The Solution Conformations of Ferrichrome and Deferriferrichrome Determined by H- NMR Spectroscopy and Computational Modeling

'

K. L. CONSTANTINE, A. DE MARCO," M. MADRID, C. 1. BROOKS 111, and M. LLINASt

Department of Chemistry, Carnegie Mellon University, Pittsburgh, Pennsylvania 15213-3809

SYNOPSIS

We have applied computational procedures that utilize nmr data to model the solution conformation of ferrichrome, a rigid microbial iron transport cyclohexapeptide of known x-ray crystallographic structure [D. van der Helm et al. (1980) J. Am. Chem. Soc. 102, 4224-42311. The A13+ and Ga3' diamagnetic analogues, alumichrome and gallichrome, dissolved in &-dimethylsulfoxide (4-DMSO ) , were investigated via one- and two-dimensional 'H-nmr spectroscopy at 300, 600, and 620 MHz. Interproton distance constraints derived from proton Overhauser experiments were input to a distance geometry algorithm [ T. F. Have1 and K. Wuthrich ( 1984) Bull. Math. Biol. 46,673-6911 in order to generate a family of ferrichrome structures consistent with the experimental data. These models were subsequently optimized through restrained molecular dynamics /energy minimization [ B. R. Brooks et al. (1983) J . Comp. Chem. 4, 187-2171. The resulting structures were characte :xed in terms of relative energies and conformational properties. Computations based on integration of the generalized Bloch equations for the complete molecule, which include the "N- 'H dipolar interaction, demonstrate that the x-ray coordinates reproduce the experimental nuclear Overhauser effect time courses very well, and indicate that there are no significant differences between the crystalline and solution conformations of ferrichrome. A similar study of the metal free peptide, deferriferrichrome, suggests that at least two conformers are present in &-DMSO at 23°C. Both are different from the ferrichrome structure and explain, through conformational averaging, the observed amide NH and CH" multiplet splittings. The occurrence of interconverting peptide backbone conformations yields an increased number of sequential NH-CH" and NH-NH Overhauser connectivities, which reflects the ( r - 6 ) dependence of the dipolar interaction. Our results support the idea that, in the case of structurally rigid peptides, moderately accurate distance constraints define a conformational subspace encompassing the "true" structure, and that energy considerations reduce the size of this subspace. For flexible peptides, however, the straightforward approach can be misleading since the nmr parameters are averaged over substantially different conformational states.

I NTRODUCTI 0N Recent developments in high-resolution nmr spectroscopy ','and associated computational methods 3-7 have made possible the rigorous investigation of 0 1990 John Wiley & Sons, Inc. CCC 0006-3525/90/3-40239-18 $04.00 Biopolymers, Vol. 30, 239-256 (1990) * Deceased. Dr. De Marco's permanent affiliation was with the: Instituto di Chimica delle Macromolecole, Consiglio Nazionale delle Ricerche, 20133 Milano, Italy. ' T o whom correspondence should be addressed.

polypeptide spatial structures in solution and the characterization of conformational changes they may experience. While these methods are under active development, it is important to further test their capabilities and limitations by examining extreme cases, e.g., rigid vs flexible peptides. Against this background, we have undertaken a study of the solution conformation of the microbial iron transport cyclohexapeptide ferrichrome,' and of its metal free derivative, deferriferrichrome. The primary structure is cyclo- ( Gly3-Gly'-Gly'-Orn3-Orn'-0rn'), where O m stands for N6-acetyl-N*-hydroxy-L-or239

240

CONSTANTINE ET AL.

0

0 It

and high resistance towards denaturant^.'^ In contrast, deferriferrichrome behaves dynamically like a normal, flexible cyclopeptide. These properties were revealed by previous nmr investigations, *'-15 which also demonstrated the general suitability of diamagnetic A13+and Ga3+analogues of ferrichromealumichrome and gallichrome-as useful isomorphous l6 model systems. The large structural and dynamic changes induced by metal ion binding are manifest in the ordinary one-dimensional (LD)'Hnmr spectra '' of alumichrome and deferriferrichrome (Figure 2). The increased chemical shift dispersion induced by complexation is particularly pronounced in the amide NH region, reflecting the unique chemical environments conferred by conformational stabilization. Here we extend these studies by reporting 1D transient proton Overhauser experiments performed on gallichrome a t 600 MHz and extensive two-dimensional ( 2D) 'H-nmr characterization of alumichrome and deferriferrichrome, including 2D nuclear Overhauser effect spectroscopy (NOESY) l7 experiments a t 620 MHz. All studies were performed on samples dissolved in deuterated dimethylsulfoxide (4-DMSO). This information, supplemented with previously obtained data, 11-15 is used to characterize conformation and flexibility. From the nuclear Overhauser enhancement (NOE) data, sets of interproton distance constraints are derived and initial

0

CH2-C - N - C H z I

H I

I

I

HH

c=o

C=O ---HN @H2 I NH I

I

@CH I

- CH2- CH2- CH2-N -CI

O=C

0

I

\

CH3 II

0

/

Figure 1 Diagram of ferrichrome: cyclo-triglycyl-tri(N6acetyl-N6-hydroxy-L-ornithyl). Dashed lines indicate H bonds. Residues are numbered according to previous convention, 13,34 which is nonstandard. M denotes a trivalent metal ion.

nithine residues, which bind high-spin Fe3+in a distorted octahedral geometry via the hydroxamate bidentates ( Figure 1) . Octahedral coordination was confirmed by x-ray analysis,' and was inferred earlier by electron paramagnetic spectroscopy, '' which indicated a strong crystal field of low symmetry. Ferrichrome assumes a very rigid conformation untypical of most small peptides, exhibiting markedly retarded 'H-'H exchange kinetics for four of the amide NH protons, l1 high thermal stability, ''

NH

Ha Deferriferrichrome

02

Id0

d 6 2

oh

I

1

9:O

0.0

3b 1I

7.'0

Alumichrome

I I

1.o

6.0

PPM

Figure 2 The 1D 'H-nmr spectra of deferriferrichrome ( A ) and alumichrome ( B ) at 300 MHz. Chemical shift changes induced by A13+binding are traced explicitly for all NH and Orn CH" protons. Recorded at N 53°C; impurities are denoted by an asterisk. The broad resonance at 9.6 ppm arises from the hydroxamic acid NOH protons; hence, it is absent from the alumichrome spectrum ( B ) .

-

SOLUTION CONFORMATION OF FERRICHROME

structures are generated using the metric matrix distance geometry (DG) program DISGE0.'s*'9 These structures are subsequently optimized and refined via the molecular mechanics/dynamics program CHARMM." This procedure aims at searching conformation space for energetically stable, constraint-satisfying structure^.'*'^ We find that the solution conformation of ferrichrome is virtually identical to the x-ray structure? For deferriferrichrome, the data are inconsistent with a single conformation. In this case, the most simple description is that of a dynamic equilibrium between two wellpopulated conformational states, neither of which resembles ferrichrome. In summary, the objectives of the work reported in this paper are ( a ) to determine the conformation of ferrichrome independent of the x-ray data, thereby testing the above-mentioned techniques on this model system; ( b ) to note any possible differences between ferrichrome in the solution and solid states; and ( c ) to characterize the conformational state ( s ) of deferriferrichrome in &-DMSO.

