NMR Studies of Structure and Dynamics of Isotope Enriched Proteins GERHARD WAGNER, V. THANABAL, BRIAN J. STOCKMAN, JEFFREY W. PENG, N. R. NIRMALA, SVEN G. HYBERTS, MATTHEW S. GOLDBERG, DAVID J. DETLEFSEN, ROBERT T. CLUBB, and MARC ADLER Department of Biological Chemistry and Molecular Pharmacology, Harvard Medical School, 240 Longwood Avenue, Boston, Massachusetts 021 1 5
SYNOPSIS
Structural studies of globular proteins by nmr can be enhanced by the use of isotope enrichment. We have been working with proteins enriched with 15N,and with both 15N and 13C. Due to the isotope enrichment we could assign several large proteins with up to 186 residues and could address structural questions. Furthermore, we can accurately measure heteronuclear and homonuclear vicinal coupling constants. This involves in part multidimensional multiple resonance experiments. This is important for characterization of minor conformational changes caused by mutations. We have also made use of isotope enrichment to study the internal mobility of proteins. We also have developed novel methods for measuring accurately "N relaxation parameters, in particular transverse relaxation rates. This has led us toward a method for directly mapping spectral density functions of the rotational motions of N-H bond vectors in proteins. The protein systems that are discussed include the unlabeled proteins kistrin and cytochrome c551, and the labeled proteins eglin c, a flavodoxin, and human dihydrofolate reductase.
INTRODUCTION Previously, structural studies of proteins by nmr were limited to proteins of molecular sizes of less than 10-12 kD, and the proteins were handled as rigid molecules, neglecting internal mobility due to the lack of strategies for handling internal dynamics. This situation has dramatically changed when isotope labeling became routinely available in the last couple of years. This tool opened the doors to a wide variety of heteronuclear nmr techniques. In particular, three-dimensional and four-dimensional 15N and 13C dispersed heteronuclear methods and multiple resonance multidimensional techniques have enabled us to address structural problems of much larger proteins we did not think of a few years ago. In addition, new methods for measuring homonuclear and heteronuclear vicinal coupling constants have become available, relying on isotope enrichment, providing a whole new class of conformational
Biopolymers. Vol. 32, 381-390 (1992) 0 1992 J o h n Wiley & Sons, Inc.
CCC 0006-3525/92/040381-10$04.00
parameters. Isotope labeling also provides access to studying motional aspects in proteins via measuring a large variety of relaxation parameters and will probably provide a relatively complete description of fast internal motions in proteins. Considering these new technologies, better descriptions of the dynamic conformations of larger proteins will be obtained soon. Here we compare structural studies of two small proteins for which isotope labeling was not available with investigations of "N- and 13C-labeledproteins. Assignments, structure determination, and measurements of relaxation parameters will be discussed.
STRUCTURAL STUDIES OF UNLABELED PROTElNS While isotope enrichment facilitates assignments and can provide additional structural parameters, it is not always available. In particular, it is often difficult to express small disulfide-rich proteins in Escherichia coli systems. We have recently deter381
382
WAGNER ET AL.
