- Email: [email protected]
Nuclear Magnetic Resonance Methods for Elucidation of Structure and Dynamics in Disordered States
258
PROTEINS
[ 12]
made it possible to use NMR spectroscopy to study a protein beyond 100 kDa molecular mass. 82 By the end of this year, the sequencing of the human genome will be largely completed and already there are numerous novel gene products awaiting characterization. Determining every possible structural motif within the human genome could increase our understanding of protein function and evolution since frequently proteins lacking sequence homology are structurally homologous. With this information, a map of the protein-protein interactions that occur within a human will be vital for a complete understanding of biological metabolism. The methods outlined in this review can be used to study protein-protein complexes and the effect of different proteins on preexisting complexes.
82 M. Salzmann, K. Pervushin, G. Wider, H. Senn, and K. Wiithrich, J. Biomol. NMR 14, 85 (1999).
[12] Nuclear Magnetic Resonance Methods for Elucidation of Structure a n d D y n a m i c s in Disordered States By H. JANE DYSON and PETER E. WRIGHT Functional Relevance of Disordered States Nuclear magnetic resonance (NMR) is unsurpassed in its ability to provide detailed information on the structure and dynamics of unfolded and partially folded states of proteins. Nonnative states of proteins do not adopt unique threedimensional structures in solution but fluctuate rapidly over an ensemble of conformations. Structural characterization of such states is of great interest because of their importance in protein folding and because of the growing awareness that unstructured proteins play a critical role in many cellular processes. Determination of the structure and dynamics of protein folding intermediates and denatured states of proteins is of central importance for a detailed understanding of protein folding mechanisms. In addition, it is now recognized that many proteins and protein domains are only partially structured or are unstructured under physiological conditionsJ ,2 Characterization of the conformational propensities
1E E. Wright and H. J. Dyson, J. Mol. Biol. 293, 321 (1999). 2 E. Garner, E Cannon, E Romero, Z. Obradovic, and A. K. Dunker, Genome Informatics 9, 201 (1998).
METHODSIN ENZYMOLOGY,VOL.339
Copyright© 2001by AcademicPress Allfightsof reproductionin any formreserved. 0076-6879100$35.00
[12]
NMR OF DISORDEREDSTATES
259
and function of these nonglobular protein sequences represents a major challenge but is essential to a proper understanding of their biological functions and interactions. Recent advances in NMR technology, including the advent of ultrahigh field instruments and the application of heteronuclear multidimensional experiments, have allowed backbone resonances of unfolded and partly folded proteins to be fully assigned and have led to unprecedented insights into the structure and dynamics of these states. In this article, we review that available methods for NMR characterization of disordered proteins. Resonance Assignments NMR characterization of disordered states of proteins presents special challenges because the polypepfide chain in such states is inherently flexible and rapidly interconverts between multiple conformations. Consequently, the chemical shift dispersion of most resonances, especially protons, is poor and sequence-specific assignment of resonances is difficult. A major advance in the characterization of unfolded states came with the introduction of methods for uniform isotope labeling with 13C and 15N and the development of multidimensional triple resonance NMR methods. The backbone 15N and 13C' (carbonyl) resonances are influenced both by residue type and by the local amino acid sequence and therefore remain well-dispersed, even in fully unfolded states. 3,4 Representative IH-15N HSQC and H(N)CO spectra of acid unfolded apomyoglobin are shown in Fig. 1 and demonstrate the superior dispersion available in the 15N and 13CO chemical shifts. When exchange between folded and unfolded states occurs on an appropriate time scale, at least backbone resonance assignments can often be made using homonuclear (IH) magnetization transfer methods to correlate resonances in the spectrum of the folded protein with the corresponding resonances in the denatured state. 5-7 Three-D triple resonance experiments to establish sequential connecfivities via the well-resolved 15N and 13C' resonances in uniformly 15N,13C-labeled proteins provide a particularly robust method for obtaining unambiguous resonance assignments. 8-I° Pulse sequences that are appropriate for this purpose are summarized in Table I. An assignment strategy that utilizes the superior chemical shift dispersion of the 13CO resonance in unfolded proteins is especially valuable, 3 D. Braun, G. Wider, and K. Wiithrich, J. Am. Chem. Soc. 116, 8466 (1994). 4 O. Zhang, J. D. Forman-Kay, D. Shortle, and L. E. Kay, J. Biomol. NMR 9, 181 (1997). 5 p. A. Evans, K. D. Topping, D. N. Woolfson, and C. M. Dobson, Proteins 9, 248 (1991). 6 D. Neri, G. Wider, and K. Wiithrich, Proc. Natl. Acad. Sci. USA 89, 4397 (1992). 70. Zhang, L. E. Kay, J. P. Olivier, and J. D. Forman-Kay, J. Biomol. NMR 4, 845 (1994). 8 T. M. Logan, Y. Thrrianlt, and S. W. Fesik, J. Mol. Biol. 236, 637 (1994). 9 A. T. Alexandrescu, C. Abeygunawardana, and D. Shortle, Biochemistry 33, 1063 (1994). 10 D. Eliezer, J. Yao, H. J. Dyson, and P. E. Wright, Nature Struct. Biol. 5, 148 (1998).
260
PROTEINS
[ 12]
o
0
.
O
O0 ° 0(:~
.d
oO°o
0 o
D O
0
E
8' o
E
8 0
o
o a'
0 0
lip
o
0
g
o
c~°°®
a.~
~
0
~. 0
O
gc~ o
T-
z
00 Q
°
~
~0 O
~?~oo
e~ o
* o~$%-
o Q,
O
8.1
1H (ppm)
o
o
°~o o
~ o o
6
e (~' o o oo °0 8'.4
00
•
O
8.7
_¢6 b.
~ o
oO ,r-
o o
/.8
7.s
a.7
;.4
o
;.1
1N (ppm)
FIG. 1. NMR spectra (750 MHz) of 15N,t3C-labeled apomyoglobin, pH 2.3, 10 mM acetate-d6, 5 °. (A) 15N-1H HSQC spectrum. (B) HNCO spectrum. (Reproduced from Ref. 11 with permission.)
given the excellent resolution and long T2 relaxation time of these resonances.ll A set of high resolution constant-time triple resonance experiments that transfer magnetization sequentially along the amino acid sequence using carbonyl ]3C homonuclear isotropic mixing have been developed specifically for assignment of unfolded proteins. 12 These experiments have the advantage that the slow ~3CO relaxation rates enable correlations to be established across proline residues. One of the advantages of working with unfolded proteins is that the intrinsic flexibility of the polypeptide generally causes the resonances to be much narrower than they would be in a folded protein of comparable molecular weight, i.e., T2 is longer than for the natively folded protein. This characteristic gives special advantages in the implementation of pulse sequences such as (HCA)CO(CA)NH, in which there are a large number of delays. For folded proteins of molecular weight greater then 1 5 - 2 0 kDa, this experiment is often rather insensitive because of decay of magnetization during the pulse sequence. For unfolded proteins, however, since 11 j. Yao, H. J. Dyson, and E E. Wright, FEBSLett. 419, 285 (1997). n2A. Liu, R. Riek, G. Wider, C. Von Schroetter, R. Zahn, and K. Wiithrich, J. Biomol. N M R 16, 127
(2OOO).
[ 12]
NMR OF DISORDEREDSTATES
261
TABLE I EXPERIMENTSDESIGNEDTO OBSERVESCALARCONNECTWITIES Experiment 1H-lSN HSQC HNCACB CBCA(CO)NH HNCO (HCA)CO(CA)NH HNCA TOCSY-HSQC C(CO)NH-TOCSY (H)N(CO-TOCSY)NH (H)CA(CO-TOCSY)NH (H)CBCA(CO-TOCSY)NH
Connectivity 15N~1Hi (lSN-1H)i-(13CetJ3Cfl)i
( 15N-1H)~(13Cot-13C/~)i_1 (15N-1H)i-13COi_I
Refs.a 1 2,3 3,4 5
(15N-1a)i-13COi,13COi_l
6
(15N-IH)i-13foq_l (15N-IH)i-Ha~(H/~,H~/...)i
5 1
(15N-IH)~Cai_I-(Cfl,Cy...)i_I
7
15N~NHi,NHi+I, NHi+2 C%-NHi,NHi+I, NHi+2 C~/,Ca~NHi, NHi+I, NHI+2
8 8 8
Key to References: (1) O. Zhang, L. E. Kay, J. E Olivier, and J. D. Forman-Kay, J. Biomol. NMR 4, 845 (1994); (2) M. Wittekind and L. Mueller, J. Magn. Reson. 101, 201 (1993); (3) D. R. Muhandiram and L. E. Kay, J. Magn. Reson. Series B 103, 203 (1994); (4) S. Grzesiek and A. Bax, J. Am. Chem. Soc. 114, 6291 (1992); (5) S. Grzesiek and A. Bax, J. Magn. Reson. 96, 432 (1992); (6) E L6hr and H. Rtiterjans, J. Biomol. NMR 6, 189 (1995); (7) S. Grzesiek, J. Anglister, and A. Bax, J. Magn. Reson, Series B 101, 114 (1993);
(8) A. Liu, R. Riek, G. Wider, C. Von Schroetter, R. Zahn, and K. Wiithrich, J. Biomol. NMR 16, 127 (2000).
