In this paper, we propose a method to extract unambiguous experiments (19, 20) have been proposed to resolve ambigu-NOE distance restraints from new 3D doubly filtered ities due to overlap in one or both proton dimensions. Further-NOESY experiments using automated assignment routines. more, ambiguities not solvable at the spectral level may be With the ability of preparing 15 N/ 13 C-labeled protein sameliminated during the iterative refinement procedure, on the ples and the development of triple-resonance experiments basis of inconsistencies with a preliminary structure. For this (1-3) and 13 C-TOCSY transfer experiments (4-6), NMR purpose, an initial structure, even ill-defined, is essential, and spectroscopy has become a method of choice for the study in the absence of structural homology with a known protein, of proteins of molecular weights up to 30 kDa. As far as this structure must be determined from an initial set of longproteins larger than 120 residues are concerned, these experirange NOE-derived distance constraints. Often, only the doubly ments have led to resonance assignments for numerous filtered 4D versions provide the unambiguous identification of cases, but, so far, to only few solution structures (7, 8). The a sufficient number of cross peaks. However, this is at the NMR data do not directly represent the protein structure (as expense of (i) an important loss of sensitivity and (ii) a coarse in the case of X-ray crystallography density maps), so that digital resolution: (i) The sensitivity drops by a factor 2 with the 3D structure determination involves a step of molecular each indirect dimension (for amplitude modulation of the demodeling using distance geometry or restrained molecular tected signal) and additional sensitivity is lost during each dynamics. A large number of significant NMR constraints is heteronuclear transfer step because of relaxation effects. (ii) necessary for the convergence to a well-defined 3D structure. Even with the use of pulsed field gradients to replace phase Despite many recent attempts to make use of spin-coupling cycling for coherence-transfer-pathway selection, the recording (9, 10) or chemical-shift information (11)(12)(13)(14), the main of a 4D experiment in a reasonable time (a few days) requires part of this structural information is still obtained from NOEextensive folding or the limitation to few time points per dimenderived 1 H-1 H distances.
sion. In both cases, the result is poor spectral resolution. The use of a computer for automated assignment and Here, we will show that new 3D doubly filtered NOESY quantification of the NOEs is very appealing as the number experiments yield the same information as the corresponding of cross peaks is in the range of several thousands for larger 4D versions with at least equal sensitivity and higher digital proteins, and the procedure of identifying potential assignresolution. In these experiments, the two heteronuclear frements should be straightforward. This requires an especially quencies are labeled together in one indirect dimension as high accuracy on the peak positions which can only be obpreviously described for other applications (21-24). The tained on data with sufficient signal-to-noise ratio and specpulse sequences for the 13 C/ 15 N-(insets A / B) and 15 N/ tral resolution. Therefore, the use of a high sample concen-15 N-filtered (insets A / C) experiments are shown in Fig. tration and the most sensitive pulse schemes are recom-1. A 3D 13 C/ 13 C-filtered NOESY experiment could be easily mended.
derived from (A / B) by replacing the final 1 H-15 N correla-For 15 N/ 13 C-labeled proteins, the 1 H-1 H NOE cross peaks tion step by a 1 H-13 C correlation (HSQC or HMQC). The which extensively overlap in 2D NOESY spectra can be dis-13 C/ 15 N-filtered experiment proves to be the most helpful persed in additional dimensions, as a function of the chemical and further discussion will be based on this case. shift of the directly bound heteronucleus (X, Y ), using HSQC In contrast to previously described applications of the re-(15) or HMQC (16) transfer steps. Three-dimensional 1 Hduced dimensionality approach, quadrature detection is ap-NOESY-( 1 H, X ) (17, 18) or 4D ( 1 H, X )-NOESY-( 1 H, Y ) plied to both heteronuclei to allow the separation of the sum and difference frequencies (v 2 Γ
v N { v C /l) in two different spectra. Therefore, the carriers (demodulation frequen-* To whom correspondence should be addressed.