mined the structures of two proteins for which isotope enrichment was not available. These proteins are kistrin and cytochrome c551. Kistrin is a protein from the Malaysian pit viper Agkistrodon rhodostoma.' It is a potent inhibitor of platelet aggregation, and functions by inhibiting the platelet-bound receptor GPIIbIIIa, which is a member of the integrin family. Platelet-bound GPIIbIIIa recognizes RGD sequences of fibrin networks, which leads to formation of blood clots. Kistrin contains an RGD sequence that is believed to be the central part of the site interacting with GPIIbIIIa. Binding of kistrin leads to inactivation of the receptor. Kistrin consists of 68 residues and contains six disulfide bonds. We have assigned the proton nmr spectrum2and determined the three-dimensional (3D) structure, relying on conventional two-dimensional (2D) nmr techniques. A set of 549 nuclear Overhauser effect (NOE ) constraints were used for the structure calculations. In addition, 40 dihedral angle constraints ( 4 and x ) were used as derived from homonuclear coupling constants and intraresidue NOES. Since only limited chemical analysis of the disulfide bonds was available, the disulfide pairing was essentially established in the course of structure calculations where the NOE distance constraints forced cysteine side chains together so that the disulfide pairing became obvious. The overall fold of the polypeptide backbone is essentially the same without and with using the disulfide bonds as constraints. Figure 1 shows a superposition of eight structures calculated with the distance geometry algorithm DSPACE (Hare Research, Inc., Woodinville ) . Residues 4-46 and 56-64 are well determined. The heavy atoms of the backbone have an average rms distance (rmsd) to the mean coordinates of 1.0 A. The N-terminus and the active site (residues 47-55) that contains the sequence R49-G50-D51are less well defined. The higher variability of the active site appears to be characteristic for many inhibitors of receptors or proteases. Kistrin contains only three aromatic residues-Phe 38, Tyr 67, and His 68. These residues are not in central regions of the protein and as a consequence, there are no large ring current shifts, so that we have little chemical shift dispersion. Assignment and quantification of the NOE cross peaks is more demanding than for other proteins of this size. Isotope labeling with the goal to disperse the nuclear Overhauser effect spectroscopy ( NOESY ) spectra in a third dimension in 3D-nmr experiments has not been possible so far because the protein, although cloned, could not yet be expressed in bacterial systems, probably because the disulfides do not fold in the native conformation in this environ-
ment. The lack of isotope-labeled protein also prohibits measurements of heteronuclear coupling constants that are essential for characterizing torsion angles. Similar limitations prevail for a structure analysis of cytochrome c551 from Pseudomonas aeruginosa. We had no access to isotope-labeled protein. However, there is a much larger spread of chemical shifts due to the ring current shifts originating from many aromatic side chains and the heme group. We have assigned this 82-residue protein4 and determined the solution structure5 on the basis of 479 NOE constraints derived from homonuclear 2D-nmr spectra. Figure 2 shows a bundle of 10 structures that have been calculated with the distance geometry program DG-11.' The rmsd for the main-chain atoms for this ensemble is 1.30 A. In contrast t o kistrin, the protein contains a significant amount of regular secondary structure, including several a-helices, a 0-sheet, and a polyproline helix. Figure 3 shows this structure, which is rather unusual for a globular protein. At present, we are working on a solution structure of cytochrome c551 where the heme iron is replaced with a zinc. This destroys the normal electron transfer function of the protein. Whether the structural basis for cytochrome function can be elucidated remains to be seen.
ASSIGNMENTS OF LARGE PROTEINS USING ISOTOPE LABELING AND HETERONUCLEAR 3D-NMR SPECTROSCOPY Assignments and structure determination of proteins with molecular weights above 10 kD are difficult without isotope labeling. With 15N labeling, assignments can be achieved using heteronuclear 3D nmr spectroscopy, in particular 3D total correlation spectroscopy/ heteronuclear multiple quantum correlation (TOCSY /HMQC) and 3D nuclear Overhauser effect spectroscopy/HMQC (NOESY / HMQC). We have used these techniques for assignments of flavodoxin ( F d ) from Anacystis nidulans7 (169 residues) and the metothrexate complex of human dihydrofolate reductase (hDHFR)' (186 residues, 21.5 kD) . Figure 4 shows a 2D-NOESY spectrum of Fd compared with a cross plane from a 3D NOESY/HMQC spectrum that shows only NOESY cross peaks to amide protons that are coupled to 15Nnuclei with a chemical shift around 125.3 ppm. The spectrum is significantly simplified. Figure 5 shows slices from the 3D NOESY-HMQC spectrum of hDHFR. They were
Figure 1. Solution structure of the anticoagulant protein k i ~ t r i nThe . ~ reactive site containing the RGD sequence is at the apex of the loop on the upper left of the structure.
W
(x,
W
384
WAGNER ET AL.
STRUCTURE AND DYNAMICS OF ISOTOPE ENRICHED PROTEINS
385
386
WAGNER ET AL.
'N = 125.3 pprn
.* *
Y125 C'HS
-.-.
.* t; . 4 * .
...
a '
. . ,ii. I
Y125 CaH
:
.
,**.
..
..
