Single-step determination of protein substructures using dipolar couplings: aid to structural genomics.

The wealth of genomic data that has recently become available with completion of the sequencing of both the human and a variety of other genomes1 has created a need for rapid and efficient determination of three-dimensional structures of the corresponding proteins. So far, most effort in this so-called structural genomics project has focused on X-ray crystallography, but NMR also shows considerable potential.2,3 Conventionally, NMR structure determination consists of a resonance assignment phase, which usually relies on analysis of an extensive set of triple resonance J-connectivity data, followed by a structure determination phase that relies on interpretation of NOE spectra and measurement of scalar and dipolar couplings.3,4 Here, we describe and demonstrate a more integrated approach, where the assignment and structural data are derived from the same experiment. This offers the opportunity to greatly accelerate the NMR structure-determination process. The method demonstrated here relies on the 3D (HA)CANH triple resonance experiment,5 which correlates amide 1H and 15N chemical shifts with those of the intraresidue and preceding 13CR nuclei. Because the 13CR chemical shift is usually insufficiently unique, such a spectrum alone cannot be used for determining complete sequential assignments. Therefore, separate experiments that correlate the amide with intraresidue and preceding residue 13C′ or 13Câ resonances are frequently used to complete the assignment process. Here, we demonstrate that if the (HA)CANH spectrum is recorded in the 1HR-coupled mode, both the size of the JCRHR splitting and the asymmetry in doublet intensity6 resulting from relaxation interference between 13CR chemical shift anisotropy (CSA) and 13CR-1HR dipolar coupling can be used to resolve the ambiguities caused by the non-uniqueness of the 13CR chemical shift. When such an experiment is conducted in both an isotropic and a liquid crystalline medium, the resulting spectra not only yield complete assignment for the backbone 1HN, 15N, and 13CR resonances, but also contain important structural information in the form of 13CR-1HR dipolar coupling, 13CR CSA, and deviations from random-coil 13CR isotropic chemical shifts. Together, this provides sufficient information to obtain complete backbone 1HN, 15N, and 13CR assignments of small and mediumsized proteins, such as ubiquitin and calmodulin, and to determine the 3D structure of fragments of such proteins. To avoid increased crowding in the HR-coupled (HA)CANH spectrum relative to the regular (HA)CANH spectrum, the two 13CR-{1HR} doublet components are separated into two separate spectra in the usual manner7 by calculating the sum and difference of an in-phase and an anti-phase 13CR-{1HR} (HA)CANH spectrum. The pulse sequence for this so-called IPAP-(HA)CANH experiment is available as Supporting Information. The 13CR{1HR} coupling remains active during the 28-ms 13CR constanttime evolution period, resulting in two highly resolved 13CR doublet components that are split by the 13CR-{1HR} coupling and have an intensity ratio that depends on the 13CR CSA. Correlations from a given 13CR-{1HR} atom pair to the intraresidue and sequential amide not only will exhibit identical 13CR shifts, but also the same 13CR-{1HR} splitting and the same 13CR CSAinduced intensity ratio of the two 13CR-{1HR} doublet components. This greatly alleviates the problem of 13CR chemical shift degeneracy The method has been applied to samples of U-13C/15N ubiquitin (3 mM), pH 6.5, 25 °C and U-13C/15N C-terminal Ca2+-ligated calmodulin (1 mM), pH 7, 25 °C, 100 mM NaCl, without and with 15.5 mg/mL Pf1 bacteriophage.8 For the isotropic 3 mM ubiquitin sample perfect conditions are available, and the