| Glycobiology | Pages |
New conformational constraints in isotopically (13C) enriched oligosaccharides
Introduction
Results and discussion
Materials and methods
Acknowledgments
Abbreviations
References
New conformational constraints in isotopically (13C) enriched oligosaccharides
Multidimensional heteronuclear NMR studies have been applied to the resonance assignment and conformational analysis of 13C-enriched Neu5Ac[alpha]2-3Gal[beta]1-4Glc. It is demonstrated that three-dimensional ROESY-HSQC experiments provide through-space distance restraints which cannot be observed with conventional homonuclear 1H techniques due to resonance overlap. In particular, connectivities demonstrating the existence of the 'anti" conformation about the Gal[beta]1-4Glc glycosidic linkage are unambiguously observed. It is shown that 13C isotopic enrichment of the trisaccharide at a level >95% enables straightforward measurement of trans-glycosidic 1H-13C and 13C-13C coupling constants and a Karplus-type relation is derived for the latter. In total 15 conformational restraints were obtained for the trisaccharide in aqueous solution, all of which were in excellent agreement with theoretical parameters computed from a 5 ns molecular dynamics simulation of the glycan. Key words: NMR/oligosaccharide/structure/resonance/ conformational analysis
Introduction
Structural studies on oligosaccharides using high-resolution 1H NMR are plagued by the very severe overlap of the majority of proton resonances (Vliegenthart et al., 1983). Apart from rendering complete resonance assignments difficult if not impossible, there remains the problem of accurate quantitation of crosspeak intensities in experiments such as NOESY or ROESY for use as structural restraints. Furthermore, the number of such restraints is usually small, thus frustrating efforts to obtain a detailed picture of oligosaccharide structure and dynamics.
By analogy with structural studies on proteins (Marion et al., 1989), the resonance overlap problem can in principle be overcome by application of heteronuclear three-dimensional NMR methods. These offer improved spectral dispersion in comparison with their homonuclear counterparts in view of the fact that one or more dimensions in the 3D spectrum can represent the chemical shift of the heteronucleus (Xu et al., 1996). Here, we show that 13C enrichment at a high level (>95%) offers a high-sensitivity approach to the resonance assignment and conformational analysis of oligosaccharides, revealing additional conformational restraints that are not observed in conventional homonuclear NMR spectra. In addition we demonstrate that isotopic enrichment offers significant advantages for the measurement of trans-glycosidic scalar coupling constants. We illustrate these approaches in application to 13C enriched Neu5Ac[alpha]2-3Gal[beta]1-4Glc.
Results and discussion
Chemoenzymatic synthesis of Neu5Ac[alpha]2-3Gal[beta]1-4Glc
13C enriched Neu5Ac[alpha]2-3Gal[beta]1-4Glc was prepared by enzymatic sialylation of 13C -enriched Gal[beta]1-4Glc acceptor using Trypanosoma cruzi trans-sialidase with p-nitrophenyl-[U-13C- Neu]5Ac as donor. Full details of this chemoenzymatic strategy will be presented elsewhere (Probert et al.,1997).
Resonance assignments in Neu5Ac[alpha]2-3Gal[beta]1-4Glc
Complete 1H and 13C resonance assignments in the trisaccharide have been determined by Lerner and Bax (1987), by application of three heteronuclear two-dimensional NMR techniques at natural abundance 13C. As anticipated, in 13C enriched form the trisaccharide can be assigned fully using a single two-dimensional (2D) HCCH-COSY experiment (Bax et al., 1990; Yu et al., 1993), (data not shown). These data confirm the assignments proposed by Lerner and Bax (1987).
