Conformational Analysis of Ochratoxin A by NMR Spectroscopy and

Ochratoxins constitute a group of toxic metabolites produced by several species (fungi and molds) of Aspergillus and Penicillium.1 These toxins contam...
0 downloads 0 Views 367KB Size
Download PDF
16926

J. Phys. Chem. B 2005, 109, 16926-16936

Conformational Analysis of Ochratoxin A by NMR Spectroscopy and Computational Molecular Modeling Photis Dais,*,† Irine Stefanaki,† Georgia Fragaki,† and Emannuel Mikros‡ NMR Laboratory, Department of Chemistry, UniVersity of Crete, 71409 Iraklion, Crete, Greece, and Department of Pharmacy, DiVision of Pharmaceutical Chemistry, UniVersity of Athens, 15771 Athens, Greece ReceiVed: January 28, 2005

Two-dimensional NMR spectroscopy has been used for a complete assignment of the proton and carbon-13 spectra of the metabolite from Aspergillus ochraceus, ochratoxin A. In addition, phase-sensitive nuclear Overhauser effect spectrometry experiments and computational molecular modeling (MM2* and MMFF force field programs) have been employed to examine the conformational properties of ochratoxin A in chloroform solutions. Particular attention has been given to intramolecular hydrogen-bonding formation involving the phenolic group on dihydroisocoumarin, which may be responsible for the toxic mechanism of ochratoxin A.

Introduction Ochratoxins constitute a group of toxic metabolites produced by several species (fungi and molds) of Aspergillus and Penicillium.1 These toxins contaminate a wide variety of foodstuffs and represent a serious threat to both humans and animals.2,3 The most toxic member of these toxins is ochratoxin A (OTA), which has been implicated primarily in nephrotoxic syndromes in several animal species (especially pigs). Also, it is considered as a causative agent of several other biological activities, including carcinogenic, hepatotoxic, teratogenic, genotoxic, and immunotoxic effects toward animal species, and has been classified as a potent carcinogen to humans.2-6 OTA contaminates a diversity of foods in the normal diet, including cereals and cereal-made foods, dried fruits, beans, coffee, cocoa, groundnuts, spices, and foodstuffs of animal origin, mainly poultry eggs, pork, and milk, including human breast milk.2-8 Recently, significant contamination of wine and beer has been reported.9-11 Chemically, OTA is composed of a substituted dihydroisocoumarin linked to L-phenylalanine (Figure 1). The significant toxicity observed for OTA relative to other metabolites of the same family has been attributed to its free hydroxyl group at C-8 (Figure 1), which forms a phenoxide ion capable of interfering with biochemical processes.12-14 It has been speculated that deprotonation of the phenolic group on dihydroisocoumarin may be involved in the toxic mechanism of OTA, which in turn may be related to the either absorption, elimination, or binding properties of OTA. Recently, it has been proposed15 that the conformation of OTA may play a role in its toxicity. Depending on the conformation of the peptide linkage connecting the phenyl ring and the dihydroisocoumarin moiety (Figure 1), the phenol hydroxyl group may interact through intramolecular hydrogen bonding, with either the amide carbonyl group (form a) or the lactone carbonyl (form b), as shown in Figure 2. OTA in form a requires that the amide carbonyl and the phenol hydroxyl * Author to whom correspondence should be addressed. E-mail: [email protected]. † University of Crete. ‡ University of Athens.

Figure 1. Chemical structure of OTA.

Figure 2. Two forms of OTA indicating possible hydrogen-bond formation involving the phenol hydroxyl group of the dihydroisocoumarin moiety.

group be syn with respect to each other, whereas form b is favored when the amide carbonyl and the phenol groups are anti. Moreover, form b allows the formation of a hydrogenbonding network involving the amide group and the phenol hydroxyl group, the latter acting as a hydrogen acceptor, as well as the hydrogen of the carboxyl group and the amide carbonyl group (Figure 2). In either case, the existence of such an intramolecular hydrogen-bonding network would have a significant influence on the pK of the phenol group and thus on the ability of OTA to interact with vital substrates. In this respect, it is interesting to study the conformational behavior of OTA to draw conclusions about the formation of intramolecular hydrogen bonding and to contribute to the efforts of understanding the toxic mechanism of OTA. In recent years, two-dimensional nuclear Overhauser effect spectrometry (NOESY) has been used in combination with

10.1021/jp058035e CCC: $30.25 © 2005 American Chemical Society Published on Web 08/11/2005

