Journal of Molecular Structure 556 (2000) 303–313 www.elsevier.nl/locate/molstruc B3LYP/6-31G ⴱ, RHF/6-31G ⴱ and MM3 heats of formation of disaccharide analogs A.D. French a,*, A.-M. Kelterer b, G.P. Johnson a, M.K. Dowd a a Southern Regional Research Center, Agricultural Research Service, US Department of Agriculture, P.O. Box 19687, New Orleans, LA 70179-0687, USA b Institut für Physikalische und Theoretische Chemie, Technische Universität Graz, Technikerstrasse 4, A-8010 Graz, Austria Dedicated to Professor N.L. Allinger on the occasion of this special issue Received 11 October 1999; accepted 27 October 1999 Abstract DHf values were calculated with the bond- and group-enthalpy method for analogs of 13 disaccharides to learn the relative stabilities for various configurations about the carbon atoms connected to the linkage oxygen atom. The analogs were based on tetrahydropyran and, in the case of the sucrose analogs, tetrahydrofuran. The molecular mechanics program MM3 calculates these values as an option. The method was also used with RHF/6-31G ⴱ and B3LYP/6-31G ⴱ quantum mechanics (QM) theory. DHf values had ranges of 15–17 kcal (all energies herein are molar) by the different methods, with the analogs of the three trehaloses and sucrose having the lowest values. DHf values for the isomeric analogs of non-reducing sugars with two anomeric centers are about ⫺150 kcal, about 8 kcal lower than for the analogs of the reducing dimers (nigerose, maltose, laminarabiose, cellobiose and galabiose (a-d-galactosyl-d-galactose)) by all three types of calculation. The analogs of the di-axial, axial– equatorial and di-equatorial, non-glycosidic pseudo-disaccharides had DHf values within 1.0 kcal by each of the three methods. They were about 8 kcal higher than for the molecules that contain one glycosidic sequence. The relative DHf values by QM were within 1.0 kcal of their corresponding relative electronic energies, except for the methylated sucrose analog, for which the discrepancy was about 1.85 kcal at both QM levels. Compared to our B3LYP results, the RHF-based calculations overestimated the stability of all molecules by about 2.85 kcal. The MM3 values were close to the B3LYP numbers, with the largest discrepancy for the cellobiose analog, 1.76 kcal. The stabilization embodied in the group enthalpy increment for anomeric centers is another manifestation of the anomeric effect. 䉷 2000 Elsevier Science B.V. All rights reserved. Keywords: Carbohydrates; Cellobiose; Exo-anomeric effect; Galabiose; Laminarabiose; Maltose; Nigerose; Sucrose; Tetrahydropyran; Tetrahydrofuran; Trehalose 1. Introduction Once standard states are defined, the heat of formation, DHf, is a quantity that can be calculated by a * Corresponding author. Tel.: ⫹1-504-286-4410; fax: ⫹1-504286-4217. E-mail address: [email protected] (A.D. French). variety of methods as well as determined experimentally. Thus it provides a useful connection between methods of predicting relative energies and structures and observable reality. We have previously found it useful to calculate DHf in our studies of carbohydrates. For example, lattice energies give useful information on the forces that might be distorting molecules in crystals compared to their shapes in an 0022-2860/00/$ - see front matter 䉷 2000 Elsevier Science B.V. All rights reserved. PII: S0022-286 0(00)00648-7 304 A.D. French et al. / Journal of Molecular Structure 556 (2000) 303–313 Fig. 1. The compounds used in this paper, with the conformations determined by B3LYP/6-31G ⴱ theory. The oxygen atoms are shown in gray, the hydrogen atoms are not shown. All carbon atoms, shown in black are nominally tetrahedral. The atoms are numbered according to the corresponding numbers for the disaccharides with standard carbohydrate nomenclature. A.D. French et al. / Journal of Molecular Structure 556 (2000) 303–313 isolated state [1]. In that work, the lattice energy is the difference between the calculated DHf of the isolated molecule and the DHf determined by the experimental heat of combustion for the crystalline solid. Another example of our use of DHf calculations concerns the relative stabilities of monosaccharide isomers such as fructofuranose and fructopyranose that have different sizes of rings [2]. DHf values give information on the relative stability that is not inherent in the simple steric energies. However, calculations of correct DHf values for molecules such as carbohydrates that have anomeric