Molecular Dynamics Simulations and the Conformational Mobility of Blood Group Oligosaccharides LHEN-YI YAN* and C. ALLEN BUSH*

C k p S ~ r r nt w of C h ~ r n i s t r y ,lllinoi5 lnstitutc of Tec hnologc Chicago I L 60616 U S A

SYNOPSIS

Molecular dynamics simulations were carried out without explicit consideration of solvent to explore t h e conformational mobility of blood group A and H oligosaccharides. T h e potential energy force field of Rasmussen and co-workers was used with the CHAHMM program on a number of disaccharide and trisaccharide models composed of fucose, galactose, glucose, N-acetyl glucosamine, and N-acetyl galactosamine chosen t o represent various fragments of blood group oligosaccharides. In agreement with results of earlier studies, stable chair conformations were found for each pyranoside from which no transitions were detected in simulations as long as 800 ps. Exocyclic dihedral angles, including t h a t of C5-C6, generally show numerous transitions on a time scale of approximately 5-30 ps. T h e dihedral angles of some but not all glycosidic linkages of blood group ohgosaccharides show transitions on t h e time scale of 30-50 ps, implying that t h e extent of internal motion in blood group ohgosaccharides depends strongly on linkage stereochemistry. For certain blood group A and H ohgosaccharides t h a t show limited internal motion in these simulations, we argue that the calculations are Consistent with our previous analysis of ' H nuclear Overhauser enhancement (NOE) data t h a t imply single conformations over a wide range of temperature and solvent conditions. While the trajectories are consistent with '"C Ti values t h a t have been interpreted as indicating rigid conformations, measurements of "C-NOE and TI as a function of magnetic field strength are proposed as an improved method for experimental detection of the internal motion that is suggested for certain oligosaccharides in these simulations. T h e results of these simulations differ substantially from those of peptides of a similar molecular weight in that t h e ohgosaccharides show much less internal motion.

I NTRODUCT 1 0N T h e forces stabilizing the conformations of peptides, proteins, and nucleic acids have been the subject of intense study by both theoretical techniques of computer modeling and by various experimental methods. Several important forces have been identified and each is represented in the terms of t h e various force fields t h a t are used in computer modeling studies. Foremost are the van der W a d s interactions, the repulsive part of which ' 1990 .John Wiley & Sons, Inc.

(YX'000~-3525,/90/040799~13 $04.00 Biopolymers, Vol. 29, 799-81 1 (1990) .Present address: Department of Chemistry and Hiochemistry. I.niversity of Maryland, Baltimore County, Baltimore, MI> 21228.

represents steric exclusion. Other important contributions include electrostatic interactions and hydrogen bonding. Although hydrophobic effects are thought t o play a substantial role in stabilizing many peptide conformations, their explicit dependence on the solvent structure makes it difficult t o include them in a force field representing the peptide alone. In the case of nucleic acids, the effects of polarizability of the aromatic heterocyclic bases are also added to the electrostatic dipole effect. Solvent plays an important role in modulating the strength of hydrogen bonding as well as of the hydrophobic effect. The importance of solvent can be experimentally demonstrated in solvent denaturation of nucleic acid and protein structures. Details of the influence of electrostatic effects in stabilizing the a-helical conformation in small pep799

800

YAN A N D RUSH

tides has been recently demonstrated in the experimental results of Baldwin and co-workers.' While the interplay of forces stabilizing the conformations of complex carbohydrates of glycoproteins and glycolipids has been less studied than have the interactions stabilizing peptide conformation, the dominant forces in oligosaccharides seem to be somewhat different from those in proteins and nucleic acids. Experimental nmr and CD studies and computational modeling indicate that van der Waals effects are dominant in carbohydrates, and that electrostatics, hydrogen bonding, and hydrophobic effects are less important. For blood group oligosaccharides, direct experimental evidence concerning the dependence of the conformation on solvent and on temperature support these Although the applicability of these generalizations to oligosaccharides other than blood groups remains to be demonstrated, there are indications that they may be valid for the complex oligosaccharides of N-Asn linked glyc~proteins.~ Not only are the forces that stabilize the conformations of complex oligosaccharides somewhat different from those that stabilize peptides and nucleic acids, but there are also differences in the distribution of conformations contributing to an ensemble average conformation. While it is generally thought that a typical small peptide exists in a distribution of conformations rather than in a single rigid conformation, this appears not to be the case for a t least some of the complex oligosaccharides. In contrast to the distribution of conformations observed experimentally and in computer modeling of small peptides, there is evidence that certain complex oli gosaccharides can have relatively rigid well-defined conformations. I t was originally proposed by Lemieux and co-workers that blood group oligosaccharides have single rigid conformations that are primarily determined by van der Waals interactions and torsional potentials.s Recent nmr studies based on complete 'H assignments for blood group A and H oligosaccharides support these proposals. Nuclear Overhauser effect (NOE) data on blood group oligosaccharides having four to six sugar residues can be reconciled with single low-energy conformations determined mainly by van der Waals interactions rather than by hydrogen bonding and electrostatic effects.6 Nuclear magnetic resonance and CD experiments have shown that these conformations are not very sensitive to solvent and t e m p e r a t ~ r e . ~ The - " , ~results of nmr experiments on asparagine-N linked oligosaccharides of glycoproteins give a somewhat similar picture. Brisson and Carver8b9 have used proton