,i\ .r'
..&. ,
;?B'a-,.
.
,
.
I
N126 NH
Figure 4. Comparison of a 2D NOE spectrum of flavodoxin from Anacystis nidulans with a cross plane from a 3D NOESY-HMQC spectrum. Clearly, most of the overlap of the 2D spectrum is resolved in the slice of the 3D spectrum facilitation assignment of cross peaks7 The 3D data were recorded on a R-500 spectrometer and processed with FELIX (Hare Research, Inc., Woodinville) .
plotted and arranged in a way that only NOEs to a single amide proton are seen in the center of each slice, and the slices are ordered according to their sequence position. Although the number of the amino acids in both proteins differs only by 17 residues, the assignments in hDHFR were significantly more demanding, mainly due to the larger line widths and shorter transverse relaxation times. Concomitantly, poorer coherence transfer in TOCSY spectra was obtained. The assignments of hDHFR will provide the possibility of characterizing the conformation of this protein in solution. The availability of I5N- and I3C-labeledprotein is being used to measure many homonuclear and heteronuclear coupling constant^.^-^^ We expect that these additional constraints will lead to relatively welldefined structures, despite the large molecular weight of these proteins.
DETERMINATION OF THE SOLUTION STRUCTURE OF THE ELASTASE INHIBITOR EGLlN C USING MANY NOE AND COUPLING CONSTANT CONSTRAINTS We have determined the solution structure of the 70-residue protein eglin c.13 More than 950 inter-
residue NOEs were identified, and ca. 550 NOE cross peaks were measured quantitatively. The NOEs were calibrated against intraresidue and sequential NOEs with known distance ranges. In addition, homonuclear H"-H@coupling constants were estimated from comparison with simulated correlated spectroscopy cross peak patterns, l4 and heteronuclear l5N-H@ vicinal coupling constants were measured from 2D TOCSY and 3D NOESY spectra of 15N-labeledprotein.g These coupling constant measurements provided an almost complete characterization of the dihedral angles xl. In addition, some constraints on the dihedral angle qi were obtained from cases where -the HN-Ha coupling constants had extreme values so that the range of qi angles could be identified unambiguously. Figure 6 shows a bundle of 10 structures calculated with the distance geometry algorithm DG-11.' They are representative examples from a larger set of calculations. Out of 50 calculations, only one did not converge since the polypeptide chain was trapped in an incorrect topology. This became evident from the fact that this structure had much higher error functions and large residual violations of constraints. For the other 49 structures, almost all constraints were fulfilled nearly perfectly. There is no residual violation of distance constraints
STRUCTURE AND DYNAMICS OF ISOTOPE ENRICHED PROTEINS
387
"p"
Q
Q
s
I175 K176 Y177 K178 F175
Figure 5. Assignments in hDHFR using slices from a 3D NOESY-HMQC spectrum. The narrow slices from 3D tiers are plotted so that they show only NOE cross peaks to a certain amide proton. The slices are arranged in an array corresponding to their sequential order. Intraresidue cross peaks are labeled with boxes, sequential cross peaks are connected with solid lines.8
larger than 0.2 A. The structure is well defined between residues 8 and 38 and between 50 and 68 (Figure 6). The average rmsd of the backbone atoms is around 0.6 8, between the different structures. The N-terminal heptapeptide (right-hand side in Figure 6 ) is completely disordered with respect to the backbone conformation, and the protease binding loop (bottom of Figure 6) has a large variability with an average rmsd of ca. 1.5 A. Analysis of the dihedral angles shows that the 6 and $ angles are well defined in the entire protein except for the N-terminal heptapeptide and the residues at the beginning and the end of the binding loop. In the middle of the binding loop these angles are also well defined. This indicates that the binding loop undergoes hinge bending mo-
tions. This may be relevant for function of the proteinase inhibitor. The binding loop has to have a certain adaptability to fit the target protein.
ATTEMPTS TO CHARACTERIZE MOBILITY BY MEASURING RELAXATION PARAMETERS Variability of structure as manifested in high rmsds in structures calculated from NOE data does not necessarily mean that these parts of the structure are mobile. The variability can also be just due to lack constraints. Mobility can be directly analyzed via measurements of relaxation parameters. In particular, if 15N-labeledprotein is available, the rota-
388
WAGNER ET AL.