centers are more difficult than for other molecules [3], and further verification of those calculated values is especially important. In some recent work [4,5], we used both RHF/6-31G ⴱ quantum mechanics (QM) theory and MM3 molecular mechanics (MM) theory to construct hybrid energy surfaces for a number of disaccharides. Those results, which covered the full range of linkage conformations, depend on compatibility between the QM and MM methods. One aspect of that compatibility would be demonstrated if the calculated DHf values for those molecules were comparable when calculated by both methods. Determination of the degree of compatibility of both the QM and MM theory was also a prime motivation for carrying out the present work. This paper presents the DHf values for analogs of 13 disaccharides (Fig. 1) that were calculated with MM3(96) 1 molecular mechanics software and with RHF/6-31G ⴱ and B3LYP/6-31G ⴱ levels of QM theory. As far as we know, these molecules exist at present only as computer models. They have various configurations of the linkages between the tetrahydropyran (THP) rings (or, in two cases, tetrahydrofuran (THF) rings) that are used to represent the participating monosaccharides. Addition of a methyl group to the C2 0 of the THP–O–THF sucrose analog gave a second, more representative analog of sucrose that has the same number of atoms and bonds of each type as the analogs of the other disaccharides that are based on THP. These THP–O–THP and THP–O–THF models have most of the types of two-bond, interring linkages found in disaccharides. The configura1 Academic users may obtain MM3 from the Quantum Chemistry Program Exchange, Indiana University, Bloomington, IN 47405 USA. All others may obtain it from Tripos Associates, 1699 South Hanley Road St. Louis, MO 63144. 305 tions include di-axial, axial-equatorial, and di-equatorial variations of typical glycosidic linkages (such as cellobiose or maltose), head-to-head glycosidic linkages (such as sucrose and the trehaloses) and non-glycosidic linkages (for our tail-to-tail, 3,3-linked pseudo-disaccharide analogs). These last three molecules are analogs of disaccharide molecules that do not themselves exist, either. The conformations and energies of disaccharide linkages are thought to be affected substantially by the general anomeric effect [6,7]. Anomeric carbon atoms are attached to both endo- and exo-cyclic oxygen atoms, and the bond between the anomeric carbon and the exo-cyclic oxygen atom is called a glycosidic bond. At our current level of understanding, the stereoelectronic, general anomeric effect causes C–O–C–O sequences to tend to take gauche conformations, rather than the trans conformation that is preferred for the C–C–C–C sequences. In structures such as 2-hydroxytetrahydropyran, the gauche conformation of the C–O–C–O sequence results in an axial disposition of the glycosidic bond (the a-gluco configuration), while the trans conformation corresponds to an equatorial orientation (the b-gluco configuration). However, the configurational energy is also governed by other torsional energies, such as C–C–C–O or C–C–O–C, as well as other factors, and the gauche conformation may not predominate. Therefore, the anomeric effect is, in practice, an unexpected degree of preference for gauche conformations, if our expectations were based on the behavior of C–C–C–C sequences. When the gauche conformation refers to the favored orientation of the C1– O1 glycosidic bond to the O5–C5 bond (Fig. 1), the cause is said to be the endo-anomeric effect. When it concerns the preferred orientation of the bond to the aglycon, e.g. the O5–C1–O1–C4 0 sequence in maltose, the cause is described as the exo-anomeric effect. 2. Methods Both MM and QM (Hartree–Fock and Density Functional Theory) can use the bond- and groupenthalpy increment scheme to calculate DHf. Even when economical levels of theory are employed, the method has demonstrated accuracy within 306 A.D. French et al. / Journal of Molecular Structure 556 (2000) 303–313 experimental error for a wide variety of typical organic compounds. This method, championed by Prof. Allinger for both MM and QM [3,8–10] is based on the premise that the change in enthalpy of the elements in their standard states to the enthalpy of a compound (i.e. DHf) can be reduced to a summary of the enthalpy of formation of the individual bonds, along with a measure of the strain induced by constructing the molecule. Certain additional