nmr to study the conformation of the complex and high mannose glycopeptides, with special emphasis on the conformation of the oligomannosides." They demonstrated that oligosaccharide chains of these glycopeptides show certain distinct conformational features and that they are not completely flexible random coils. Paulsen et al.' l2 have reported conformational energy calculations and proton NOE on a series of synthetic oligosaccharides closely related to the N-asparagine linked glycopeptides. The general conclusion reached by both research groups was that the conformations about the (1 + a), (1 3), and (1 4) glycosidic linkages in these glycopeptides show well-defined conformations. The Man ( a - l 6) Man linkage, which occurs as a branch point in these glycopeptides, is a special case as a result of the rotation of three single bonds in the intersaccharide linkage leading to great flexibility. While there continues to be some controversy about the details of just exactly how broad a distribution of conformations must be considered for each glycosidic linkage in the oligomannosides, there is a consensus that certain of the oligomannoside linkages can have single conformations while others may have varying degrees of conformational freedom, especially about the Man ( a - 1 + 6) Man linkage.1 9 - 15 In addition to the studies cited above, Kochetkov and co-workers concluded from energy calculations, NOE data, coupling constants, and optical linkage rotations and in aqueous solutions of cellobiose and maltose derivatives there was a complicated conformational equilibrium.'"'8 It would appear that some oligosaccharides can be represented as single conformations or an ensemble of very closely similar conformations, while others may have a broader distribution of conformations contributing to the ensemble. Apart from the question of how broad a distribution of conformations contributes to the conformational ensemble of a biopolymer, one must consider the time scale for exchange among the conformations that contribute to the ensemble. Molecular dynamics (MD) simulation provides a useful tool for detailed descriptions of the fluctuations of individual atoms and variations with time of structural parameters such as torsion angles and interatomic distances. Simulations have been used widely in recent years to study the internal motions of proteins and of peptides. Substantial internal mobility has been found for small peptides, which can show numerous transitions among conformational energy minima over a time scale of tens of picoseconds.'9 Although the method has been applied extensively to proteins and nucleic --f

+

MOLECULAR DYNAMICS ANI) CONFOftMA'FIONAI. MOHI 1,ITY

acids, only a few reports of MD simulations of carbohydrates have appeared. The experimental results for carbohydrates discussed above suggest that MD simulations of oligosaccharide conformations might give results that differ qualitatively from those found for peptides. Since MD simulation has been very instructive concerning the dynamics of peptides and proteins, application to blood group oligosaccharides might provide some insight concerning exactly which types of glycosidic linkage are most likely to have rigid conformations. Identification of points of flexibility in a complex oligosaccharide could contribute to the understanding of the biological significance of the complex carbohydrate side chains of glycoproteins. In spite of our primitive understanding of the details of the potential energy functions for sugars, such dynamics calculations might assist in classification of relative flexibilities of various glycosidic linkages. Previous attempts in this direction include the work of Brady,20x21who has reported MD simulations of the monosaccharides, a- and b-D-glucopyranose, showing that the mean dynamical structure was acceptably close to the crystal structure. Homans et a1.22 performed an MD simulation of two mannose disaccharides for 10 ps. The average structures were reported without a detailed analysis of this brief trajectory. Ha et aL2" have reported studies of the disaccharide, maltose, in which they carefully studied the potential energy constructing a relaxed energy surface containing a number of distinct energy minima. They reported short trajectories started a t each of the minima, some of which showed conformational transit ions.

METHODS The general molecular mechanics code, CHARMM, developed by Karplus and co-workers, was used in this s t ~ d y . ' ~Newton's equation of motion was integrated by a Verlet scheme with a step size of 5 x 10- l 6 s. All bond lengths were kept constant during the simulations using the SHAKE algorithm with an error tolerance of lop6. Following Brady's work,") the parameters for the force field developed by Rasmussen and co-workers25,26are used for the oligosaccharide simulations. Since the parameters for atoms in the amide group are not defined in Rasmussen's force field, the acetamido sugars, GlcNAc and GalNAc, were accommodated in the MD simulations with parameters for the

801

Table I Oligosaccharide Fragments Used in the MD Study No.