T2is significantly longer, the (HCA)CO(CA)NH experiment is extremely sensitive and can be utilized with great success. 1° The same is, unhappily, not true for partly folded proteins, or molten globule states; the NMR spectra of these species often contain a mixture of sharp (from unstructured regions) and broad resonances (from the structured subdomains). Indeed, many molten globules are extremely difficult to study by NMR because their lines are exceptionally broad as a result of exchange processes on intermediate time scales. For example, the molten globule state of a-lactalbumin gives rise to extremely broad resonances that largely preclude direct high-resolution NMR analysis.13,14 High-resolution NMR studies of unfolded proteins have been largely enabled by the advent of high-field spectrometers and multidimensional heteronuclear NMR experiments. High sensitivity is a critical factor in the study of unfolded and partly folded proteins, since NMR experiments must often be performed at very low concentration (of the order of 50-300/zM) to prevent aggregation. 13j. Baum, C. M. Dobson, E A. Evans, and C. Hanley, Biochemistry 28, 7 (1989). 14 S. Kim, C. Bracken, and J. Baum, J. Mol. Biol. 294, 551 (1999).
262
PROTEINS
[ 12]
NMR Parameters Once backbone resonance assignments have been made, a number of NMR parameters can be used to characterize residual structure in unfolded and partly folded states. However, in interpreting the NMR data, it is important to bear in mind that unfolded and partly folded states of proteins are highly dynamic and that all parameters are a population-weighted average over all structures in the conformational ensemble. Conformational preferences are always identified, therefore, by comparison of experimental NMR parameters to those expected for a random coil state, in which the polypeptide backbone dihedral angles adopt a Boltzmann distribution over the ~b, ~ energy surface.15' 16 In addition, it is implicitly assumed that the dominant conformers in unfolded or partly folded polypeptides will have backbone dihedral angles that lie within the broad ot and/3 minima on the q~, 7z conformational energy surface.
Chemical Shifts The deviations of chemical shifts from random coil values, especially for X3C~, 13C~, and 1I-I~,provide a convenient and sensitive probe of secondary structural propensities.17,18 The 13CO chemical shifts also provide information on secondary structure content and dihedral angle preferences, provided they are corrected for sequence dependence. 18a Theoretical calculations 19 show that the 13Ca and 13Cfl chemical shifts are determined primarily by the backbone ~, ~ dihedral angles, and empirical correlations confirm that these resonances are reliable indicators of secondary structure. 17,18,20 In ot helices, the 13C~ resonances are shifted downfield by an average of 3.1 + 1.0 ppm. 17 The 13CO resonances are also shifted downfield, while the 13C~ and 1H~ resonances are shifted upfield from their positions in random coil states. In/3 sheets, the 13Ct~resonance is shifted upfield (AS = - 1.5 + 1.2 ppm 17) with respect to the random coil chemical shift, while the lt-Ia and tac~ resonances are shifted downfield. Rapid conformational averaging between the ct and/3 regions of ~b, ~p space reduces the secondary structure shifts below these extreme values in unfolded or partly folded proteins. Indeed, the deviations of the chemical shifts from random coil values can be used to calculate the relative population of dihedral angles in the ot and/3 regions or the population of helix in defined regions of the polypeptide. 1° For example, the fractional helicity of a 15 H. J. Dyson and E E. Wright, Ann. Rev. Biophys. Biophys. Chem. 20, 519 (1991). 16 L. J. Smith, K. M. Fiebig, H. Schwalbe, and C. M. Dobson, Fold. Design. 1, R95 (1996). 17 S. Spera and A. Bax, J. Am. Chem. Soc. 113, 5490 (1991). 18 D. S. Wishart and B. D. Sykes, Methods Enzymol. 239, 363 (1994). 18a S. Schwarzinger. G. J. A. Kroon, T. R. Foss, J. Chung, P. E. Wright, and H. J. Dyson, J. Am. Chem. Soc. 123, in press. 19 A. C. de Dios, J. G. Pearson, and E. Oldfield, Science 260, 1491 (1993). 20 D. S. Wishart and A. M. Nip, Biochem. Cell Biol. 76, 153 (1998).
[121
NMR OF DISORDERED STATES
~ 2~
--
263
---~7
,~
13CO
1 0
-1
A5 (ppm)
0.0
-0.2 -0.4
llta -0.6 1
0, '11
-1
20
""'""' " "'"" IIIIIIII IIr'' "11"illl,I
'P
13CI3
40
60
80
100
120
140
Residue N u m b e r
FIG.2. Secondarychemical shifts for ]3C~, 13CO, 1Ha, and ]3CP as a functionof residue number in apomyoglobinat pH 4.1. Bars at the top indicatethe presenceof HNi-HNI+INOEs; the smallerbars indicatethat the NOE was ambiguousbecauseof resonanceoverlap.Back rectanglesat the base of the top panel indicatethe locationsof helices in the nativeholomyoglobinstructure.55 Hatchedrectangles indicate putativeboundariesfor helicalregions in the pH 4 intermediate,based on the chemical shift and NOE data. (Reproducedfrom Ref. 21 with permission.) polypeptide between residues i to j can be determined from the 13C~ secondary 13C7)/3.1. Similar expressions are used to shifts using the expression
~~ij(A~
calculate helicity from the secondary 1H~, 13C~, or 13CO chemical shifts. A n example of the use of chemical shifts to delineate residual secondary structure is given in Fig. 2 for the molten globule state of apomyoglobin.l°,21 Combined use of 13C~, laa, 13Cfl, and 13CO secondary shifts gives a more precise definition of secondary structure boundaries than use of 13C~ shifts alone. 2]
264
PROTEINS
[ 121
Coupling Constants Coupling constants also provide insights into the backbone conformational preferences in partly folded or unfolded proteins. 22 The most commonly measured coupling constant, 3JHNa,varies from about 8-10 Hz in 15 strands to 4 - 5 Hz in helix; conformational averaging in unfolded states frequently results in intermediate values of 3jrlN,~ that are not particularly valuable as diagnostics of secondary structural propensities. Thus, for unfolded and partly folded proteins, as for peptides, 15 coupling constants are rarely as useful as chemical shifts in the detection of residual structure. It is of course clear that if a nonaveraged coupling constant is observed, this is an indication that the population of structured forms is quite high. 23 Temperature Coefficients Amide proton temperature coefficients and hydrogen exchange rates can provide valuable information about hydrogen bonding interactions and solvent sequestration in unfolded or partly folded proteins. 15 Temperature coefficients have the advantage that they are easy to measure. Abnormally low temperature coefficients, relative to random coil values, are a clear indication of local structure and interactions. Amide proton exchange rates tend to be very rapid in unfolded proteins. However, observation of amide protons that are protected from exchange is an especially convenient method for identifying structured regions in partly folded proteins and molten globules. 24 Nuclear Overhauser Effect As in folded proteins, the nuclear Overhauser effect (NOE) provides valuable information on secondary structure formation and long-range interactions in unfolded or partly folded proteins. However, in these disordered states, the NOE is difficult to interpret quantitatively because of the ubiquitous conformational averaging. Nevertheless, the d~N(i, i + 1), dNN(i, i + 1), anddtm(i, i + 1) NOEs between sequential amino acid residues do provide information on the local polypeptide backbone conformational preferences, i.e., on the relative population of dihedral angles in the ot and/3 regions of ~b, 7t space) 5 Such NOEs constitute a valuable supplement to chemical shifts in the analysis of backbone conformational preferences. It is important to note that the d~N(i, i + 1) and dry(i, i + 1) sequential NOEs provide information only on local ~b and ~p dihedral angle
21 D. Eliezer, J. Chung, H. J. Dyson, and E E. Wright, Biochemistry 39, 2894 (2000). 22L. J. Snlith, K. A. Bolin, H. Schwalbe, M. W. MacArthur, J. M. Thornton, and C. M. Dobson, J. Mol. Biol. 255, 494 (1996). 23j. Yao, H. J. Dyson, and E E. Wright. J. Mol. Biol. 243, 754 (1994). 24E M. Hughson, E E. Wright, and R. L. Baldwin, Science 249, 1544 (1990).