Solution structure of Neu5Ac[alpha]2-3Gal[beta]1-4Glc
The solution structure of Neu5Ac[alpha]2-3Gal[beta]1-4Glc was initially investigated by use of 1H-1H rotating frame Overhauser effect (ROESY) measurements in the conventional manner (Bothner-by et al., 1984; Bax and Davis, 1985). Although Neu5Ac[alpha]2-3Gal[beta]1-4Glc is only a trisaccharide, it has a very complex 1H NMR spectrum, with most resonances concentrated within a 0.4 ppm shift range. Despite the availability of complete 1H resonance assignments (Lerner and Bax, 1987), the interpretation of the conventional two-dimensional 1H-1H ROESY spectrum is not straightforward. To illustrate this point, Figure 1a shows a 1D slice parallel to F2 at the frequency of the Gal H-1 resonance (4.56 ppm) from the 2D ROESY spectrum of unlabelled Neu5Ac[alpha]2-3Gal[beta]1-4Glc. A few peaks can be identified unambiguously, but severe overlap prevents both unambiguous identification and quantitation of certain resonances. By contrast, the observed crosspeaks were readily separated by their 13C chemical shift in a two-dimensional F2/F3 (13C/1H) slice from the 3D-ROESY-HSQC spectrum at the F1 (1H) resonance frequency of Gal H-1 (Figure 1b), and all ROE connectivities from Gal H-1 can be assigned unambiguously. In particular, the connectivity between Gal H-1 and Neu5Ac H-8 proposed in earlier work (Siebert et al., 1992), and found to be of similar intensity to the connectivity between Gal H-1 and Glc[alpha] H-6s, is absent in the present study. The reason for this discrepancy is unclear. The ROE between Gal H-3 and Neu5Ac H-8 is observed however (not shown), which is consistent with the orientation of the glycerol sidechain of Neu5Ac ([thetas]1 = -60°, [thetas]2 = 180°) suggested in previous work (Poppe and van Halbeek, 1991a; Siebert et al., 1992), and is similarly supported by 1H-1H coupling constants in the present study (data not shown). An intriguing aspect of the data shown in Figure 1b is the presence of weak crosspeaks from Gal H-1 to Glc[alpha]/[beta] H-3 and H-5. These connectivities were not observable in the two-dimensional 1H-1H ROESY spectrum due to essentially complete overlap with other resonances, and indeed to our knowledge have not been observed previously for this oligosaccharide in aqueous solution. In this regard, caution must be exercised in interpreting weak crosspeaks in ROE spectra, since these can arise from a strong ROE connectivity (in this case Gal H-1 to Glc H-4) followed by relayed coherence transfer (to Glc H-3 and Glc H-5) to other coupled protons via a HOHAHA mechanism (Neuhaus and Keeler, 1986). However, the intensities of these crosspeaks were found to be independent of the position of the spin-lock carrier, confirming that they correspond to direct connectivities.
Figure
In total seven trans-glycosidic ROEs were observed (Table I">), and these were utilized in restrained dynamical simulated annealing calculations, using as input ten pseudo-random geometries generated by a dynamical quenching procedure (Homans and Forster, 1992). This resulted in three families of structures termed 'A," 'B," and 'C" (Figure 2a) that differ principally in the conformation about the Neu5Ac[alpha]2-3Gal glycosidic linkage; 3 of the 10 starting structures annealed to conformer 'A" ([phiv],[psi] = -70°,+5°), five annealed to conformer 'B" ([phiv],[psi] = -165°,-20°), and one annealed to conformer 'C" ([phiv],[psi] = -95°,-45°). The final structure had large restraint violations and was discarded. Similar conformers about the Neu5Ac[alpha]2-3Gal linkage were predicted in the original study of Neu5Ac[alpha]2-3Gal[beta]1-4GlcNAc[beta]-Asn by Breg et al. (1989). In order to probe the extent of torsional fluctuation about each of these conformers, the lowest energy structure of each family was arbitrarily chosen as input for a 5 ns restrained molecular dynamics simulation in vacuo. Since all MD simulations gave similar results (data not shown), only the MD simulation with the global minimum energy structure as input is considered further. The instantaneous values of relevant torsion angles are shown in Figure 2b. In order to determine whether this dynamic behavioris a plausible model for the solution dynamics of Neu5Ac[alpha]2-3Gal[beta]1-4Glc, theoretical ROEs were back calculated from the molecular dynamics simulations using a full relaxation matrix approach (Forster, 1991) from which it is apparent that there is very good agreement between theoretical predictions and experimental values (Table I). In particular, the intensities of the relatively weak ROEs between Gal H-1 and Glc H-3 and Gal H-1 and Glc H-5 are reproduced in the theoretical simulations. These ROEs derive from the relatively small population of conformers that adopt the 'anti" conformation (Figure 3) about the Gal[beta]1-4Glc glycosidic linkage during the molecular dynamics simulation ([psi] = 180°), a result that lends strong support to indirect evidence that a small population of such conformations exists for various oligosaccharides in solution at physiological temperature (Lipkind et al., 1985, 1987; Poppe et al., 1990; Bock et al., 1994; Dabrowski et al., 1995).
Figure
Table I
It could be argued that the excellent agreement between experimental and theoretical ROEs in Table II would be anticipated from a molecular dynamics simulation which includes ROE restraints. However, as discussed at length previously (Rutherford and Homans, 1994), the computed ROE can in principle vary over a very wide range as a result of the choice of restraint distance bounds. Moreover, the availability of 13C enriched material enables the measurement of trans-glycosidic coupling constants with which to validate further the predicted motional behavior.