Primary Structure" Fuc(a-1 + 2)Gal P-0-methyl Gal(a-1 --* 3)Gal P-0-methyl GalNAc( a-1 + 3)Gal P-0-methyl Fuc(a-l

+ 2)\ ~

Gal /3-0-methyl

GalNAc(a-1 --* 3) Gal( P-1 4 4)Glc p-0-methyl Gal( p-1 + 3)Glc B-0-methyl Gal( P-1 + 3)GlcNAc P-0-methyl Fuc( a-l + 2)Gal(p-1 + 3)GlcNAc P-0-methyl "All monosaccharide units are D sugars except for Fuc, which is an L sugar.

amide group taken from the CHARMM parameter set.24Although these two parameter sets are not strictly compatible, there is no single complete parameter set that includes both carbohydrates and amides available a t this time. The primary structures of the fragments of the blood group A and H oligosaccharides used in present study are given in Table I. There are eight structures, six of which are disaccharides and 2 are trisaccharides serving as models for the nonreducing terminal fragments of blood group A and type 1 H blood group structures. These structures were chosen because of the availability of nmr data, including NOE and T, data and related model building studies. In the blood group and related oligosaccharides, all the sugar residues exist in the pyranose form and belong to the D series, except for fucose, which is an L sugar. All atoms of the oligosaccharides were included explicitly in the simulations. The x-ray structure of the oligosaccharides"; were first refined by energy minimization using the standard CHARMM minimizers ABNR and NRAP, both of the Newton-Raphson type.24 The atoms were then assigned velocities from a Maxwellian distribution at a low temperature (0-10 K) and heated in increments of 30 K to a final temperature of 300 K. After the heating period, the residual overall transitional and rotational motion was removed to simplify the analysis of the subsequent conformational fluctuations. Each trajectory was equilibrated for 20-40 ps with periodic temperature scalings during the equilibration period to bring the system to a temperature acceptably close to the desired temperature of 300 K. The equilibration period was terminated when n o systematic

''

802

YAN AND RUSH

changes in t h e temperature were evident over a time period of about 10 ps. The actual dynamical simulation results, which are coordinates and velocities for all the atoms as a function of time, were then obtained by continuing t o integrate the equations of motion for 120 ps or longer for each oligosaccharide. The coordinates of atoms were saved at every 25 time steps for subsequent analysis. Standard nomenclature and conventions for dihedral angles used throughout this study are those adopted by a commission of the International Union of Pure and Applied Chemistry and International Union of biologist^.^" The glycosidic dihedral angles designated by + and 4,which are measured in degrees, are defined as follows. The glycosidic dihedral angle is defined by the four atoms Oring-Cl-Ol-C; and 4 is defined by C1- 01- -C; . The initial conformation, (+, 4 ) = (O",O"), is taken to be the one in which the Oring-C1 bond eclipses the 01-C; bond ( + rotation) and the C1-01 bond eclipses the Cd- C.,! bond ( 4 rotation) with a clockwise rotation taken as positive.

+

~

~,

RESULTS lime-Averaged Structures

Following the MD simulations, coordinate sets as a function of time were used for the calculation of time-averaged structures. Selected internal coordinates (bond angles and dihedral angles) and the rms fluctuations in these internal coordinates were calculated for the dynamically averaged structures of all the oligosaccharides listed in Table I over trajectories a t least 120 ps in length. Comparison of these calculated coordinates (data not shown) with the x-ray diffraction data for crystals of the monosaccharides, p-D-galactose and a-L-fucose,2s and those for residue GalP in Gal(0-1 + 4)GlcNAcaZ9and glucose2' with the dynamical average values of the pyranoside internal coordinates shows t h a t the dynamical structure of each pyranose ring is quite close to that for the crystal. The dynamical values differ from the experimental data within the range of differences between the two sets of experimental data for Gal. The dynamical averages for the bond angles are in most cases larger than the crystal values, and also larger than the equilibrium values selected for the energy function, which agrees with Brady's results.z"~2'The agreement of the dynamically averaged structures

with crystal data confirms t h e suitability of the Rasmussen potential energy functions for these sugars. Internal Ring Dihedral Angles

Plots of the histories of the internal dihedral angles of the Fuc, Gal, and GlcNAc pyranose ring in the triscaccharide, Fuc[ a-1 -+ 2)Gal( p-1 + 3)GlcNAc ,&]-methyl, are typical for all ring internal dihedral angles of the eight oligosaccharides. Our results (data not shown) are similar to those of Brady2' for glucose, and show that the rms dynamical fluctuation of the internal ring dihedral angles are fairly uniform among the residues in the oligosaccharides and are typically in the range of 5"-8". The small oscillations about fixed pyranoside chair conformations, i.e., 'C, for Fuc and 'C, for Gal and Glc, have a subpicosecond time scale, and tend to be uniform over all the structures. Motions with more collective character and longer time scale characterizing transitions between distinctly different conforniations of the pyranoside ring were not observed for the internal ring dihedral angles of the eight oligosaccharides. The absence of conformational transitions of the rings from the initial chair form to other chair or boat-like forms is consistent with the interpretation of nmr coupling constant data. Exocyclic Dihedral Angles