STRUCTURE AND DYNAMICS OF ISOTOPE ENRICHED PROTEINS
tional diffusion of the N-H bond vectors relative to the external magnetic field is manifested in the relaxation parameters. Relaxation times are related to the so-called spectral density function J ( w ) . This spectral density function is essentially a function describing the distribution of rotational diffusional motions with the frequency w of the N-H bond vector. Traditionally, longitudinal relaxation times, TI, transverse relaxation times, T 2 ,and heteronuclear NOES are measured. If we consider relaxation of 15N nuclei, these parameters sample the spectral density function at five frequencies: o = 0 , w N , OH CON, W H , and OH - O N . Measurement of only three relaxation parameters cannot determine the spectral density function at these five frequencies. Therefore, we developed a strategy l5 to measure six relaxation parameters, the longitudinal relaxation rate R (N,) , the transverse relaxation rates of inphase and antiphase coherence, R ( Nx,y)and R (N,,H,) , the heteronuclear cross-relaxation rate R (H, --* N,) , the relaxation rate of longitudinal two-spin order R (N,H,) , and the longitudinal proton relaxation
+
389
rate R ( H,) . These parameters are linearly related to the spectral density function at the frequencies described above. Once these parameters are measured, solving a system of linear equations provides the values of the spectral density function at these frequencies. Figure 7 shows which relaxation parameters sample the spectral density function at which frequencies. Measurements of a complete set of such relaxation parameters for the protein eglin c are in progress.
This work was supported by grants from NIH (project GM38608 and T32-GM08270 to J W P ) , NSF (project DMB9007878), and Daymon Runyon-Walter Winchell Cancer Research Fund (fellowship to BJS). We thank Dr. M. GrCitter and Dr. D. Heinz, CIBA-GEIGY, Basel, Switzerland, for providing a sample of "N-labeled eglin c, and Drs. R. Lazarus and M. Dennis for providing kistrin. We are grateful to Dr. Dennis Hare for providing DSPACE and the data processing software FELIX. We thank Dr. T. Have1 for providing the distance geometry program DG-11.
390
WAGNER ET AL.
REFERENCES 1. Dennis, M. S., Henzel, W. J., Pitti, R. M., Lipari, M. T., Napier, M. A., Deisher, T. A., Bunting, S.& Lazarus, R. A. (1990) Proc. Natl. Acud. Sci. U S A 87, 2471-2475. 2. Adler, M. & Wagner, G. Biochemistry, in press. 3. Adler, M., Lazarus, R. A., Dennis, M. S. & Wagner, G. (1991) Science 2 5 3 , 445-448. 4. Detlefsen, D. J., Thanabal, V., Pecoraro, V. L. & Wagner, G. (1990) Biochemistry 29,9377-9386. 5. Detlefsen, D. J., Thanabal, V., Pecoraro, V. L. & Wagner, G. (1991) Biochemistry 30,9040-9046. 6. Havel, T. F. (1991) Prog. Biophys. MoZ. Biol. 5 6 , 4 3 78. 7. Clubb, R. T., Thanabal, V., Osborne, C. & Wagner, G. (1991) Biochemistry 3 0 , 7718-7730. 8. Stockman, B. J., Nirmala, N. R., Wagner, G., Delcamp, T. J., DeYarman, M. T. & Freisheim, J. H. (1991)
Biochemistry, in press.
9. Montelione, G. T., Winkler, M. E., Rauenbuehler, P. & Wagner, G. (1989) J. Magn. Reson. 82,198-204. 10. Montelione, G. T. & Wagner, G. (1989) J.Am. Chern. S O C . 111,5474-5475. 11. Wagner, G., Thanabal, V. & Schmieder, P. (1991) J. Magn. Reson. 93,436-440. 12. Schmieder, P., Thanabal, V. & Wagner, G. (1991) J. Am. Chem. SOC.113,6323-6324. 13. Hyberts, S. G., Goldberg, M. S., Havel, T. F. & Wag-
ner, G., Protein Science, submitted. 14. Hyberts, S. G., Marki, W. & Wagner, G. (1987) Eur. J. Biochem. 164,625-635. 15. Peng, J. W. & Wagner, G. (1992) J. Magn. Reson.,
in press.