ROE connectivity
Restraint
Experimental ROE
intensity (%)aTheoretical ROE
intensity (%)b
Neu5Ac H-3ax - Gal H-3
W
1.0
1.5
Neu5Ac H-8 - Gal H-3
W
0.4
0.6
Gal H-1 - Glc H-4
S
4.2
3.9
Gal H-1 - Glc H-6
W
0.4
0.6
Gal H-1 - Glc H-6[prime]
W
0.5
0.7
Gal H-1 - Glc H-3
W
0.4
0.4
Gal H-1 - Glc H-5
W
0.5
0.6
Long-range 13C-13C coupling constants
Conventionally, trans-glycosidic 1H-13C coupling constants have been utilized for the conformational analysis of oligosaccharides (Poppe and van Halbeek 1991b,c; Rutherford et al., 1993, 1994). In the present study carbon-carbon coupling constants are particularly valuable since only one 13C-1H coupling is available across the Neu5Ac[alpha]2-3Gal glycosidic linkage. However, although experiments have been described for the measurement of long-range carbon-carbon couplings in 13C enriched material (Bax et al., 1992, 1994), and while 13C-13C coupling constants have been reported in carbohydrates (Wu and Serianni, 1992; Wu et al., 1992; Duker and Serianni, 1993; Church et al., 1996), a suitable parametrization of the Karplus relationship has not been reported for the C-C-O-C fragment. We have therefore undertaken such a parametrization, by use of model compounds in 13C enriched form containing C-C-O-C fragments with known, fixed dihedral angles.
Karplus parametrization
Parametrization of the Karplus relation ideally requires experimental 3JCC values for C-C-O-C fragments with known dihedral angles distributed over a wide range. On grounds of sensitivity such coupling constants must be determined in compounds with a high level of enrichment with 13C, and in this regard the information that can be obtained from readily available compounds is limited. For example, the only usable coupling constant for a C-C-O-C fragment in uniformly 13C-enriched glucose is that for the C1-O-C5-C6 fragment, where the dihedral angle is 180°. The coupling constant for the C1-O-C5-C4 fragment (+60°) cannot be utilized since it occurs simultaneously with the coupling via C1-C2-C3-C4, and the two contributions cannot readily be separated. However, dihedral angle values of +60°,-60°, and ~+120° in C-C-O-C fragments can be obtained in two simple sugar derivatives, namely, methyl 4,6-O-(1-methylbenzylidene) [alpha]-D-glucopyranoside and 2,3-O-isopropylidene 1,6-anhydro-D-mannopyranose (Table II). The 3JCC values listed in Table II were fitted to a Karplus relation of the form 3JCC = Acos2[thetas] + Bcos[thetas] + C, giving rise to the curve shown in Figure 4. It is appreciated that with the limited experimental data used in fitting and neglect of electronegativity effects, this parametrization is semiquantitative at best, and is probably unreliable in the region 0° < [thetas] < 60° (not shown). However, the experimental datapoints lie within the region of the curve that is likely to be populated by trans-glycosidic 3JCC values, and thus should predict these values with reasonable accuracy.
Figure
Table II
Three-bond scalar carbon-carbon couplings can be measured using 2D long-range carbon-carbon correlation spectroscopy (LRCC, Bax et al., 1992). However, Neu5Ac[alpha]2-3Gal[beta]1-4Glc suffers from poor spectral dispersion in critical regions of the 13C spectrum. For example, Figure 5 (left) shows an extract from the 2D LRCC spectrum of the trisaccharide, illustrating severe overlap between Gal C-3,C-5 and Glc[beta] C-3,C-5. The three bond coupling between Gal C-1 and Glc[beta] C-3/C-5 clearly cannot be determined from this spectrum. It is therefore necessary to resort to the three-dimensional analog of this experiment (Bax et al., 1994), where an F2/F3 plane (Figure 5, right) at an F1 frequency of ~75 ppm (Gal C-3, C-5, Glc[beta] C-3, C-5) allows 3JCC between Gal C-1 and Glc[beta] C-5 to be measured. In total, four long-range carbon-carbon coupling constants could be determined in the trisaccharide and two others could be inferred as being < 1 Hz, and these values could be compared with those predicted theoretically from the MD simulation of the trisaccharide using the above Karplus relation. It can be seen from Table III that the theoretical values are in very good agreement with those measured experimentally, despite the fact that no angular restraints were applied during the MD simulation. These data therefore support the validity of the MD simulations as a plausible model for the solution dynamics of the oligosaccharide.