During these MD simulations, the hydroxyl groups on the ring carbon atoms of all the sugar residues underwent frequent transitions among their allowed conformations due to extensive freedom of motion about the respective C- 0 bonds. Moreover, the exocylic CH,OH groups show several reorientations during the simulations of each oligosaccharide. Unlike the ring dihedral angles, this side-chain dihedral angle on each residue has three low-energy values accessible t o it, +6O" and 180". Although all three allowed values were observed for Gal and GalNAc and for 3-substituted Glc, for 4-substituted Glc in Gal p-(1 + 4) Glc the conformer with +60° is not seen. Since reorientation involves an exchange of hydrogen-bond partners, hydroxyl rotations are expected t o be strongly influenced by the presence of solvent. Although solvent was not explicitly included in this simulation, the empirical force field of Rasmussen et al. includes some condensed phase effects. Therefore the accuracy of this aspect of the simulation is questionable because the barrier t o rotation is effec-

MOLECULAR DYNAMICS AND CONFORMATIONAL MOBIIJTY

tively zero in the torsional energy term. While no experimental data are available t o determine the transitions of the CH,OH groups of these oligosaccharides, Brady 20,21 has examined the population distribution for the three main orientations of the hydroxyl methyl group of glucose in MD simulation and found the results differed from the distribution in crystal structures, a result that he has interpreted as indicating inaccuracy in the potential energy surface. Clycosidic Dihedral Angles

T h e main emphasis of this study is on the glycosidic dihedral angles and their evolution in time. Since the pyranoside rings maintain a relatively rigid conformalion, i t is the angles $I and q5 that are the internal coordinates most important in determining the conformation of the oligosaccharides. In contrast to the rotation of hydroxyl groups about (' -- 0 bonds, the rotations about the C -0 bonds where the glycosidic oxygen is involved are restncted. Seven distinct simulations of the disaccharide Fuc(a-1 -+ 2)Gal P-0-methyl were started from a minimized conformation $I = -72" and 4 = 126", and run for 40 ps after equilibration. Three o f the trajectories showed evidence of conformational transitions and two trajectories were extended t o 160 ps. The history of + and # for a typical trajectory given in Figure 1 shows modera t e flexibility, the torsion angle q5 fluctuating about the starting values for most of time with several transitions to about 50". The motion of angle + is correlated with that of 4.

T o confirm that the molecule is traversing different conformational states rather than undergoing fluctuations about a single conformation, instantaneous configurations were selected a t 75, 95, and 120 ps along the trajectory, and their energies were minimized. They all converged t o a distinctly different conformational energy minimum with glycosidic dihedral angles of q5 = 56.5" and $ = -85.6'. The energy minimizations of some other points-50, 100, and 150 ps-give the starting conformations of $ = 125.7' and = -72.3". These minimum-energy structures characterize the nature of the structural features about which the dynamics fluctuations take place for the potential functions used. In Figure 2, the trajectory shown in Figure 1 is mapped onto the energy surface of the disaccharide, represented by the contour plot of rnergy as a function of + and I/ calculated a t 10" intervals, with all other degrees of freedom fixed. Superimposed on the contours at intervals of 1 kcal/mol is the trajectory represented by the dotted line that connects sequential conformations as a function of time. Thus, the trajectory represents the excursions of the disaccharide in +, 4 space and the transitions among conformational states are readily apparent from this representation. I t can be seen in Figure 2 that one of the energ) minima a t about # = 90" and 0 = -160" is no: touched b:7 the trajectory. Using this point as the starting Tonformation for energy minimization, where all degrees of freedom are relaxed, results in 4 = 125.9' and 9 = -72.7", which is same as the major conformation on the trajectory revealing that this

+

zoo. 150

-

W M

G

O

0

25

50

803

100

75

125

150

Time in p s

Figure 1. Histories of $I (lower trace) a d P-0-methyl.

I$

(upper trace) for Fuc(a-1

+

2)Gal

804

YANANDBUSH

150.

100.