Conformational Analysis of Ochratoxin A geometry and energy refinement calculations to determine molecular structures in solution.16-19 These experiments are based on proton-proton dipolar relaxation, yielding a correlation map between protons of close proximity (e6 Å). The intensities (or volumes) of cross peaks of the NOESY spectrum are related to the distances between the protons (the proton-proton crossrelaxation rates depend on the sixth power of the reciprocal of the distance) and can therefore be used to estimate interproton distances. The present study describes the conformational behavior of this important mycotoxin as assessed by phase-sensitive NOESY experiments and the calculation of the peak intensities by using the complete relaxation matrix analysis (CORMA) method.20,21 A critical feature of CORMA is the explicit treatment of the relaxation network and in particular spin diffusion while calculating the NOESY intensities. Also, this work reports the assignment of all proton and carbon chemical shifts in the corresponding NMR spectra of OTA with a complete set of 1H-1H coupling constants at a high magnetic field strength by employing modern twodimensional NMR techniques. Experimental Section NMR. NMR experiments were conducted on a Bruker DRX400 spectrometer operating at 400.13 and 100.6 MHz for proton and carbon nuclei, respectively. The probe temperature was 30 ( 1 °C as measured by precalibrated thermocouples in the probe insert. Samples of OTA (5 mg) were dissolved in deuterated chloroform (0.5 mL) in 5-mm NMR tubes. Some details of the two-dimensional NMR experiments for OTA are given below. For a review of gradient correlation spectroscopy (COSY), heteronuclear single quantum coherence (HSQC), heteronuclear multiple-bond correlation (HMBC), and phasesensitive NOESY experiments, see ref 22. Gradient COSY (gCOSY). The H,H-COSY spectrum22 was recorded with gradient pulses for selection with spectral widths of 4000 Hz in both dimensions, using 256 increments of 1000 real data points, 16 transients for each free induction decay, and a recycle delay of 1.0 s. The data set was zerofilled to a 2000 × 2000 matrix prior to Fourier transformation. A sine-bell weighting function was used in both dimensions. Gradient HSQC (gHSQC). The phase-sensitive gradientselected hydrogen-carbon HSQC22 experiment was acquired with sweep widths of 15 ppm for 1H and 240 ppm for 13C using 128 increments, 24 transients of 1000 data points, and a delay time τ ) 1.8 ms. The relaxation delay was 1.5 s. Data were processed using a π/4-shifted sine-square bell function and zero-filled along F1 to 256 points prior to Fourier transformation. Gradient HMBC (gHMBC). The phase-sensitive gradientselected hydrogen-carbon HMBC22 experiment was performed using a low-pass J-filter (3.4 ms) and delays of 65 and 36 ms to observe long-range C-H couplings optimized for 3 and 7 Hz with 256 increments and 86 transients of 2000 data points. The relaxation delay was 2.0 s. Zero-filling to a 2000 × 2000 matrix and π/2-shifted sine-square bell multiplication were performed prior to Fourier transformation. NOESY. The phase-sensitive 1H NOESY spectrum was recorded in the time proportional phase increment (TPPI) mode using the (90°-t1-90°-τm-90°-acquire)n pulse sequence,23 with spectral widths of 6000 Hz in both dimensions and mixing times τm ) 0.8, 1.0, 1.2, 1.4, 1.6, and 1.8 s. The relaxation delay was set to 1.3 s. A total of 512 complex t1 data and 2000 real data points in the t2 dimension were acquired with 16

J. Phys. Chem. B, Vol. 109, No. 35, 2005 16927 transients for each free induction decay. The spectra were zerofilled to a final size of 2000 × 2000. A π/2-shifted sine-squared weighting function was applied prior to Fourier transformation. The peak volumes were measured using the software AURELIA provided by Bruker, which sums the intensity in the area around the peak of interest. 13C Relaxation Measurements. 13C relaxation measurements were conducted at the 100.5 MHz 13C Larmor frequency on a Bruker DRX400 spectrometer under broadband proton decoupling. The temperature was controlled to within (1° C by means of precalibrated thermocouples in the probe inserts. Spin-lattice relaxation times (T1) were measured by the standard Inversion Recovery Fourier Transform (IRFT) method with a repetition time longer than 5T1. A total of 400 transients were accumulated, for a set of 10 arrayed t values. The 180° pulse width was 15.0 µs. Values of T1 were determined by a three-parameter nonlinear procedure with root-mean-square error of (5%. The reproducibility of each T1 value was (5%. 13C NOE experiments were carried out by the inverse gated decoupling technique. At least three experiments have been performed at each temperature value. Delays of at least 10 times the longest T1 were used between 90° pulses. NOE values were estimated to be accurate to within (10%. Conformational Calculations. All calculations were carried out using Macromodel 6.5 (Schrondinger)24 running on a Silicon Graphics O2 R5000 computer under IRIX 6.3. Intramolecular energy calculations were performed using the MM2* and MMFF force field (modified MM2)25 programs incorporated in the Macromodel software with the Generalized Born equation/ Surface Area (GB/SA) solvation model.26 The truncated Newton conjugate gradient (TNCG)27 minimization method with an energy convergence criterion of 0.01 kJ/mol was used for geometry optimizations. The conformational energy map in terms of the dihedral angles C15-C14-C13-N12 and C14-C13N12-C11 (Figure 1) was computed with the DRIVE mode of Macromodel at 15° increments. When the potential energy surface of OTA was mapped, two possible conformations about the C7-C11 bond were taken into consideration. One of these conformations assumes a dihedral angle C8-C7-C11-N12 of 0° and is identical to form b (Figure 2), whereas the other (C8-C7-C11-N12 ) 180°) is identical to form a (Figure 2). Calculations using the first conformation resulted in lowerenergy structures by about 5 kcal mol-1 than those obtained adopting the second conformation. It is clear that the conformation with C8-C7-C11-N12 ) 0° is stabilized through the formation of a hydrogen-bonding network with N12-H‚‚‚O-C8 and C8-O-H‚‚‚OdC1. Accordingly, conformers corresponding to an angle C8-C7-C11-N12 of 180° were not taken into account in the present analysis. The initial geometry of OTA was based on the atomic coordinates obtained from the crystal structure,15 in which the amide carbonyl and phenol hydroxyl groups are anti with respect to each other. Macroscopic coupling constants were calculated using the electronegativity-modified Karplus relationship28 as implemented in Macromodel. Calculation of the Relaxation Matrix and Coupling Constants. The time dependence of the cross peaks in the NOESY spectrum is given by eq 120

V(τm) ) V0 exp(-Rτm)

(1)

where V(τm) is the peak-volume matrix at the mixing time τm, V0 is the peak-volume matrix at zero mixing time, and R is the relaxation rate matrix containing the cross-relaxation rates as off-diagonal elements and the longitudinal decay rates as the

16928 J. Phys. Chem. B, Vol. 109, No. 35, 2005

Dais et al.