MATERIALS A N D METHODS Experimental

Alumichrome and gallichrome were generated from deferriferrichrome as previously de~cribed.''~~~ Deferriferrichrome was obtained by extracting Fe3+ from ferrichrome overnight using a 40-fold excess of 8-hydroxyquinoline (Sigma, St. Louis, MO) in a methanol-water solution. The free and complexed 8-hydroxyquinoline was subsequently removed by repeated chloroform extraction followed by evaporation under reduced pressure to yield a dry powder. All samples were dissolved in 0.5 mL &-DMSO (Merck, St. Louis, MO) to concentrations of ca. 100 m M (gallichrome) or 70 m M (alumichrome and deferriferrichrome) . No evidence of aggregation is found at these concentrations. Tetramethylsilane was used as an internal reference standard. 'H-nmr spectra were recorded at 300 MHz with a Bruker WM-300 spectrometer, and at 600 and 620 MHz with the spectrometer of the National NMR Facility for Biomedical Research at Carnegie Mellon University. The 1D transient NOE spectra of gallichrome were recorded at 600 MHz by selective inversion of the signal of interest, and percent NOE values were determined for delay times of 0.07,0.16, 0.30, 0.55, 0.97, and 1.50 s. The 2D chemical shift correlated (COSY) spectra were recorded at 620

241

MHz in both the absolute value and phase sensitive mode^.'^.^^ RELAYED-COSY experimentsz6 with several different delay times were recorded in the absolute value mode at 300 MHz. NOESY experiments 17*'427 were performed a t 620 MHz with mixing times ( 7 , ) of 0.12, 0.20, 0.28, and 0.44 s. Cross-peak volumes were estimated by evaluating peak base areas from contour plots and peak shapes and heights from stacked plots. We estimate that volumes measured this way are accurate to f 25%. The 1D spectra were acquired with quadrature detection and, when appropriate, resolution enhanced via Gaussian multiplication. Deferriferrichrome subspectra were simulated using the Bruker PANIC program. The 2D nmr experiments were acquired with 512 equally spaced evolution time ( t') periods with 16 or 32 transients averaged along t2. The time domain data were zero filled and multiplied by an unshifted sine bell in both dimensions (300 MHz) or the Varian PSEUDO weighting function (620 MHz) to yield frequency domain data matrices of 1024 X 1024 (absolute value) or 2048 X 2048 (phase-sensitive ) data points. The spectral widths in the tzdimension were generally about 3500 Hz at 300 MHz and 7000 Hz at 620 MHz. The 2D spectra are shown symmetrized. Interproton distance information was derived from NOE data2,'8 by assuming isotropic motion with a single correlation time for all interacting spins. Initial NOE buildup rates are proportional to r l 6 , where ri, is the internuclear distance between protons i and j. With this approach, rG is regarded as an upper distance limit,' which can be determined from

where NOEM is a calibration NOE based on a known or estimated internuclear distance ru. For the rigid ferrichrome analogues, the distance between the O m 2 CH" and NH protons is estimated to be 2.8 f 0.1 8, using the measured value O f 3 J , (6.0 ~ ~Hz) and the appropriate Karplus relationship, 29,30 which indicates -60" 2 6 2 -180". Over this range, raN varies between ca. 2.7 and 2.9 A. The corresponding NOE intensity was measured accurately from 1D transient NOE experiments. For deferriferrichrome, the Gly3 NH and Om' CH" resonances manifest a very strong NOE at all mixing times; this was taken as evidence of an interproton distance 5 2.3 A, affording a reasonable and convenient calibration standard. Overlap of NOESY cross peaks between geminal protons prevents their use in establishing a calibration standard.

242

CONSTANTINE E T AL.

Computational

All calculations were performed on microVAX I1 or VAX 11/780 computers. DG calculations were carried out with the DISGEO p r ~ g r a r n . " For ~ ~ ~DISGEO computations, interproton distance constraints were cast in the appropriate pseudoatom represent a t i ~ n . Stereospecific ~' assignments for every methylene proton in ferrichrome analogues have been previously obtained 1 4 ; however, these were based in part on x-ray data. Since we are interested in generally applicable methods, constraints to methylene protons were taken to pseudoatoms located midway between the geminal pair, using appropriate increases in the upper distance b o u n d ~Hydroxamate .~~ oxygen atoms belonging to different side chains were constrained t o lie between 2.8 and 4.2 A apart, corresponding, respectively, to the smallest cis and largest trans distance expected for a distorted octahedral hydroxamate-Fe3+ complex. Note that these 1 2 constraints do not in themselves assume any absolute configuration for the metal binding site. Distance constraints involving identified H bonds were included in most of the ferrichrome calculations. Restrained energy minimization ( E M ) and restrained molecular dynamics ( M D ) were carried out using CHARMM.20T h e CHARMM-19 parameter and topology files were modified to incorporate the O m residues, treating the Orn CH" and all amide N H protons explicitly. Bond lengths and angles for the hydroxdmate moiety were taken from the x-ray s t r ~ c t u r e Iiiternuclear .~ distance constraints were incorpoiated into the CHARMM potential energy function by including terms of the following form:

.

kcal mol-1A-2. T h e hydroxamate partial atomic charges were set equal t o zero for this part of the calculation. In stage B, Fe3+was placed in the center of geometry defined by the hydroxamate oxygens, charges were scaled to idealized small values, the interoxygen constraints were removed, and six Fe0 constraints with r ; = 2.0 A and kl = k2 = 180.0 kcal- m o l - ' k 2 were incorporated. All energy minimizations were performed with the steepest descent alogorithm followed by the adopted basis NewtonRaphson ( ABNR) routine.20 In principle, discontinuous second derivatives due to the half-harmonic constraint terms can interfere with the restrained ABNR minimizations; in practice, no evidence for such interference was found. Also, ABNR has a n advantage over first-derivative techniques in that it can effectively avoid saddle points on the potential surface.20A distance-dependent dielectric constant numerically equal to the internuclear separation in A was used for all calculations. Restrained MD simulations were carried out using the method of Verlet32t o integrate Newton's equations of motion. A time step = 1 fs was employed, and all covalent bonds involving explicit hydrogens were constrained by the SHAKE routine.33Consistency between the crystalline and solution conformations of ferrichrome was verified via simulation of the time courses of the 1D NOE spectra. Deferriferrichrome structures were generated following a protocol similar to that outlined above as well as by a n alternative strategy whereby a consistent structure was proposed, built, and subsequently characterized with respect to the experimental data.

RESULTS AND DISCUSSION Far NOE and H-bond constraints the potential was made effectively half harmonic by setting k2 = 22.0 k c a l - m o l - ' A k 2and kl = 2.2 X kcal-mol~'~-2. T h e value of k2 allows upper bound violations of 0.2-0.3 A t o easily occur, reflecting the uncertainty in our distance estimations. All distances involving pseudoatoms were referred to the corresponding extended carbon. T h e r i values were consistent with the upper bounds used in the DG calculations. T h e optimization of ferrichrome structures was carried out in two stages: the first without (stage A ) , and the second with (stage B ) metal bound. Hydroxamate oxygens on different side chains were constrained in stage A by interoxygen potential terms with r i = 3.5 A, and k1 = k2 = 1.8

Ferri zhrome

NMR Analysis. Phase-sensitive and absolute value COSY experiments a t 620 MHz allowed for a complete identification of the alumichrome spin systems. Connectivities between the O m CH" (4.12-4.76 p p m ) , CH"."' (1.12-2.67 p p m ) , CHYsY'(1.51-1.75 p p m ) , and CH6,&'(3.20-4.03 ppm) resonances are readily located (Figure 3 ) . Figure 4 shows the results of a RELAYED-COSY experiment a t 300 MHz optimized to detect multiplets coupled through 3 J E 5 Hz. All NH-CH" scalar and NH-CHP relayed connectivities appear in the 2D display. Thus, the combined COSY /RELAYED-COSY data overdetermine the spin-system identifications. The NH-CH " and NH-NH NOE connectivities (discussed below)

SOLUTION CONFORMATION OF FERRICHROME

243

+H a , H 8 4 +HP, Hy+

10

15 :

2 0

8 8 : .