Figure Table
Coupling
Experimental (Hz)a
Theoretical (Hz)
3JCC
Gal C-3 - Neu5Ac C-3
1.9
1.6
Gal C-2 - Neu5Ac C-2
1.8
1.8
Gal C-4 - Neu5Ac C-2
<1
1.1
Gal C-1 - Glc[beta] C-5
1.9
1.6
Gal C-2 - Glc[alpha]/[beta] C-4
2.6
2.6
Gal C-1 - Glc[alpha]/[beta] C-3
<1
1.1
3JHC
Gal H-1 - Glc C-4
3.5
3.7
Gal C-1 - Glc H-4
n.d.
4.9
Neu5Ac C-2 - Gal H-3
4.7
4.5
Long-range 1H-13C coupling constants
Trans-glycosidic 13C-1H coupling constants can conveniently be measured in 13C enriched oligosaccharides (Gitti et al., 1994; Lippens et al., 1996; Xu and Bush 1996), and in the present work we have utilized a heteronuclear constant-time COSY sequence (Ionides et al., 1995) and iterative fitting strategy described by Titman and Keeler (1990). Two trans-glycosidic 13C-1H coupling constants were obtained with this approach (data not shown), and were again compared with theoretical predictions based on the relevant Karplus relationship for coupling pathways of this type (Mulloy et al., 1989; Tvaroska et al., 1989). The 13C-1H coupling constant between Gal C-1 and Glc H-4 could unfortunately not be obtained due to resonance overlap and strong coupling. As with trans-glycosidic 13C-13C coupling constants, simulated 13C-1H coupling constants were in excellent agreement with these experimental values.
Conclusions
Isotopic enrichment of oligosaccharides at high levels (>95%) 13C offers considerable advantages for structural and conformational analysis. A major advantage is the additional spectral dispersion offered in heteronuclear experiments by virtue of an additional 13C dimension. Of course, heteronuclear experiments are possible at natural abundance provided sufficient material is available, but it is unlikely that sensitivity would be sufficient to observe some of the weaker connectivities detected in the present study. For example, weak ROE connectivities are observed which prove the existence of an 'anti" conformation about the Gal[beta]1-4Glc linkage, which to our knowledge have not been observed previously for this disaccharide in aqueous solution but have, however, been observed very recently in GlcNAc[beta]1-4GlcNAc (Espinosa et al., 1996). Moreover, trans-glycosidic 13C-1H and 13C-13C coupling constants, which are measurable with difficulty or not at all at natural abundance, have been shown to provide a valuable semiquantitative assessment of the quality of dynamic models derived from restrained molecular dynamics simulations. Given that 13C-enriched oligosaccharides can now be synthesized at high yield in milligram amounts using chemoenzymatic strategies, we anticipate that further developments in the structural and conformational analysis of oligosaccharides will be apparent in the future.
Materials and methods
Sample preparation
N-Acetyl [13C] Neu5Ac[alpha]2-3Gal[beta]1-4Glc. NH4+ (~7.5 mg) was dissolved and lyophilized into 99.96% D2O three times followed by dissolution into 700 µl D2O.
NMR experiments
NMR spectra were obtained at 300K with a 1H reference frequency of 500 MHz on a Varian Unity+ spectrometer equipped with a self shielded z gradient triple resonance probe. All spectra were recorded in the phase-sensitive mode with use of the States (States et al., 1982) method for quadrature detection.
2D HCCH-COSY experiments were recorded using the three-dimensional (3D) pulse scheme of Bax et al. (1990). A total of 1K complex and 2K complex points were acquired in the t1 and t2 dimensions, respectively, with spectral widths of 2 kHz and 10 kHz. Two- (Bax et al., 1992) and three-dimensional (Bax et al., 1994) gradient-enhanced long range carbon-carbon correlation (LRCC) experiments were recorded with a proton sweep width of 2 kHz, consisting of 1024 complex points and a 13C sweep-width of 11 kHz in t1 (128 complex points) and a 6.5 kHz (32 complex points) in t2 for the 3D experiment. A total of 64 and 8 scans were acquired per increment for the 2D and 3D experiments, respectively. 13C-13C couplings evolve and are refocused during the delay (2T) of 22.2 ms. Total acquisition times for the 2D and 3D experiments were ~12 h and ~48 h, respectively. The values of the long range coupling constants are derived from the ratios of cross-peaks obtained in the spectrum in the manner described by Bax et al. (1992). Long-range 13C-13C couplings in the compounds methyl 4,6-O-(1-methylbenzylidene) [alpha]-d-glucopyranoside and 2,3-O-isopropylidene 1,6-anhydro-d-mannopyranose were obtained by conventional homonuclear 13C spin-decoupling experiments.