Figure 3. (a) 1H NMR 400 MHz and (b) 13C NMR 100 MHz spectra of OTA in deuterated chloroform solutions.

diagonal elements. The cross-relaxation rate for protons i and j is given by

( )

1 γ2p 2 σij ) {-J0(0) + 6J2(2ω)} 10 r 3 ij

(2)

where γ and ω are the gyromagnetic ratio and the Larmor frequency of protons and rij is the interproton distance. For isotropic motion described by the molecular correlation time τC, the spectral densities have the form

J(nω) )

2τC 1 + (nω)2τC2

(3)

Since cross-relaxation rates depend on the interproton distance and correlation time through eqs 2 and 3, NOESY experiments can be used to determine the dynamics and internuclear distances. The approach used here is to simulate the experimental NOESY spectrum, i.e., the peak-volumes V(τm) and V0, from a model structure using the program CORMA.21 Comparison between measured volumes and volumes calculated from the proposed structure allows an assessment of the validity of the model structure. The simulation of NOESY volumes with CORMA is based on the ensemble-averaged relaxation rates over the number of conformers resulting from the conformational analysis. Conformations corresponding to the energy minima in the conformational energy map of the OTA molecule were taken into account in the averaging procedure of these calculations. The relative population Pi of each conformer i with energy Ei, distributed over n conformational states, has

been calculated by a Boltzmann distribution

Pi ) exp

( ) ( ) ∆Ei kBT

∆Ei

/

∑n exp k T

(4)

B

where kB is Boltzmann’s constant and ∆Ei is the relative energy of each conformer from the lowest energy in the conformational energy map. The calculation of the macroscopic coupling constant requires consideration of the balanced contribution from each conformation in the form of an ensemble average

〈3JH,H〉 )

∑n Pi3JiH,H

(5)

Pi as before represents the relative population of the ith conformer with energy Ei and is governed by a Boltzmann distribution according to eq 4. The theoretical NOESY spectrum was computed using a new form of the CORMA program,20,29 which has been modified30 to incorporate (1) internal motions modeled by a model-free approach, (2) chemical exchange described by a kinetic matrix of exchange rates, and (3) quality factors R, R2, Rx, Rx2, Qx, and Qx2, which reflect the match of experimental and calculated NOESY peak intensities or volumes. The program uses a model structure to generate a theoretical NOESY spectrum, which is to be compared with the experimental one. This process is repeated iteratively until the error between the calculated and the observed peak volumes reaches a minimum value. The computation of the theoretical spectrum incorporates all of the effects of network relaxation and multiple spin effects (full relaxation rate matrix).

Conformational Analysis of Ochratoxin A

J. Phys. Chem. B, Vol. 109, No. 35, 2005 16929

Figure 4. H,H-gCOSY spectrum of OTA in deuterated chloroform solution.

The R-factor is equivalent to the crystallographic R-factor, and Rx is a variation of the R-factor involving sixth-root weighting of intensities to avoid domination by short relaxation pathways, that is

Rx )

∑i |Ie1/6(i) - Ic1/6(i)|/∑i |Ie1/6(i)|

∑i |Ie1/6(i) - Ic1/6(i)|/∑i |Ie1/6(i) + ∑i Ic1/6|

proton H-3

δ 1H (ppm) 4.69

(6)

where Ie(i) is the experimental and Ic(i) is the corresponding calculated intensity of the cross peak i for a particular structure.29,31 The Qx factor is given by the following equation

Qx )

TABLE 1: Proton and Carbon-13 Nuclear Magnetic Resonance Data for OTA in Chloroform-d Solutions

(7)

In this procedure, the R, Rx, Q, and Qx factors are calculated for a rapidly interchanging ensemble of structures, the relaxation rates of individual snapshots are averaged, and the resulting theoretical intensity matrix is compared with experimental intensities. This is distinct from arithmetically comparing averaged intensities for a set of structures. Results Assignment of Proton and Carbon NMR Spectra of OTA. The successful interpretation of a NOESY spectrum requires first the complete assignment of the corresponding 1H NMR spectrum. The establishment of all of the proton and carbon assignments with a complete set of 1H-1H coupling constants requires a more detailed investigation at high magnetic fields using two-dimensional NMR techniques. The one-dimensional 1H and 13C NMR spectra at 400 MHz are shown in Figure 3

H-4a H-4b H-6 H-8 (OH) H-12 (NH) H-13 H-14a H-14b H (phenyls) H-21 (methyl)

3.22 2.79 8.36 12.69 8.43 4.96

3

JH,H (Hz)

JH-3,H-4a ) 11.5 JH-3,H-4b ) 3.5 JH-3,H-21 ) 6.3 JH-4a,H-4b ) 17.5

carbon

δ 13C (ppm)