enthalpies for formation of rings, for terminal methyl groups etc., which are systematically not calculated correctly are added to this information. In the most correct calculations of DHf, there are also compensations for conformational flexibility and the overall translational and rotational degrees of freedom. The bond- and group-enthalpy effects are determined for various classes of molecules by optimizing their values so as to best reproduce the known DHf values for a number of different molecules. Typically, the molecules used for these calibrations are small, and systematic errors can be magnified for compounds as large as studied in the present work. As an option, the MM3 program [11], draws upon built in tabulations of such bond- and group-enthalpies to calculate the DHf for the compound in question. The remaining term is the strain in the molecule induced by its formation. In the formalism of MM3, this strain corresponds to the “final steric energy” term, composed of the bond stretching, the angle bending, the van der Waals terms, the torsional term, the dipole–dipole energy and a few smaller contributions. As a practical note for users of MM3, the steric energy values for isomers with different ring sizes do not correctly indicate their relative enthalpies but their DHf values should. This is precisely because the group effects for different ring sizes must also be included in the comparison. The QM calculations are quite similar, but are not available in the usual QM software. Instead, they may be carried out expeditiously with a spreadsheet program. Here the computed quantity that is modified by the sum of the bond- and group-enthalpy increments is the total electronic energy [8]. That total contains the calculated DHf information, and the task for the bond enthalpy increments is to convert the total electronic energy to the DHf by removing the contributions arising from the electronic energy not only for the formation of bonds between the atoms in their standard states but also the formation of the atoms themselves from the electrons and nuclei. This makes the bond-enthalpy increments for the QM calculations very large. The total calculated electronic energy of a molecule depends very much on the particular level of QM theory that was used, and so the individual bond- and group-enthalpy contributions will be applicable to only one level of theory. For example, the RHF/6-31G ⴱ and B3LYP/6-31G ⴱ total electronic energies for our a,a-trehalose analog differ by 3.82 Hartrees (almost 2400 kcal, all energies herein are molar), and the C–C bond-enthalpy increments for the two levels of theory differ by 74 kcal/ mol of bonds. Ideally, the group-enthalpy increments would not be required with QM methods, but the limitations of affordable levels of theory require these additional empirical corrections. Besides the ring-size effects, our molecules require group increments for anomeric centers with various substitutions, O–C–C–O sequences, and secondary oxygen atoms such as those that are part of the ring when they are not adjacent to an anomeric center. The QM calculations of DHf values now also include group enthalpy increments for six-membered rings [10]. Besides values for C–C and C–H bonds and applicable groups from that work on alkanes, our QM-derived values used preliminary, unpublished new parameters for the bond and group enthalpies of alcohols and ethers from the Allinger group [12]. 2 MM3 values of DHf were calculated for isolated molecules with its optional routine and the default dielectric constant of 1.5, after constructing the complete f , c energy surfaces and determining the global minimum energy conformations. The RHF/631G ⴱ values were for the global minima from the relaxed conformational surfaces calculated with that level of theory using Gaussian 94 [13] or gamess [14]. The locations of the global minima for the laminarabiose and cellobiose analogs are different on the RHF/6-31G ⴱ and MM3 surfaces. However, these minima correspond to positions that are just slightly higher in relative energy on the surfaces calculated by the other method. The B3LYP/6-31G ⴱ values, computed using Gaussian 94, were from 2 These provisional parameters are believed to be final but are still subject to last-minute corrections. A.D. French et al. / Journal of Molecular Structure 556 (2000) 303–313 307 Table 1 Linkage torsion angles and energies of the global minima computed with B3LYP/6-31G ⴱ, RHF/6-31G ⴱ and