[12]
NMR OF DISORDEREDSTATES
265
preferences and do not by themselves indicate the presence of folded conformations. Definitive identification of folded elements of secondary structure requires additional information, either from medium-range NOEs [e.g., d~N(i, i + 2), d~r~(i, i + 3), d ~ ( i , i + 3) NOE connectivities] or from other forms of spectroscopy such as circular dichroism. 15 Long-range NOEs are generally difficult to observe and assign in unfolded or partly folded proteins, because of their intrinsic flexibility and poor resonance dispersion. While observation of a long-range NOE between two protons provides a definitive indication that they are in close proximity in at least some structures in the conformational ensemble, determination of the nature of the folded structures is difficult unless an extensive network of NOEs can be observed. This has so far been achieved in only one case. 25 Although NOE connectivities in unfolded proteins can be observed in 15N-edited 3D NOESY-HSQC spectra, the severe overlap of the lH resonances often reduces the utility of these experiments and seriously limits the number of NOE peaks that can be assigned. Improved resolution can be achieved for NOEs between NH groups by labeling both protons involved with their attached 15N frequencies, using the 3D 15N-HSQC-NOESY-HSQC experiment. 4'26'z7 Although some long-range interactions can be observed in favorable casesY this experiment primarily provides information about backbone conformational preferences. Detailed characterization of structured conformers requires identification and assignment of NOE connectivities involving side-chain protons, which are of course poorly dispersed in disordered states. To overcome this problem, Kay and co-workers have developed an elegant series of triple-resonance based NOESY experiments that exploit the dispersion of the 15N and 13CO resonances to resolve ambiguities in the aliphatic 1H and 13C chemical shifts. 4'28 These experiments, which are summarized in Table II, transfer magnetization from aliphatic protons to the 15N or laCt nuclei before or after the NOE mixing period, so that the NOEs involving aliphatic protons are observed at well-resolved nuclei with minimal resonance overlap. Paramagnetic Relaxation Probes
Given the difficulty of detecting and interpreting long-range NOEs in unfolded and partly folded proteins, an excellent alternative is to utilize paramagnetic nitroxide spin labels to probe the global structure. The method utilizes the r -6 distancedependent enhancement of nuclear spin relaxation by the unpaired electron in a paramagnetic center, where r is the distance between the unpaired electron and the nuclear spin. In native proteins, paramagnetic relaxation enhancement provides
25y. K. Mok, C. M. Kay,L. E. Kay,and J. Forman-Kay,J. Mol. BioL 289, 619 (1999). 26T. Frenkiel, C. Bauer, M. D. Cart, B. Birdsall, and J. Feeney,J. Magn. Reson. 90, 420 (1990). 27M. Ikura, A. Bax, G. M. Clore, and A. M. Gronenbron,J. Am. Chem. Soc. 112, 9020 (1990). 28O. Zhang, L. E. Kay,D. Shortle, and J. D. Forman-Kay,J. Mol. Biol. 272, 9 (1997).
PROTEINS
266
[ 1 21
TABLE II EXPERIMENTS DESIGNED TO OBSERVE N O E CONNECTIVITIESIN UNFOLDED PROTEINS
Experiment
NOE connectivity
Observed nuclei
Refs. a
Ci-NOESY-TOCSY-Nj+IHj+I (HN)COi-NOESY-Nj+IHj+I
CHi --->CHj aliphatic CHi --~ CHj aliphatic
13Ci, Nj+I,Hj+I 13C'i, Nj+I,Hj+I
1 1
Ni+I-TOCSY-NOESY-NjHj (HN)COi-TOCSY-NOESY-NjHj (HCA)COi-TOCSY-NOESY-NjHj
CHi ~ NHj CHi ~ NHj CHi ~ NHj
Ni+I, Nj,Hj 13C'i, Nj,Hj 13C'i, Nj,Hj
1 1 1
Ni+I-NOESY-Nj+IHj+I Ci-NOESY-Nj+IHj+I Ni-NOESY-NjHj (HCA)CO-NOESY-NjHj
Hai --->I-Iaj Hal ~ Haj H'~i --~ NHj H"i ~ NHj
15NI, Nj+I, Hj+I 13t.. . . . i,. .l'~j+l, . nj+l 15Ni, Nj,Hj 15Cti, Nj,Hj
15N-NOESY-HSQC 15N-HSQC-NOESY-HSQC
NHi ~ NHj, CHj NHi ~ NHj
15Ni, Hi, CHj, NHj 15Ni, Hi, NHj
1 1
1 1 2 3,4
a Key to References: (1) O. Zhang, J. D. Forman-Kay, D. Shortle, and L. E . Kay, J. Biomol. NMR 9, 181 (1997); (2) D. Marion, P. C. Driscoll, L. E. Kay, P. T. Wingfield, A. Bax, A. M. Gronenborn, and G. M. Clore, Biochemistry 28, 6150 (1989); (3) T. Frenuiel, C. Bauer, M. D. Carr, B. Birdsall, and J. Feeney, J. Magn. Reson. 90, 420 (1990); (4) M. Ikura, A. Bax, G. M. Clore, and A. M. Gronenborn, J. Am. Chem. Soc. 112, 9020 (1990).
measurements of long-range distances that are in good agreement with those calculated from crystal structures. 29 The method has been successfully to characterize long-range structure in a denatured state of staphylococcal nuclease. 3°,31 To utilize this method, nitroxide spin labels must be introduced at specific sites in the polypeptide. This can readily be accomplished by site-directed mutagenesis, substituting a single amino acid residue with cysteine to provide a site for coupling of an iodoacetamide derivative of the spin label. By inserting spin labels at multiple sites, a sufficient number of long-range distance constraints can be obtained to allow determination of the global topology. 3°,31 If the protein contains natural cysteine residues in its sequence, it is necessary to mutate these to avoid multiple sites of paramagnetic labeling. The PROXYL spin label is well suited to such studies, although care must be taken to select sites for labeling that are unlikely to perturb the residual structure in the denatured state. 31 The problems inherent in the application of this method to a dynamic ensemble of conformations such as found in disordered states of proteins have been previously described. 3° The presence of conformational averaging is likely to introduce a bias toward those conformations with the shortest contact distances, even though 29 p. A. Kosen, R. M. Scheek, H. Naderi, V. J. Basus, S. Manogaran, P. G. Schmidt, N. J. Oppenheimer, and I. D. Kuntz, Biochemistry 25, 2356 (1986). 30 j. R. Gillespie and D. Shortle, J. Mol. BioL 268, 158 (1997). 31 j. R. Gillespie and D. Shortle, J. Mol. Biol. 268, 170 (1997).
[12]
NMR OF DISORDEREDSTATES
267
their populations within the ensemble may be quite small. Nevertheless, use of paramagnetic spin label probes can provide exceptionally important insights into the global topology and long-range interactions in unstructured states and partly folded proteins. M e a s u r e m e n t of D y n a m i c s NMR relaxation and diffusion experiments have the potential to provide valuable insights into the intemal molecular dynamics and the overall hydrodynamic behavior of unfolded and partly folded states. Local variations in backbone dynamics are correlated with propensities for local compaction of the polypeptide chain that results in constriction of backbone motions. 10,21 This can occur through formation of local hydrophobic clusters, through formation of elements of secondary structure, or through long-range tertiary interactions in a compact folding intermediate. These measurements also provide insights into the effects of denaturants both on the overall hydrophobic behavior of a polypeptide in solution and on its local structural elements. Backbone dynamics are most commonly investigated by measurement of 15N T1 and T2 relaxation times and the {1H}-ISN NOE in uniformly 15N-labeled protein. The relaxation and NOE data are generally measured using 2D HSQC-based methods. Optimized pulse sequences have been reported by Kay and co-workers .32 These incorporate the sensitivity enhancement method developed by Rance and co-workers, 33,34 in which two orthogonal in-phase proton magnetization components are detected, and pulse field gradients for coherence pathway selection and to minimize artifacts. The pulse sequences have also been designed to minimize water saturation35'36; this is especially important for unfolded proteins since the rapid solvent exchange could lead to errors in the relaxation times and NOEs if water is partially saturated during the pulse train. 1H decoupling is applied, using shaped 180° pulses, 32 during the relaxation period of the T1 experiment to eliminate the effects of dipolar cross-relaxation and cross-correlation between dipolar and CSA relaxation mechanisms. The effects of cross-correlation on T2 are eliminated by application of 180° pulses synchronously with every second echo in the 15N CPMG pulse sequence. 37'38 If necessary, amino acid labeling can be used to obtain additional data in overlapped regions of the spectrum. 32 N. A. Farrow, R. Muhandiram, A. U. Singer, S. M. Pascal, C. M. Kay, G. Gish, S. E. Shoelson, T. Pawson, J. D. Forman-Kay, and L. E. Kay, Biochemistry 33, 5984 (1994). 33 A. G. Palmer, J. Cavanagh, P. E. Wright, and M. Rance, J. Magn. Reson. 93, 151 (1991). 34 M. J. Stone, W. J. Fairbrother, A. G. Palmer HI, J. Reizer, M. H. Saier, Jr., and P. E. Wright, Biochemistry 31, 4394 (1992). 35 S. Grzesiek and A. Bax, J. Am. Chem. Soc. 115, 12593 (1993). 36 L. E. Kay, G. Y. Xu, and T. Yamazaki, J. Magn. Reson. Ser. A 109, 129 (1994). 37 A. G. Palmer, N. J. Skelton, W. J. Chazin, E E. Wright, and M. Rance, Mol. Phys. 75, 699 (1992). 38 L. E. Kay, L. K. Nicholson, E Delaglio, A. Bax, and D. A. Torchia, J. Magn. Reson. 97, 359 (1992).
