H igh R esolution N M R in the D eterm ination of S tructure in C … · 2010-04-16 ·...

Vol. 10, No. 3/4 73

High Resolution NMR in the Determination ofStructure in Complex Carbohydrates

C. Allen Bush

Department of Chemistry,Illinois Institute of Technology

Chicago, IL 60616 U.S.A.

ContentsI. Introduction: A Description of Complex Carbohydrates and their Biological Function 73

II. The Determination of Covalent Structure

III. The "Structural Reporter" Method in JH NMR

IV. Structure Determination by Complete Assignment of the *H NMR Spectrum

V. Prospects for Carbon NMR

VI. NMR Spectroscopy of Other Nuclei: 31P, 17O, 15N

VII. High Resolution NMR of Intact Glycoproteins

VIII. References

75

76

79

88

91

93

94

I. Introduction: A Descrip-tion of Complex Carbohy-drates and their BiologicalFunction

For the purposes of this review, we will definecomplex carbohydrates as polymers composed ofsuch monosaccharides as mannose, galactose, N-Acetyl glucosamine, N-Acetyl galactosamine, rham-nose, fucose and sialic acid connected in highly var-ied linkages and anomeric configurations. Thesemacromolecules, which contain glucose only rarelyand are not used as energy sources by an organism,are found in glycoproteins, glycolipids and in theextracellular polysaccharides of bacteria. Proteo-glycans containing such polysaccharides as heparin,chondroitin and dermatan covalently linked to pro-tein could also be considered as glycoproteins but

the glycosaminoglycans, including hyaluronic acidwill not be discussed in detail in this review. Thisexclusion along with that of starch, glycogen, chitinand plant cell wall polysaccharides does not resultfrom any lack of their biological importance butrather from the necessity to place some reasonableboundaries on this enormously varied subject.

The complex carbohydrate structures in glyco-proteins, glycolipids and bacterial polysaccharideshave generally distinct features and each class canbe separately defined. On the other hand, they havemany similarities and there is considerable biologicalevidence, such as immunological or lectin receptoractivity which interrelates them. Typical oligosac-charide structures may have from two to 15 sugarresidues in a non-repeating sequence. While therepetition of a common structural subunit is usu-ally associated with the lipopolysaccharides and the

74 Bulletin of Magnetic Resonance

capsular polysaccharides of bacteria, this repeatingstructure motive has also been found in glycopro-teins (cf. polylactosamines) as well as in lipids suchas polyglycosyl ceramides.

These complex carbohydrates have highly variedstructures and they occur in rather small propor-tions in the cells of higher organisms. It is thoughtthat they do not have a storage function in energymetabolism as do the glucose polymers, starch andglycogen. The complex structures, whose functionappears to be in control of cell metabolism, showremarkable differences among closely related animalspecies. These differences stand in contrast to suchproteins as hemoglobins and cytochromes, extensivestudies of whose amino acid sequences have revealedsurprising homologies among distantly related or-ganisms. The carbohydrate structures of glycopro-teins show substantial differences among closely re-lated species. For example, studies of the carbohy-drate chains of the red blood cell membrane glyco-protein, glycophorin, in horse, sheep, cattle, mouseand man show quite different structures (1). In factsubstantial genetically determined polymorphism inthe carbohydrate structures of glycoproteins withina given species is well recognized, perhaps the bestknown example being case of blood group activitywhich is determined by specific complex oligosac-charide antigens.

Although the biosynthesis of complex carbohy-drates is not understood in complete detail, it isclear that higher plants and animals expend sub-stantial metabolic energy in both the anabolism andcatabolism of these molecules. The biosynthesis ofthe carbohydrate chain of asparagine N-linked gly-coproteins, which has been intensively studied in re-cent years has been found to be quite complicatedwith buildup of a tetradecasaccharide chain on alipid donor followed by transfer to sites on the pro-tein. This transfer is followed rapidly by cleavage ofthe tetradecasaccharide by a series of specific gly-cosidases and, in some glycoproteins, by subsequentresynthesis of a complex antenna or hybrid carbohy-drate chain (2). A biosynthetic scheme of this com-plexity suggests that there must be some rather im-portant function for the complex asparagine-linkedoligosaccharides of the glycoproteins of eucaryotes.