3 50.

approximation the dynamics of GalNAc( a-1 + 3)Gal P-0-methyl are not distinguishable from those of Gal(a-1 + 3)Gal P-0-methyl suggesting the addition of amide group at C2 has little influence on the conformation of this disaccharide. For the blood group A trisaccharide Fuc(a-1 -+ 2)[GalNAc( a-1 + 3)IGal P-0-methyl, two 60 ps trajectories were similar to the 120 ps trajectory given in Figure 4, which shows that the glycosidic linkage of fucose is more flexible than that of GalNAc(a-1 3). The two different linkage conformations that occurred in Figure 1 for Fuc( a-1 + 2)GalP are also seen in Figure 4 for the trisaccharide. However, for the GalNAc( a-1 -+ 3)GalP linkage, the minor conformation in Figure 3 does not appear in any of the trisaccharide trajectories because the conformational energy is over 500 kcal/mol as a result of the repulsive interactions of 0 5 and H1 of fucose with C5 and 0 5 of GalNAc. In spite of the limited length of this trajectory, t h e results suggest t h a t t h e GalNAc(cu-1 + 3)GalP linkage in the trisaccharide is relatively rigid. A stereopair diagram of the superimposed average structures for successive 10 ps time steps during the first 80 ps of molecular dynamics simulation of the blood group A trisaccharide is shown in Figure 5 where a slight rotation of the fucose residue and some conformational transitions of exocyclic groups can be seen. To compare the results of MD simulations with experimental NOE data, the distance between several pairs of protons were plotted as a function of time, (data not shown). The averaged distance between GalNAc H1 and Gal H4 in the trisaccharide is 2.394 A, which is smaller than that between GalNAc H1 and Gal H3 (2.996 A) in agreement with the experimental observations that the NOE between GalNAc H1 and Gal H4 is larger than that between GalNAc H1 and Gal H3.6 The distance between Fuc H1 and Gal H2 is important for comparison of the simulation of the fucosidic linkage conformation with experimental NOE. The two conformations for the fucose linkage along the trajectories show essentially the same value for this distance, an average of 2.484 A, which gives NOE consistent with experimental values. Therefore, experimentally observed NOE of Gal H2 on irradiating Fuc H1 cannot be used to distinguish the major and minor conformations of the fucosidic linkage in solution. However, the internuclear distances between Fuc H1 and GalNAc H3 or H5 are distinguishable for the two conformations. The NOE data reported by Bush et al.6 suggests that the major conformation must dominate in aqueous so+

0. - 150.

- 100.

-50.

0.

0 Figure 2. Energy contour map (not relaxed) of Fuc (a-1 -+ 2)Gal /3-0-methyl plotted on the trajectory shown by dots.

apparent energy minimum is an artifact of the rigid body approximation for energy surface calculations. T o test the importance of the initial conformation for the trajectory, the simulation was repeated by starting from the other minimized conformation # = 56" and C$ = -85". During the equilibration period, the I+L and + angles transited to the values corresponding to the major conformation in Figure 1 and a similar trajectory was observed, demonstrating the insensitivity of the dynamic trajectories to the initial conditions. For the disaccharide Gal(a-1 + 3)Gal p-0methyl, three different 40 ps trajectories were started from the lowest energy conformation with C$ = 60" and I+L = -160", which corresponds to conformation D in Figure 2 of the paper of Bush e t al.' Although the 40 ps trajectories did not give clear evidence of conformational transitions, extension t o 120 ps (Figure 3) shows that the glycosidic dihedral angles did transit several times to a different conformation with + = 80" and II, = -60" approximately, which is marked as conformation E in Figure 2 Bush et al.' T o test our assumption that the amide group has little influence on the trajectory of the disaccharide, a simulation for GalNAc(a-1 -+ 3)Gal p0-methyl was carried out. The resulting trajectory (data not shown) shows mostly small oscillations about a single minimum energy conformation with a few conformational transitions that are very similar t o those illustrated in Figure 3. At this level of

MOLECULAR DYNAMICS AND CONFORMATIONAL MOB1 IATY

I

1

805

I

I,

300. v1

W b. W

M a,

W

.-

200.

W

0. !

I

0.

25.

1

1

50.

75.

100

Time in ps

Figure 3. Histories of the dihedral angles Q (lower trace) and Gal( a-1 + 3)Gal P-0-methyl.

+

(upper trace) for

significant. P-Methyl lactoside, Gal( p-1 4)Glc P-0-methyl, was chosen as the simplest model of a type 2 chain. Five trajectories of 40 ps were run and one trajectory extended to 120 ps shows most of the features t h a t were observed. The histories of the glycosidic dihedral angles @ and 4 in Figure 6 show two conformations that fall in two of the three local minima on the potential energy map. While the conformation a t 4 = 115.8 and @ = -50.7 is 1.0 kcal/mol higher in energy than the one a t 4 = 132.4 and @ = 37.7 when energy minimized, the former gives a much broader area on the

lution while the minor conformation may be favored in pyridine solution.3 In addition t o the blood group H and A terminal fragments, simulations were also carried out on the more internal part of the blood group oligosaccharides of type 1 and type 2 chains. Since blood group determinants on type 1 and type 2 chains can be immunologically distinguished, the antibody-combining sites are thought to be large enough to accommodate several sugar residues and conformational differences between type 1and type 2 A and H determinants are probably biologically

--f

100. 50.