268
PROTEINS
[121
Spin relaxation data for folded proteins is commonly interpreted within the framework of the model-free formalism, in which the dynamics are described by an overall rotational correlation time Zm, an internal correlation time re, and an order parameter S2 describing the amplitude of the internal motions. 39,4° Modelfree analysis is popular because it describes molecular motions in terms of a set of intuitive physical parameters. However, it is unlikely that the dynamics of unfolded or partly folded proteins can be described accurately by the model-free approach, since the basic assumption ofisotropic tumbling with a single molecular correlation time is of questionable validity. Nevertheless, qualitative insights into the dynamics of unfolded states can still be obtained by model-free analysis. 41-43 An extension of the model-free analysis to incorporate a spectral density function that assumes a distribution of correlation times on the nanosecond time scale has been reported. 44 The authors show that this model better explains the experimental 15N relaxation data for an unfolded protein than does the conventional model-free approach. The assumptions inherent in model-free analysis can be avoided by direct mapping of the spectral density, either by measurement of additional relaxation parameters 45 or by making simplifying approximations. 46'47 The accuracy of the dynamics information obtained by spectral density mapping depends only on the accuracy of the experimental relaxation data, not on the assumptions about molecular motions (which are of dubious validity for unfolded proteins) made in modelfree analysis. Assuming that J(~o) c< 1/o92 between J(wa + WN) and J(WH -- CON),the values of J(0), J(wN), and J(0.87tOH), i.e., the spectral densities at the frequencies 0, CON, and 0.87wn, are given by Eqs. (1)-(4)48'49: CrNH= R I ( N O E - 1)~/~, H
(1)
J(0) = (6Rz - 3R1 - 2.72trNr0/(3d 2 + 4c 2)
(2)
J(WN) = (4R1
- 5 o ' N n ) / ( 3 d 2 + 4 c 2)
J(0.87ogn) = 4trNH/(5d 2)
(3)
(4)
where d = (IzohyNYH/87r2)(r-3) and c = OgNAO'/~/3. J(0) is sensitive both to fast internal motions on a picosecond-nanosecond (ps-ns) time scale and to slow motions on the millisecond-microsecond (ms-/zs) time scale. Rapid internal 39 G. Lipari and A. Szabo, J. Am. Chem. Soc. 104, 4546 (1982). 40 G. Lipari and A. Szabo, J. Am. Chem. Soc. 104, 4559 (1982). 41 N. A. FalTOW,O. Zhang, J. D. Forman-Kay, and L. E. Kay, Biochemistry 34, 868 (1995). 42 A. Z. Alexandrescu and D. Shortle, J. Mol. Biol. 242, 527 (1994). 43 M. Buck, H. Schwalbe, and C. M. Dobson, J. Mol. Biol. 257, 669 (1996). 44 A. V. Buevich and J. Baun, J. Am. Chem. Soc. 121, 8671 (1999). 45 j. W. Peng and G. Wagner, J. Magn. Reson. 98, 308 (1992). 46 N. A. FalTOW,O. Zhang, A. Szabo, D. A. Torchia. and L. E. Kay, J. Biomol. NMR 6, 153 (1995). 47 R. Ishima and K. Nagayama, Z Magn. Reson. Series B 10g, 73 (1995). 48 N. A. Farrow, O. Zhang, A. Szabo, D. A. Torchia, and L. E. Kay, J. Biomol. NMR 6, 153 (1995). 49 C. Bracken, E A. Carr, J. Cavanagh, and A. G. Palmer II/, J. Mol. Biol. 285, 2133 (1999).
[ 12]
NMR OF DISORDEREDSTATES .
.
.
.
,
.
.
.
.
,
.
.
.
.
,
.
.
.
.
,
.
.
.
.
,
.
.
.
.
,
.
.
.
.
,
.
.
.
.
,
.
.
.
.
,
.
.
.
.
,
.
.
.
.
.
.
.
.
269 .
,
.
.
.
.
,
.
.
.
.
.
.
.
.
.
.
3.0 2.0
1.0
u . m m
.... m l u
u . n . m . ,
0.3 0.2
0.1 ::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
J(750) (ns) 0.03 0.02
0.01
20
40
60
80
100
120
140
Residue Number
FIG.3. Spectraldensities of backbone motions in the pH 4.1 apomyoglobinintermediateat three different frequencies. (Reproducedfrom Ref. 21 with permission.) motions tend to reduce the value of J(0), while slow motions lead to anomalously large values of J(0). In contrast, the high-frequency spectral density J(0.87ogH) is sensitive only to fast internal motions, on a subnanosecond time scale; fast motions are reflected in relatively large values of J(OgH). The J(WN) and J(con) spectral densities are insensitive to the slow m s - # s time scale motions and so may be used to identify contributions of these motions to the J(0) spectral densities. Reduced spectral density analysis has been applied to investigate the backbone dynamics in the pH 4 molten globule state of apomyoglobin, 21 as shown in Fig. 3. Both J(0) and J(0.87coi0 are highly sensitive to variations in backbone motion, although J(ogN) is much less informative. The largest values of J(0) and the smallest values of J(0.87O~H) are found in the A, B, G, and H helices, indicating
270
PROTEINS
[ 12]
restriction of motions on the subnanosecond time scale and packing to form a compact core in the molten globule state. Anomalously large values of J(0) for certain residues in this core suggest the presence of additional motions on a slower/zsms time scale. J(0.87WH) is exquisitely sensitive to motions on fast time scales and reveals large differences in motional behavior in different regions along the protein sequence. Reduced spectral density mapping suggests that motions are generally more uniform in urea- or guanidine-denatured states, although values of J(0) tend to be increased in regions of high hydrophobicity, probably because of local hydrophobic collapse of the polypeptide chain. 5°,51
Translational Diffusion Measurements The compactness of unfolded states of proteins can be determined by measurement of the translational diffusion coefficient, from which the Stokes radius can be calculated. Translational diffusion coefficients can be readily measured using the pulsed field gradient water-suppressed longitudinal encode-decode (water-sLED) experiment. 52 The method can easily be used to monitor unfolding transitions 53 or to directly measure hydrodynamic radii. 53,54 The effective dimensions of a polypeptide chain are strongly dependent on the extent of local compaction, through formation of local hydrophobic clusters or local elements of secondary structure. Outlook An exceptionally powerful set of NMR tools is now available for characterization of the structure and dynamics of unfolded and partly folded polypeptides. As these tools become more widely applied, we can anticipate a major advance in our understanding of the fundamental interactions involved in the initiation of protein folding, and of the changes in structure and dynamics that occur as a protein progresses across the complex energy landscape that links the folded and unfolded states. In addition, we can look forward to a more detailed understanding of intrinsically unstructured proteins and the mechanism by which they function in important biological processes. Acknowledgments This work was supported by grants DK34909 and GM57374 from the National Institutes of Health. 50 N. A. Farrow, O. Zhang, J. D. Forman-Kay, and L. E. Kay, Biochemistry 36, 2390 (1997). 51 A. E. Meekhof and S. M. V. Freund, J. Mol. Biol. 286, 579 (1999). 52 A. S. Altieri, D. P. Hinton, and R. A. Byrd, Z Am. Chem. Soc. 117, 7566 (1995). 53 j. A. Jones, D. K. Wilkins, L. J. Smith, and C. M. Dobson, J. Biomol. NMR 1O, 199 (1997). 54 D. K. Wilkins, S. B. Grimshaw, V. Receveur, C. M. Dobson, J. A. Jones, and L. J. Smith, Biochemistry 38, 16424 (1999). 55 j. Kuriyan, S. Wilz, M. Karplus, and G. A. Petsko, J. Mol. Biol. 192, 133 (1986).