2 5

.. ... .

*"

8

PPM

%# 11'

3 0

3 5

40

4 5

,c

0

4 5

4 0

35

30

25

2 0

I5

to

PPM

Figure 3 Absorption mode 'H-nmr COSY spectrum of alumiciirome at 620 MHz: aliphatic CH connectivities. Cross-peak regions are indicated on the right, and 1D spectral regions are labeled above the spectrum. Gly CH"-CH" geminal cross Feaks are labeled explicitly. The symbols 0denote cross peaks that are not detected in this spectrum, but were observed in a magnitude mode COSY spectrum of pre-exchanged ('H-'H) alumichrome at 620 MHz. Recorded a t -23OC.

allowed for a straightforward sequential assignment2 of all resonances (excluding the conformationally uninformative O m N*-acetyl methyl peaks). The 2D results corroborate the proton assignments for ~ ferrichrome analogues previously r e p ~ r t e d . ' The latter, which included stereospecific assignments of all geminal protons, were based on 1D homonuclear decoupling experiments, comparison of homologues, and reference to the crystal structure of the related peptide ferrichrome A.34Chemical shifts for alumichrome, measured directly from the 2D spectra a t 23"C, are listed in Table I. For completeness, we retain the stereospecific labeling3s of prochiral protons, but only nonstereospecific information is used for modeling. Figure 5 shows the NH-NH ( A ) and NH-CH" ( B ) cross-peak region of a 620-MHz NOESY experiment with a mixing time 7, = 0.44 s. Strong '

NH-NH cross peaks between Glyl and O m 3 and between O m 1 and O m 2 ,along with NH-CH NOEs linking Gly3 to O m ' , Gly' to Gly3, Gly' to Gly', and O m ' to O m 3 reveal the presence' of a type I1 0-turn containing Gly and Gly' and a type I 0-turn structured by O m ' and Om'. In addition, the formation of a small antiparallel &sheet containing Gly3 and O m 3 is evidenced by the NOEs between Orn3 NH and Gly3 NH [shown boxed; also present in transient 1D NOE experiments (not shown) ] and between O m 3 NH and Gly2 CHa. The experiment also reveals a cross peak between the Orn2 NH and CH* protons (shown circled). This combined evidence, in conjunction with DG modeling without Hbond constraints and consideration of other criteria (amide temperature coefficients" and 'H-*H exchange rates" ), serves to establish the existence of a n intraresidue H bond between the O m 2 NH and

'

244

CONSTANTINE ET AL.

-z0

-8.0

PPM

9.0

10.0

5!0

I

4.0

" . '

I

3.0

"

"

210

I

1.0

PPM

Figure 4 The 'H-nmr RELAYED-COSY spectrum of alumichrome at 300 MHz. NHCH" (direct) and NH-CH@(relayed) connectivities are indicated by solid and dashed lines, respectively. The experiment was optimized for 3J = 5.0 Hz. Reference 1D spectra are shown along the b1 and b2 axes. Recorded at e 5 3 " C .

hydroxamate NO" atoms and the existence of a transannular H bond between the Om3NH and Gly3 0' atoms. Transient NOE difference spectra were recorded at 600 MHz following selective inversion of the Om3 CH" and of the Gly', Gly3, and all Orn amide NH magnetizations. As an example, Figure 6 shows the Table I Alumichrome Proton Chemical Shifts ( f O . O 1 ppm) at 23°C in 4-DMSO"

NH CH" CHN2 CHU3 CH" CHP3 CHY2 CHr3 CH62 CHS3 a

Gly'

Glf

Gly3

Orn'

Om2

Om3

9.01

8.93

6.89

6.46 4.18

10.07 4.12

7.87 4.76

3.49 3.80

3.70 3.45

3.79 3.68 1.12 2.08 1.51 1.71 3.73 3.42

Stereospecific labeling follows Ref. 14.

2.67 1.69 1.95 1.60 4.03 3.65

1.78 1.61 1.28 1.52 3.63 3.20

NOE difference spectrum recorded with a delay I = 0.55 s after irradiation of the Om' NH transition. A strong NOE is apparent at the Om3 CHm resonance; according to the x-ray s t r u ~ t u r ethis , ~ proton is only 2.00 A from the Om' NH. Figure 7 illustrates the time course of the transient NOE response for those protons spatially close to the O m 2NH proton. Most NOE buildups exhibit reasonably linear behavior up to delay times of ca. 0.2 s. Additional NOE intensities and buildup rates were estimated from NOESY cross-peak volumes. Explicit interproton distance estimates, derived from the 1D and 2D NOE data, are listed in Table 11. Modeling. The ferrichrome structure generation

and optimization protocol is summarized in Scheme I. The derived ferrichrome structures were characterized by root mean square deviations (RMSDs) to the x-ray coordinates and to average coordinates calculated a t each optimization step. Selected structures were further characterized by constraint violations, dihedral angle differences, internal energies, additional RMSDs, consistency with J-coupling information, and overall resemblance to the x-ray

SOLUTION CONFORMATION OF FERRICHROME

246

60

/

/*'

61 PPM

/

/ /

/ /

/ /

/

/

/ /

I

/

1 00

d/

00'0'

00

00

,-.

o2dQ

1

3:.

60

40

62 PPM

Figure 5 The 'H-nmr NOESY spectrum of alumichrome at 620 MHz. ( A ) NH-NH connectivities. ( B ) NH-CH" connectivities ( fingerprint region ) . The diagonal is indicated by a dashed line ( - - - ) . Cross peaks are labeled X "Y m , where X and Y refer to the resonances along b1 and b2, respectively. A dashed box is used to highlight a weak NH-NH connectivity, and the Orn' NH-Om2 CHa cross peak is circled with dashed trace. The mixing time was 0.44 s. Conditions are the same as for Figure 3. ~

structure. The x-ray structures before and after restrained EM were evaluated using relaxation theory.

tions. Constraints reflecting the identified H bonds were also employed. The O m 2 NH * * 0 '' O m 2 H bond was considered to be "strong" in view of the low-field chemical shift ( 10.07 ppm) , the weak chemical shift temperature dependence12 ( A 6 / A T

-

Distance Geometry. Using pseudo atom^,^^ 18 NOEderived constraints were used for the DG calcula-

NOE

10

0

4

6

2

0

PPm

Figure 6 Transient NOE experiment on gallichrome at 600 MHz. ( A ) The 1D reference spectrum. ( B ) NOE difference spectrum obtained from inversion of the O m 2 NH magnetization. Selected peaks are labeled. The delay time was 0.55 s. Recorded at ~ 2 5 ° C .

246

CONSTANTINE ET AL.

Intraresidue NOES

Yo

NOE

Interresidue NOEs

8.0

6.0

4.0

2.0

0.0 0.0

0.2

0.6

0.4

0.8

1.0

1.2

1.4

Time (sec.)

