i” identified in the HSQC spectrum (here, we have taken it as residue X), corresponding to four frequency labels: (Ni − ΔHi, Ni+1 − ΔHi+1, Ni + ΔHi, and Ni+1 + ΔHi+1) in the F1 (15N ± Δ1HN)-dimension of HN(C)NH spectrum are identified. Then, knowing the position of (15HiN ± 15Ni) peak, sequential amide (1HNi+1, 15Ni+1) resonances can be calculated from Ni+1 − ΔHi+1 and Ni+1 + ΔHi+1 chemical shifts.

amide 1H and 15N chemical shifts coevolve during the incrementable delays tH1 and tN1 , where tH1 = ktN1 and “k” is a scaling factor. The 15N chemical shifts are detected in quadrature, while 1HN chemical shifts modulate the transfer amplitude, which results in four peaks with the linear combinations of amide 1H and 15N chemical shifts in the F1dimension. As a result of a 10-fold higher gyromagnetic ratio of 1 H relative to 15N, the chemical shifts of 1H are scaled 10 times over 15N. Chemical shifts scaling along the F1 dimension can be tuned using the factor k,21,22 which further enhances the dispersion of the peaks. The peak coordinates in the F1−F3 and F2−F3 planes of the (4,3)D-HN(C)NH spectrum are F1 = Ni ± ΔHi ,

triplets of residues, covering the general and all the special situations, the expected peak patterns in the F1−F3 planes of the spectrum are shown as schematic strips in Figure 3A. Here, each panel represents the F1−F3 plane at the 15N chemical shift of the central residue in the triplet. The signs of self (i) and sequential (i + 1) peaks in this plane are dictated by the nature of the residues at positions i and i − 1. For residues other than glycine and proline, the self (i) and sequential (i + 1) peaks have positive (red) and negative (black) signs, respectively (Figure 3A, first panel). For glycines, different patterns arise depending on whether the ith or (i − 1)th residue is a glycine or otherwise (Figure 3A, middle panels). For example, in a stretch −XGYZ−, in the F1−F3 plane at the 15N chemical shift of G, all peaks appear with same sign (positive) and, similarly, at the 15N chemical shift of Y, all four peaks will also have the same but opposite sign (i.e., negative). For the same reason, in a double-glycine stretch such as GG′Z, the inter-residue and intraresidue correlation peaks have distinct patterns. Similarly, in the case of proline at position i + 1, the absence of the amide proton results in the absence of (Hi, Ni+1 + ΔHi+1) and (Hi, Ni+1 − ΔHi+1) peaks (Figure 3A, last panel). These special peak patterns at glycines, proline, and the residues present next to

(F3 , F2) = (Hi , Ni), (Hi − 1, Ni − 1)

F2 = Ni , (F3 , F1) = (Hi , Ni + ΔHi), (Hi , Ni − ΔHi), (Hi , Ni + 1 + ΔHi + 1), (Hi , Ni + 1 − ΔHi + 1)

The phases of the peaks are calculated following standard product operator formalism, and the resultant peak patterns are schematically shown in Figure 2B. The F2−F3 plane at F1 = Ni ± ΔHi shows a self-peak (Hi, Ni) and a sequential peak (Hi−1, Ni−1). The F1−F3 plane at F2 = Ni shows four peaks, two of which are downfield while the other two are upfield from the offset (15Ncarrier), because of the addition and subtraction, respectively, of amide proton frequencies. One peak from each set (up-field/down-field) is a self-peak, (Hi, Ni + ΔHi/Hi, Ni − ΔHi) and the other is a sequential peak, (Hi, Ni+1 + ΔHi+1/Hi, Ni+1 − ΔHi+1). The inter-residue and intraresidue correlation peaks in different planes appear in opposite signs, except in special situations; in situations where a glycine/proline residue is involved, the actual sign patterns of the two types of peaks in each set would vary as in HN(C)N.14 Considering different 10406