C-1

169.72

C-3 C-4 C-5 C-6 C-7 C-8

75.16 31.32 141.05 138.99 120.05 159.04

JH-13,H-12 ) 6.9 JH-13,H-14a ) 5.4 JH-13,H-14b ) 7.4 3.30 JH-14a,H-14b ) 14.2 C-9 3.15 C-10 7.13-730 multiplet C-11 1.53 C-13 C-14 C-15 C-16, C-20 C-17, C-19 C-18 C-21 C-22

110.06 123.24 163.38 54.37 37.15 135.65 129.31 128.71 127.31 20.67 174.13

along with the labeling of the various protons and carbon nuclei, whose chemical shifts have been assigned unequivocally. Most of the proton peaks are clearly resolved, except perhaps the aromatics, which form a second-order subspectrum. The doublet at δ 1.53 in the proton spectrum is assigned to a methyl proton (H-21), whereas the singlets at δ 8.36 and 12.69 are assigned to the isolated proton H-6 and the hydroxyl proton H-8,

16930 J. Phys. Chem. B, Vol. 109, No. 35, 2005

Dais et al.

Figure 5. H,C-gHSQC spectrum of OTA in deuterated chloroform solution.

respectively. (For the numbering system, see Figure 1.) From the cross-peak pattern in the gradient-selected COSY spectrum shown in Figure 4, protons H-3 and H-4 of the lactone ring, the amide proton H-12, and protons H-13 and H-14 of the peptide chain are assigned. Protons H-3, H-4a, and H-4b form an AMX system (inset in Figure 3), which is readily analyzed, allowing thus the determination of the coupling constants and the assignment of the two diastereotopic methylene protons. The coupling constants 3JH-3,H-4a ) 11.5 Hz and 3JH-3,H-4b ) 3.5 Hz advocate that the lactone ring assumes a single conformation with the H-3 proton in a pseudoaxial position, anti and syn to protons H-4a and H-4b, respectively. However, the assignment of protons H-14a and H-14b is not an easy task, since the coupling constants of these protons with H-13 do not differ significantly (3JH-13,H-14 ) 5.4 and 7.4 Hz). Neither the coupling constants of these protons with the C-22 carboxyl carbon in the coupled 13C NMR spectrum nor the 3JH,N couplings with the nitrogen atom offered a solid indication of the identity of the protons H-14a and H-14b. However, as it will be shown later, the NOE values of these protons with the neighboring H-13 proton and the calculated intensities of the theoretical NOESY spectrum facilitate the assignment of the corresponding chemical shifts. Table 1 lists the proton chemical shifts and coupling constants of OTA in deuterated chloroform, which agree closely with those reported previously,15 in DMSO-d6 solvent, except the H-14 protons.

The protonated aromatic carbons of OTA are easily assigned through the phase-sensitive gradient-selected HSQC (gHSQC) experiment, which detects one-bond 13C-1H connectivities. The gHSQC spectrum obtained with a delay time τ ) 1.8 ms is illustrated in Figure 5. The assignment of the signals of the nonprotonated aromatic carbons for OTA has been achieved by using the phase-sensitive gradient-selected HMBC (gHMBC) experiment, which correlates proton and carbon nuclei via longrange proton-carbon couplings. The gHMBC spectrum of the aromatic region presented in Figure 6 allows the assignment of the nonprotonated carbons. It is easily seen that each nonprotonated carbon correlates with protons via 3JC,H and a few 2JC,H and 4JC,H couplings. The aromatic carbons C-15 through C-20 have been assigned from cross peaks in the gHMBC spectrum (Figure 6), 13C NMR relaxation experiments (discussed below), and information from the literature. Table 1 lists the carbon chemical shifts and coupling constants of OTA in deuterated chloroform. Conformational Analysis of OTA. As mentioned in the Experimental Section, the energy map of OTA has been obtained by using the MM2* force field program of Macromodel and applying a systematic perturbation of torsion angles about the bonds Φ ) C15-C14-C13-N12 and Ψ ) C14-C13-N12-C11, which constitute the flexible part of the molecule, over the whole angular range. The variation of the torsional angles resulted in seven energy minima summarized in the conformational energy

Conformational Analysis of Ochratoxin A

J. Phys. Chem. B, Vol. 109, No. 35, 2005 16931

Figure 6. H,C-gHMBC spectrum of OTA in deuterated chloroform solution.

map of Figure 7. The geometries of these minima are listed in Figure 8. The energy of the lowest minimum in the map, corresponding to conformer I, is assigned arbitrarily to zero. Table 2 summarizes the relative energies, relative populations, and the values of the dihedral angles for each conformer. It is interesting to note that all conformers adopt the form b, where the amide carbonyl and the phenol groups are anti, favoring the hydrogen-bonding interaction between the hydroxyl group OH-8 with the lactone carbonyl group. The relative energy of the resulting conformers was found to be in range from 0 to 2.56 kcal mol-1, with conformers I and VII having the lowest and highest relative energy, respectively. Moreover, the seven conformers can be divided into two families in terms of their relative energies and populations; the low relative energy (0.000.82 kcal mol-1) and high relative population (0.287-0.072) family of conformers I-V, which are more probable to occur, and the high relative energy (2.47 and 2.57 kcal mol-1) and low relative population (0.005 and 0.004) family of conformers VI and VII. It appears, that conformers I-V favor the proximity of the phenyl ring and the amide bond of the molecule as reflected in the observed NOE between the phenyl protons H-16 and H-20 and the amide proton H-12. Dynamics of OTA. Reproducing the theoretical NOESY volumes for flexible molecules using CORMA (see below), relative rates of overall and internal motions should be considered. Usually, molecular correlation times describing the various modes of reorientation can be obtained from 13C NMR spin-lattice relaxation experiments in combination with theoretical time-correlation functions and/or with their Fourier pair, the spectral density functions.32 Overall and internal motions are taken into consideration by CORMA through the spectral