Figure 7 NOE time course: experimental (symbols) and computed (solid lines) responses to gallichrome O m 2 NH magnetization inversion a t 600 MHz. The theoretical curves were calculated using the XMIN coordinates. ( A ) and ( B ) reflect intraresidue and interresidue NOEs, respectively. The experimental error for the NOE values, estimated from the signalto-noise ratio, is f 0.2%. Different correlation times (those giving the best individual fits) were used for the computed curves, with the following correspondences: O m 2 CHa3,T , = 0.58 ns; O m 2 CH", 7, = 1.60 ns; O m 2 CHo2,T , = 1.70 ns; O m 3 CHP2,T , = 1.26 ns; Orn' NH, T~ = 1.70 ns; Orn3 CH", T~ = 0.65 ns; and Orn3 CHY3,7, = 0.80 ns. Experimental conditions are given in the caption to Figure 6.

=

-1.90 X

ppm.K-'), and the retarded 'H-

2H exchange kinetics" ( tlIP = 210 min in D20, pD 5.14,297 K ) of the Om2amide NH resonance. Thus, the H * * 0 distance constraint ( r & = 2.05 A ) was assigned a value reflecting a short H bond. For the Om3NH * * O=C Gly3H bond, a looser constraint, r& = 2.30 A, was employed. Twelve different ferrichrome DISGEO structures were obtained with the constraints derived above. The average all-atom RMSD value of the 1 2 DG coordinate sets (DGO1-DG12 ) to the crystal structure (henceforth referred to as XRAY) is 1.17 A, with high and low values of 1.32 A (DG05) and 0.73 A (DG06), respectively. The 1 2 DG structures were averaged to obtain DGAV; the RMSD between this structure and XRAY is 0.82 A. Between the individual DG structures and DGAV, the RMSD ranges between 1.04 A (DG12) and 0.67 A (DGOl) with a mean value of 0.85 A. ( A s discussed below, it turns

-

-

out that DG12 leads to the "best" individual structure after EM/MD optimization and refinement; this demonstrates the danger of relying on DG alone to produce nmr-derived structures.) Stereoviews of 6 representative DG structures and 2 extreme structures (relative to the x-ray coordinates) are presented in Figure 8 (A and B ) , respectively. Structure Optimization. Using the stage A potential (no explicit metal ion), the DG structures were subjected to 100 steps of steepest descent followed by 600 steps of ABNR minimization. Most calculations converged to an energy decrease per step of less than lo-' kcal- mol-'. The minimized structures (EMAO1-EMA12) were then used as starting coordinates for 6 ps of restrained MD at 600 K, and the average coordinates from the last 2 ps of each run were computed. This yielded 12-stage A dynamics structures (RDAO1-RDA12). The simulation

SOLUTION CONFORMATION OF FERRICHROME

Table I1 Interproton Distance Estimates Obtained from 1D and 2 D NOE Experiments on Diamagnetic Ferrichrome Analogues Proton Pairs

Distance

1

2

Gly' CH"'

Gly' NH Gly' NH Gly2 NH Gly2 CHU3

orn3NH ~

NH ~ 3 NH ~ 3 ~ 1 NH ~ 3 ~ 1 NH ~ 3 ~ 1 NH ~ 3 ciY3 NH Om' NH Om' NH Om' CHY3 O m 2 NH Orn2 NH Om2 NH Om2 NH Om2 NH Orn' CH" orn3NH

~

1

~

1

1

CH ~ "~ 3

orn3NH om3NH Om' NH Om' CH" Om' CHo3 orn3C H Y ~ Om3 CHY3 Om2 CH" Om2 NH Om2 CHm om3CH" om3C H ~ Om3 CHo3 Om3 CHY3 Om2 CH63 O m 2 CH63 om3C H T ~

(A)

NOE

X-Ray"

2.5 2.8 2.3 4.0 3.7 2.4 3.1 4.0 3.1 2.8 3.5b 2.7 3.5 3.1 2.3 3.5 4.0 2.6 3.5 2.9

2.14 2.69 2.32 3.71 3.81 2.37 2.77 3.98 3.51 2.62 3.46 2.62 3.01 2.58 2.00 3.43 3.06 2.27 3.57 2.76

247

For continued optimization we focused on two sets of coordinates: ( a ) the one produced by averaging the RDB structures (RDBAV), and ( b ) the "best" individual structure. In order to identify the latter independent of the x-ray data, the RDB structures were subjected to unconstrained EM so as to ascertain the most energetically favorable structure in the absence of constraints. (All RDB structures satisfy the experimental constraints so well that constraint violations turn out not to be suitable criteria for selecting particular structures.) The average internal energy after unconstrained EM is -53.82 kcal mol-', with RDBl2 having the lowest energy (-58.84 kcale mol-') . Reassuringly, this structure has the smallest RMSD to both RDBAV and XRAY, and was taken to be the best structure. Using stage B constraints, RDBAV, RDB12, and XRAY were energy minimized to produce RDBAVM, RDBlBM, and XMIN, respectively. RMSD values between these ,three structures and XRAY are reported in Table IV. The stereoview of XRAY and XMIN are shown in Figure 9A, and Figure 9B shows RDBAVM and RDBl2M fitted to XRAY. The constrained internal energies of XMIN, RDBlBM, and RDBAVM are -60.05, -58.71, and -55.26 kcal mol-', respectively. The largest con+

a H atom positions were built on to the x-ray structure using standard bond lengths and bond angles (D. J. States, unpublished). NOE not observed; 3.5 8, is approximately the maximum allowed d m N ( i , 1).

+

I

NOE data

I DG structures

1

length was based on the observation that the rms fluctuations settled after 2-3 ps of restrained dynamics; i.e., stable ( or metastable) conformations are attained very quickly. Since we are interested in structure optimization only, and not in computing physically meaningful dynamical average properties, 2 ps of averaging was deemed sufficient. Stage B optimization was initiated by placing an iron ion in the center of geometry of the hydroxamate oxygens. Restrained EM was then applied with the stage B constraints, producing structures EMBO1-EMB12. At this point, two of the structures (EMB10 and EMB11) located the same local minimum, with an RMSD between the two of 1 X A. The resulting 11EMB structures were subjected to restrained MD (as outlined above) with the stage B constraints, yielding RDBO1-RDB12. (The original numbering is retained to avoid confusion. There is no RDBlO structure.) Relevant RMSD values between structures produced to this point are listed in Table 111.

restrained EM with stage A constraints (EMA structures)

1 restrained MD with stage A constraints (RDA structures)

1

restrained EM with stage B constraints (EMB structures)

1 restrained MD with stage B constraints

unrestrained EM

Scheme I

248

CONSTANTINE E T AL.

(A)

m1

Figure 8 Stereoviews of ferrichrome distance geometry structures. (A) Six representative DG outputs. ( B ) The ferrichrorne x-ray structure (heavy trace), and the DG structures with the minimum (DG6) and maximum (DG5) RMSDs to the x-ray. Residues and mainchain carbonyl oxygens are labeled.

straint violation is found in RDBAVM, for which the O m 2internal NH-0'' distance is 2.19 A (constraint distance = 2.05 A).