Received June 10, 1991 Accepted July 25, 1991
SYNOPSIS
Structural studies of globular proteins by nmr can be enhanced by the use of isotope enrichment. We have been working with proteins enriched with 15N,and with both 15N and 13C. Due to the isotope enrichment we could assign several large proteins with up to 186 residues and could address structural questions. Furthermore, we can accurately measure heteronuclear and homonuclear vicinal coupling constants. This involves in part multidimensional multiple resonance experiments. This is important for characterization of minor conformational changes caused by mutations. We have also made use of isotope enrichment to study the internal mobility of proteins. We also have developed novel methods for measuring accurately "N relaxation parameters, in particular transverse relaxation rates. This has led us toward a method for directly mapping spectral density functions of the rotational motions of N-H bond vectors in proteins. The protein systems that are discussed include the unlabeled proteins kistrin and cytochrome c551, and the labeled proteins eglin c, a flavodoxin, and human dihydrofolate reductase.
INTRODUCTION Previously, structural studies of proteins by nmr were limited to proteins of molecular sizes of less than 10-12 kD, and the proteins were handled as rigid molecules, neglecting internal mobility due to the lack of strategies for handling internal dynamics. This situation has dramatically changed when isotope labeling became routinely available in the last couple of years. This tool opened the doors to a wide variety of heteronuclear nmr techniques. In particular, three-dimensional and four-dimensional 15N and 13C dispersed heteronuclear methods and multiple resonance multidimensional techniques have enabled us to address structural problems of much larger proteins we did not think of a few years ago. In addition, new methods for measuring homonuclear and heteronuclear vicinal coupling constants have become available, relying on isotope enrichment, providing a whole new class of conformational
Biopolymers. Vol. 32, 381-390 (1992) 0 1992 J o h n Wiley & Sons, Inc.
CCC 0006-3525/92/040381-10$04.00
parameters. Isotope labeling also provides access to studying motional aspects in proteins via measuring a large variety of relaxation parameters and will probably provide a relatively complete description of fast internal motions in proteins. Considering these new technologies, better descriptions of the dynamic conformations of larger proteins will be obtained soon. Here we compare structural studies of two small proteins for which isotope labeling was not available with investigations of "N- and 13C-labeledproteins. Assignments, structure determination, and measurements of relaxation parameters will be discussed.
STRUCTURAL STUDIES OF UNLABELED PROTElNS While isotope enrichment facilitates assignments and can provide additional structural parameters, it is not always available. In particular, it is often difficult to express small disulfide-rich proteins in Escherichia coli systems. We have recently deter381
382
WAGNER ET AL.
mined the structures of two proteins for which isotope enrichment was not available. These proteins are kistrin and cytochrome c551. Kistrin is a protein from the Malaysian pit viper Agkistrodon rhodostoma.' It is a potent inhibitor of platelet aggregation, and functions by inhibiting the platelet-bound receptor GPIIbIIIa, which is a member of the integrin family. Platelet-bound GPIIbIIIa recognizes RGD sequences of fibrin networks, which leads to formation of blood clots. Kistrin contains an RGD sequence that is believed to be the central part of the site interacting with GPIIbIIIa. Binding of kistrin leads to inactivation of the receptor. Kistrin consists of 68 residues and contains six disulfide bonds. We have assigned the proton nmr spectrum2and determined the three-dimensional (3D) structure, relying on conventional two-dimensional (2D) nmr techniques. A set of 549 nuclear Overhauser effect (NOE ) constraints were used for the structure calculations. In addition, 40 dihedral angle constraints ( 4 and x ) were used as derived from homonuclear coupling constants and intraresidue NOES. Since only limited chemical analysis of the disulfide bonds was