dx.doi.org/10.1021/ac302656k | Anal. Chem. 2012, 84, 10404−10410

Analytical Chemistry

Article

+ ΔHi, and Ni+1 + ΔHi+1) in F1 (15N ± Δ1HN) dimension of (4,3)D-HN(C)NH spectrum, the HSQC peak (1HNi , 15Ni) can be readily connected to the sequential peak, (1HNi+1, 15Ni+1). Sequence-Dependent Peak Identification in F2−F3 Planes. Once the sequentially connectivities are established between the amide peaks, the next step is identification of the amino acid type corresponding to each 1H and 15N frequency pair in the 15N-HSQC spectrum. As we see in Figure 3A, the different triplets of residues produce different patterns along the F1 (15N ± Δ1HN)-dimension at amide (1Hi, 15Ni) chemical shifts of central residue. The simple projection down along the F1 dimension leads to a HSQC type spectrum where we can only observe the glycines and their following residues. This is because when an F1 projection is taken, peaks belonging to glycines and following residues are added together, self-and sequential peaks are of same sign (see Figure 3A, middle panels), whereas for other residues, self-and sequential peaks which are in opposite sign (see Figure 3A, first panel), the intensities cancel each other. The peak intensities for the residues next to proline (self-peaks only, see Figure 3A, last panel) also add up but they have approximately half the intensity of other peaks. Hence, we can selectively adjust the contour level to remove these peaks and observe only checkpoint information. Considering different triplets of residues, covering the general and all the special situations, the expected peak patterns in the F2−F3 projection plane of the (4,3)D-HN(C)NH spectrum are schematically shown in Figure 3B. This HSQC type 2D-spectrum recorded using (4,3)DHN(C)NH pulse sequence by avoiding “t1” evolution and selectively edited to show glycines and their following residue is named here as “G-HSQC” spectrum. Similarly, special patterns around serines/threonines and glycines/alanines can be generated by tuning the offset and bandwidth of the pulse marked by circle in Figure 1, as has been described in previous publications.17,23−25 The HSQC-type spectra showing serine/ threonines and alanine/glycine are called “S/T-HSQC” and “A/G-HSQC”, respectively. Thus, depending on the type of tuning of pulse sequence, the HSQC-type spectra provide amino acid type identification of glycines, alanines, serines/ threonines and the residues following them in the sequence directly on the 15N-HSQC spectrum. This is experimentally demonstrated on folded M-Crystallin (Holo form) in Figure S1 in the Supporting Information and on unfolded Human SUMO protein in Figure S-2 in the Supporting Information. After making all the possible i → (i + 1) connectivities in different stretches, using the checkpoint information of glycines, alanines, serines/threonines, and following residues from HSQC-type spectra, residue-specific identification of certain peaks was done and these stretches were then mapped on the primary sequence for final assignment. Advantages from RD Incorporation in the HN(C)N Experiment. The basic HN(C)N experiment16 has been proven to be very useful and provides identification of sequentially connected peaks in the 15N-HSQC spectrum from its different F1−F3 planes [i.e., (Hi, Ni) → (Hi, Ni+1)]. However, sometimes, problems arise when the HSQC spectrum becomes highly crowded and several peaks have the same amide 1H or 15N chemical shifts [Δδ(1H) < 0.1 ppm or [Δδ(15N) < 0.5 ppm]. In such a case, the identification of sequential correlations for a particular self-peak becomes ambiguous, and the assignment protocol may fail. This is illustrated in Figure S-3 in the Supporting Information (left panel). Even worse will be the case when the self-peak and the

Figure 4. Illustrative stretches of sequential connection through F1(15N ± Δ1HN)−3(1H) strips of the (4,3)D-HN(C)NH spectrum of unfolded Human SUMO protein for residues Glu76-Thr85. The red and black contours indicate positive and negative peaks, respectively. The strips provide identification of self (i) and sequential (i + 1) amide 15 N ± Δ1HN chemical shifts. The labels and numbers at the top in each panel identify the residue and the respective F2 (15N) chemical shifts. A horizontal line connects a sequential peak (black) in one plane to a self-peak (red) in the adjacent plane on the right, except at the check points, where the signs will be different. Note that labels on the top of the panels shown in red constitute the check points. The dashed line (gray) is the frequency offset (118 ppm) for the F1 (15N ± Δ1HN)-dimension, dividing the spectrum into the “plus” and “minus” domains. Sequential connections are shown (in cyan) in both domains. The plus sign (“+”) in red in each panel indicates the 15N chemical shift self (i) residue, which falls exactly in the middle of the two self-peaks. The absence of the sequential peaks confirms that the next residue is a proline (Pro86)a break point.