MM3(96) for the disaccharide analogs (kcal) fa Sugar analog ca DHf (kcal) Rel. DHf (kcal) QM Rel. b Eelect. (kcal) a,a-Trehalose Axial–axial 1,1-Linked B3LYP/6-31G ⴱ RHF/6-31G ⴱ MM3 70.14 71.39 74.50 70.14 71.38 76.50 ⫺149.86 ⫺153.40 ⫺150.03 0.36 0.00 0.35 0.00 0.00 –- a,b-Trehalose Axial–equatorial 1,1-Linked B3LYP/6-31G ⴱ RHF/6-31G ⴱ MM3 58.85 61.90 76.91 ⫺96.11 ⫺95.33 ⫺79.53 ⫺147.38 ⫺150.43 ⫺147.79 2.84 2.95 2.59 2.48 2.95 –- b,b-Trehalose Equatorial–equatorial 1,1-Linked B3LYP/6-31G ⴱ RHF/6-31G ⴱ MM3 ⫺71.50 ⫺70.09 ⫺76.87 ⫺71.50 ⫺70.09 ⫺76.87 ⫺147.45 ⫺150.89 ⫺148.60 2.77 2.50 1.78 2.41 2.50 –- Sucrose (THF–O–THP) Axial–pseudo-axial 1,2-Linked (28 atoms) B3LYP/6-31G ⴱ RHF/6-31G ⴱ MM3 58.38 60.77 78.70 ⫺97.22 ⫺98.16 ⫺78.03 ⫺140.18 ⫺143.57 ⫺141.67 10.04 9.82 8.71 Sucrose (Methyl–THF–O– THP) Axial–pseudo-axial 1,2-Linked (31 atoms) B3LYP/6-31G ⴱ 98.67 ⫺36.51 ⫺150.22 0.00 1.88 100.72 80.05 ⫺40.16 ⫺63.67 ⫺152.76 ⫺150.38 0.63 0.00 2.44 – Nigerose Axial–equatorial 1,3-Linked B3LYP/6-31G ⴱ RHF/6-31G ⴱ MM3 ⫺48.53 ⫺46.77 ⫺41.77 ⫺36.46 ⫺33.28 ⫺40.52 ⫺142.59 ⫺145.16 ⫺141.97 7.64 8.22 8.41 7.29 8.25 – Laminarabiose Equatorial–equatorial 1,3-Linked B3LYP/6-31G ⴱ RHF/6-31G ⴱ MM3 51.12 54.33 35.66 ⫺20.83 ⫺14.43 ⫺51.21 ⫺141.69 ⫺145.08 ⫺140.57 8.54 8.31 9.81 8.19 8.33 – Maltose Axial–equatorial 1,4-Linked B3LYP/6-31G ⴱ RHF/6-31G ⴱ MM3 ⫺47.15 ⫺47.00 ⫺45.46 ⫺30.70 ⫺32.91 ⫺40.41 ⫺141.31 ⫺143.79 ⫺140.06 8.92 9.60 10.32 7.99 9.37 – Cellobiose Equatorial–equatorial 1,4-Linked B3LYP/6-31G ⴱ RHF/6-31G ⴱ MM3 50.94 53.84 34.74 ⫺19.33 ⫺13.59 ⫺51.72 ⫺140.62 ⫺143.94 ⫺138.86 9.60 9.45 11.52 8.67 9.21 – Galabiose Axial–axial 1,4-Linked B3LYP/6-31G ⴱ RHF/6-31G ⴱ MM3 ⫺48.87 ⫺47.97 ⫺44.66 ⫺37.12 ⫺35.83 ⫺38.49 ⫺140.44 ⫺143.64 ⫺140.06 9.78 9.75 10.32 8.85 9.51 – Pseudo dimer d Di-axial 3,3-Linked B3LYP/6-31G ⴱ HF/6-31 ⴱ MM3 35.26 32.91 33.36 35.26 32.92 35.65 ⫺134.14 ⫺136.63 ⫺133.59 16.08 16.76 16.79 15.75 16.80 – Pseudo dimer d Axial–equatorial 3,3-Linked B3LYP/6-31G ⴱ HF/6-31 ⴱ MM3 35.56 32.34 42.14 35.47 33.87 41.24 ⫺134.47 ⫺136.54 ⫺133.55 15.75 16.84 16.83 15.42 16.89 – Pseudo dimer d Di-equaotrial 3,3-Linked B3LYP/6-31G ⴱ HF/6-31 ⴱ MM3 35.91 37.29 38.88 35.91 37.29 38.38 ⫺134.76 ⫺136.32 ⫺133.57 15.46 17.07 16.81 15.14 17.11 – RHF/6-31G ⴱ MM3 c c – a For the trehalose and sucrose analogs, f is defined as O5–C1–O1–C1 0 (O5–C1–O1–C2 0 for sucrose) and c is defined by O5 0 –C1 0 –O1– C1 (O5 0 –C2 0 –O1–C1 for sucrose). For nigerose, laminarabiosse, maltose and cellobiose analogs, f is defined by H1–C1–O1–C4 0 (H1–C1– O1–C3 0 for nigerose and laminarabiose), and c is defined by H4 0 –C4 0 –O1–C1 (H3 0 –C3 0 –O1–C1 for nigerose and laminarabiose). b Energies calculated with gamess and Gaussian 94 were the same to seven decimal places. The lowest electronic energies were for the a,atrehalose analog, at ⫺613.7542486 a.u. (RHF/6-31G ⴱ) and ⫺617.5741626 a.u. (B3LYP/6-31G ⴱ). c Relative energies are affected because the sucrose analog without the methyl group has fewer atoms than the other analogs. Its electronic energies are ⫺574.7118895 (RHF) and ⫺578.25309698 (B3LYP) a.u. Therefore, these relative values are not included. d These molecules have an equivalent minimum at ⫺f , ⫺c . 308 A.D. French et al. / Journal of Molecular Structure 556 (2000) 303–313 minimizations of the RHF/6-31G ⴱ structures. No other basis sets were used. In some calculations with the B3LYP level of theory, satisfactory termination of the minimization required the use of a step size of 0.05 [IOP(1/8 5)]. This was used with the SCF Tight keyword. All DHf values are for 298⬚C. For the molecules with head-to-head glycosidic bonds, the f torsion angle was defined by O5–C1– O1–C1 0 (O5–C1–O1–C2 0 for the sucrose analogs), and the c torsion angle was defined by C1–O1–C1 0 – O5 0 (C1–O1–C2 0 –O5 0 for the sucrose analogs). See Fig. 1 for the atom numbering. Torsion angles for the other structures were defined with the hydrogen atoms attached to the carbon atoms that participate in the inter-ring linkages (e.g. f H1–C1–O1–C4 0 and c C1–O1–C4 0 –H4 0 for galabiose (a-galactosyl(1,4)-galactose), maltose and cellobiose). 