Trans-glycosidic carbon-proton coupling constants (CT-LRCH) were measured using the 2D constant-time pulse sequence of Ionides et al., (1995). Proton and carbon sweep-widths were identical to those used in the 2D LRCC spectrum with 4096 complex points in t2 and 256 t1 increments consisting of 16 scans per increment.
Three-dimensional 13C-edited ROESY (Bothner-by et al., 1984; Bax and Davis, 1985) experiments were acquired using a conventional ROESY-HSQC pulse sequence with offset compensation (Griesinger and Ernst, 1987) and with spectral widths of 2 kHz, 6.5 kHz, and 2 kHz and 128, 32, and 1024 complex points in t1, t2, and t3, respectively. The proton transmitter offset was placed 500 Hz downfield of the lowest field resonance during the spin-lock period to minimize coherence transfer effects. To optimize digital resolution the Neu5Ac C5 and C3 resonances (~54 ppm and ~40 ppm, respectively) were folded-in once. The effective field for spin-locking was 2 kHz and was applied for a mixing time of 250 ms. Prior to Fourier transformation, data were apodized with cosine-bell functions, and the t1, and t2 dimensions were zero filled to 256 and 64 complex points, respectively. Total acquisition time for the 3D spectrum was ~55 h.
Molecular modeling
Dynamical simulated annealing calculations and restrained molecular dynamics simulations were computed in vacuo using described procedures (Homans and Forster, 1992; Rutherford and Homans, 1994). Theoretical ROE intensities were computed from molecular dynamics simulations using the in-house written software package MDNOE. This package incorporates a heteronuclear full relaxation matrix approach including a formalism appropriate for the computation of NOE and ROE data due to fluctuating internuclear distances arising from internal motions which are fast with respect to the rate of molecular tumbling (Tropp, 1980; Homans and Forster, 1982). A single overall isotropic correlation time for the molecule was assumed, and was obtained by fitting the ratio of the theoretical diagonal peak to crosspeak (intraresidue ROE) intensities of the Neu5Ac H-3 axial and H-3 equatorial protons to the experimentally measured values. The experimental ROE intensities (i.e., crosspeak volumes) were then normalized to the theoretical ROE intensities to generate the absolute magnitudes of the experimental ROEs. The torsion angles [phiv] and [psi] are defined as H-1 - C-1 - O-1 - C-X (C-1 - C-2 - O-2 - C-X in the case of Neu5Ac) and C-1 - O-1 - C-X - H-X, respectively, where C-X and H-X refer to the aglyconic atoms. The torsion angles [thetas]1 and [thetas]2 for the glycerol sidechain of Neu5Ac are defined as H-6 - C-6 - C-7 - H-7 and H-7 - C-7 - C-8 - H-8, respectively.
Acknowledgments
This work was supported by Wellcome Trust Grant Ref. 041973/Z/94/Z/JRS/RF to S.W.H. and 042472/Z/94/Z to R.A.F. We thank Dr. P. Lightfoot for performing the crystallographic analysis of 2,3-O-isopropylidene 1,6-anhydro-d-mannopyranose.
Abbreviations
COSY, correlated spectroscopy; HCCH-COSY, 1H-13C-13C-1H correlation via 1Jcc couplings; ROESY-HSQC, three-dimensional rotating-frame Overhauser effect heteronuclear single-quantum correlation spectroscopy; NOESY, nuclear Overhauser effect spectroscopy; ROE, rotating frame Overhauser effect; ROESY, rotating frame Overhauser effect spectroscopy; 2D, two-dimensional; 3D, three-dimensional.
References
This page is run by Oxford University Press, Great Clarendon Street, Oxford OX2 6DP, as part of the OUP Journals
Comments and feedback: www-admin{at}oup.co.uk
Last modification: 27 Jan 1998
Copyright© Oxford University Press, 1998.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  M. L DeMarco and R. J Woods Atomic-resolution conformational analysis of the GM3 ganglioside in a lipid bilayer and its implications for ganglioside-protein recognition at membrane surfaces Glycobiology, April 1, 2009; 19(4): 344 - 355. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||