NOESY water line

MAGNETIC RESONANCE IN CHEMISTRY Magn. Reson. Chem. 2007; 45: 925–928 Published online 18 September 2007 in Wiley InterScience (www.interscience.wile y.com) DOI: 10.1002/mrc.2068 Residual water suppression by indirect covariance NMR Yanbin Chen,1Fengli Zhang2and Rafael Br ¨uschweiler1,2∗ 1Department of Chemistry and Biochemistry, Flo rida State University, Tallahassee, FL 32306, USA 2National High Magnet Field Lab oratory, Florida State Univer sity, Tallahassee, FL 32310, USA Received 9 May 2007; Revised 20 July 2007; Accepted 25 July 2007 Residual water solvent signals in 2D NMR experim ents adversely affect appearance and subsequent analysis of spectra. A method for water suppression that is based on indirect covariance processing is described. It produces a symmetric spectrum with a water signal that is substantially decreased or completely absent. The method, which can be com bined with other water suppression schemes, is demonstrated for 2D TOCSY, NOESY, and ROESY spec tra of the protein, ubiquitin in aqueous solution. Copyright 2007 John Wiley & Sons, Ltd. KEYWORDS: water suppression; 2D NMR; indirect covariance NMR INTRODUCTION Multidimensional1H NMR spectra of molecules in aqueous solution require efficient suppression of the water signal. Since the full water signal is typically of the order of 103to 105times larger than the solute signal, it is reduced by one or the combination of several different solvent suppression schemes. Residual water is manifested in 2D FT spectra in the form of a vertical strip along ω1indicative of t1-noise along the water resonance (Fig. 1(a)) that may obscure cross peaks that are close to the water signal and thereby impede spectral analysis. By contrast, the horizontal waterline alongω 2is typically much less intense and does not cause major problems for data analysis. Some water suppression methods such as solvent pre- saturation, the WATERGATE sequence,1and excitation sculpting2are an integral part of the applied NMR pulse sequence, while others reduce the water signal at the dataprocessing stage. Processing methods include low-pass fil-ter deconvolution, 3symmetrization, principal component analysis,4singular value decomposition,5congruent matrix pencils,6and wavelet transform.7It is demonstrated here how water t1-noise signals of common 2D NMR spectra can be easily and efficiently removed by indirect covariance transform. METHOD Indirect covariance spectroscopy is a variant of covari-ance spectroscopy 8–1 0and was originally introduced for establishing13C–13C connectivities from 2D heteronuclear ŁCorrespondence to: Rafael Br ¨uschweiler, Department of Chemistry and Biochemistry, National High Magnetic Field Laboratory, Florida State University, Tallahassee, FL 32306, USA. E-mail: [email protected]–TOCSY experiments.11Blinov et al.i n t r o d u c e da s y m - metrical processing of indirect covariance spectra to elimi- nate potential artifacts.12,13The basic idea behind indirect covariance is the following. A 2D FT spectrum, represented by the real N 1ðN2matrix F,i sc o n v e r t e dt oas y m m e t r i c N1ðN1matrix CD⊿FÐFT/triangleleft1/2⊿1/triangleleft where superscript T refers to the matrix transpose and the square-root denotes the matrix square-root (for nondiago- nal matrices, the matrix square-root generally differs from the matrix that contains the square-root of the individual matrix elements). Note that matrix FÐFTh a sa si t se l e - ments the dot products between rows of Fand the indirect covariance transform of Eqn (1) has the effect that both frequency axes of spectrum Ccorrespond to the indirect frequency axis ω1of spectrum F.11Therefore, the spec- tral resolution of Calong both dimensions is determined by the spectral resolution