PROTEINS
[ 12]
made it possible to use NMR spectroscopy to study a protein beyond 100 kDa molecular mass. 82 By the end of this year, the sequencing of the human genome will be largely completed and already there are numerous novel gene products awaiting characterization. Determining every possible structural motif within the human genome could increase our understanding of protein function and evolution since frequently proteins lacking sequence homology are structurally homologous. With this information, a map of the protein-protein interactions that occur within a human will be vital for a complete understanding of biological metabolism. The methods outlined in this review can be used to study protein-protein complexes and the effect of different proteins on preexisting complexes.
82 M. Salzmann, K. Pervushin, G. Wider, H. Senn, and K. Wiithrich, J. Biomol. NMR 14, 85 (1999).
[12] Nuclear Magnetic Resonance Methods for Elucidation of Structure a n d D y n a m i c s in Disordered States By H. JANE DYSON and PETER E. WRIGHT Functional Relevance of Disordered States Nuclear magnetic resonance (NMR) is unsurpassed in its ability to provide detailed information on the structure and dynamics of unfolded and partially folded states of proteins. Nonnative states of proteins do not adopt unique threedimensional structures in solution but fluctuate rapidly over an ensemble of conformations. Structural characterization of such states is of great interest because of their importance in protein folding and because of the growing awareness that unstructured proteins play a critical role in many cellular processes. Determination of the structure and dynamics of protein folding intermediates and denatured states of proteins is of central importance for a detailed understanding of protein folding mechanisms. In addition, it is now recognized that many proteins and protein domains are only partially structured or are unstructured under physiological conditionsJ ,2 Characterization of the conformational propensities
1E E. Wright and H. J. Dyson, J. Mol. Biol. 293, 321 (1999). 2 E. Garner, E Cannon, E Romero, Z. Obradovic, and A. K. Dunker, Genome Informatics 9, 201 (1998).
METHODSIN ENZYMOLOGY,VOL.339
Copyright© 2001by AcademicPress Allfightsof reproductionin any formreserved. 0076-6879100$35.00
[12]
NMR OF DISORDEREDSTATES
259
and function of these nonglobular protein sequences represents a major challenge but is essential to a proper understanding of their biological functions and interactions. Recent advances in NMR technology, including the advent of ultrahigh field instruments and the application of heteronuclear multidimensional experiments, have allowed backbone resonances of unfolded and partly folded proteins to be fully assigned and have led to unprecedented insights into the structure and dynamics of these states. In this article, we review that available methods for NMR characterization of disordered proteins. Resonance Assignments NMR characterization of disordered states of proteins presents special challenges because the polypepfide chain in such states is inherently flexible and rapidly interconverts between multiple conformations. Consequently, the chemical shift dispersion of most resonances, especially protons, is poor and sequence-specific assignment of resonances is difficult. A major advance in the characterization of unfolded states came with the introduction of methods for uniform isotope labeling with 13C and 15N and the development of multidimensional triple resonance NMR methods. The backbone 15N and 13C' (carbonyl) resonances are influenced both by residue type and by the local amino acid sequence and therefore remain well-dispersed, even in fully unfolded states. 3,4 Representative IH-15N HSQC and H(N)CO spectra of acid unfolded apomyoglobin are shown in Fig. 1 and demonstrate the superior dispersion available in the 15N and 13CO chemical shifts. When exchange between folded and unfolded states occurs on an appropriate time scale, at least backbone resonance assignments can often be made using homonuclear (IH) magnetization transfer methods to correlate resonances in the spectrum of the folded protein with the corresponding resonances in the denatured state. 5-7 Three-D triple resonance experiments to establish sequential connecfivities via the well-resolved 15N and 13C' resonances in uniformly 15N,13C-labeled proteins provide a particularly robust method for obtaining unambiguous resonance assignments. 8-I° Pulse sequences that are appropriate for this purpose are summarized in Table I. An assignment strategy that utilizes the superior chemical shift dispersion of the 13CO resonance in unfolded proteins is especially valuable, 3 D. Braun, G. Wider, and K. Wiithrich, J. Am. Chem. Soc. 116, 8466 (1994). 4 O. Zhang, J. D. Forman-Kay, D. Shortle, and L. E. Kay, J. Biomol. NMR 9, 181 (1997). 5 p. A. Evans, K. D. Topping, D. N. Woolfson, and C. M. Dobson, Proteins 9, 248 (1991). 6 D. Neri, G. Wider, and K. Wiithrich, Proc. Natl. Acad. Sci. USA 89, 4397 (1992). 70. Zhang, L. E. Kay, J. P. Olivier, and J. D. Forman-Kay, J. Biomol. NMR 4, 845 (1994). 8 T. M. Logan, Y. Thrrianlt, and S. W. Fesik, J. Mol. Biol. 236, 637 (1994). 9 A. T. Alexandrescu, C. Abeygunawardana, and D. Shortle, Biochemistry 33, 1063 (1994). 10 D. Eliezer, J. Yao, H. J. Dyson, and P. E. Wright, Nature Struct. Biol. 5, 148 (1998).
260
PROTEINS
[ 12]
o
0
.
O
O0 ° 0(:~
.d
oO°o
0 o
D O
0
E
8' o
E
8 0
o
o a'
0 0
lip
o
0
g
o
c~°°®
a.~
~
0
~. 0
O
gc~ o
T-
z
00 Q
°
~
~0 O
~?~oo
e~ o
* o~$%-
o Q,
O
8.1
1H (ppm)
o
o
°~o o
~ o o
6
e (~' o o oo °0 8'.4
00
•
O
8.7
_¢6 b.
~ o
oO ,r-
o o
/.8
7.s
a.7
;.4
o
;.1
1N (ppm)
FIG. 1. NMR spectra (750 MHz) of 15N,t3C-labeled apomyoglobin, pH 2.3, 10 mM acetate-d6, 5 °. (A) 15N-1H HSQC spectrum. (B) HNCO spectrum. (Reproduced from Ref. 11 with permission.)
given the excellent resolution and long T2 relaxation time of these resonances.ll A set of high resolution constant-time triple resonance experiments that transfer magnetization sequentially along the amino acid sequence using carbonyl ]3C homonuclear isotropic mixing have been developed specifically for assignment of unfolded proteins. 12 These experiments have the advantage that the slow ~3CO relaxation rates enable correlations to be established across proline residues. One of the advantages of working with unfolded proteins is that the intrinsic flexibility of the polypeptide generally causes the resonances to be much narrower than they would be in a folded protein of comparable molecular weight, i.e., T2 is longer than for the natively folded protein. This characteristic gives special advantages in the implementation of pulse sequences such as (HCA)CO(CA)NH, in which there are a large number of delays. For folded proteins of molecular weight greater then 1 5 - 2 0 kDa, this experiment is often rather insensitive because of decay of magnetization during the pulse sequence. For unfolded proteins, however, since 11 j. Yao, H. J. Dyson, and E E. Wright, FEBSLett. 419, 285 (1997). n2A. Liu, R. Riek, G. Wider, C. Von Schroetter, R. Zahn, and K. Wiithrich, J. Biomol. N M R 16, 127
(2OOO).
[ 12]
NMR OF DISORDEREDSTATES
261
TABLE I EXPERIMENTSDESIGNEDTO OBSERVESCALARCONNECTWITIES Experiment 1H-lSN HSQC HNCACB CBCA(CO)NH HNCO (HCA)CO(CA)NH HNCA TOCSY-HSQC C(CO)NH-TOCSY (H)N(CO-TOCSY)NH (H)CA(CO-TOCSY)NH (H)CBCA(CO-TOCSY)NH
Connectivity 15N~1Hi (lSN-1H)i-(13CetJ3Cfl)i
( 15N-1H)~(13Cot-13C/~)i_1 (15N-1H)i-13COi_I
Refs.a 1 2,3 3,4 5
(15N-1a)i-13COi,13COi_l
6
(15N-IH)i-13foq_l (15N-IH)i-Ha~(H/~,H~/...)i
5 1
(15N-IH)~Cai_I-(Cfl,Cy...)i_I
7
15N~NHi,NHi+I, NHi+2 C%-NHi,NHi+I, NHi+2 C~/,Ca~NHi, NHi+I, NHI+2
8 8 8
Key to References: (1) O. Zhang, L. E. Kay, J. E Olivier, and J. D. Forman-Kay, J. Biomol. NMR 4, 845 (1994); (2) M. Wittekind and L. Mueller, J. Magn. Reson. 101, 201 (1993); (3) D. R. Muhandiram and L. E. Kay, J. Magn. Reson. Series B 103, 203 (1994); (4) S. Grzesiek and A. Bax, J. Am. Chem. Soc. 114, 6291 (1992); (5) S. Grzesiek and A. Bax, J. Magn. Reson. 96, 432 (1992); (6) E L6hr and H. Rtiterjans, J. Biomol. NMR 6, 189 (1995); (7) S. Grzesiek, J. Anglister, and A. Bax, J. Magn. Reson, Series B 101, 114 (1993);
(8) A. Liu, R. Riek, G. Wider, C. Von Schroetter, R. Zahn, and K. Wiithrich, J. Biomol. NMR 16, 127 (2000).