3. Results and discussion DHf values and f , c coordinates of the global minima for the 13 analogs are in Table 1. Fits of the QM DHf values to the relative electronic energies are shown in Fig. 2, and the fits of the RHF/6-31G ⴱ and MM3 DHf results to the B3LYP/6-31G ⴱ DHf data are shown in Fig. 3. The THF–O–THP analog of sucrose is not included in Fig. 2 and its relative electronic energy is not in Table 1. Its electronic energy is much smaller than for the other molecules because it has three fewer atoms. Alternative treatments of the data based on subtracted averages instead of relative energies give essentially the same results. The most important results are the large, 15– 17 kcal ranges for these isomers, and the fact that the DHf values computed with all three levels of theory and the relative electronic energies are roughly parallel. Also, the energies are in three distinct clusters. The lowest values of about ⫺150 kcal are for the analogs of the trehaloses and the methylated THF–O– THP analog of sucrose. Those molecules have two glycosidic bonds that connect to the same oxygen atom. In a second cluster, about 8 kcal higher, are the analogs with a single glycosidic bond (nigerose, laminarabiose, maltose, cellobiose and galabiose). The THF–O–THP sucrose analog, with two anomeric centers and three fewer atoms and bonds, coincidentally falls in that group, too. Highest, by another 7–8 kcal, are the molecules having simple ether linkages with no glycosidic bonds. The a,a-trehalose analog with di-axial linkages had the lowest electronic energy by either QM method. This molecule, wherein all of its O–C–O–C sequences are in the preferred gauche conformations, is stabilized by both endo- and exo-anomeric effects. The lowest DHf values for the B3LYP and MM3 levels were found for the methylated sucrose analog. It has an axial linkage bond from the pyranosyl ring, and a pseudo-axial linkage bond from the furanosyl ring. Along with its f and c values, the O1–C2 0 – O5 0 –C5 0 torsion angle deviates substantially from the theoretically preferred gauche conformations because of the furanosyl ring. These non-ideal conformations, as well as slightly higher electronic energies, suggest that the MM3 and B3LYP DHf values might be somewhat lower than the correct value for Me–THF–O– THP. Labanowski et al. found that the B3LYP/6-31G ⴱ calculations gave statistically more accurate calculations of DHf than did the RHF calculations with the 6/ 31G ⴱ basis set [10]. Also, the magnitude of the groupenthalpy increments is generally considerably smaller with B3LYP theory than for the RHF theory. This is especially so for the group-enthalpies for the ring size, with the value for six-membered rings being 5.97 kcal by RHF/6-31G ⴱ theory and only 1.37 kcal by B3LYP theory. The five-membered ring enthalpy is similarly reduced, from 4.60 to 0.70 kcal. The reduction in group-enthalpy increment size is probably due to the accounting for more of the electron correlation in B3LYP theory. The main exception is for the O–C– C–O group (see below). The other B3LYP group enthalpies are smaller than their RHF counterparts by about one third. This suggests that the B3LYP theory has less need of supplemental information to predict DHf values [10]. However, a recent study of monosaccharides indicates that B3LYP/6-31G ⴱ theory should not be used on compounds with many hydroxyl groups such as sugars [15]. The average difference between the B3LYP relative electronic energies and the relative DHf values (0.33 kcal) is slightly larger than the comparable difference for the RHF calculations (⫺0.10 kcal). Table 1 and Fig. 2 show that for both QM methods, the biggest deviation between the relative electronic energy and the relative DHf is for the methylated A.D. French et al. / Journal of Molecular Structure 556 (2000) 303–313 309 Fig. 2. Agreement between the relative electronic energy and the relative DHf for our compounds, as calculated by the RHF and B3LYP levels of theory. The points corresponding to the methylated sucrose analog (Me–THF–O–THP) are indicated, showing the relative electronic energy to be 1.85 kcal/mol greater than the relative DHf while the other compounds agree quite well. Relative DHf and electronic energies are relative to the lowest values found for any of the compounds—see Table 1. sucrose analog (B3LYP 1.88 kcal, RHF 1.81 kcal). This suggests that methyl groups attached to the anomeric carbon may require a different groupenthalpy increment, compared to the methyl group attached to non-anomeric carbon atoms. Table 1 and Fig. 3 show that the disagreement between the B3LYP and MM3 DHf values is strongest for the cellobiose analog (1.76 kcal) and for THF–O–THP analog of sucrose (1.49 kcal). Other than the offset shown by Fig. 3, there is no large discrepancy between the RHF and the B3LYP DHf