along the indirect ( ω1)d i m e n - sion of F. By contrast, the direct covariance spectrum8–1 0 Cdirect D⊿FTÐF/triangleleft1/2has its resolution along both dimensions determined by the resolution of Falong the direct dimen- sionω2. Figure 1 compares a 2D FT NOESY spectrum of ubiquitin (Fig. 1(a)) with the direct covariance (Fig. 1(b)) and indirect covariance spectrum (Fig. 1(c)). The peaks in the boxed area illustrate the intrinsic power of indirect covariancespectroscopy to reconstruct cross peaks in the vicinity of the waterline. To further reduce the water signal, matrix Fcan be preprocessed by setting the columns of Fthat are strongly affected by the water signal to zero. The selected columns constitute a vertical strip, which is centered at the water signal and which covers the full spectral width along ω 1 and width H2Oalong ω2,w h e r e b y H2Ocan be adjusted if necessary. This modified 2D FT spectrum is then subjected to indirect covariance processing according to Eqn (1) resulting Copyright 2007 John Wiley & Sons, Ltd. 926 Y. Chen, F. Zhang and R. Br ¨uschweiler Figure 1. 2D NOESY spectrum of ubiquitin in 95% H 2O and 5% D 2Ou s i n gW A T E R G A T E1for solvent suppression: (a) 2D FT spectrum; (b) direct covariance spectrum; (c) indirect covarian ce spectrum; (d) indirect covari ance spectrum wit h water strip removed prior to indirect covariance processing. in the indirect covariance spectrum C0, which is essentially free of the water signal (Fig. 1(d)). The water streak in Falong ω1leads to an unwanted con- tribution to the dot product of any pair of rows in FÐFT.T h i s is the reason why removal of the water streak prior to appli-cation of Eqn (1) improves the quality of the spectrum. Asnoted above, the indirect covariance NMR spectrum depictsonly frequencies along both axes sampled during the t 1evo- lution period; therefore, removal of proton signals in Funder the water signal does not cause a major loss of spin correlationinformation in the indirect covariance spectrum C 0. MATERIALS AND EXPERIMENTS All NMR experiments were performed on a 0.7 m Msample of (unlabeled) ubiquitin in 95% H 2Oa n d5 %D 2Oa t pH 5.0 and carried out at 800 MHz NMR field strengthat room temperature. 2D TOCSY, NOESY, and ROESYNMR experiments were conducted with spectral widths of14.0 ppm, 11.0 ppm, and 10.2 ppm , respectively, along both dimensions. The 2D TOCSY experiment with 1024 t 2and 256 t1(complex) data points was collected using the MLEV-17 mixing sequence14with a 60 ms mixing time using excitation sculpting for water suppression.2The 2D NOESY experiment with 1024 t2and 512 t1(complex) data points had a 120 ms mixing time and it used the WATERGATE 3-9-19 pulsesequence element with gradients for water suppression. 1The 2D ROESY experiment with 1024 t2and 512 t1(complex) data points was collected with a 125 ms cw-spinlock pulse formixing and with presaturation for water suppression. States-TPPI was used for quadrature detection along the indirectdimension for all the three experiments. The time-domaindata were apodized using a cosine function, zero-filled tofinal size of 2048 ð2048 data points, 2D Fourier transformed, and polynomially baseline corrected along both dimensionsusing NMRPipe. 15All covariance processing was performed using in-house Matlab programs. For indirect covarianceprocessing Eqn (1) was either applied directly to the 2DFT spectrum, or after a strip along ω 1around the water resonance of width H2OD0.69 ppm (for TOCSY), 1.45 ppm (for NOESY), or 1.77 ppm (for ROESY) was set to zero. RESULTS AND DISCUSSION The method is demonstrated in Fig. 2 for all the threeexperiments. All panels show the region of the spectrumindicated by the dashed box in Fig. 1. The first columnfrom the left shows the spectral region obtained by 2D FTprocessing, the 2nd column shows the region calculatedby direct covariance processing, the 3rd column shows thespectral region obtained by indirect covariance processing,and the 4th column shows the spectral region obtained byindirect covariance processing after removal of the waterstrip. Copyright 2007 John Wiley & Sons, Ltd. Magn. Reson. Chem. 2007; 45: 925–928 DOI: 10.1002/mrc Residual water suppression by indirect covariance NMR 927 Figure 2. 