available, the disulfide pairing was essentially established in the course of structure calculations where the NOE distance constraints forced cysteine side chains together so that the disulfide pairing became obvious. The overall fold of the polypeptide backbone is essentially the same without and with using the disulfide bonds as constraints. Figure 1 shows a superposition of eight structures calculated with the distance geometry algorithm DSPACE (Hare Research, Inc., Woodinville ) . Residues 4-46 and 56-64 are well determined. The heavy atoms of the backbone have an average rms distance (rmsd) to the mean coordinates of 1.0 A. The N-terminus and the active site (residues 47-55) that contains the sequence R49-G50-D51are less well defined. The higher variability of the active site appears to be characteristic for many inhibitors of receptors or proteases. Kistrin contains only three aromatic residues-Phe 38, Tyr 67, and His 68. These residues are not in central regions of the protein and as a consequence, there are no large ring current shifts, so that we have little chemical shift dispersion. Assignment and quantification of the NOE cross peaks is more demanding than for other proteins of this size. Isotope labeling with the goal to disperse the nuclear Overhauser effect spectroscopy ( NOESY ) spectra in a third dimension in 3D-nmr experiments has not been possible so far because the protein, although cloned, could not yet be expressed in bacterial systems, probably because the disulfides do not fold in the native conformation in this environ-
ment. The lack of isotope-labeled protein also prohibits measurements of heteronuclear coupling constants that are essential for characterizing torsion angles. Similar limitations prevail for a structure analysis of cytochrome c551 from Pseudomonas aeruginosa. We had no access to isotope-labeled protein. However, there is a much larger spread of chemical shifts due to the ring current shifts originating from many aromatic side chains and the heme group. We have assigned this 82-residue protein4 and determined the solution structure5 on the basis of 479 NOE constraints derived from homonuclear 2D-nmr spectra. Figure 2 shows a bundle of 10 structures that have been calculated with the distance geometry program DG-11.' The rmsd for the main-chain atoms for this ensemble is 1.30 A. In contrast t o kistrin, the protein contains a significant amount of regular secondary structure, including several a-helices, a 0-sheet, and a polyproline helix. Figure 3 shows this structure, which is rather unusual for a globular protein. At present, we are working on a solution structure of cytochrome c551 where the heme iron is replaced with a zinc. This destroys the normal electron transfer function of the protein. Whether the structural basis for cytochrome function can be elucidated remains to be seen.
ASSIGNMENTS OF LARGE PROTEINS USING ISOTOPE LABELING AND HETERONUCLEAR 3D-NMR SPECTROSCOPY Assignments and structure determination of proteins with molecular weights above 10 kD are difficult without isotope labeling. With 15N labeling, assignments can be achieved using heteronuclear 3D nmr spectroscopy, in particular 3D total correlation spectroscopy/ heteronuclear multiple quantum correlation (TOCSY /HMQC) and 3D nuclear Overhauser effect spectroscopy/HMQC (NOESY / HMQC). We have used these techniques for assignments of flavodoxin ( F d ) from Anacystis nidulans7 (169 residues) and the metothrexate complex of human dihydrofolate reductase (hDHFR)' (186 residues, 21.5 kD) . Figure 4 shows a 2D-NOESY spectrum of Fd compared with a cross plane from a 3D NOESY/HMQC spectrum that shows only NOESY cross peaks to amide protons that are coupled to 15Nnuclei with a chemical shift around 125.3 ppm. The spectrum is significantly simplified. Figure 5 shows slices from the 3D NOESY-HMQC spectrum of hDHFR. They were
Figure 1. Solution structure of the anticoagulant protein k i ~ t r i nThe . ~ reactive site containing the RGD sequence is at the apex of the loop on the upper left of the structure.
W
(x,
W
384
WAGNER ET AL.
STRUCTURE AND DYNAMICS OF ISOTOPE ENRICHED PROTEINS
385
386
WAGNER ET AL.
'N = 125.3 pprn
.* *
Y125 C'HS
-.-.
.* t; . 4 * .
...
a '
. . ,ii. I
Y125 CaH
:
.
,**.
..
..
,i\ .r'
..&. ,
;?B'a-,.
.
,
.