values. The preference for the axial glycosidic-bonded members of the nigerose/laminarabiose and maltose/ cellobiose pairs is small by the B3LYP method and smaller by RHF theory. Either on the basis of the relative electronic energies or the relative DHf values, the axial–equatorial analog of nigerose is preferred by 0.90 kcal (B3LYP) or 0.08 kcal (RHF) over the diequatorial laminarabiose analog. The axial–equatorial maltose analog is preferred by 0.68 kcal (B3LYP) or 0.15 kcal (RHF) over the di-equatorial cellobiose analog. The axial–equatorial members of these pairs are favored more with MM3, by 1.40 kcal and 1.20 kcal, respectively. By MM3, the DHf and f and c values for the galabiose analog are roughly comparable to those for maltose. However, the Table 2 Relative values of the most variable MM3 energy components for the trehaloses at the respective minimum-energy conformations Trehalose Total Bending Dipole–dipole Torsion Sum of major components a,a a,b b,b 0.00 2.23 1.43 1.68 1.23 0.00 0.00 1.60 0.95 0.00 1.24 2.39 1.68 (0.00) 4.07 (2.39) 3.34 (1.66) 310 A.D. French et al. / Journal of Molecular Structure 556 (2000) 303–313 Fig. 3. Comparison of the MM3 and RHF heats of formation to the B3LYP DHf. The RHF values are systematically higher than the B3LYP values by 2.85 kcal, while the MM3 values and the B3LYP values agree very well, except for the cellobiose analog and THF–O–THP molecule, for which the agreement is only fair. axial–axial configuration in the former has a less stable B3LYP DHf value by 0.9 kcal, compared to the DHf value for the axial–equatorial configuration in maltose. As shown in Fig. 3, the RHF values of DHf are offset from the B3LYP values by ⫺2.85 kcal on the average, a rather large systematic discrepancy. On the other hand, the MM3 values are fairly similar to the B3LYP values, being 0.34 kcal higher on average. Indeed, differences between our RHF and MM3 calculated DHf values are relatively large, compared to the mean errors of less than 1 kcal reported for these methods [3,11]. Still, confidence in each method is enhanced by the parallel results. The relative QM electronic energies strongly support the ranking of the DHf values for these isomeric analogs. Work to be published elsewhere that involved hybrids of RHF/ 6-31G ⴱ and MM3 energy surfaces [4,5] depends to some extent on compatibility between these two levels of theory. The comparable DHf values from both the QM and MM3 methods imply the validity of the assumption of comparable bond stretching, angle bending and van der Waals forces to a first approximation. Table 2 shows the MM3 components of the energy that differ the most among the minimum energy structures of the three trehaloses. (Our previous energies for the methylated analogs of the trehaloses [16] were calculated at a dielectric constant of 4.0, not appropriate for comparison with QM calculations.) Although not expected to be quantitatively meaningful (because of coupling between terms during parameterization), such component analysis offers qualitative insight into dominant energetic interactions when differences are large. Di-axial a,a-trehalose has the most bending energy and di-equatorial b,b-trehalose has the least. The dipole–dipole interactions are the highest (most destabilizing) for a,btrehalose. In its minimum-energy conformation, the C1–O5 and C1 0 –O5 0 bonds are adjacent and roughly parallel (Fig. 1) while the same bonds are more distant and directed antiparallel in the other two trehalose analogs. The relative torsion angle energies of 0.00, 1.24 and 2.38 kcal are consistent with the notion of A.D. French et al. / Journal of Molecular Structure 556 (2000) 303–313 311 Table 3 Relative energy components of a,a-trehalose and cellobiose analogs from MM3 DHf calculations (kcal) Component Compression Bending Bend–bend Stretch–bend van der Waals 1,4 van der Waals Other Torsion Torsion–stretch Dipole–dipole Total steric energy Bond enthalpy increments C–C C–O C–H Group enthalpy increments O–CHR–O O-Sec sp 3 O–C–C–O Population, torsion &translation/rotation Total DHf and D DHf a,a-Trehalose Cellobiose Difference 0.40 4.02 ⫺0.07 0.25 13.59 ⫺0.60 ⫺2.17 ⫺0.02 ⫺5.62 9.76 0.52 2.97 ⫺0.11 0.20 13.20 ⫺0.22 2.76 ⫺0.03 ⫺1.80 17.50 0.12 ⫺1.05 ⫺0.04 ⫺0.05 ⫺0.39 0.38 4.93 ⫺0.01 3.82 7.74 8 × 2:45 19:58 6 × ⫺14:32 ⫺85:91 18 × ⫺4:59 ⫺82:62 8 × 2:45 19:58 6 × ⫺14:32 ⫺85:91 18 × ⫺4:59 ⫺82:62 0 0 0 2 × ⫺6:618 ⫺13:236 – – 2.40 1 × ⫺6:618 ⫺6:62 1 × ⫺5:03 ⫺5:03 1 × 1:84 1:84 2.40 