2D spectral region, indicated by the boxed area in Fig. 1, for d ifferent types of experiments and different processing schemes. Comparison of the same spectral region of 2D FT sp ectrum (1st column from left), d irect covariance spectrum (2nd column), indirect covariance sp ectrum using Eqn (1) (3rd column), and indirect cova riance spectrum with wat er strip removed prior to application of Eqn (1) (4th column) of TOCSY (1st row), NOESY (2nd row) and ROESY (3rd row) experiments of ubiquitin. Positive contours are in blue and negative ones in red. The widths H2Oof the water strips that are zeroe d prior to indirect covariance processing are 0.69 ppm (for TOCSY), 1.45 ppm (for NOESY) , and 1.77 ppm (for ROESY). Selected cross peaks indicated by squares and letters belong to the following resonances (see BMRB entry 6816): (U) T55HB-HG2, (V) K6HA-HB2, (W) L71HA-HB2, (X) L67HA-HG, (Y) E16HA-HB3, (Z) P37HA-HB3. Depending on the experimental water suppression scheme, in both the 2D FT and the direct covariance spectrathe vicinity of the waterline is either largely void of crosspeaks or contaminated by strong t 1-noise. Indirect covariance processing (3rd column of Fig. 2) largely eliminates the water signal in the TOCSY and NOESY spectra, but fails to do so in the case of the solvent presaturated ROESY spectrum whosewaterline is significantly stronger (and broader) than for theother two spectra. Elimination of the vertical water strip inthe 2D FT spectrum prior to indirect covariance processing(4th column) improves the situation in all three spectra and provides reconstruction of a large number of cross peaks (assignments for selected cross peaks, labeled by letters, areg i v e ni nt h ec a p t i o nt oF i g .2 ) . The spectral regions further apart from the waterline are remarkably insensitive to the width of the zeroed waterstrip,  H2O. Since peak reconstruction in the strip area is based at least in part on the spectral information contained on the other side of the diagonal, H2Oshould be kept reasonably small. Generally, no useful spectral informationcan be gained within the square of width  H2Oaround the water diagonal peak of the indirect covariance spectrum. Forthe above examples,  H2Ovalues between 0.6 and 1.8 ppm provide good water suppression results. However, otherchoices for H2Oprovide equivalent results: the TOCSY spectrum (1st row) spectrum is essentially unaffected when H2Ois varied between 0 and 0.69 ppm; the NOESY spectrum (2nd row) does hardly change when H2Ovaries between 0.1 and 1.45 ppm; the ROESY spectrum (3rd row) is virtually the same when H2Ois varied between 1.16 and 1.77 ppm. This illustrates the robustness of the water suppression schemewith respect to the particular choice of the  H2Oparameter. While ‘vertical’ cutting (along ω1) prior to indirect covariance processing improves the spectral quality, ‘hor-izontal’ cutting (along ω 2) will only result in a cross- shaped zero-intensity region centered on the diagonal peak of the water resonance. Conversely, direct covari-ance spectra gain only little when the horizontal waterstrip is removed prior to covariance processing. Verti-cal cutting can not only be applied to the water strip,but also to other resonances that show strong t 1noise, such as methyl