I
N126 NH
Figure 4. Comparison of a 2D NOE spectrum of flavodoxin from Anacystis nidulans with a cross plane from a 3D NOESY-HMQC spectrum. Clearly, most of the overlap of the 2D spectrum is resolved in the slice of the 3D spectrum facilitation assignment of cross peaks7 The 3D data were recorded on a R-500 spectrometer and processed with FELIX (Hare Research, Inc., Woodinville) .
plotted and arranged in a way that only NOEs to a single amide proton are seen in the center of each slice, and the slices are ordered according to their sequence position. Although the number of the amino acids in both proteins differs only by 17 residues, the assignments in hDHFR were significantly more demanding, mainly due to the larger line widths and shorter transverse relaxation times. Concomitantly, poorer coherence transfer in TOCSY spectra was obtained. The assignments of hDHFR will provide the possibility of characterizing the conformation of this protein in solution. The availability of I5N- and I3C-labeledprotein is being used to measure many homonuclear and heteronuclear coupling constant^.^-^^ We expect that these additional constraints will lead to relatively welldefined structures, despite the large molecular weight of these proteins.
DETERMINATION OF THE SOLUTION STRUCTURE OF THE ELASTASE INHIBITOR EGLlN C USING MANY NOE AND COUPLING CONSTANT CONSTRAINTS We have determined the solution structure of the 70-residue protein eglin c.13 More than 950 inter-
residue NOEs were identified, and ca. 550 NOE cross peaks were measured quantitatively. The NOEs were calibrated against intraresidue and sequential NOEs with known distance ranges. In addition, homonuclear H"-H@coupling constants were estimated from comparison with simulated correlated spectroscopy cross peak patterns, l4 and heteronuclear l5N-H@ vicinal coupling constants were measured from 2D TOCSY and 3D NOESY spectra of 15N-labeledprotein.g These coupling constant measurements provided an almost complete characterization of the dihedral angles xl. In addition, some constraints on the dihedral angle qi were obtained from cases where -the HN-Ha coupling constants had extreme values so that the range of qi angles could be identified unambiguously. Figure 6 shows a bundle of 10 structures calculated with the distance geometry algorithm DG-11.' They are representative examples from a larger set of calculations. Out of 50 calculations, only one did not converge since the polypeptide chain was trapped in an incorrect topology. This became evident from the fact that this structure had much higher error functions and large residual violations of constraints. For the other 49 structures, almost all constraints were fulfilled nearly perfectly. There is no residual violation of distance constraints
STRUCTURE AND DYNAMICS OF ISOTOPE ENRICHED PROTEINS
387
"p"
Q
Q
s
I175 K176 Y177 K178 F175
Figure 5. Assignments in hDHFR using slices from a 3D NOESY-HMQC spectrum. The narrow slices from 3D tiers are plotted so that they show only NOE cross peaks to a certain amide proton. The slices are arranged in an array corresponding to their sequential order. Intraresidue cross peaks are labeled with boxes, sequential cross peaks are connected with solid lines.8
larger than 0.2 A. The structure is well defined between residues 8 and 38 and between 50 and 68 (Figure 6). The average rmsd of the backbone atoms is around 0.6 8, between the different structures. The N-terminal heptapeptide (right-hand side in Figure 6 ) is completely disordered with respect to the backbone conformation, and the protease binding loop (bottom of Figure 6) has a large variability with an average rmsd of ca. 1.5 A. Analysis of the dihedral angles shows that the 6 and $ angles are well defined in the entire protein except for the N-terminal heptapeptide and the residues at the beginning and the end of the binding loop. In the middle of the binding loop these angles are also well defined. This indicates that the binding loop undergoes hinge bending mo-
tions. This may be relevant for function of the proteinase inhibitor. The binding loop has to have a certain adaptability to fit the target protein.
ATTEMPTS TO CHARACTERIZE MOBILITY BY MEASURING RELAXATION PARAMETERS Variability of structure as manifested in high rmsds in structures calculated from NOE data does not necessarily mean that these parts of the structure are mobile. The variability can also be just due to lack constraints. Mobility can be directly analyzed via measurements of relaxation parameters. In particular, if 15N-labeledprotein is available, the rota-
388
WAGNER ET AL.