backbone can completely and fully automatically be assigned using the IPAP-(HA)CANH experiment, even in the absence of residual alignment. Owing to the high 13CR resolution and the high signal-to-noise ratio, the isotropic 13CR-{1HR} splitting and the 13CR CSA are sufficient for resolving any 13CR chemical shift ambiguities (data not shown). For less concentrated samples of R-helical proteins, the accuracy at which the small variation in vicinal 13CR-{1HR} splitting can be measured is generally insufficient to resolve 13CR chemical shift degeneracy. For example, for the 1 mM calmodulin sample the 13CR-{1HR} splitting measured from the intraresidue correlation differs by a root-mean-square (rms) value of 2.4 Hz from the same splitting measured through the cross-peak to the sequential amide (this rmsd is only 0.4 Hz for the ubiquitin sample). Similarly, for isotropic calmodulin the rms difference in the reduced CSA (CSAred),6 as measured from the intraresidue and sequential amides, becomes rather large (11.3 ppm vs 5.6 ppm for ubiquitin). For aligned calmodulin, the signal-to-noise ratio is even lower, and the pairwise rms differences between the intraresidue and sequential measurements of the 13CR-{1HR} splitting and CSAred increase to 3.2 Hz and 14.8 ppm, respectively. Note that these high rmsd’s primarily reflect the large uncertainty in the weak Ci-Hi+1 correlation. (The uncertainty in the CSAred and JCH values, as measured for the intraresidue correlation, are much lower.) However, the introduction of a liquid crystalline medium, also increases the range of 13CR-{1HR} splittings from a few Hz to (50 Hz (Figure 1). This greatly increased variation in splitting, together with the remaining dispersion in CSAred, enables long contiguous stretches of residues to be assigned. Common assignment approaches rely heavily on 13Câ shifts for positioning assigned fragments within the primary sequence. With the present approach, secondary structure is already available for many residues (see below) and CR chemical shifts alone are then quite characteristic of the amino acid type.9 Combination of this information with uniquely identified glycines (from the CR chemical shift and the 13CR-{1HR} splitting) relaxes the need for 13Câ chemical shifts when positioning assigned fragments within the primary sequence. This made it possible to completely assign the backbone of the 148-residue, R-helical protein calmodulin using only the IPAP-(HA)CANH data. In addition to backbone assignment, the IPAP-(HA)CANH data can be used to define the secondary structure. First, R-helices (1) Venter, J. C.; Adams, M. D.; Myers, E. W.; Li, P. W.; Mural, R. J.; et al. Science 2001, 291, 1304-1351. Lander, E. S.; Linton, L. M.; Birren, B.; Nusbaum, C.; Zody, M. C.; et al. Nature 2001, 409, 860-921. (2) Abbott, A. Nature 2000, 408, 130-132. (3) Moseley, H. N. B.; Montelione, G. T. Curr. Opin. Struct. Biol. 1999, 9, 635-642. Fowler, C. A.; Tian, F.; Al-Hashimi, H. M.; Prestegard, J. H. J. Mol. Biol. 2000, 304, 447-460. Montelione, G. T.; Zheng, D.; Huang, Y. J.; Gunsalus, K. C.; Szyperski, T. Nature Struct. Biol. 2000, 7, Suppl., 982985. (4) Tjandra, N.; Bax, A. Science 1997, 278, 1111-1114. (5) Montelione, G. T.; Wagner, G. J. Magn. Reson. 1990, 87, 183-188. Kay, L. E.; Ikura, M.; Bax, A. J. Magn. Reson. 1991, 91, 84-92. (6) Tjandra, N.; Bax, A. J. Am. Chem. Soc. 1997, 119, 9576-9577. (7) Yang, D. W.; Nagayama, K. J. Magn. Reson., Ser. A 1996, 118, 117121. Ottiger, M.; Delaglio, F.; Bax, A. J. Magn. Reson. 1998, 131, 373-378. (8) Hansen, M. R.; Mueller, L.; Pardi, A. Nat. Struct. Biol. 1998, 5, 10651074. Zweckstetter, M.; Bax, A. J. Biomol. NMR 2001, 20, 365-377. (9) Spera, S.; Bax, A. J. Am. Chem. Soc. 1991, 113, 5490-5492. Grzesiek, S.; Bax, A. J. Biomol. NMR 1993, 3, 185-204. 9490 J. Am. Chem. Soc. 2001, 123, 9490-9491