density function obtained from the so-called model-free approach developed by Lipari and Szabo.33 The program calculates the spectral density J(ω, A, S2, τ1, τ2, τe) as a function of frequency, ω, the anisotropy parameter, A, the order parameter S2, the overall correlation times τ1 and τ2 for anisotropic motion, and the correlation time τe for internal motion. In the program, the ratio τ1/τ2 is used. For τ1/τ2 ) 1 or A ) 1, isotropic overall motion is obtained. If the order S2 parameter is set to unity, then isotropic motion with no internal motion is considered. The experimental 13C spin-lattice relaxation times (T1) for each protonated carbon of OTA in chloroform-d solutions were C-3 (0.78 s), C-4 (0.36 s), C-6 (0.77 s), C-13 (0.74 s), C-14 (0.39 s), C-18 (0.77 s), and C-16, C-17, C-19, and C-20 (average value of 1.00 s). All 13C NOE values attained their maximum values 3.00 ( 0.01, indicating that the relaxation of OTA is dominated by the dipolar 13C-1H interactions and that the overall and internal motions are fast, within the extreme narrowing condition τC(ωC + ωH) , 1. The ortho and meta carbons of the phenyl ring are characterized by higher T1 values than that of the para carbon, indicating internal phenyl rotation about the C2 symmetry axis of the ring.35 The T1 value of the para carbon is similar to those of the C-3, C-6, and C-13 carbons, with those of the C-4 and C-14 carbons both moderated by two protons within experimental error. The correlation time calculated from the average T1 values (0.76 s) of these carbons reflects the rate of the overall motion, which is isotropic. On the basis of these relaxation times, the following parameters were calculated and used as input parameters for CORMA: τ1/τ2 ) 1, A ) 1, S2 ) 0.56, τ1 ) τ2 ) 0.062 ns, and τe ) 0.052 ns. Values of 1.09 and 1.08 Å were used for the C-H bond lengths of the aliphatic and aromatic carbons, respectively. The cor-

16932 J. Phys. Chem. B, Vol. 109, No. 35, 2005

Dais et al.

Figure 7. Conformational energy map of OTA (in kcal mol-1) obtained with the MM2* force field program. The energies are expressed relative to the lowest-energy minimum corresponding to conformer I.

relation times for the overall and internal motions were calculated by fitting the experimental 13C relaxation data to Woessner’s equation34 for internal rotation superimposed on isotropic overall motion and the order parameter from eq 1 of ref 33. S2 values of 1.00 were used for each individual NOE interaction in the rigid part of the molecule and 0.56 for NOESY cross-peak volumes connecting protons situated in the rigid and mobile segment of the molecule (e.g., between protons H-12 and H-16). It should be noted that the model-free approach is exact when the internal motion is much faster than the overall motion and lies within the extreme narrowing limit.33 This is true for large molecules (e.g., random coil polymers) for which the overall and internal motions are independent, and therefore the total time-correlation function can be factored out. In the present analysis, where the internal motion lies within the extreme narrowing limit, but its time scale is comparable with that of the overall motion, the use of the model-free approach is an approximation. The calculated order parameter S2 of this size (0.56) is likely to be lower by 15-25% than the exact value.33 NOESY Volumes and J Coupling Constants. Figure 8 illustrates the experimental 400 MHz NOESY spectrum of OTA in chloroform solution at 303 K with a mixing time of 1.0 s. This spectrum shows several cross peaks between protons in the same ring and more interestingly between the various ring protons and protons of the amide linkage. In particular, the cross peak connecting the hydroxyl proton H-8 with the amide proton

H-12 supports the preponderance of form b as mentioned earlier.15 The normalized experimental cross-peak volumes from the NOESY spectrum (Figure 9) along with the network of dipolar interactions is reported in Table 3. Normalization was achieved by using the normalization factor fnorm

fnorm )

∑i Iijc ∑i

(8) Iije

This normalization procedure was preferred over other choices in CORMA, because it provides the opportunity to test the validity of the calculated volumes between other pairs of protons in the OTA molecule. The simulated cross-peak volumes upon applying CORMA and using the conformational structures I-VII obtained by the MM2* force field are summarized in Table 3. Dipolar connections, which were observed in the NOESY spectrum (Figure 9) between the magnetically equivalent H-16 and H-20 protons of the phenyl ring and protons H-14a, H-14b, H-8, and H-12 of the peptide linkage, were taken into consideration in the present calculations. Nevertheless, the phenyl protons were not resolved in the 1H NMR spectrum of OTA, and therefore their cumulative NOESY volumes could not be compared with the separate volumes calculated by CORMA. Accordingly, these cross-peak volumes were not included in Table 3. As it can be seen, the normalized