0.

E

-50. - 100.

2-150.

W

c -200.

150 0

'

I

25.

I

50

75

1

I

100

Time in p s

Figure 4. Histories of the glycosidic dihedral angles for the blood group A trisaccharide Fuc( a-1 + P)[GalNAc(a-1 + 3)lGal P-0-methyl.

806

YAN AN11 RUSH

2)[GalNAc(n-l Figure 5. Stereodiagram of blood group A trisaccharide Fuc(n-l 3)lGal 8-0-methyl, for successive 10 ps steps during t h e first 80 ps of MD simulation. +

potential energy surface and therefore is the major dynamical conformation in Figure 6. The conformation a t I/J = 99.5 and C#I = - 145.4 on the potential energy map, which did not appear on the dynamical trajectories, has a minimized energy of 2.2 kcai/mole higher than the lowest energy conformation. The trajectory shows transitions between conformations differing by 100" in the coordinate with smaller fluctuations in consistent with the potential energy surface in which the energy well in the C#I dimension is three times wider than in the I/ dimension. Although we have no suitable NOE and model building data for /%methyl

+,

I

lactoside, Lipkind et a1." INhave reported data for &methyl cellobioside, a disaccharide whose conformational dynamics should be quite similar. Since the potential energy surface of Scott and Scheraga" used by Lipkind et al." is quite different from that used in our simulations, our results differ substantially, and although their method of averaging the NOE over the conforniational states is not formally correct, ', nevertheless our results are in agreement on the qualitative conclusion that their NOE and 3cx,, data are best interpreted by an ensemble average over several conformations of these (p-1+ 4) linked disaccharides. The experi-

+

''

I

I

I

I

a,

a d ..+

50.

:

I

0.

c

d

-50.

c.

a

2a

-100.

150

1

1 0.

25.

50

75

100

Time in p s

Figure 6. Histories of t h e glycosidic dihedral angles 9 (lower trace) and )I (upper trace) for Gal( 8-1 4)Glc P-0-methyl. +

MOLECU1,AH DYNAMICS AN I ) CONFOIIMATIOh Al, MOHI I , I T Y

807

50 - 100 n

-150 w

:-2oc 50 4

M

C

-2

O

50

a

2- -100 a

150

0

50

25

75

100

Time in p s

Figure 7. Histories of t h e glycosidic dihedral angles 0 (lower trace) and for Gal( P-1 + 3)Glc P-0-methyl.

(upper trace)

for a typical 120 ps trajectory for Gal( /GI 3)Glc P-0-methyl, showing that the disaccharide is more flexible than Gal(P-1 -+ 4)Glc P-0-methyl. The flexibility of the disaccharide can be explained hy the niultimininium potential surface for van der Waals interactions. The disaccharide oscillates in each potential well and exhibits numerous transitions among seven distinct minima that [vere identified by niinimizations against all degiws of freedom on the energy surface. Addition of the amide group to C2 of' Glc yields the type 1 core disaccharide, Gal(P-1 3)GlcNAc P-0-methyl, found in many blood group glycoproteins. A typical molecular dynamics trajectory in

mental d at a used by Lipkind e t al." in their analysis are probably inadequate t o distinguish between t h e force fields of Scott and Scheraga." and that of Rasmussen and co-workers.'"''" All of our simulations of the disaccharide Gal( p1 :l)Glc P-0-methyl, which serves as a model for thr : ? p e 1 chain, show flexibility greater than that seen in ,&methyl lactoside. Three 120 ps simulations, a 440 ps simulation, and an 800 ps simulation were started from the minimum energy conformation a t = -60" and I) = -120", and several conformational transitions t o new values of and 4 oc~ui.i.etlduring each of the simulations. Figure 7 gives thcl histories of the glycosidic angles I) and 0 -j

6 1

4

+

-+

50 100 n 3

?

150

M 3

a

-

"

200

I

200

I

0

25

I

50

I1 75

100

Time in p s

Figure 8. Histories of t h e glycosidic dihedral angles 0 (lower trace) and I+L (upper trace) for Gal( P-1 + 3)GlcNAc P-0-methyl.

808

YAN AND BUSH

0.

-50.

;- 100. c. W M

V

-150. I

I

C d

150 100

0

50

150

100

200

Time in ps

Figure 9. Histories of the glycosidic dihedral angles for Fuc(a-1 + 2) Gal(P-1 + 3) GlcNAc P-0-methylover 240 ps.