T2is significantly longer, the (HCA)CO(CA)NH experiment is extremely sensitive and can be utilized with great success. 1° The same is, unhappily, not true for partly folded proteins, or molten globule states; the NMR spectra of these species often contain a mixture of sharp (from unstructured regions) and broad resonances (from the structured subdomains). Indeed, many molten globules are extremely difficult to study by NMR because their lines are exceptionally broad as a result of exchange processes on intermediate time scales. For example, the molten globule state of a-lactalbumin gives rise to extremely broad resonances that largely preclude direct high-resolution NMR analysis.13,14 High-resolution NMR studies of unfolded proteins have been largely enabled by the advent of high-field spectrometers and multidimensional heteronuclear NMR experiments. High sensitivity is a critical factor in the study of unfolded and partly folded proteins, since NMR experiments must often be performed at very low concentration (of the order of 50-300/zM) to prevent aggregation. 13j. Baum, C. M. Dobson, E A. Evans, and C. Hanley, Biochemistry 28, 7 (1989). 14 S. Kim, C. Bracken, and J. Baum, J. Mol. Biol. 294, 551 (1999).
262
PROTEINS
[ 12]
NMR Parameters Once backbone resonance assignments have been made, a number of NMR parameters can be used to characterize residual structure in unfolded and partly folded states. However, in interpreting the NMR data, it is important to bear in mind that unfolded and partly folded states of proteins are highly dynamic and that all parameters are a population-weighted average over all structures in the conformational ensemble. Conformational preferences are always identified, therefore, by comparison of experimental NMR parameters to those expected for a random coil state, in which the polypeptide backbone dihedral angles adopt a Boltzmann distribution over the ~b, ~ energy surface.15' 16 In addition, it is implicitly assumed that the dominant conformers in unfolded or partly folded polypeptides will have backbone dihedral angles that lie within the broad ot and/3 minima on the q~, 7z conformational energy surface.
Chemical Shifts The deviations of chemical shifts from random coil values, especially for X3C~, 13C~, and 1I-I~,provide a convenient and sensitive probe of secondary structural propensities.17,18 The 13CO chemical shifts also provide information on secondary structure content and dihedral angle preferences, provided they are corrected for sequence dependence. 18a Theoretical calculations 19 show that the 13Ca and 13Cfl chemical shifts are determined primarily by the backbone ~, ~ dihedral angles, and empirical correlations confirm that these resonances are reliable indicators of secondary structure. 17,18,20 In ot helices, the 13C~ resonances are shifted downfield by an average of 3.1 + 1.0 ppm. 17 The 13CO resonances are also shifted downfield, while the 13C~ and 1H~ resonances are shifted upfield from their positions in random coil states. In/3 sheets, the 13Ct~resonance is shifted upfield (AS = - 1.5 + 1.2 ppm 17) with respect to the random coil chemical shift, while the lt-Ia and tac~ resonances are shifted downfield. Rapid conformational averaging between the ct and/3 regions of ~b, ~p space reduces the secondary structure shifts below these extreme values in unfolded or partly folded proteins. Indeed, the deviations of the chemical shifts from random coil values can be used to calculate the relative population of dihedral angles in the ot and/3 regions or the population of helix in defined regions of the polypeptide. 1° For example, the fractional helicity of a 15 H. J. Dyson and E E. Wright, Ann. Rev. Biophys. Biophys. Chem. 20, 519 (1991). 16 L. J. Smith, K. M. Fiebig, H. Schwalbe, and C. M. Dobson, Fold. Design. 1, R95 (1996). 17 S. Spera and A. Bax, J. Am. Chem. Soc. 113, 5490 (1991). 18 D. S. Wishart and B. D. Sykes, Methods Enzymol. 239, 363 (1994). 18a S. Schwarzinger. G. J. A. Kroon, T. R. Foss, J. Chung, P. E. Wright, and H. J. Dyson, J. Am. Chem. Soc. 123, in press. 19 A. C. de Dios, J. G. Pearson, and E. Oldfield, Science 260, 1491 (1993). 20 D. S. Wishart and A. M. Nip, Biochem. Cell Biol. 76, 153 (1998).
[121
NMR OF DISORDERED STATES
~ 2~
--
263
---~7
,~
13CO
1 0
-1
A5 (ppm)
0.0
-0.2 -0.4
llta -0.6 1
0, '11
-1
20
""'""' " "'"" IIIIIIII IIr'' "11"illl,I
'P
13CI3
40
60
80
100
120
140
Residue N u m b e r
FIG.2. Secondarychemical shifts for ]3C~, 13CO, 1Ha, and ]3CP as a functionof residue number in apomyoglobinat pH 4.1. Bars at the top indicatethe presenceof HNi-HNI+INOEs; the smallerbars indicatethat the NOE was ambiguousbecauseof resonanceoverlap.Back rectanglesat the base of the top panel indicatethe locationsof helices in the nativeholomyoglobinstructure.55 Hatchedrectangles indicate putativeboundariesfor helicalregions in the pH 4 intermediate,based on the chemical shift and NOE data. (Reproducedfrom Ref. 21 with permission.) polypeptide between residues i to j can be determined from the 13C~ secondary 13C7)/3.1. Similar expressions are used to shifts using the expression
~~ij(A~
calculate helicity from the secondary 1H~, 13C~, or 13CO chemical shifts. A n example of the use of chemical shifts to delineate residual secondary structure is given in Fig. 2 for the molten globule state of apomyoglobin.l°,21 Combined use of 13C~, laa, 13Cfl, and 13CO secondary shifts gives a more precise definition of secondary structure boundaries than use of 13C~ shifts alone. 2]
264
PROTEINS
[ 121
Coupling Constants Coupling constants also provide insights into the backbone conformational preferences in partly folded or unfolded proteins. 22 The most commonly measured coupling constant, 3JHNa,varies from about 8-10 Hz in 15 strands to 4 - 5 Hz in helix; conformational averaging in unfolded states frequently results in intermediate values of 3jrlN,~ that are not particularly valuable as diagnostics of secondary structural propensities. Thus, for unfolded and partly folded proteins, as for peptides, 15 coupling constants are rarely as useful as chemical shifts in the detection of residual structure. It is of course clear that if a nonaveraged coupling constant is observed, this is an indication that the population of structured forms is quite high. 23 Temperature Coefficients Amide proton temperature coefficients and hydrogen exchange rates can provide valuable information about hydrogen bonding interactions and solvent sequestration in unfolded or partly folded proteins. 15 Temperature coefficients have the advantage that they are easy to measure. Abnormally low temperature coefficients, relative to random coil values, are a clear indication of local structure and interactions. Amide proton exchange rates tend to be very rapid in unfolded proteins. However, observation of amide protons that are protected from exchange is an especially convenient method for identifying structured regions in partly folded proteins and molten globules. 24 Nuclear Overhauser Effect As in folded proteins, the nuclear Overhauser effect (NOE) provides valuable information on secondary structure formation and long-range interactions in unfolded or partly folded proteins. However, in these disordered states, the NOE is difficult to interpret quantitatively because of the ubiquitous conformational averaging. Nevertheless, the d~N(i, i + 1), dNN(i, i + 1), anddtm(i, i + 1) NOEs between sequential amino acid residues do provide information on the local polypeptide backbone conformational preferences, i.e., on the relative population of dihedral angles in the ot and/3 regions of ~b, 7t space) 5 Such NOEs constitute a valuable supplement to chemical shifts in the analysis of backbone conformational preferences. It is important to note that the d~N(i, i + 1) and dry(i, i + 1) sequential NOEs provide information only on local ~b and ~p dihedral angle
21 D. Eliezer, J. Chung, H. J. Dyson, and E E. Wright, Biochemistry 39, 2894 (2000). 22L. J. Snlith, K. A. Bolin, H. Schwalbe, M. W. MacArthur, J. M. Thornton, and C. M. Dobson, J. Mol. Biol. 255, 494 (1996). 23j. Yao, H. J. Dyson, and E E. Wright. J. Mol. Biol. 243, 754 (1994). 24E M. Hughson, E E. Wright, and R. L. Baldwin, Science 249, 1544 (1990).