Dihedral angles are compared in Table V. All C#J angles in XRAY, XMIN, RDBAVM, and RDB12M are consistent with experimental J a ~a1ues.l~ ~ ~ The

Table I11 Selected RMSD Values Between Various Ferrichrome Structures RMSD

(A)

Set"

Aveb

Largest'

Smallestd

Avee

Largest'

Smallestg

Aveh

DG EMA RDA EMB RDB

1.17 1.07 0.95 0.98 0.95

1.32 [5]' 1.37 [7] 1.29 [7] 1.35 [7] 1.36 [7]

0.73 0.70 0.51 0.48 0.42

0.85 0.77 0.73 0.76 0.74

1.04 [12] 0.97 [4] 0.98 [4] 0.99 [7] 1.00 [7]

0.67 0.58 0.56 0.56 0.58

0.82 0.76 0.64 0.65 0.65

(61 [6] [6] [12] [12]

[l] [8] [6] [12] [12]

'There are 11or 12 structures for each set (see text). Average structure for each set is the coordinate set obtained by averaging the individual structures. Average RMSD between the set structures and the XRAY structure. ' Largest RMSD relative to the XRAY structure. Smallest RMSD relative to the XRAY structure. 'Average RMSD between the set structures and the average structure for the set. Largest RMSD relative to the average structure for the set. Smallest RMSD relative to the average structure for the set. RMSD between the XRAY structure and the average structure for the set. Numbers in brackets indicate particular structure with corresponding high or low RMSD value, e.g., [ 5 ] = DG5 in this case.

SOLUTION CONFORMATION OF FERRICHROME

Table IV RMSDs (A) Between the Crystallographic and Computed Ferrichrome StructuresP

XRAY XMIN RDBAVM RDB12M a

XRAY

XMIN

RDBAVM

RDBIPM

0

0.26 0

0.54 0.45 0

0.45 0.35 0.27 0

Described in the text.

largest differences between the nmr-derived structures and the x-ray results are the conformations of the Orn' and Orn' side chains of RDBAVM, and the conformation of the Om' side chain of RDB12M. This correlates with the energy ordering XMIN < RDBlZM < RDBAVM. We also note that the major difference between the backbone conformations of XRAY and the energy minimized structures involves a Gly' lc/ angle rotation of ca. 35". Although the resulting angles are still within the

+

249

expected range for a type 11 (lc/ = 120" f 40" ) ,this difference may reflect a minor departure of the solid state from the solution conformation, as crvstal Dacking forces may have an influence on cyclohexapeptide conformation^.^^ The main source of ambiguity in our nmr-derived structures are the O m side-chain conformations. We therefore examined the effects of constraining the Orn x angles to one of the three sterically allowed values (60°, -60", or 180"). These constraints were derived from the gallichrome 3Ja0 coupling constants14 and the NOE data in accordance with established procedure.38O m 3 x1 is constrained to 60" since both JaOcoupling constants are < 4 Hz. The X' angles of both Orn' and Orn' are constrained to -60" as both residues have one JmO > 1 2 Hz and the other < 4 Hz, and only one strong intraresidue NH-CHBNOE is observed for each residue. A final round of restrained EM/ MD calculations was performed starting with RDBl2M and RDBAVM in which the X angles were constained using potential terms k ( x ' - x A ) 2 , with k = 80 kcal-mol-' rad-'. I

'

'

Figure 9 Stereoviews of optimized ferrichrome structures. ( A ) XRAY (heavy trace) and XMIN, which was obtained by restrained EM of XRAY. ( B ) XRAY (heavy trace), RDBlPM (the best optimized structure after restrained E M ) , and RDBAVM (the average optimized structure after restrained EM). The latter structure is denoted by 63 at the O m 2side chain.

250

CONSTANTINE E T AL.

Table V Comparison of Dihedral Angles Between the Crystallographic and Computed" Ferrichrome Structures Dihedral Angle Differences (deg) Dihedral Angleb

XRAY (deg)

XRAY-

XMIN

RDBAVM

RDB12M

G' # G' J. G'G2 w G2# G2J. G2G3w G3# G3J.

77.0 11.0 176.4 -62.5 129.7 -175.3 165.0 -162.7 179.6 -119.6 21.7 -64.4 159.4 -59.5 141.6 -167.5 -67.1 -32.5 -64.7 77.8 44.9 90.1 -173.0 -163.1 -175.6 60.6 -179.7 -51.2 112.6 -168.5

-10.0 -14.9 -0.6 -0.6 35.1 -1.0 -8.5 -1.8 0.5 0.2 -0.8 3.0 -5.6 4.5 9.2 3.9 -2.4 2.0 -2.5 3.0 -6.9 3.4

-8.9 -17.3 -1.3 0.1 39.6 -0.1 -23.1 4.9 2.3 32.5 -5.1 -114.9 -47.4 147.7 41.3 13.4 2.4 -24.8 -14.4 -74.0 107.2 -69.9 -11.1 -6.9 4.2 -8.5 7.4 6.0 -11.7 13.1

-10.2 -17.6 -1.2 -0.2 39.1 0.1 -21.4 7.0 -1.7 28.2 -7.8 -114.9 -47.2 145.2 37.7 7.2 -4.5 -10.9 2.0 3.2 -9.8 4.3 -6.5 -7.7 0.5 -6.8 -0.1 10.8 -7.5 14.0

0 1 ~ 3

*

0' 4 0'

0'x' 0'x 2 0'

x3

0' x4 0'O2 w o2#

o2J. 0 2

XI

0 2 x2

o2x3 0 2 x4

0203

o34 o3

*

0 3

XI

0 3 ~2 0 3 ~3 0 3

~4

~ 3 0 1

a

0.5

-7.3 0.0 -7.1 3.1 8.4 -5.3 14.1

XRAY-

XRAY-

Described in the text. G and 0 label Gly and Orn residues, respectively.

All of the data presented thus far indicate that the solution and crystalline conformations of ferrichrome are essentially identical. In order to rigorously check for consistency between XRAY and XMIN and the NOE data, their respective coordinates were used to solve the generalized Bloch equati on^.^' The initial conditions were chosen to obtain the time dependence of the magnetization after selective 44% inversion of the O m 2 NH resonance, thus simulating the experimental results for gallichrome (Fig. 7 ) . The spin-lattice and cross-relaxation rates were calculated assuming dipole-dipole interaction as the principal relaxation me~hanism.~' For the amide protons, appropriate terms were included to account for the effects of the 'H-I4N dipolar intera~tion.'~ Neglect of this interaction results in a ca. 14% increase in the computed NOEs at 450 ms. The 69Gaand 71Ganuclear spins ( I = 3/2 ) produced negligible effects on the computed NOEs. The calculated curves are found to be in excellent agreement with experiment (Figure 7).Both coordinate sets yield nearly identical results. By varying the effective local correlation time ( 7 , ) values between 0.58 ns and 1.70 ns, all NOE time courses measured experimentally could be reproduced using either the XRAY or the XMIN coordinates. These 7 , values, which reflect molecular motions at 25"C, are in good agreement with the previously reported value of ca. 0.4 ns at 45°C.41Since local 7 , values are not known to high accuracy, it is not possible using the available data to determine which of the very similar XRAY and XMIN structures is in better agreement with experiment; however, it is certain that no significant differences between the crystalline and solution conformations of ferrichrome exist.

Deferriferrichrorne NMR Analysis. In contrast to ferrichrome, deferri-

RDBl2M was subjected to restrained EM to yield RDBlBF, which has an RMSD to XMIN of 0.07 A. RDBAVM was subjected to restrained MD followed by restrained EM to yield RDBAVF with an RMSD to XMIN of 0.29 A. Constraining Om' X is sufficient to correct the conformation of the entire side chain by restrained EM only, while constraining O m 2 x' failed to correct the remaining X angles of the RDBAVM O m 2 side chain (Table V ) , even when constrained MD was employed. Little change was expected to result from constraint of the O m 2 and Om3 x1 angles, as they were already near their optimal values.