Conformational Analysis of Ochratoxin A

J. Phys. Chem. B, Vol. 109, No. 35, 2005 16933 summarized in Table 3. It is seen that the theoretical values are not in good agreement with the experimental values. The discrepancy between calculated and experimental values is over 1 Hz, and the calculated quality factor (see footnote of Table 3) k2 ) 0.2073 is quite high. There may be two sources of errors that contribute to the observed inconsistency of the J couplings: (a) The calculated relative energies of conformers I-VII by using the MM2* force field may lead to erroneous relative populations, and (b) the calculated geometry of conformers I-VII upon energy minimization by MM2* may not be consistent with the actual ones. Small changes in geometry lead to large differences in the calculated J couplings. To check the first possibility, all possible combinations of the relative populations of the seven conformers were generated with a pace of 10%. We ended up with 8008 such combinations, which were sorted out upon minimization of the quality factors of the calculated NOESY volumes and J couplings for each combination. Twenty combinations exhibiting the lowest quality factors were taken into account, and the calculated averaged populations of the seven conformers are now 0.13 (I), 0.15 (II), 0.38 (III), 0.08 (IV), 0.14 (V), 0.05 (VI), and 0.07 (VII). These relative populations were used by CORMA to calculate the final NOESY volumes and J couplings, which are reported in Table 4. The agreement between the theoretical and the experimental NOESY volumes is good as before, whereas the theoretical J couplings are very close to the experimental values, except perhaps the coupling constant between the H-12 and H-13 protons. To explore the second possibility of error, we performed energy calculations by using the MMFF force field implemented in Macromodel. Initial calculations showed the possibility of an intramolecular hydrogen-bond formation involving the hydrogen of the carboxyl group (C-22). Contrary to MM2*, the MMFF program takes into consideration the formation of intramolecular hydrogen bonding. Further calculations were performed taking into account the relative orientation of the O-H moiety of the carboxyl group. Two new energy conformational energy maps were calculated upon fixing the dihedral angle OdC-22-O-H-8 at 0° and 180°. The final map was constructed by selecting the lowest-energy structure for a given Φ/Ψ point in the map. This procedure led us to the same number of conformers as those using MM2*. Details of these calculations are summarized in Table 5. Conformers I, II, III, and IV exhibit relative energies within 1 kcal mol-1. Among these, conformer IV is characterized with the lowest relative energy and the highest relative population, although its relative energy is only 0.05 kcal mol-1 lower than that of conformer II. It appears that conformer IV is stabilized by the formation of an intramolecular hydrogen bond between the hydrogen of the carboxyl group (C-22) and the amide carbonyl group (C-11), whereas conformer II, the relative energy of which is barely higher that that of conformer IV, shows a weak interaction between the hydrogen of the carboxyl group (C-22) and the amide proton. It is interesting to note that remaining conformers I, III, VI, and VII show the same hydrogen-bond

Figure 8. Geometries of the low-energy conformations I-VII for OTA.

experimental volumes of pairs of protons are in good agreement with the theoretical values, indicating that the calculations of the NOESY volumes are fairly reliable. The match of theoretical to experimental data, reflected on the quality factors R ) 0.0584, R2 ) 0.0389, Rx ) 0.0177, Rx2 ) 0.0230, Qx ) 0.0089, and Qx2 ) 0.0163, is satisfactory regarding the experimental error (20%) involved in the determination of the Vij volumes. A slightly better match between theoretical and experimental crosspeak volumes was obtained by using an order parameter value of 0.70, which is 25% higher that that calculated by the modelfree approach as discussed previously. The ensemble-averaged coupling constants Jave over the seven conformers using their relative populations from Table 2 are

TABLE 2: Relative Energies, Relative Populations, and Torsional Angles of the Seven Conformers Calculated Using the MM2* Force Field Program conformers mol-1

energy (kcal ) population C6-C7-C11-N12 (deg) C7-C11-N12-C13 C11-N12-C13-C14 (deg) N12-C13-C14-C15 (deg) C13-C14-C15-C16/20 (deg)

I

II

III

IV

V

VI

VII

0.0 0.287 -177.6 176.0 144.4 60.7 -94.0/85.8

0.02 0.277 -170.6 -177.4 94.6 57.8 -88.5/90.7

0.13 0.230 -170.2 178.8 149.6 -54.9 -79.6/99.6

0.49 0.125 164.0 172.2 -77.6 -49.9 -81.2/97.5

0.82 0.072 178.7 176.7 148.3 -178.2 -98.2/81.3

2.47 0.005 169.5 176.8 -81.0 -166.1 -113.8/66.9

2.56 0.004 169.6 174.3 -91.7 54.4 -80.2/100.7

16934 J. Phys. Chem. B, Vol. 109, No. 35, 2005

Dais et al.

Figure 9. Phase-sensitive 400 MHz NOESY spectrum of OTA in deuterated chloroform solution. The spectrum was obtained with a mixing time of 1.0 s.

TABLE 3: Normalized Experimental (Ve) and Theoretical (Vc) Cross-Peak Volumes and Experimental (Je) and Theoretical (Jc) Coupling Constantsa (in Hz) for OTAb proton 1

proton 2

Ve

Vave

H-3 H-4b H-4a H-8 H-13 H-14a H-14a H-14b H-14b

H-4a H-4a H-21 H-12 H-12 H-12 H-13 H-12 H-13

0.00852 0.06219 0.00276 0.00510 0.00588 0.00169 0.00891 0.00398 0.00600

0.01015 0.06297 0.00264 0.00576 0.00556 0.00167 0.00838 0.00281 0.00510

a

Je

TABLE 4: Normalized Experimental (Ve) and Theoretical (Vc) Cross-Peak Volumes and Experimental (Je) and Theoretical (Jc) J Coupling Constants (in Hz) for OTAa