them provide important start or check points during the course of the sequential assignment process. Thus, for a given protein with a known amino acid sequence, it is possible to identify several special triplet sequences simply by inspecting the various F1−F3 planes in the (4,3)D-HN(C)NH spectrum. Strategy for Sequential Assignment. Sequential i → i + 1 Correlations in F1−F3 planes. Sequential correlations of peaks in the HSQC spectrum using (4,3)D-HN(C)NH spectrum has been schematically described in Figure 3C. First, the backbone amide spin systems are identified in the 15 N-HSQC spectrum. For each spin system i, four peaks are observed in F1 (15N ± Δ1HN)−F3 (1H) plane of the (4,3)DHN(C)NH spectrum at the F2 (15Ni) chemical shift. In each set of two peaks (upfield/downfield), considering the general situation, we have one self-peak (positive sign) and one sequential peak (negative sign) shown in red and black color, respectively. Each correlation peak in the 15N -HSQC spectrum creates chemical shift labels, of 1HN in the F3-dimension, and of 15 N in the F1- and F2-dimensions of the (4,3)D-HN(C)NH spectrum, since both dimensions are centered around 15N chemical shifts. This 15N label in the F1 dimension falls exactly in the middle of the two self-peaks, which, in fact, helps to identify the self-peaks in the F1−F3 planes of the (4,3)DHN(C)NH spectrum in some special situations. After identifying the four spin labels (Ni − ΔHi, Ni+1 − ΔHi+1, Ni 10407

dx.doi.org/10.1021/ac302656k | Anal. Chem. 2012, 84, 10404−10410

Analytical Chemistry

Article

Figure 5. (A) Overlay of 15N−HSQC spectra of M-Crystallin in calcium-bound −Holo form (red contours) and free −Apo form (green contours). Assigned stretch of Gly64-Ile74 (shown in panel (B)) is annotated in the Holo form 15N−HSQC spectrum. (B) F1 (15N ± Δ1HN)−F3 (1H) 2Dprojection plane of the (4,3)D-HN(C)NH spectrum of M-Crystallin (Holo form), showing the sequential connection of the Gly64-Ile74 stretch (shown as the blue dashed line). Red and black contours indicate positive and negative peaks, respectively. Two peaks joined by vertical line provides self (i) and sequential (i + 1) amide (1H and 15N) chemical shifts. From these connections and the chemical shifts in the “plus” and/or “minus” domains, the proton and nitrogen chemical shifts of the (i + 1) residue can be calculated (see Figure 3C) and, thus, the HSQC peak of residue (i + 1) can be identified.

is that there are two different sets of sequential correlations: one in the 15N + Δ1HN domain (plus domain) and the other in the 15N − Δ1HN domain (minus domain). Thus, any ambiguity in the identification of i → (i + 1) sequential correlation in upfield set of peaks can be resolved using the down-field set of peaks and vice versa. However, the strategy has one limitation: for those proteins that exhibit degeneracy in both amide 1HN and 15N shifts, the strategy may lead to ambiguities in backbone assignment. In such a case, residue-type identifications will help to remove the ambiguity. The features of the (4,3)DHN(C)NH spectrum described above also enable easy sequential backbone walk. Illustrative stretches of such walks through the (4,3)D-HN(C)NH spectra of unfolded Human SUMO are shown in Figure 4. A sequential (i + 1) correlation peak in one plane connects to the self (i) peak in the adjacent

sequential peaks for a particular residue have almost the same 15 N chemical shift [Δδ < 0.5 ppm]. In such a situation, the sequential peaks can be absent (because of opposite sign and higher intensity of self-peaks in these spectra) and assignment process may break or may be wrongly considered as the break point because of proline. However, in (4,3)D-HN(C)NH, the incorporation of the RD ideas circumvents some of these problems. The ambiguity arising because of degenerate 15N chemical shift can be resolved in the F1 (15N ± Δ1HN)− F3(1HN) planes. In the F1 (15N ± Δ1HN)-dimension, the chemical shifts are linear combinations of amide 1H and 15N chemical shifts; so, even if there is degeneracy in 15N chemical shifts, the sequential correlations can be resolved by amide 1H chemical shifts and vice versa (see Figure S-3, right panel, in the Supporting Information). Furthermore, a particular advantage 10408

dx.doi.org/10.1021/ac302656k | Anal. Chem. 2012, 84, 10404−10410

Analytical Chemistry

Article

methods, such as L-optimization28 or BEST techniques,29 and nonlinear sampling schemes.30 Taken together, we believe that the experiment presented here will provide an important analytical tool for proteomics, protein folding, and drug discovery research by NMR.