'

ferrichrome is a highly flexible peptide.l2z4'Figure 10 shows the 3.2 < 6 < 4.4 ppm CH" region of the deferriferrichrome 1D spectrum ( A ) and the corresponding NH-CH" connectivity regioh of a COSY spectrum ( B ) , at 620 MHz. Figures 1OC and 11 show, respectively, the NH-CH" and NH-NH regions of a 620 MHz NOESY spectrum (7, = 0.44 s ) . At this mixing time, all 9 sequential NH-CH" and all 6 sequential NH-NH connectivities are observed. Although flexible, the range of NOE intensities and J coupling constants (Table VI) observed indicate a nonrandom conformation. On the basis of these data and RELAYED-COSY experiments (not shown), sequence-specific resonance assignments were obtained for all protons except the de-

SOLUTION CONFORMATION OF FERRICHROME

(B) 80

85

I

8, PPM

' 80

85

PPM Figure 10 'H-NMR spectra of deferriferrichrome at 620 MHz: NH-CH" connectivities. ( A ) The 1D reference spectrum of the CH" and CH6regions. ( B ) COSY fingerprint region. Solid lines link geminal Gly cross peaks arising from the same residue. ( C ) NOESY fingerprint region. Interresidue NH-CH" cross peaks are labeled as in Figure 5 . The complete sequential pathway is traced out with horizontal solid lines connecting cross peaks arising from the NH resonances and vertical dashed lines joining those arising from the CH" resonances. Experimental conditions are as given in Figures 3 and 5 .

generate O m CHPY',CH$&',and acetyl CH3 groups (Figure 2A). Observed chemical shifts and J values, refined by spectral simulations, are reported in Table VI. Although all sequential connectivities are present in the T, = 0.44 s NOESY spectrum, a number of these are relatively weaker a t shorter mixing times, reflecting slow initial NOE buildups. T o remain within the two-spin approximation [ Eq. ( 1) 3 , only NOES observed a t short mixing times (7, I0.20 s ) were used t o derive distance constraints. This conservative approach greatly reduced the size of the data set suitable for use in modeling. T h e NOEderived distance estimates are listed in Table VII.

25 1

Modeling. Given the potential complicating effects of internal dynamics, the following approach was adopted First, a structure evaluation was performed using a computational strategy similar to that used for ferrichrome. Second, a n independent model was rationally proposed and tested for consistency with the nmr observables. Since no constraints were obtained for the Orn side chains, the DG computations were performed on cyclo- ( Gly3-Gly2-Gly l-Ala3-Ala2-Ala1). Out of 20 attempts, no calculations successfully completed the final minimization of chirality and constraint error functions; however, 1 2 complete structures were embedded.3 Although these structures generally satisfy the experimental constraints poorly, and many even have the wrong chirality a t the Ala ( O m ) C" positions, their backbone conformations afforded suitable starting structures for further refinement. For the restrained E M / M D calculations, two dihedral angle constraints were employed in addition t o NOE distance constraints. Based on the 3 J , ~ ~ coupling constants (Table VI) , the 6 angles of O m 2 and Orn3 were constrained to near 120" with harmonic terms ( k = 22.0 kcal mol-' rad-' ) . Orn side chains, with the hydroxamic acid 0 atoms protonated and with small idealized charges, were incorporated into the starting structures using the BUILD routine of CHARMM.20 The resulting structures were subjected to restrained EM (100 steps of steepest descent followed by 600 steps of ABNR), followed by 5 ps of restrained dynamics a t 500 K. Coordinates were averaged over the last 2 ps of each run, and the resulting structures were optimized by a second round of restrained EM. A t this point, the average RMSD between the backbone atoms of each individual structure and the average coordinate set was 1.28 A; the corresponding allatom value was 2.61 A. Of the 1 2 final structures, 9 satisfy the distance constraints with no upper bound distance violations > 0.2 A. The average internal energy of the 9 constraint-satisfying structures is 18.39 kcal. mol-', with high and low values of 25.90 and 6.05 kcal mol-', respectively. These results indicate that the backbone conformation is underdetermined by the distance and dihedral angle constraints. Additional information was taken into account in order to lower the number of possible conformations. The reduced temperature dependences of the O m 2and Gly2N H resonances in &-DMSO (A6/ AT = -3.17 X and -2.61 X ppm/K, respectively ) suggest that these protons are involved in intramolecular H bonding." (On the basis of 1D experiments, these resonances had previously been

''

-

252

CONSTANTINE ET AL.

8.5

62

8.0

PPM Figure 11 Deferriferrichrome NOESY spectrum at 620 MHz: NH-NH region. Cross peaks are labeled as in Figure 5; the sequential connectivities are traced with solid lines.

misassigned to O m 3 NH and Gly3 NH, respectively.'2) Most likely, these are internal, transannular H bonds, in agreement with previous data obtained for d e f e r r i f e r r i c r ~ i n( ~ ~ substitutes Gly2 Ser2 in deferriferrichrome) in 4-DMSO that indicated H bonds bridging O m 2 and Ser2. Inspection of the 9 remaining structures revealed that two had formed transannular H bonds between O m 2 and Gly2. Reassuringly, these are the two lowest energy struc-

tures: 6.05 and 8.57 kcal-mol-'. The RMSD between the backbone atoms of these structures is only 0.12 A. The backbone conformations of the two structures are reminiscent of the cyclohexane boat conformation; therefore, the structures will be referred to as BOAT1 and BOAT2, respectively. The backbone dihedral angles for both structures are listed in Table VIII, and a stereoview of the BOAT2 structure is presented in Fig. 12A.

Table VI Deferriferrichrome Proton Chemical Shifts (6) and Spin-Spin Coupling Constants ( 3 J )at 23OC in 4 - D M S O

6 (ppm)" Residue

NH

CH"

CH"'

Gly' GI3 Gly" Om' Om2

8.455 8.089 8.306 8.037 7.857 8.252

3.80 3.78 3.83 3.983 4.257 4.107

3.65 3.68 3.55

orn3

3J (Hz)",~

CHB

1.64 1.75 1.81

CHB'

1.54 1.57 1.51

3 J u ~ ~

5.0 k 0.9 4.8 k 0.4 6.6 f 0.5 5.4 f 0.2 8.3 +_ 0.1 8.1 f 0.1

3J0wi

Jasl

Ja,z

8.1 k 0.9 5.3 k 0.4 5.3 f 0.5 6.0 f 0.3 4.2 f 0.1 4.4 k 0.1

6.0 k 0.3 8.4 k 0.1 8.6 f 0.1

a The spectral parameters are best estimates based on spectral simulations of experimental data. All amide NH and Om CH" 6 values are accurate to f0.002 ppm. Orn CH8 and Gly CH" 6 values are accurate to fO.O1 ppm. 3J4 and 'J=pvalues are obtained from simulation of the CH" multiplets; the correspondence between CHB,CH8' and stereospecific CH", CH&are not established.

253

SOLUTION CONFORMATION OF FERRICHROME

Table VII Interproton Distance Estimates Obtained from Deferriferrichrome NOESY Data

Proton Distance’ 1

Gly’ NH Gly’ NH Gly’ CH” Gly’ CH“’ Gly2 NH ciy3NH O m ’ NH Om’ NH Om‘ NH a

2

Gly2 CH“ Gly2 CH”’ orn3NH orn3NH ~ 1 NH ~ 3 Om’ CH” Om2N H Om2CH“ orn3NH

(A) 3.0 2.8 2.3 3.1 3.2 2.3 2.9 2.4 2.7

Accurate to 210% if due to a single conformation.