Figure 8 is similar to that for Gal(P-1 --+ 3)Glc j3-0-methyl and substantially more flexible than the core of the type 2 chain. In contrast to the histories of the glycosidic dihedral angles of the core disaccharides, the trajectories for the type 1 blood group H trisaccharide Fuc( a-1 2)Gal( p-1 3)GlcNAc P-0-methyl did not show any conformational transitions during a 120 ps trajectory or in the 240 ps MD simulations, given in Figure 9. The absence of conformational transitions, even in this long trajectory, implies a fundamental difference in flexibility of the type 1 chain introduced by the Fuc(a-1 + 2) substituent. The decreased flexibility of this j3-(1+ 3) linkage on addition of the Fuc(a-1 + 2) substituent has also been pointed out in modeling studies by Biswas and Rao,;j2 although our results differ in the detail from the conformation proposed in this early work. The pronounced decrease in the conformational space available to the Gal (P-1 + 3) GlcNAc linkage caused by the fucosyl substituent results mainly from the van der Waals repulsive interactions between fucose and GlcNAc that keep the trisaccharide oscillating in a single potential well in a tightly folded conformation.2 The major conformation found for this trisaccharide in these simulations is essentially the same as that deduced from NOE data and modeling studies of Rao et a1.2

-

--f

DISCUSSION A N D CONCLUSIONS Our results demonstrate that it is very practical to carry out MD simulations for small oligosaccha-

rides over short time ranges with limited computational resources. Simulation of a trisaccharide requires approximately 1 h of computer time on a VAX 11/750 per picosecond of simulation. While simulation times of 100-200 ps are too short to reliably simulate such experiments as T,measurements in nmr, our observation that the disaccharide trajectories are insensitive to starting values of @ and shows that these trajectories do sample all important parts of the conformational space. Since a similar demonstration for trisaccharides was not done, our results for those cases may be only qualitatively valid. I t has been shown that analysis of a dynamic trajectory obtained from a molecular dynamics simulation can provide some insight into the conformational fluctuations of oligosaccharides. A t room temperature, the dynamic structure of oligosaccharides are rich in fluctuations and sometimes conformational transitions, and differ considerably from the static structures represented by conventional molecular models. During the molecular dynamics simulations of the oligosaccharides, conformational transitions for the sugar ring dihedral angles are not observed; the pyranosides oscillate rapidly about miniaum energy chair conformations, 4 C , for the D sugars and 'C, for the L sugars. This result is similar to Brady's2'I2' for a- and j3-glucose, and it shows that the addition of the equatorial substituents at C4 in Fuc and Gal does not induce transitions among chair forms. These oscillatory motions are on the subpicosecond time scale, which is much faster than the rotational correlation time of a typical tetrasaccharide, which is hundreds of picoseconds.