6.62 ⫺5.03 1.84 0 ⫺150.03 ⫺138.86 11.17 two, one and zero anomeric stabilizations for the a,a-, a,b- and b,b-trehaloses, respectively, while the exoanomeric stabilizations remain constant. These exoanomeric stabilization energies may not actually be constant, however. For one thing, the c torsion for the a,b-trehalose analog deviates substantially from the gauche conformation. Another consequence of the preferred conformation of a,b-trehalose affects the ability of the lone pairs of electrons on the linking oxygen atom to delocalize into the s ⴱ orbitals for the C1–O5 and C1 0 –O5 0 bonds. For a,a- and b,btrehalose analogs in their minimum-energy conformations, each of the lone pairs of electrons on the linkage oxygen is positioned to stabilize the molecule by being trans to a different C1–O5 bond. In the a,btrehalose analog, one of the lone pairs is trans to both C1–O5 bonds and the other lone pair is trans to neither. This analysis is based on the very rough approximation that the lone pairs on oxygen have sp 3 orientations. Still, less stabilization would be expected under this situation. Research continues in this area. Table 3 shows the different components of the MM3 DHf calculation for the analogs of a,a-trehalose and cellobiose which, except for the analogs of the imaginary pseudo disaccharides, gave the low and high MM3 DHf values. The differences between the two molecules include 7.8 kcal of steric energy (MM3’s standard “Total Steric Energy”) and 3.4 kcal of group enthalpy additions. The main contributions to the steric energy difference include 1.05 kcal for bending, favoring the di-equatorial cellobiose analog; 4.93 kcal of torsional energy favoring the trehalose analog (two anomeric effects and one extra exo-anomeric effect); and 3.82 kcal of dipole–dipole interactions, also favoring the trehalose analog. The extra anomeric group effect in trehalose is offset to some extent by the presence of a group effect for the secondary sp 3 oxygen atom in the cellobiose analog. The nigerose analog has a considerably lower MM3 DHf (Table 1, 1.91 kcal) than the maltose analog at least partly because of a group increment of 1.83 kcal for the O5 0 –C5 0 –C4 0 –O1 sequence in the maltose analog model. The nigerose/maltose difference of 1.91 kcal by MM3 can be compared with the smaller 1.37 and 1.28 kcal differences calculated by RHF and B3LYP theory for the same molecules. The O–C–C–O group 312 A.D. French et al. / Journal of Molecular Structure 556 (2000) 303–313 enthalpy increments for the RHF and B3LYP calculations are only 0.26 and 0.58 kcal, respectively (the only group-enthalpy increment that is bigger for B3LYP theory than for RHF theory). A special O–C–C–O group enthalpy increment was not developed for when one of the oxygen atoms was part of an anomeric center. The laminarabiose analog also has a lower MM3 DHf (by 1.71 kcal) than the cellobiose analog. The laminarabiose analog is favored by the RHF and B3LYP DHf values by the smaller amounts of 1.14 and 1.07 kcal. Because our compounds are isomeric (except for the THP–O–THF analog of sucrose), there are 18 H–C, 8 C–C and 6 C–O bonds in all of the molecules. Despite the different group effects that apply, their DHf values track their electronic energies fairly well as shown in Fig. 2. That property makes the present compounds a valuable and interesting set, even though they have not yet been synthesized and no experimental DHf values exist. Of course, synthesis and DHf experiments on the different members of this set would be interesting to confirm the very large differences in DHf. Limiting the study to only the tetrahydropyran and tetrahydrofuran models has several advantages over the studying the actual sugars when QM is used. For one, the computer time is reduced. Secondly, the hydrogen bonding, a serious matter for true carbohydrates, is avoided in these molecules. In particular, basis set superposition errors are fairly large (e.g. 2.3 kcal per hydrogen bond) at the B3LYP/6-31G ⴱ level for hydrogen-bonded water and methanol clusters [15], so molecular energies of actual disaccharides which have 8 or so hydroxyl groups might be badly exaggerated. In compensation for ignoring the hydrogen bonding, we were able to focus on the impact of various sorts of glycosidic linkages on molecular energies. While these calculations suggest that it is feasible to calculate heats of formation for these compounds and thus the disaccharides themselves with some expectation of accuracy, our analog DHf values cannot be extrapolated to the parent disaccharides themselves. For example, the maltose molecule would not be expected to have a DHf value some 8 kcal above that of a,a-trehalose. The real disaccharide has two anomeric centers, but the analog has only one. The additional group effect plus the intrinsically calculated enthalpy for the second anomeric center would increase the calculated stability of the maltose molecule. 