3 J , coupling ~ ~ constants for the BOAT coordinate sets were calculated with the two more established Karplus” relationships, the “ferrichrome curve”14 and the “BPTI c ~ r v e , ’ which ’ ~ ~ differ slightly in the values of the empirical parameters. The results of this analysis are presented in Table IX. We restrict our analysis to the Om3J a N H coupling constants. The BOAT-type structures consistently yield computed 3 J avalues ~ ~ that are somewhat larger than the experimental values for all three Om residues with both parameter sets. The BOAT-type structures also fail to explain the fact that all backbone proton sequential NOE connectivities are observed in the 0.44 s NOESY spectrum of deferriferrichrome. This is unlikely to arise solely from spin diffusion since the alumichrome NOESY data indicates minimal such effects under identical conditions. We suggest that the observations reflect conformational averaging. Recent studies 44-46 of cyclic peptides have demonstrated the existence of complex internal motions (e.g., amide plane rotations) that are slow with resDect to overall tumbling. but are fast on the chemical shift and relaxation time scales. These motions may be reflected in nmr data, such as NOE values and J coupling constants, that are incompatible with a single stable conform a t i ~ n This . ~ ~ results from the ( rP6) dependence of the dipolar interaction, and from the averaging of the 3 J values. In such a situation, it is judicious to attempt to identify the simplest conformer interconversion scheme that can account for all the available data.44 A classic47 model for deferriferrichrome is a Schwyzer-type structure* with two @-turnscon-

nected by an antiparallel @-sheet.We explored the possible occurrence of such a conformer in dynamic equilibrium with the BOAT-type structure. By considering the short interproton distances expected in different secondary structures* and the range of dihedral angles associated with these conformations, 36 we postulate a deferriferrichrome conformer that contains two type I1 @-turns-one occupied by Om1 and Gly3, the other by Om3 and Glyl-and an antiparallel @-sheetformed by Gly’ and Om’. Coordinates consistent with this model were constructed, starting from the ferrichrome backbone, by using restrained EM to constrain dihedral angles and distances associated with the proposed transannular H bonds to their desired values. Dihedral angles for the resulting structure (hereafter referred to as SWZR) are listed in Table VIII, and a stereoview is presented in Figure 12B. As expected, this structure satisfies all of the NOE distance constraints except for the one between Om NH and Om NH, which is violated by ca. 1.5 A. Table IX lists the experimental and computed 3 J adata. ~ ~With both Karplus parameter sets, the 3 J avalues ~ ~ calculated on the basis of the SWZR structure for all three Orn residues are smaller than the experimental values. This contrasts the results obtained for the BOAT-type structures. If a rapid interconversion between the BOAT-type and SWZR structures is assumed (which is fast compared to the lifetimes of these states), a [BOAT]/ [ SWZR] population ratio of 0.69/0.31 (0.47/0.53) gives an excellent fit between the experimental coupling constants and those calculated using the ferrichrome

’

Table VIII Backbone Dihedral Angles (Degrees) of the BOAT1, BOATS, and SWZR Deferriferrichrome Models Dihedral Angle

BOAT1

BOAT2

SWZR

69.4 -85.1 -74.2 74.8 97.0 -46.3 56.0 69.3 -113.2 62.5 -118.3 -0.8

67.9 -85.6 -75.5 72.0 105.7 -48.4 52.7 65.1 -114.5 63.1 -115.3 0.5

-62.0 102.9 -160.4 -168.0 86.9 1.1 -60.0 113.8 -158.6 167.9 62.7 34.9

I ,

GlY’ 4 GlY’ $ GlY24 Gly2$

Gly3 ~ 1$ ~ Orn’ 4 Om’ $ Om2 4 Om2 $ Orn3

orn3G

3

254

CONSTANTINE E T AL.

G'

-2

Figure 12 Stereoviews of computed deferriferrichrome structures. ( A ) BOAT2 model and ( B ) SWZR model. Residues are labeled.

( B P T I ) parameters. Finally, we note that, since the O m ' NH-Om2 N H distance in the BOAT-type conformation is 3.0 A, a 0.6/0.4 [BOAT]/ [ SWZR] ratio yields a n effective interproton distance of 3.2 A. This is, within the error, consistent with the experimental estimate (2.9 A ) .

CONCLUSIONS We have derived the solution conformation offerrichrome in a manner independent of crystallographic data by combining nmr data with computational structure generation and optimization techniques. The value of optimization via restrained E M / MD has been demonstrated and corroborates previous such studies on structurally less defined polypeptide^.^ Our analysis underscores the value of reasonably accurate NOE distance constraints. The effects of incorporating x dihedral angle constraints have been evaluated. In favorable cases, e.g., O m ' ,

constraining this single dihedral angle to the correct value can propagate a conformational change to the remaining side-chain degrees of freedom. Incorrect side-chain conformations led to structures of slightly greater energies than the energy-minimized x-ray structure. Thus, for the rigid ferrichrome molecule, the CHARMM 19 empirical potential energy function is adequate to distinguish correct from incorrect conformational features. Based on solution of the generalized Bloch equations,39 which include the 14N-'H dipolar contribution, l5 no differences between the crystalline and solution conformations of ferrichrome were found. Minor differences could in principle be identified if precise local 7, values could be obtained. Regarding the nmr-derived structures, both the polypeptide backbone conformation and the correct configuration of the metal binding site were found-a result of direct functional signifi~ a n c e . ~It' is noteworthy that one of the structures incorporating x' constraints (RDB12F) is essentially identical to the energy-minimized x-ray structure.

SOLUTION CONFORMATION OF FERRICHROME

Table IX Experimental and Calculated 3 J , ~ ~ Spin-Spin Coupling Constants (in Hertz)

255

hydroxamate groups and the metal ion completely overwhelm the conformational preferences of the metal-free state.

3JaNH

Residue Om' Orn' Om' Om' Om' Orn' Orn2 Orn2 Om2 Om2 Om2 Orn2

orn3 orn3

Orn" Om:' Om:' Om:'

Structure or Parameter Set BOATl BOAT2 BOATAVE' SWZR .69/.31d .47/.53d

BOATl BOAT2 BOATAVE SWZR .69/.31 .47/.53

BOATl BOAT2 BOATAVE SWZR .69/.31 .47/.53

Exp"

fcmb

BPTIb

5.4 5.4 5.4 5.4 5.4 5.4 8.3 8.3 8.3 8.3 8.3 8.3 8.1 8.1 8.1 8.1 8.1 8.1

6.3 6.2 6.& 4.2 5.6 5.2 8.8 8.8 8.8 6.5 8.1 7.6 8.9 8.9 8.9 6.3 8.1 7.5

6.9 6.8 6.85 4.2 6.0 5.4 9.6 9.6 9.6 6.9 8.8 8.2 9.7 9.7 9.7 6.9 8.8 8.2

Experimental. The calculated values are based on the Karplus relationz9

+

@+

3J,N~= A.COS% B - C O S C where 0 = 1 I$ - 60" 1. The two parameter sets used are the ferrichrome (fcm) curve," with A = 5.4 Hz, B = -1.3 Hz, and C = 2.2 Hz, and the bovine pancreatic trypsin inhibitor with A = 6.4 Hz, B = - 1.4 Hz, and C = 1.9 Hz. ' BOATAVE denotes the average theoretical coupling constant between BOATl and BOAT 2. Denotes a 0.69/0.31 (or 0.47/0.53) weighted average between the BOATAVE and SWZR theoretical coupling constants, respectively.