+

MO1,ECULAR DYNAMICS AND CONFORMATIONAL MOHI 1,ITY

809

McCain and M a r k l e ~have ~ ~ analyzed the 13C TI of only a single minimum energy conformation show the carbon atoms of sucrose, which they take to be essentially no conformational transitions a t all. We a rigid disaccharide with little motional freedom have found two especially interesting cases in which about the glycosidic linkage. Their results on I3C the addition of a third residue to a disaccharide greatly restricts the motion of the glycosidic linkT , and NOE data can be interpreted by the age. The flexibility of the a-(1 + 3) linkage is model-independent analysis of Lipari and S ~ a b o ~ ~ in which rapid motions on time scale fast compared greatly diminished by the addition of Fuc a-(1 + 2) t o t he rotational correlation time T ~ contribute , to to GalNAc a-(1 + 3) Gal P-0-methyl to make the a lengthening of the observed T I .Measurements of blood group A trisaccharide. (See Figures 3 and 4.) 1,3 C NOE and TI as a function of static magnetic A second example th a t is even more clear is that of the addition of Fuc a-(1 + 2) to the type 1 core field strength can be used to estimate a true rCand a n order parameter that represents the extent to disaccharide to make the type 1 H blood group which rapid internal motion contributes to the structure. The disaccharide is very flexible (Figure spectral density. For sucrose their data indicate 8) but the same linkage in the type 1 H trisacchat ha t only about 15% of the spectral density is ride is extremely rigid, and neither of the two glycosidic linkages shows any conformational tranrepresented by rapid internal motions, a result sitions (Figure 9). consistent with the decay of the autocorrelation function observed in our simulations of oligosacWhile it would be desirable to test various ascharides whose trajectories do not show conformapects of these simulations, existing experimental tional transitions in the glycosidic dihedral angles. data are not completely adequate. Simulations of the 'H-NOE data for the blood group A trisacchaT h e histories of C4 -C5 -C6 -0 6 dihedral ride in a single conformation essentially the same angles of most residues show three orientations as the major conformation of Figure 5 agree well (i60" and 180"), while only two of them (-60" with experimental NOE data for model oligosacand 180") are seen for Glcp in Gal(0-1 + 4)Glc charides in water solution. While NOE data for P-0-methyl. Although many conformational transitions are observed for the exocyclic dihedral ansamples in DMSO solution also fit this single congles of the hydroxyl groups in these simulations, formation well, NOE data for pyridine solution the significance of this observation in the absence give a better fit for the minor conformation of of water may be minimal. One would expect that Figure 5.3 In fact, these two conformations, which there should be strong hydrogen bonding to the differ mainly in the value of #I.'uc, are not very hydroxyl groups, which must modify their motion. different. 13C T I data a t a single field strength for a We suspect t hat inclusion of water molecules in the blood group A tetrasaccharide in aqueous solution simulation would be necessary for accurate predicshow the TI for all the methine carbon atoms are tion of the '"C TI of the C6 resonances. similar, suggesting that the oligosaccharide moves T he histories of the glycosidic dihedral angles as a single unit with single rotational correlation trajectories show that there are transitions among time T ~ . Although , ~ ~ these experiments are consisdistinct conformations for some but not all linktent with a single, essentially rigid conformation, ages. T he histories of dihedral angles # and show they do not rule out other interpretations. Perhaps the most interesting test case for these t ha t the dynamical conformations are essentially simulations arises in the contrasting results for the among the local minima shown on potential energy surfaces computed by van der Waals interactions type 1 core disaccharide, Gal( p-1 + 3)GlcNAc 0only, suggesting the van der Waals term is predom0-methyl, and the type 1 H structure, Fuc a-(1 + inant in the potential energy function. The energy 2) Gal(0-l + 3)GlcNAc 0-0-methyl. The observavalues of the structures along the trajectories have tion of many conformational transitions for the been minimized to confirm th at the molecule is former and of very few for the latter suggests a passing through distinct conformational energy marked difference in flexibility. These two fragminima about which fluctuations are taking place. ments occur a t the nonreducing terminal of the Consistent with this interpretation is the observahuman milk oligosaccharides lacto N-tetraose tion of a correlation between the dynamic behavior (LNT) and lacto N-fucopentaose 1 (LNF-l), which of the angles C$ and #. Some of the glycosidic have been previously studied by nmr methods2 linkages of the blood group oligosaccharides show While a rigorous analysis of the NOE data for numerous transitions among distinct minimum enLNF-1 is impossible, it may be reasonably conergy conformations, as was observed by Ha et a1.23 cluded th a t a conformation with dihedral anfor maltose. But other oligosaccharides th a t have = 140 f lo", C$Fuc = -80 f lo", $Gal = gles

+

810

YAN AND BUSH

- 110 f. 10" and = -70 10" is most likely t he one existing in water for the type 1 blood group H This interpretation of the experimental results is consistent with the data of Figure 9. Since signal overlap prevents a rigorous analysis of the NOE data for LNT, no fit of the NOE data t o a model conformation was attempted. "C T , data for both LNT and LNF-1, which have been reported a t a single frequency (20 MHz), showed that I3C Tl for all the methine carbon atoms were similar.37 This result was interpreted as a n indication that both molecules rotate a s a unit with a single rotational correlation time and negligible internal motion. A more sensitive test for internal motion on the TItime scale (hundreds of picoseconds) is offered by measurements of TI and NOE as a function of field strength such as those reported by McCain and Markley for sucrose with the model-independent analysis of Lipari and S z a b ~We . ~ anticipate ~ th at it should be possible t o detect the 30 ps internal motions in the Gal(/3-1 + 3)GlcNAc ,&linkage in LNT and that they would be absent in LNF-1. Verification of this prediction would provide a valuable experimental and theoretical method for detection and classification of internal motion in complex oligosaccharides. T he results of our simulations of oligosaccharide dynamics differ in a basic way from those of small peptides by Hagler et al.19 The dihedral angles of t he peptide linkages show much more motion on a time scale of 10 ps than do those of the oligosaccharides. Some oligosaccharides show conformational transitions but some show few or none a t all suggesting t hat experimental approaches t o detect this motion might be fruitful. Internal motion in oligosaccharides could have some biological consequences. T he presence of conformational isomerism in complex oligosaccharides has been proposed as a mechanism for control in glycoprotein biosynthesis and may be responsible for other biological activities.'

'J

',

This research was supported by NIH grant GM 31449.

REFERENCES 1. Shoemaker, K. R., Kim, P. S., York, E. J., Stewart, J. M. & Baldwin, R. L. (1987) Nature 326, 563-567. 2. Rao, B. N. N., Dua, V. K. & Bush, C. A. (1985) Biopolymers 24, 2207-2229. 3. Yan, Z.-Y., Rao, €3. N. N. & Bush, C. A. (1987) J . Am. Chem. Soc. 109, 7663-7669.

4. Homans, S. W., Dwek, R. A. & Rademacher, T. W. (1987) Biochemistry 26, 6571-6578. 5 . Lemieux, R. U., Bock, K., Delbaere, L. T.