[12]
NMR OF DISORDEREDSTATES
265
preferences and do not by themselves indicate the presence of folded conformations. Definitive identification of folded elements of secondary structure requires additional information, either from medium-range NOEs [e.g., d~N(i, i + 2), d~r~(i, i + 3), d ~ ( i , i + 3) NOE connectivities] or from other forms of spectroscopy such as circular dichroism. 15 Long-range NOEs are generally difficult to observe and assign in unfolded or partly folded proteins, because of their intrinsic flexibility and poor resonance dispersion. While observation of a long-range NOE between two protons provides a definitive indication that they are in close proximity in at least some structures in the conformational ensemble, determination of the nature of the folded structures is difficult unless an extensive network of NOEs can be observed. This has so far been achieved in only one case. 25 Although NOE connectivities in unfolded proteins can be observed in 15N-edited 3D NOESY-HSQC spectra, the severe overlap of the lH resonances often reduces the utility of these experiments and seriously limits the number of NOE peaks that can be assigned. Improved resolution can be achieved for NOEs between NH groups by labeling both protons involved with their attached 15N frequencies, using the 3D 15N-HSQC-NOESY-HSQC experiment. 4'26'z7 Although some long-range interactions can be observed in favorable casesY this experiment primarily provides information about backbone conformational preferences. Detailed characterization of structured conformers requires identification and assignment of NOE connectivities involving side-chain protons, which are of course poorly dispersed in disordered states. To overcome this problem, Kay and co-workers have developed an elegant series of triple-resonance based NOESY experiments that exploit the dispersion of the 15N and 13CO resonances to resolve ambiguities in the aliphatic 1H and 13C chemical shifts. 4'28 These experiments, which are summarized in Table II, transfer magnetization from aliphatic protons to the 15N or laCt nuclei before or after the NOE mixing period, so that the NOEs involving aliphatic protons are observed at well-resolved nuclei with minimal resonance overlap. Paramagnetic Relaxation Probes
Given the difficulty of detecting and interpreting long-range NOEs in unfolded and partly folded proteins, an excellent alternative is to utilize paramagnetic nitroxide spin labels to probe the global structure. The method utilizes the r -6 distancedependent enhancement of nuclear spin relaxation by the unpaired electron in a paramagnetic center, where r is the distance between the unpaired electron and the nuclear spin. In native proteins, paramagnetic relaxation enhancement provides
25y. K. Mok, C. M. Kay,L. E. Kay,and J. Forman-Kay,J. Mol. BioL 289, 619 (1999). 26T. Frenkiel, C. Bauer, M. D. Cart, B. Birdsall, and J. Feeney,J. Magn. Reson. 90, 420 (1990). 27M. Ikura, A. Bax, G. M. Clore, and A. M. Gronenbron,J. Am. Chem. Soc. 112, 9020 (1990). 28O. Zhang, L. E. Kay,D. Shortle, and J. D. Forman-Kay,J. Mol. Biol. 272, 9 (1997).
PROTEINS
266
[ 1 21
TABLE II EXPERIMENTS DESIGNED TO OBSERVE N O E CONNECTIVITIESIN UNFOLDED PROTEINS
Experiment
NOE connectivity
Observed nuclei
Refs. a
Ci-NOESY-TOCSY-Nj+IHj+I (HN)COi-NOESY-Nj+IHj+I
CHi --->CHj aliphatic CHi --~ CHj aliphatic
13Ci, Nj+I,Hj+I 13C'i, Nj+I,Hj+I
1 1
Ni+I-TOCSY-NOESY-NjHj (HN)COi-TOCSY-NOESY-NjHj (HCA)COi-TOCSY-NOESY-NjHj
CHi ~ NHj CHi ~ NHj CHi ~ NHj
Ni+I, Nj,Hj 13C'i, Nj,Hj 13C'i, Nj,Hj
1 1 1
Ni+I-NOESY-Nj+IHj+I Ci-NOESY-Nj+IHj+I Ni-NOESY-NjHj (HCA)CO-NOESY-NjHj
Hai --->I-Iaj Hal ~ Haj H'~i --~ NHj H"i ~ NHj
15NI, Nj+I, Hj+I 13t.. . . . i,. .l'~j+l, . nj+l 15Ni, Nj,Hj 15Cti, Nj,Hj
15N-NOESY-HSQC 15N-HSQC-NOESY-HSQC
NHi ~ NHj, CHj NHi ~ NHj
15Ni, Hi, CHj, NHj 15Ni, Hi, NHj
1 1
1 1 2 3,4
a Key to References: (1) O. Zhang, J. D. Forman-Kay, D. Shortle, and L. E . Kay, J. Biomol. NMR 9, 181 (1997); (2) D. Marion, P. C. Driscoll, L. E. Kay, P. T. Wingfield, A. Bax, A. M. Gronenborn, and G. M. Clore, Biochemistry 28, 6150 (1989); (3) T. Frenuiel, C. Bauer, M. D. Carr, B. Birdsall, and J. Feeney, J. Magn. Reson. 90, 420 (1990); (4) M. Ikura, A. Bax, G. M. Clore, and A. M. Gronenborn, J. Am. Chem. Soc. 112, 9020 (1990).
measurements of long-range distances that are in good agreement with those calculated from crystal structures. 29 The method has been successfully to characterize long-range structure in a denatured state of staphylococcal nuclease. 3°,31 To utilize this method, nitroxide spin labels must be introduced at specific sites in the polypeptide. This can readily be accomplished by site-directed mutagenesis, substituting a single amino acid residue with cysteine to provide a site for coupling of an iodoacetamide derivative of the spin label. By inserting spin labels at multiple sites, a sufficient number of long-range distance constraints can be obtained to allow determination of the global topology. 3°,31 If the protein contains natural cysteine residues in its sequence, it is necessary to mutate these to avoid multiple sites of paramagnetic labeling. The PROXYL spin label is well suited to such studies, although care must be taken to select sites for labeling that are unlikely to perturb the residual structure in the denatured state. 31 The problems inherent in the application of this method to a dynamic ensemble of conformations such as found in disordered states of proteins have been previously described. 3° The presence of conformational averaging is likely to introduce a bias toward those conformations with the shortest contact distances, even though 29 p. A. Kosen, R. M. Scheek, H. Naderi, V. J. Basus, S. Manogaran, P. G. Schmidt, N. J. Oppenheimer, and I. D. Kuntz, Biochemistry 25, 2356 (1986). 30 j. R. Gillespie and D. Shortle, J. Mol. BioL 268, 158 (1997). 31 j. R. Gillespie and D. Shortle, J. Mol. Biol. 268, 170 (1997).
[12]
NMR OF DISORDEREDSTATES
267
their populations within the ensemble may be quite small. Nevertheless, use of paramagnetic spin label probes can provide exceptionally important insights into the global topology and long-range interactions in unstructured states and partly folded proteins. M e a s u r e m e n t of D y n a m i c s NMR relaxation and diffusion experiments have the potential to provide valuable insights into the intemal molecular dynamics and the overall hydrodynamic behavior of unfolded and partly folded states. Local variations in backbone dynamics are correlated with propensities for local compaction of the polypeptide chain that results in constriction of backbone motions. 10,21 This can occur through formation of local hydrophobic clusters, through formation of elements of secondary structure, or through long-range tertiary interactions in a compact folding intermediate. These measurements also provide insights into the effects of denaturants both on the overall hydrophobic behavior of a polypeptide in solution and on its local structural elements. Backbone dynamics are most commonly investigated by measurement of 15N T1 and T2 relaxation times and the {1H}-ISN NOE in uniformly 15N-labeled protein. The relaxation and NOE data are generally measured using 2D HSQC-based methods. Optimized pulse sequences have been reported by Kay and co-workers .32 These incorporate the sensitivity enhancement method developed by Rance and co-workers, 33,34 in which two orthogonal in-phase proton magnetization components are detected, and pulse field gradients for coherence pathway selection and to minimize artifacts. The pulse sequences have also been designed to minimize water saturation35'36; this is especially important for unfolded proteins since the rapid solvent exchange could lead to errors in the relaxation times and NOEs if water is partially saturated during the pulse train. 1H decoupling is applied, using shaped 180° pulses, 32 during the relaxation period of the T1 experiment to eliminate the effects of dipolar cross-relaxation and cross-correlation between dipolar and CSA relaxation mechanisms. The effects of cross-correlation on T2 are eliminated by application of 180° pulses synchronously with every second echo in the 15N CPMG pulse sequence. 37'38 If necessary, amino acid labeling can be used to obtain additional data in overlapped regions of the spectrum. 32 N. A. Farrow, R. Muhandiram, A. U. Singer, S. M. Pascal, C. M. Kay, G. Gish, S. E. Shoelson, T. Pawson, J. D. Forman-Kay, and L. E. Kay, Biochemistry 33, 5984 (1994). 33 A. G. Palmer, J. Cavanagh, P. E. Wright, and M. Rance, J. Magn. Reson. 93, 151 (1991). 34 M. J. Stone, W. J. Fairbrother, A. G. Palmer HI, J. Reizer, M. H. Saier, Jr., and P. E. Wright, Biochemistry 31, 4394 (1992). 35 S. Grzesiek and A. Bax, J. Am. Chem. Soc. 115, 12593 (1993). 36 L. E. Kay, G. Y. Xu, and T. Yamazaki, J. Magn. Reson. Ser. A 109, 129 (1994). 37 A. G. Palmer, N. J. Skelton, W. J. Chazin, E E. Wright, and M. Rance, Mol. Phys. 75, 699 (1992). 38 L. E. Kay, L. K. Nicholson, E Delaglio, A. Bax, and D. A. Torchia, J. Magn. Reson. 97, 359 (1992).