STRUCTURE AND DYNAMICS OF ISOTOPE ENRICHED PROTEINS
tional diffusion of the N-H bond vectors relative to the external magnetic field is manifested in the relaxation parameters. Relaxation times are related to the so-called spectral density function J ( w ) . This spectral density function is essentially a function describing the distribution of rotational diffusional motions with the frequency w of the N-H bond vector. Traditionally, longitudinal relaxation times, TI, transverse relaxation times, T 2 ,and heteronuclear NOES are measured. If we consider relaxation of 15N nuclei, these parameters sample the spectral density function at five frequencies: o = 0 , w N , OH CON, W H , and OH - O N . Measurement of only three relaxation parameters cannot determine the spectral density function at these five frequencies. Therefore, we developed a strategy l5 to measure six relaxation parameters, the longitudinal relaxation rate R (N,) , the transverse relaxation rates of inphase and antiphase coherence, R ( Nx,y)and R (N,,H,) , the heteronuclear cross-relaxation rate R (H, --* N,) , the relaxation rate of longitudinal two-spin order R (N,H,) , and the longitudinal proton relaxation
+
389
rate R ( H,) . These parameters are linearly related to the spectral density function at the frequencies described above. Once these parameters are measured, solving a system of linear equations provides the values of the spectral density function at these frequencies. Figure 7 shows which relaxation parameters sample the spectral density function at which frequencies. Measurements of a complete set of such relaxation parameters for the protein eglin c are in progress.
This work was supported by grants from NIH (project GM38608 and T32-GM08270 to J W P ) , NSF (project DMB9007878), and Daymon Runyon-Walter Winchell Cancer Research Fund (fellowship to BJS). We thank Dr. M. GrCitter and Dr. D. Heinz, CIBA-GEIGY, Basel, Switzerland, for providing a sample of "N-labeled eglin c, and Drs. R. Lazarus and M. Dennis for providing kistrin. We are grateful to Dr. Dennis Hare for providing DSPACE and the data processing software FELIX. We thank Dr. T. Have1 for providing the distance geometry program DG-11.
390
WAGNER ET AL.
REFERENCES 1. Dennis, M. S., Henzel, W. J., Pitti, R. M., Lipari, M. T., Napier, M. A., Deisher, T. A., Bunting, S.& Lazarus, R. A. (1990) Proc. Natl. Acud. Sci. U S A 87, 2471-2475. 2. Adler, M. & Wagner, G. Biochemistry, in press. 3. Adler, M., Lazarus, R. A., Dennis, M. S. & Wagner, G. (1991) Science 2 5 3 , 445-448. 4. Detlefsen, D. J., Thanabal, V., Pecoraro, V. L. & Wagner, G. (1990) Biochemistry 29,9377-9386. 5. Detlefsen, D. J., Thanabal, V., Pecoraro, V. L. & Wagner, G. (1991) Biochemistry 30,9040-9046. 6. Havel, T. F. (1991) Prog. Biophys. MoZ. Biol. 5 6 , 4 3 78. 7. Clubb, R. T., Thanabal, V., Osborne, C. & Wagner, G. (1991) Biochemistry 3 0 , 7718-7730. 8. Stockman, B. J., Nirmala, N. R., Wagner, G., Delcamp, T. J., DeYarman, M. T. & Freisheim, J. H. (1991)
Biochemistry, in press.
9. Montelione, G. T., Winkler, M. E., Rauenbuehler, P. & Wagner, G. (1989) J. Magn. Reson. 82,198-204. 10. Montelione, G. T. & Wagner, G. (1989) J.Am. Chern. S O C . 111,5474-5475. 11. Wagner, G., Thanabal, V. & Schmieder, P. (1991) J. Magn. Reson. 93,436-440. 12. Schmieder, P., Thanabal, V. & Wagner, G. (1991) J. Am. Chem. SOC.113,6323-6324. 13. Hyberts, S. G., Goldberg, M. S., Havel, T. F. & Wag-
ner, G., Protein Science, submitted. 14. Hyberts, S. G., Marki, W. & Wagner, G. (1987) Eur. J. Biochem. 164,625-635. 15. Peng, J. W. & Wagner, G. (1992) J. Magn. Reson.,
in press.
Received June 10, 1991 Accepted July 25, 1991
