Removal of Dispersive Baseline Distortions Caused by Strong Water Signals

A baseline correction routine that eliminates dispersive tails extending from the water tr noise band in 2D NMR spectra has been developed. Streaks along the w2 axis (Fig. la) spreading from the w2 position of the water resonance are among the most perturbing artifacts in 2D NMR spectra where the water signal is reduced by presaturation. In our experience, these streaks have dispersive lineshapes and cannot be eliminated with conventional baseplane correction routines, such as polynomial fits. They also add significantly to the apparent noise in the entire 2D spectra. To properly correct for these distortions, an w;’ term has been added to the polynomial baseline correction routine of our 2D processing software (FTNMR, Hare Research, Inc.). This term estimates the amount of the dispersive component. The actual correction baseplane is constructed as a sum of a constant offset, a linear term, and a dispersive component where the weight of the latter is calculated from the contribution of the w;’ term. With this procedure we have achieved major improvements of the quality of the spectra and of the apparent signal-to-noise in 2D protein spectra in water recorded with presaturation. Presaturation of the water signal was the first method to record 2D NMR spectra in H20 (1, 2), and it is still the most common technique for recording 2D NMR spectra of proteins in H20. During the recycle delay between accumulations, a long selective pulse is applied to the water resonance. Usually, the transmitter and the decoupler are set to the water frequency in order to minimize distortions in 2D spectra. Phase coherence of transmitter and decoupler is important for good water suppression. Nevertheless, this technique leaves residual artifacts. An example is shown in Figs. 1 and 2. In the 1 D spectra of a 5 mM solution of the protein kistrin (3) collected with this technique, the water peak was roughly 50 times higher than the largest methyl signals. A NOESY spectrum acquired under these conditions has large ridges parallel to w2 (Fig. la). As shown in Fig. 2, these ridges are due to large dispersive artifactual signals centered at the water frequency. A conventional polynomial baseline correction cannot accurately reproduce a dispersive lineshape. Also, this technique is very sensitive to the choice of reference points for the polynomial fit. Since the reference points for the polynomial fit can only be used from regions of the spectrum where there are no resonances, this procedure is not very practical for crowded protein spectra. An alternative procedure has been