plane on the right. Note that the G77-Q78 panels in Figure 4, showing sequence-dependent special patterns, act as check points in the sequential walk and the absence of sequential peaks in the last strip proves the presence of proline in the (i + 1) position. Implications to a Drug Discovery Program. NMR screening methods for drug discovery and design can be divided into those that detect interactions by observation of either macromolecule NMR parameters or small-molecule NMR parameters. In the case of proteins (macromolecules) as targets, amide resonances are sensitive reporters of changes in their local chemical environments. Therefore, target-based NMR screening methods most commonly involve the acquisition of two-dimensional (2D) 1HN−15N correlation spectra.3,26,27 Binding sites of ligands can be located on the basis of chemical-shift changes induced by the complex formation, provided that 1HN and 15N resonance frequencies are assigned and the structure of the target protein is known. This has to be done thousands of times to find the best drug candidate. At times, chemical-shift changes induced by the complex formation are so large that the sequential assignment will have to be done ab initio. In particular, when high-affinity ligands are in a slow exchange with their receptor, there will be two sets of peaks and the identity of shifted resonances of the complex cannot be unraveled simply by titration. In such a case, the assignment protocol explained above is most amenable to re-establishing sequential assignments of target proteins perturbed by the binding of ligands. Here, we have demonstrated this by taking the example of largely perturbed calcium bound form (Holo form) of M-Crystallin whose structure has been solved earlier.37 The perturbation of amide chemical shifts upon calcium binding has been shown as an overlay of 15N-HSQC spectra in Figure 5A. The sequential connectivity between the HSQC peaks was established directly from the F1 (15N ± Δ1HN)−F3 (1HN) projection plane of the (4,3)D-HN(C)NH experiment, as described in Figure 5B. The set of two sequential correlations observed due to linear combination of amide (1H and 15N) chemical shifts helps in removing the ambiguities and provide rapid sequential correlation. On the other hand, F2 (15N)−F3 (1HN) projection planes (see Figures S-1B, S-1C, and S-1D in the Supporting Information) help in identification of the residue type, which can be mapped onto the protein primary sequence for sequence specific assignment. The complete assigned 15N-HSQC spectrum of M-Crystallin (Holo form) is shown in Figure S1A in the Supporting Information. We believe that such rapid analysis will be of great value in drug discovery programs, where it is necessary to identify the binding pockets of thousands of different candidates rather quickly.

■

ASSOCIATED CONTENT

S Supporting Information *

15

N-HSQC spectra of M-Crystallin (Holo form), showing complete assignment and residue-type identification of glycines, alanines, and serines/threonines. HSQC-type spectra for identification of glycines, alanines, and serines/threonines residues and a comparison of (4,3)D-HN(C)NH and 3DHN(C)N spectra recorded on unfolded Human SUMO were shown. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*Tel. (lab): +91-22-22782795 Tel. (off): +91-22-22782488. Fax: +91-22-22804610. E-mail: [email protected]. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS We thank Government of India for providing financial support to the National Facility for High Field NMR at Tata Institute of Fundamental Research, India. We would like to acknowledge Prof. K. V. R. Chary and A. L. Susmitha (Dept. of Chemical Sciences, TIFR, Mumbai) for providing 13C/15N-labeled MCrystallin protein.

■

REFERENCES

(1) Dyson, H. J.; Wright, P. E. Nat. Rev. Mol. Cell Biol. 2005, 6, 197. (2) Galea, C. A.; Wang, Y.; Sivakolundu, S. G.; Kriwacki, R. W. Biochemistry 2008, 47, 7598. (3) Shuker, S. B.; Hajduk, P. J.; Meadows, R. P.; Fesik, S. W. Science 1996, 274, 1531. (4) Frueh, D. P.; Sun, Z. Y.; Vosburg, D. A.; Walsh, C. T.; Hoch, J. C.; Wagner, G. J. Am. Chem. Soc. 2006, 128, 5757. (5) Frueh, D. P.; Arthanari, H.; Koglin, A.; Walsh, C. T.; Wagner, G. J. Am. Chem. Soc. 2009, 131, 12880. (6) Shirakawa, M.; Walchli, M.; Shimizu, M.; Kyogoku, Y. J. Biomol. NMR 1995, 5, 323. (7) Weisemann, R.; Ruterjans, H.; Bermel, W. J. Biomol. NMR 1993, 3, 113. (8) Ikegami, T.; Sato, S.; Walchli, M.; Kyogoku, Y.; Shirakawa, M. J. Magn. Reson. 1997, 124, 214. (9) Matsuo, H.; Kupce, E.; Li, H.; Wagner, G. J. Magn. Reson., Ser. B 1996, 111, 194. (10) Grzesiek, S.; Anglister, J.; Bax, A. J. Magn. Reson., Ser. B 1993, 101, 114. (11) Bracken, C.; Palmer, A. G., III; Cavanagh, J. J. Biomol. NMR 1997, 9, 94. (12) Chandra, K.; Jaipuria, G.; Shet, D.; Atreya, H. S. J. Biomol. NMR 2012, 52, 115. (13) Ohki, S. Y.; Kainosho, M. Prog. Nucl. Magn. Reson. Spectrosc. 2008, 53, 208. (14) Chatterjee, A.; Kumar, A.; Hosur, R. V. J. Magn. Reson. 2006, 181, 21. (15) Chugh, J.; Kumar, D.; Hosur, R. V. J. Biomol. NMR 2008, 40, 145. (16) Panchal, S. C.; Bhavesh, N. S.; Hosur, R. V. J. Biomol. NMR 2001, 20, 135. (17) Kumar, D.; Paul, S.; Hosur, R. V. J. Magn. Reson. 2010, 204, 111.

