Andrzej M. Krezel

Direct cross-relaxation NOESY (D.NOESY). A method for removing spin-diffusion cross peaks from two-dimensional NOE spectra of macromolecules

Spin-diffusion effects compromise the interpretation of nuclear Overhauser effects in 1 D ( 1, 2) and 2D (3-5) spectroscopy. Problems arise when multistep (two or more) magnetization exchange occurs during the mixing time. The geometrical arrangement of protons in macromolecules provides many such pathways, which lead (i) to the appearance of multistep cross-relaxation cross peaks from pairs of spins that are not near neighbors and (ii) to the reduction of volumes of the direct cross-relaxation peaks. A further complication arises when cross relaxation is mediated by local mobility of parts of the molecule; such effects, although important ( 6)) are not addressed here. Spin-diffusion effects can be particularly pronounced when geminal protons are involved in the pathway. Geminal protons ( j, k), which are separated by a short fixed distance (about 1.8 A), are found to exhibit cross-relaxation rates from 2 to 100 times faster than those of other pairs of protons (k, I) located in the same molecule but separated by 2-4 A (7 ) . Consider a linear chain (j . * k. . . I) with direct cross-relaxation rates uk( and aj/. Misinterpretation of two-step spin diffusion ( ajk * uki) as direct cross relaxation u,/ may lead to underestimation of the distance r,/ by as much as 30% (when (Tk/’ 7, = 0.1 and ffjk/ok/ = 100). With longer mixing times, magnetization transfer may involve additional steps and lead to an even larger underestimation of the distance. One normally attempts to include distances derived from as many cross peaks as possible as input data for algorithms used in structure calculations (e.g., distance geometry). This is because a larger number of distance constraints leads to a tighter convergence of the calculated structures. The problem is that many of the weaker peaks do not represent direct cross relaxation and some direct peaks may have perturbed intensities. The danger is that the erroneous distances, derived from these cross peaks, will contribute to a set of convergent three-dimensional structures which, although consistent with the input distances, would not model the molecule accurately.

An evaluation of least-squares fits to COSY spectra as a means of estimating proton—proton coupling constants II. Applications to polypeptides

SummaryA new computational method for simultaneously estimating all the proton-proton coupling constants in a molecule from COSY spectra [Yang, J.-X. and Havel, T.F. (1994) J. Biomol. NMR, 4, 807–826] is applied to experimental data from two polypeptides. The first of these is a cyclic hexapeptide denoted as VDA (-d-Ala1-Phe2-Trp3-Lys(Z)4-Val5-Phe6-), in deuterated DMSO, while the second is a 39-residue protein, called decorsin, in aqueous solution. The effect of different data processing strategies and different initial parameter values on the accuracy of the coupling constants was explored. In the case of VDA, most of the coupling constants did not depend strongly on the initial values chosen for the optimization or on how the data were processed. This, together with our previous experience using simulated data, implies strongly that these values are accurate estimates of the coupling constants. They also differ by an average of only 0.36 Hz from the values of the 14 coupling constants that could be measured independently by established methods. In the case of decorsin, many of the coupling constants exhibited a moderate dependence on their initial values and a strong dependence on how the data were processed. With the most successful data processing strategy, the amide-α coupling constants differed by an average of 1.11 Hz from the 21 values that could be measured by established methods, while two thirds of the three-bond coupling constants fell within 1.0 Hz of the ranges obtained by applying the Karplus relation to an independently computed ensemble of distance geometry structures. The averages of the coupling constants over multiple optimizations using random initial values were computed in order to obtain the best possible estimates of the coupling constants. Most clearly incorrect averages can be identified by large standard deviations in the coupling constants or the associated line widths and chemical shifts, and can be explained by strong coupling and/or overlap with the water signal, the diagonal peaks or other cross peaks.

Computational Aspects of Multinuclear NMR Spectroscopy of Proteins at NMRFAM

In recent studies, we have explored various strategies for using isotope labeling in conjunction with 1D, 2D, and 3D NMR spectroscopy to obtain information about how proteins work. With larger proteins, a concerted approach to the assignment of protein spectra that makes use of all available information about through-bond and through-space connectivities appears to provide the most reliable results. We summarize here various ways in which we have been using computers to process and analyze data and to derive information about protein structure and dynamics. We also describe in brief our protein NMR database project.

NMR Studies of Proteins Involved in Cell Adhesion Processes

Many biological processes on the molecular level are associated with membrane-bound receptors or other integral membrane proteins. We have been working with proteins that are domains of receptors, or inhibit their function. The first topic are the T-cell surface glycoprotein receptor CD2 and its counter receptor CD58. The other topic is about antagonists of the integrin adhesion receptor glycoprotein IIbIIIa (GPIIbIIIa) which is found on platelet surfaces. Human CD2 is a glycoprotein, and the carbohydrate of its adhesion domain is crucial for adhesion function. The platelet receptor GPIIbIIIa, a Ca2+ dependent heterodimeric glycoprotein from the integrin family, binds fibrinogen and mediates the aggregation of platelets to form a blood clot. Natural protein antagonists of this receptor have primarily been found in the venum of various snakes, which have been termed disintegrins, and in the saliva of blood-sucking leeches. They contain an Arg-Gly-Asp (RGD) sequence in their active site. Due to the potent antiplatelet effect of these RGD proteins, the structures of their active sites have been of considerable interest for the design of antithromotic drugs.