The deferriferrichrome conformation was not nearly as well defined by the experimental constraints-a result that can generally be anticipated when going from a highly rigid to a flexible molecule. Nonetheless, useful and detailed information regarding conformational states has been obtained, which led us t o propose a dynamic equilibrium between BOAT and SWZR backbone conformations. This simple picture accounts for all the available nmr data. I t should be stressed that more complex conformer equilibria may also be consistent with the data. Nevertheless, it is apparent that none of the major deferriferrichrome conformations resemble ferrichrome in &-DMSO. This result may have functional significance as the large conformational differences may be important for receptor specificity. A major rearrangement accompanies metal ion binding. T h e strong ionic interactions between the

This research was supported by NIH grants GM37554 and HL29409. The 600-620 MHz NMR facility is supported by NIH grant RR00292. KLC is a recipient of a NIH predoctoral traineeship.

REFERENCES 1. Ernst, R. R., Bodenhausen, G. & Wokaun, A. (1987) Principles of Nuclear Magnetic Resonance in One and Two Dimensions, Oxford University Press, New York. 2. Wuthrich, K. (1986) N M R of Proteins and Nucleic Acids, John Wiley & Sons, New York. 3. Havel, T. F., Kuntz, I. D. & Crippen, G. M. (1983) Bull. Math. Biol. 4 5 , 665-720. 4. Braun, W. (1987) Quart. Rev. Biophys. 19, 115-157. 5. Karplus, M. & McCammon, J. A. (1983) Ann. Rev. Biochem. 5 3 , 263-300. 6. Brooks, C. L. 111, Karplus, M. & Pettitt, B. M. ( 1988) Adu. Chem. Phys. 71. 7. Clore, G. M., Briinger, A. T., Karplus, M. & Gronenborn, A. M. (1986) J. Mol. Biol. 191,523-551. 8. Neilands, J. B. ( 1974) in Microbial Iron Metabolism, Neilands, J. B. Ed., Academic Press, New York, pp. 4-34. 9. van der Helm, D., Baker, J. R., Eng-Wilmot, D. L., Hossain, M. B. & Loghry, R. A. ( 1980) J. Am. Chem. SOC.102,4224-4231. 10. Wickman, H. H., Klein, M. P. & Shirley, D. A. (1965) J. Chem. Phys. 42,2113-2117. 11. Llinis, M., Klein, M. P. & Neilands, J. B. (1973) J. Biol. Chem. 248,915-923. 12. Llinhs M., Klein, M. P. & Neilands, J. B. (1970) J. Mol. Biol. 5 2 , 399-414. 13. Llinis, M. & Klein, M. P. (1975) J. A m . Chem. SOC. 97,4731-4737. 14. DeMarco, A., Llinhs, M. & Wuthrich, .K. ( 1978) Biopolymers 17,617-636,637-650. 15. Llinhs, M., Klein, M. P. & Wuthrich, K. (1978) Biophys. J. 24,849-862. 16. van der Helm, D., Baker, J. R., Loghry, R, A. & Ekstrand, J. D. ( 1981 ) Acta Cryst. B 3 7 , 323-330. 17. Macura, S. & Ernst, R. R. (1980) Mol. Phys. 4 1 , 9 5 117. 18. Havel, T. F. & Wuthrich, K. (1984) Bull. Math. Biol. 46,673-698. 19. Havel, T. F. (1986) DISGEO, Quantum Chemistry Program Exchange, No. 507, Indiana University. 20. Brooks, B. R., Bruccoleri, R. E., Olafson, B. D., States, D. J.,Swaminathan, S. & Karplus, M. ( 1983) J. Comp. Chem. 4, 187-217. 21. Kaptein, R., Boelens, R., Scheek, R. M. & van Gunsteren, W. F. (1988) Biochemistry 2 7 , 5389-5395.

256

CONSTANTINE E T AL.

22. Llinh, M., Klein, M. P. & Neilands, J. B. (1972) J . Mol. BWl. 68,265-284. 23. Llinhs, M. & Neilands, J. B. (1976) Biophys. Struct. Mech. 2,105-117. 24. Wider, G., Macura, S., Kumar, A., Ernst, R. R. & Wuthrich, K. (1984) J . Mag.Res. 56,207-234. 25. Nagayama, K., Kumar, A., Wuthrich, K. & Ernst, R. R. (1980) J . Mag.Res. 40,321-334. 26. Eich, G., Bodenhausen, G. & Ernst, R. R. (1982) J . Am. Chem. Soc. 104, 3731-3732. 27. Kumar, A., Ernst, R. R. & Wuthrich, K. (1980) Biochem. Bwphys. Res. Commun. 95, 1-6. 28. Olejniczac, E. T., Gampe, R. T. & Fesik, S. W. (1986) J . Mag,Res. 67,28-41. 29. Karplus, M. (1959) J . Chem. Phys. 3 0 , l l - 1 5 . 30. Bystrov, V. F. (1976) Prog. N M R Spec. 10,41-81. 31. Wuthrich, K., Billiter, M. & Braun, W. (1983) J . Mol. BWl. 169,949-961. 32. Verlet, L. (1967) Phys. Rev. 159,98-103. 33. van Gunsteren, W. F. & Berendsen, H. J. C. (1977) Mol. Phys. 34,1311-1330. 34. Zalkin, A., Forrester, J. D. & Templeton, D. H. (1966) J . Am. Chem. Soc. 88,1810-1814. 35. IUPAC-IUB Commission on Biochemical Nomenclature (1970) J . Mol. Bwl. 5 2 , l - 1 7 . 36. Robson, B. & Garner, J. (1986) Introduction to Proteins and Protein Engineering, Elsevier Science Publishers, New York, pp. 86-87. 37. Kitson, D. H. & Hagler, A. T. (1988) Biochemistry 27,5246-5257. 38. Wagner, G., Braun, W., Havel, T. F., Schaumann, T.,

39. 40. 41. 42. 43. 44.

45. 46.

47. 48.

49.

Go, N. & Wuthrich, K. (1987) J . Mol. Biol. 196,611639. Madrid, M. & Jardetsky, 0. (1988) Biochim. Biophys. Acta 953,61-69. Abragam, A. ( 1961 ) The Principles of Nuclear Magnetism, Clarendon Press, Oxford. Llinh, M., Meier, W. & Wuthrich, K. (1977) Bwchim. Biophys. Acta 492, 1-11. Llinds, M., Klein, M. P. & Neilands, J. B. (1972) Int. J . Peptide Protein Res. 4, 157-166. Pardi, A., Billiter, M. & Wuthrich, K. (1984) J . Mol. Biol. 1 8 0 , 741-751. Kessler, H., Griesinger, C., Lautz, J., Muller, A., van Gunsteren, W. F. & Berendsen, H. J. C. (1988) J . Am. Chem. Soc. 110,3393-3396. Kopple, K. D., Wang, Y., Cheng, A. G. & Bhandary, K. K. (1988) J . Am. Chem. SOC.110,4168-4176. Pepermans, P., Fourwe, D., van Binst, G., Boelens, R., Scheek, R. M., van Gunsteren, W. F. & Kaptein, R. (1988) Biopolymers 27,323-338. Ovchinnikov, Y. A. & Ivanov, V. T. (1975) Tetrahedron 31, 2177-2209. Schwyzer, R. & Wolstenholme, G. E. W., Eds. (1958) Amino Acids and Peptides with Antimetabolic Activity, J. & A. Churchill, Ltd., London. Matzanke, B. F., Muller, G. & Raymond, K. N. (1989) in Iron Carriers and Iron Proteins, Loehr, T. M., Ed., VCH Press, Weinheim.

Received June 15, 1989 Accepted November 6,1989