Jave

proton 1

proton 2

Ve

Vave

Je

Jave

6.9

8.4

5.1

6.0

0.01283 0.08676 0.00342 0.00654 0.00732 0.00226 0.00990 0.00496 0.00576

5.4

7.4

0.01134 0.08275 0.00368 0.00678 0.00783 0.00225 0.01185 0.00529 0.00798

8.1

4.2

H-4a H-4a H-21 H-12 H-12 H-12 H-13 H-12 H-13

6.9

5.4

H-3 H-4b H-4a H-8 H-13 H-14a H-14a H-14b H-14b

7.4

7.0

The quality factor for coupling constants is given by k2 )

a

x(∑i[Je(i)-Jc(i)]2)/∑iJe2(i). b Theoretical values qere Obtained using

Theoretical values were obtained using CORMA and optimizing the relative populations of conformational structures I-VII calculated with MM2*.

formation as conformer IV, although this interaction is weaker for conformer I. Nevertheless, the calculated J couplings were not improved significantly (k2 ) 0.1967) upon using the relative populations obtained from MMFF, whereas the theoretical NOESY volumes became even worse compared to those calculated with MM2*. This is also reflected in the higher values assumed by the quality factors (R ) 0.168, R2 ) 0.122, Rx ) 0.0615, Rx2 ) 0.0802, Qx ) 0.0314, and Qx2 ) 0.0576). It is important to recognize that all previous calculations mentioned were performed for an isolated molecule, omitting

possible influence of the medium on the hydrogen bonding and the relevant calculated parameters. To obtain a more realistic model, we repeated the MMFF calculations using the GB/SA program implemented in Macromodel for solvent simulation, which is parametrized for chloroform. The resulting family of conformers and their relative energies showed insignificant differences from those calculated in a vacuum, although the quality factors of CORMA were slightly improved. The experimental and calculated NOESY and coupling constants in Tables 3 and 4 support the assignment of protons H-14a (δ 3.30) and H-14b (δ 3.15) in Table 1. Pro-S proton H-14a is characterized by larger NOE and smaller coupling

CORMA and the conformational structures I-VII calculated with MM2*.

Conformational Analysis of Ochratoxin A

J. Phys. Chem. B, Vol. 109, No. 35, 2005 16935

TABLE 5: Relative Energies, Relative Populations, and Torsional Angles of the Seven Conformers Calculated Using the MMFF Force Field Program conformers

I

II

III

IV

V

VI

VII

energy (kcal mol-1) population C6-C7-C11-N12 (deg) C7-C11-N12-C13 (deg) C11-N12-C13-C14 (deg) N12-C13-C14-C15 (deg) C13-C14-C15-C16/20 (deg)

0.95 0.149 173.3 -176.6 137.0 54.1 -97.7/81.3

0.05 0.254 -178.2 -179.9 74.6 61.5 92.6/-86.9

0.53 0.191 177.4 -167.2 160.9 -58.9 -69.7/110.0

0 0.262 178.9 164.5 -60.1 -56.0 -73.4/105.6

2.81 0.050 -178.5 178.0 75.2 -147.5 -98.5/80.6

2.96 0.046 179.1 178.4 -58.8 -58.7 -77.8/101.4

2.86 0.048 177.9 178.4 151.8 -63.5 -76.7/102.5

constant with proton H-13 compared to pro-R proton H-14b. Interchanging the experimental NOESY volumes of H-14a and H-14b protons, while simulating NOESY volumes with CORMA, leads to large discrepancies between experimental and calculated NOE values for all pairs involving these protons. Discussion The family of seven OTA conformations calculated by the MM2* force field appears to consist of a good structural model in reproducing successfully the experimental NOESY crosspeak volumes and proton coupling constants. Also, this is reflected in the low values of the quality factors associated with CORMA. However, MM2* is not suitable for predicting the presence of intramolecular hydrogen bonding. Calculations performed with the MMFF force field program allowing hydrogen-bond formation resulted in seven conformers, which are different from those obtained by MM2* with respect to relative energies and populations. Although coupling constants were improved slightly relative to those calculated using MM2*, the discrepancy between experimental and calculated NOESY volumes increased with MMFF. Nevertheless, the seven conformers obtained by MMFF showed the existence of hydrogen bonding between the hydrogen of the carboxyl group and the amide carbonyl in the molecule, which is similar to that formed in the backbone of peptide chains of globular proteins, the socalled “γ-turn”, stabilizing thus the folding of the peptide chain.36 This structural pattern (Figure 10) obtained with MMFF very probably exists in solution, but it is difficult to predict to what extent. Moreover, the hydrogen-bond formation induces changes in the dihedral angles Φ and Ψ by 10-20°, which in turn leads to differences in the corresponding 3JH,H values by 1-2 Hz. It is thus explicable why the theoretical models deviate slightly from the experimental NMR data. In summary, it appears that these findings are commensurate with a solution equilibrium among conformers I, III, II, and IV, the latter two being the predominant ones. This equilibrium is expected to be influenced by the existence of the OdC-11‚‚‚H-O-C-22

Figure 10. Part of the MMFF force field structure of conformer IV showing the C-11dO‚‚‚H-O-C-22 hydrogen bond. This structure is similar to the γ-turn stabilizing the folding of a peptide chain.