■

CONCLUSIONS In conclusion, we have developed an experiment for efficient and rapid sequential assignment in unfolded proteins and folded proteins alike. The new experiment, (4,3)D-HN(C)NH, provides a unidirectional sequential correlation, i → (i + 1), using both amide 1H and 15N chemical shifts. The sequentially connected residues can be extended to longer stretches, which can then be mapped uniquely onto the primary sequence based on their amino acid types identified from F2−F3 projection spectra. The experiment provides easy identification of self-and sequential peaks, since they have different phases, and the procedure can be readily automated. The experiment can be further enhanced by combining with other rapid data collection 10409

dx.doi.org/10.1021/ac302656k | Anal. Chem. 2012, 84, 10404−10410

Analytical Chemistry

Article

(18) Keller, R. The Computer Aided Resonance Assignment Tutorial; CANTINA, Verlag: Goldau, Switzerland, 2004. (19) Szyperski, T.; Wider, G.; Bushweller, J. H.; Wuthrich, K. J. Biomol. NMR 1993, 3, 127. (20) Szyperski, T.; Yeh, D. C.; Sukumaran, D. K.; Moseley, H. N.; Montelione, G. T. Proc. Natl. Acad. Sci. U.S.A. 2002, 99, 8009. (21) Kim, S.; Szyperski, T. J. Am. Chem. Soc. 2003, 125, 1385. (22) Szyperski, T.; Atreya, H. S. Magn. Reson. Chem. 2006, 44, S51. (23) Kumar, D.; Hosur, R. V. Curr. Sci. India 2010, 99, 1581. (24) Chugh, J.; Hosur, R. V. J. Magn. Reson. 2008, 194, 289. (25) Kumar, D.; Chugh, J.; Hosur, R. J. Chem. Sci. 2009, 121, 955. (26) Hajduk, P. J.; Dinges, J.; Miknis, G. F.; Merlock, M.; Middleton, T.; Kempf, D. J.; Egan, D. A.; Walter, K. A.; Robins, T. S.; Shuker, S. B.; Holzman, T. F.; Fesik, S. W. J. Med. Chem. 1997, 40, 3144. (27) Stockman, B. J.; Dalvit, C. Prog. Nucl. Magn. Reson. Spectrosc. 2002, 41, 187. (28) Pervushin, K.; Vogeli, B.; Eletsky, A. J. Am. Chem. Soc. 2002, 124, 12898. (29) Schanda, P.; Van Melckebeke, H.; Brutscher, B. J. Am. Chem. Soc. 2006, 128, 9042. (30) Atreya, H. S.; Szyperski, T. Methods Enzymol. 2005, 394, 78. (31) Shaka, A. J.; Keeler, J.; Frankiel, T.; Freeman, R. J. Magn. Reson. 1983, 52, 335. (32) Shaka, A. J.; Keeler, J.; Freeman, R. J. Magn. Reson. 1983, 53, 313. (33) Shaka, A. J.; Barker, P. B.; Freeman, R. J. Magn. Reson. 1985, 64, 547. (34) Emsley, L.; Bodenhausen, G. Chem. Phys. Lett. 1990, 165, 469. (35) Marion, D.; Ikura, M.; Tschudin, R.; Bax, A. J. Magn. Reson. 1989, 85, 393. (36) Piotto, M.; Saudek, V.; Sklenar, V. J. Biomol. NMR 1992, 2, 661. (37) Barnwal, R. P.; Jobby, M. K.; Devi, K. M.; Sharma, Y.; Chary, K. V. J. Mol. Biol. 2009, 386, 675.

10410

dx.doi.org/10.1021/ac302656k | Anal. Chem. 2012, 84, 10404−10410