resonances, or regions that display severe baseline distortions in the 2D FT spectrum along ω1. This will reduce such artifacts in the associated indirectcovariance spectrum analogous to the suppression of sol-vent artifacts. The main drawback of indirect covariance is that its digital spectral resolution /ETB 1is restricted by the number Copyright 2007 John Wiley & Sons, Ltd. Magn. Reson. Chem. 2007; 45: 925–928 DOI: 10.1002/mrc 928 Y. Chen, F. Zhang and R. Br ¨uschweiler oft1increments N1,w h i c hi s /ETB 1D1/⊿N1Ðt1/triangleleftwhere t1is the time-increment in the evolution period t1. Because the total measurement time is inversely proportional to the spectral resolution /ETB 1, one typically sets N1< N2and therefore the spectral resolution of the indirect covariance spectrum is lower than the one of the standardcovariance spectrum, which maintains high-resolution withas little as N 1 D48 for a TOCSY experiment of a decapeptide.17Nonetheless, the ubiquitin TOCSY with only 256t1increments (Fig. 2 (1st row)) produces an indirect covariance spectrum that has reasonably good quality toallow spectral analysis inclu ding peak assignments. Like direct covariance NMR, the indirect covariance methodpresented here can be applied to most homonuclear 2Dexperiments, including 2QF-COSY, but not to COSY itself because the dispersive diagonal peaks adversely affect the matrix product in Eqn (1). CONCLUSION Indirect covariance spectroscopy is remarkably resilient towater artifacts. In addition, after deletion of the characteristicwater strip of 2D FT spectra prior to indirect covarianceprocessing, the water signal is essentially entirely removed.The procedure, which does not require iterative fitting orelaborate parameter adjustments, produces symmetric 2Dspectra whose spectral resolution is determined by the sampling scheme during the e volution period. Indirect covariance spectra are best suited for cross-peak analysisclose to the water line, whereas direct covariance spectraare preferable for all other spectral regions. Fully symmetric2D spectra that are free of solvent artifacts are particularlysuitable for automatic spectral analysis encountered in high-throughput applications, including applications in proteomics and metabolomics. Acknowledgements Useful discussions with Dr. Vladimir Sklenar are acknowledged. This work was supported by the National Institutes of Health (grant R01 GM 066041). REFERENCES 1. Piotto M, Saudek V, Sklenar V. J. Biomol. NMR. 1992; 2: 661. 2. Hwang TL, Shaka AJ. J. Magn. Reson. A 1995; 112: 275. 3. Marion D, Ikura M, Bax A. J Magn Reson . 1989; 84: 425. 4. Mitschang L, Cieslar C, Holak TA, Oschkinat H. J Magn. Reson. 1991; 92: 208. 5. Zhu G, Smith D, Hua YB. J. Magn. Reson. 1997; 124: 286. 6. Stadlthanner K, Tome AM, Theis FJ, Lang EW, Gronwald W, Kalbitzer HR. Neurocomputing 2006; 69: 497. 7. Gunther UL, Ludwig C, Ruterjans H. J. Magn. Reson. 2002; 156: 19. 8. Br ¨uschweiler R. J. Chem. Phys. 2004; 121: 409. 9. Br ¨uschweiler R, Zhang F. J. Chem. Phys. 2004; 120: 5253. 10. Trbovic N, Smirnov S, Zhang F, Br ¨uschweiler R. J. Magn. Reson. 2004; 171: 277. 11. Zhang F, Br ¨uschweiler R. J. Am. Chem. Soc. 2004; 126: 13180. 12. Blinov KA, Larin NI, Kvasha MP, Moser A, Williams AJ, Martin GE. Magn. Reson. Chem. 2005; 43: 999. 13. Blinov KA, Larin NI, Williams AJ, Zell M, Martin GE. Magn. Reson. Chem. 2006; 44: 107. 14. Bax A, Davis DG. J. Magn. Reson. 1985; 65: 355. 15. Delaglio F, Grzesiek S, Vuister GW, Zhu G, Pfeifer J, Bax A. J. Biomol. NMR 1995; 6: 277. 16. Weber PL, Brown SC, Mueller L. Biochemistry 1987; 26: 7282. 17. Chen Y, Zhang F, Bermel W, Br ¨uschweiler R. J. Am. Chem. Soc. 2006; 128: 15564. Copyright 2007 John Wiley & Sons, Ltd. Magn. Reson. Chem. 2007; 45: 925–928 DOI: 10.1002/mrc