4. Conclusions DHf values for disaccharide analogs based on tetrahydrofuran and tetrahydropyran were calculated at the B3LYP/6-31G ⴱ, RHF/6-31G ⴱ and MM3 levels. Insight was gained by studying five molecules having various configurations of glycosidic linkages with overlapping exo-anomeric sequences and five others having ordinary glycosidic linkages. Three other isomers with simple ether linkages were also studied. There was a large (15–17 kcal) range of electronic energies and DHf for the conformations at the global minima of the different isomeric analogs, and there was reasonable correspondence between the DHf values from both MM3 and QM calculations. The mostly excellent agreement between the MM3 and B3LYP/6-31G ⴱ values for these moderately large, problematic compounds having anomeric sequences is especially remarkable because the parameterization for MM3 was established more than a decade ago. Even the fit to the RHF/6-31G ⴱ results, after compensating for a 2.85 kcal offset, was quite good. Of course, synthesis of these compounds and experimental verification of the large range in their heats of formation is a good example of actual experiment suggested by computations. Acknowledgements We thank Professor N.L. Allinger for valuable discussion along with Dr K.-H. Chen for furnishing the prepublication parameters. Dr Chen also assisted with the QM DHf calculations. We also thank Professor Lothar Schäfer for suggesting the collaboration between ADF and AMK. Besides Professor Allinger, Drs J. Labanowski and W.E. Franklin read presubmission versions of this paper. References [1] A.D. French, M.K. Dowd, S.K. Cousins, R.M. Brown, D.P. Miller, ACS Symp. Ser. 618 (1995) 13–37. [2] A.D. French, M.K. Dowd, P.J. Reilly, J. Mol. Struct. (Theochem) 395/396 (1997) 271–287. [3] N.L. Allinger, L.R. Schmitz, I. Motoc, C. Bender, J.K. Labanowski, J. Phys. Org. Chem. 3 (1990) 732–736. [4] A.D. French, A.-M. Kelterer, C.J. Cramer, G.P. Johnson, M.K. Dowd, Carbohydr. Res. 326 (2000) 305–322. A.D. French et al. / Journal of Molecular Structure 556 (2000) 303–313 [5] A.D. French, A.-M. Kelterer, G.P., Johnson, M.K. Dowd, C.J. Cramer. Constructing and evaluating energy surfaces of crystalline disaccharides. J. Mol. Graphics and Modelling 18 (2000) in press. [6] G.R.J. Thatcher (Ed.), The Anomeric Effect and Associated Stereoelectronic Effects ACS Symposium Series, vol. 539, ACC, Washington, DC, 1994. [7] I. Tvaroška, T. Bleha, Adv. Carbohydr. Chem. Biochem. 47 (1989) 45–123. [8] N.L. Allinger, L.R. Schmitz, I. Motoc, C. Bender, J.K. Labanowski, J. Am. Chem. Soc. 114 (1992) 2880–2883 (See the supplemental information for this paper for many of the details of the heat of formation calculations. Copies are available from Professor Allinger). [9] N.L. Allinger, K. Sakakibara, J. Labanowski, J. Phys. Chem. 99 (1995) 9603–9610. [10] J.K. Labanowski, L. Schmitz, K.-H. Chen, N.L. Allinger, J. Comput. Chem. 19 (1998) 1421–1430. [11] N.L. Allinger, M. Rahman, J.-H. Lii, J. Am. Chem. Soc. 112 (1990) 8293–8307. 313 [12] K.-H. Chen, N.L. Allinger, personal communication. [13] M.J. Frisch, G.W. Trucks, H.B. Schlegel, P.M.W. Gill, B.G. Johnson, M.A. Robb, J.R. Cheeseman, T. Keith, G.A. Petersson, J.A. Montgomery, K. Raghavachari, M.A. AlLaham, V.G. Zakrzewski, J.V. Oritz, J.B. Foresman, J. Cioslowski, B.B. Stefanov, A. Nanayakkara, M. Challacombe, C.Y. Peng, P.Y. Ayala, W. Chen, M.W. Wong, J.L. Andres, E.S. Replogle, R. Bomperts, R.L. Martin, D.J. Fox, J.S. Binkley, D.J. DeFrees, J. Baker, J.P. Stewart, M.L. HeadGordon, C. Gonzalez, J.A. Pople, Gaussian 94. Gaussian, Inc., Pittsburgh, PA, 1995. [14] M.W. Schmidt, K.K. Baldridge, J.A. Boatz, S.T. Elbert, M.S. Gordon, J.H. Jensen, S. Koseki, N. Matsunaga, K.A. Nguyen, S.J. Su, T.L. Windus, M. Dupuis, J.A. Montgomery, J. Comput. Chem. 14 (1993) 1347–1363. [15] J.-H. Lii, B. Ma, N.L. Allinger, J. Comput. Chem. 20 (1999) 1593–1603. [16] M.K. Dowd, P.J. Reilly, A.D. French, J. Comput. Chem. 13 (1992) 102–114.
© Copyright 2026 Paperzz