hydrogen bond. Nevertheless, these conformations are more likely to be stabilized in the hydrophobic environment offered by the chloroform solvent rather than in aqueous solution, which constitutes the biological environment of where this toxin acts. Acknowledgment. We thank the Greek Ministry of Education for financial support through the B’EPEAEK Graduate Program and the Program Hrakleitos. References and Notes (1) Frisvad J. Stored Grain Ecosystems; Jayas, D., White, N., Muir, W., Eds.; Marcel Dekker: New York, 1995; pp 251-288. (2) WHO, World Health Organization. Ochratoxin A: Toxicological EValuation of Certain Food AdditiVes and Contaminants; WHO Food Additives Series 35; World Health Organization: Geneva, Switzerland, 1991; pp 365-417. (3) WHO, World Health Organization. Ochratoxin A: Toxicological EValuation of Certain Food AdditiVes and Contaminants; WHO Food Additives Series 35; World Health Organization: Geneva, Switzerland, 1996; pp 363-376. (4) IPCS, International Programme on Chemical Safety. Selected Mycotoxins: Ochratoxins, Trichothecenes, Ergot; Environmental Health Criteria 105; World Health Organization: Geneva, Switzerland, 1990. (5) Majerus, P.; Cutka, I.; Dreyer, A.; El-Dessouki, S.; Eyrich, W.; Reusch, H.; Schurer, B.; Waiblinger, H. U. Dtsch. Lebensm. Rundsch. 1993, 89, 112-114. (6) Walker, R. Mycotoxins of Growing Interest: Ochratoxins. In Proceedings of the 3rd Joint FAO/WHO/UNEP International Conference on Mycotoxins, Tunis, Tunisia, March 3-6, 1999. (7) Creppy, E. E. J. Toxicol. 1998, 17, 479-481. (8) Kuiper-Goodman, T. Food Addit. Contam. 1996, 13, 53-57. (9) Stefanaki, I.; Foufa, E.; Tsatsou-Dritsa, A.; Dais, P. Food Addit. Contam. 2003, 20, 74-83. (10) Visconti, A.; Pascale, M.; Centonze, G. J. Chromatogr., A 2000, 888, 321-326. (11) Zimmerli, B.; Dick, R. Food Addit. Contam. 1996, 13, 655-668. (12) Il’ichev, Y. V.; Perry, J. L.; Simon, J. D. J. Phys. Chem. B 2002, 106, 452-459. (13) Il’ichev, Y. V.; Perry, J. L.; Simon, J. D. J. Phys. Chem. B 2002, 106, 460-465. (14) McMasters, D. R.; Vedani A. J. Med. Chem. 1999, 42, 30753086. (15) Bredenkamp, M. W.; Dillen, J. L. M.; van Rooyen, P. H.; Steyn, P. S. J. Chem. Soc., Perkin Trans. 2 1989, 1835-1839. (16) James, T. L.; Basus, V. J. Annu. ReV. Phys. Chem. 1991, 42, 501542. (17) Borgias, B. A.; James, T. L. In Biological Magnetic Resonance; Berliner, L. J., Reuben J., Eds.; Plennum Press: New York, 1990; Vol. 9, pp 119-154. (18) Mikros, E.; Dais, P.; Sauriol, F. Carbohydr. Res. 1996, 294, 1-13. (19) Karali, A.; Dais, P.; Mikros, E.; Heatley, F. Macromolecules 2001, 34, 5547-5554. (20) Keepers, J. W.; James, T. L. J. Magn. Reson. 1984, 57, 404-426. (21) Borgias, B. A.; Thomas, P. D.; James, T. L. Complete Relaxation Matrix Analysis (CORMA); University of California: San Francisco, 1992. (22) Braun, S.; Kalinowski, H.-O.; Berger, S. 100 and More Basic NMR Experiments: A Practical Course; VCH Publishers: Weinheim, Germany, 1996. (23) Macura, S.; Huang, Y.; Suter, D.; Ernst, R. R. J. Magn. Reson. 1981, 43, 259-281. (24) Mohamadi, F.; Richards, N. G. J.; Guida, W. C.; Liskamp, R.; Lipton, M.; Caufield, C.; Hendrikson, T.; Still, W. C. J. Comput. Chem. 1990, 11, 440-467. (25) Allinger, N. L. J. Am. Chem. Soc. 1977, 99, 8127-8134. (26) Still, W. C.; Tempczyk, A.; Hawley, R. C.; Hendrickson, T. J. Am. Chem. Soc. 1990, 112, 6127-6129.

16936 J. Phys. Chem. B, Vol. 109, No. 35, 2005 (27) Ponder, J. W.; Richards, F. M. J. Comput. Chem. 1987, 8, 10161024. (28) Haasnoot, C. A. G.; de Leeuw, F. A. A. M.; Altona, C. Tetrahedron 1980, 36, 2783-2792. (29) Borgias, B. A.; James, T. L. J. Magn. Reson. 1998, 79, 493-512. (30) Schmitz, U.; Kumar, A.; James, T. L. J. Am. Chem. Soc. 1992, 114, 10654-10656.

Dais et al. (31) Gonzalez, C.; Rullmann, J. A. C.; Bonvin, A. M. J. J. R.; Boelens, R.; Kaptein, R. J. Magn. Reson. 1991, 91, 659-663. (32) Dais, P. AdV. Carbohydr. Chem. Biochem. 1995, 51, 63-131. (33) Lipari, A.; Szabo, A. J. Am. Chem. Soc. 1982, 104, 4546-4559. (34) Woessner, D. E. J. Chem. Phys. 1962, 36, 1-11. (35) Dais, P. Magn. Reson. Chem. 1987, 25, 141-146. (36) Matthews, B. W. Macromolecules 1972, 5, 818-819.