268
PROTEINS
[121
Spin relaxation data for folded proteins is commonly interpreted within the framework of the model-free formalism, in which the dynamics are described by an overall rotational correlation time Zm, an internal correlation time re, and an order parameter S2 describing the amplitude of the internal motions. 39,4° Modelfree analysis is popular because it describes molecular motions in terms of a set of intuitive physical parameters. However, it is unlikely that the dynamics of unfolded or partly folded proteins can be described accurately by the model-free approach, since the basic assumption ofisotropic tumbling with a single molecular correlation time is of questionable validity. Nevertheless, qualitative insights into the dynamics of unfolded states can still be obtained by model-free analysis. 41-43 An extension of the model-free analysis to incorporate a spectral density function that assumes a distribution of correlation times on the nanosecond time scale has been reported. 44 The authors show that this model better explains the experimental 15N relaxation data for an unfolded protein than does the conventional model-free approach. The assumptions inherent in model-free analysis can be avoided by direct mapping of the spectral density, either by measurement of additional relaxation parameters 45 or by making simplifying approximations. 46'47 The accuracy of the dynamics information obtained by spectral density mapping depends only on the accuracy of the experimental relaxation data, not on the assumptions about molecular motions (which are of dubious validity for unfolded proteins) made in modelfree analysis. Assuming that J(~o) c< 1/o92 between J(wa + WN) and J(WH -- CON),the values of J(0), J(wN), and J(0.87tOH), i.e., the spectral densities at the frequencies 0, CON, and 0.87wn, are given by Eqs. (1)-(4)48'49: CrNH= R I ( N O E - 1)~/~, H
(1)
J(0) = (6Rz - 3R1 - 2.72trNr0/(3d 2 + 4c 2)
(2)
J(WN) = (4R1
- 5 o ' N n ) / ( 3 d 2 + 4 c 2)
J(0.87ogn) = 4trNH/(5d 2)
(3)
(4)
where d = (IzohyNYH/87r2)(r-3) and c = OgNAO'/~/3. J(0) is sensitive both to fast internal motions on a picosecond-nanosecond (ps-ns) time scale and to slow motions on the millisecond-microsecond (ms-/zs) time scale. Rapid internal 39 G. Lipari and A. Szabo, J. Am. Chem. Soc. 104, 4546 (1982). 40 G. Lipari and A. Szabo, J. Am. Chem. Soc. 104, 4559 (1982). 41 N. A. FalTOW,O. Zhang, J. D. Forman-Kay, and L. E. Kay, Biochemistry 34, 868 (1995). 42 A. Z. Alexandrescu and D. Shortle, J. Mol. Biol. 242, 527 (1994). 43 M. Buck, H. Schwalbe, and C. M. Dobson, J. Mol. Biol. 257, 669 (1996). 44 A. V. Buevich and J. Baun, J. Am. Chem. Soc. 121, 8671 (1999). 45 j. W. Peng and G. Wagner, J. Magn. Reson. 98, 308 (1992). 46 N. A. FalTOW,O. Zhang, A. Szabo, D. A. Torchia. and L. E. Kay, J. Biomol. NMR 6, 153 (1995). 47 R. Ishima and K. Nagayama, Z Magn. Reson. Series B 10g, 73 (1995). 48 N. A. Farrow, O. Zhang, A. Szabo, D. A. Torchia, and L. E. Kay, J. Biomol. NMR 6, 153 (1995). 49 C. Bracken, E A. Carr, J. Cavanagh, and A. G. Palmer II/, J. Mol. Biol. 285, 2133 (1999).
[ 12]
NMR OF DISORDEREDSTATES .
.
.
.
,
.
.
.
.
,
.
.
.
.
,
.
.
.
.
,
.
.
.
.
,
.
.
.
.
,
.
.
.
.
,
.
.
.
.
,
.
.
.
.
,
.
.
.
.
,
.
.
.
.
.
.
.
.
269 .
,
.
.
.
.
,
.
.
.
.
.
.
.
.
.
.
3.0 2.0
1.0
u . m m
.... m l u
u . n . m . ,
0.3 0.2
0.1 ::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::
J(750) (ns) 0.03 0.02
0.01
20
40
60
80
100
120
140
Residue Number
FIG.3. Spectraldensities of backbone motions in the pH 4.1 apomyoglobinintermediateat three different frequencies. (Reproducedfrom Ref. 21 with permission.) motions tend to reduce the value of J(0), while slow motions lead to anomalously large values of J(0). In contrast, the high-frequency spectral density J(0.87ogH) is sensitive only to fast internal motions, on a subnanosecond time scale; fast motions are reflected in relatively large values of J(OgH). The J(WN) and J(con) spectral densities are insensitive to the slow m s - # s time scale motions and so may be used to identify contributions of these motions to the J(0) spectral densities. Reduced spectral density analysis has been applied to investigate the backbone dynamics in the pH 4 molten globule state of apomyoglobin, 21 as shown in Fig. 3. Both J(0) and J(0.87coi0 are highly sensitive to variations in backbone motion, although J(ogN) is much less informative. The largest values of J(0) and the smallest values of J(0.87O~H) are found in the A, B, G, and H helices, indicating
270
PROTEINS
[ 12]
restriction of motions on the subnanosecond time scale and packing to form a compact core in the molten globule state. Anomalously large values of J(0) for certain residues in this core suggest the presence of additional motions on a slower/zsms time scale. J(0.87WH) is exquisitely sensitive to motions on fast time scales and reveals large differences in motional behavior in different regions along the protein sequence. Reduced spectral density mapping suggests that motions are generally more uniform in urea- or guanidine-denatured states, although values of J(0) tend to be increased in regions of high hydrophobicity, probably because of local hydrophobic collapse of the polypeptide chain. 5°,51
Translational Diffusion Measurements The compactness of unfolded states of proteins can be determined by measurement of the translational diffusion coefficient, from which the Stokes radius can be calculated. Translational diffusion coefficients can be readily measured using the pulsed field gradient water-suppressed longitudinal encode-decode (water-sLED) experiment. 52 The method can easily be used to monitor unfolding transitions 53 or to directly measure hydrodynamic radii. 53,54 The effective dimensions of a polypeptide chain are strongly dependent on the extent of local compaction, through formation of local hydrophobic clusters or local elements of secondary structure. Outlook An exceptionally powerful set of NMR tools is now available for characterization of the structure and dynamics of unfolded and partly folded polypeptides. As these tools become more widely applied, we can anticipate a major advance in our understanding of the fundamental interactions involved in the initiation of protein folding, and of the changes in structure and dynamics that occur as a protein progresses across the complex energy landscape that links the folded and unfolded states. In addition, we can look forward to a more detailed understanding of intrinsically unstructured proteins and the mechanism by which they function in important biological processes. Acknowledgments This work was supported by grants DK34909 and GM57374 from the National Institutes of Health. 50 N. A. Farrow, O. Zhang, J. D. Forman-Kay, and L. E. Kay, Biochemistry 36, 2390 (1997). 51 A. E. Meekhof and S. M. V. Freund, J. Mol. Biol. 286, 579 (1999). 52 A. S. Altieri, D. P. Hinton, and R. A. Byrd, Z Am. Chem. Soc. 117, 7566 (1995). 53 j. A. Jones, D. K. Wilkins, L. J. Smith, and C. M. Dobson, J. Biomol. NMR 1O, 199 (1997). 54 D. K. Wilkins, S. B. Grimshaw, V. Receveur, C. M. Dobson, J. A. Jones, and L. J. Smith, Biochemistry 38, 16424 (1999). 55 j. Kuriyan, S. Wilz, M. Karplus, and G. A. Petsko, J. Mol. Biol. 192, 133 (1986).