Many studies have focused on the question of thefunction of complex carbohydrates yielding a diversearray of apparently unrelated functions. Specific

oligosaccharide sequences in heparin are active inangiogenesis while other sequences are active in con-trol of blood clotting. Complex carbohydrates havebeen shown to possess various cell surface receptorfunctions which include the activity of gangliosideGM1 as the receptor for cholera toxin. The non-reducing terminal galactose residue in serum glyco-proteins controls the clearance of glycoproteins fromcirculation by receptor mediated endocytosis (3).Asparagine N-linked glycosylation of newly synthe-sized glycoproteins in the golgi apparatus is knownto control intracellular migration to lysozomes (4).Antifreeze glycoprotein, a major serum glycoproteinof arctic fish results in the non-colligative lowering ofthe freezing point of the fish blood (5). The resultsfrom a new technique of blotting bacteria onto thinlayer chromatograms of glycolipids isolated from themucosal membranes implies that these glycolipidsmay serve as natural receptors for bacteria and thusplay a role in their pathogenieity (6). Blood groupantigens and a number of closely related tumor anddifferentiation antigens are known to be complexcarbohydrate structures.

While it is possible to list numerous functions forcomplex carbohydrates, what is lacking in the cat-alog given above is some unifying hypothesis con-cerning function. Apparently we have overlookedsome underlying principle which might contributesubstantially to our understanding of such impor-tant biological problems as growth, differentiation,immunology, cancer and the organization of mul-ticellular organisms in general. Perhaps the con-nection between cellular organization and the func-tion of complex carbohydrates in higher animals andplants could be illuminated by an analogy to foot-ball uniforms. The brightly colored and distinctiveuniform is detrimental to the function of an individ-ual player; it would obviously be easier for a play-erto take the ball to the goal without it. But theuniforms are essential to the function of the footballteam and the -complex rules governing the game asa whole. Therefore hypotheses which assign to thecarbohydrate of a glycoprotein a function in stabiliz-ing a protein against denaturation or in protectingthe cell surface against proteolysis might be com-pared with the hypothesis that the function of afootball uniform with colorful insignia and numeralsis just to keep the player warm. Biochemists havebeen taught to concentrate our attention on func-

Vol. 10, No. 3/4 75

tion at the cellular level and our failure to uncoverthe underlying principle which will unify our un-derstanding the function of complex carbohydratesresults from the difficulty in focusing on the organi-zation of the whole organism.

In the absence of any clear synthetic idea aboutthe biological function of complex carbohydrates,one might learn from history that the study of struc-ture and conformation might be the source of valu-able evidence concerning function. Thus there hasbeen an increase in research activity in the area witha resulting growth in the precision of our knowledgeabout the structure and to a lesser extent about thethree dimensional conformation of complex carbo-hydrates. High resolution NMR has made majorcontributions to this recent progress.

II. The Determination of Cova-lent Structure

The problem of structure determination in com-plex carbohydrates differs substantially from thatin either proteins or nucleic acids. Although thestructural subunit is generally limited to at most 15residues and may be composed of no more than fiveor six different monosaccharides, the residues are in-terconnected by many different glycosidic linkages,eg. (1-+2), (1-+3), (1-+4), (1-+6) etc., with vari-able anomeric configuration. Since there are multi-ple substitution points, branching is not only pos-sible but is in fact quite common. Determinationof the structure from genetic information (DNA se-quence) is impossible since the structure of a com-plex carbohydrate, unlike that of a protein or nucleicacid, is not under the control of a single gene; it iscontrolled by the interaction of a collection of inter-acting gene products (glycosyl transferases) whoseactivities depend not only the rate of their synthesisbut also on other environmental factors.

The problem of structure determination of com-plex oligosaccharides has been attacked by manytechniques in past years and the present state ofthe art continues to be confused by the lack of anysingle methodology which can be called upon for anew structural problem. Unlike proteins and nucleicacids for which well formulated strategies are avail-able for sequencing, a plethora of competing tech-niques exist for complex oligosaccharides, none ofwhich is truly adequate. Among the classical meth-

ods, methylation analysis involves tedious chemistryof limited reliability and its sensitivity (10 to 100nanomoles) is unimpressive by modern biochemi-cal standards. Furthermore, methylation analysisyields information only on positions of substitutionand not the anomeric configuration or sequence ofan oligosaccharide. Periodate oxidation of vicinaldiols is still widely used in spite of difficulties inthe interpretation of the data. Although enzymaticdegradation would appear to be a valuable method,the battery of known exoglycosidases is quite smalland the routine use of endoglycosidases, the "restric-tion enzymes" of complex carbohydrate chemistry,has been possible only in very recent years. Theavailability of endoglycosidases whose specificity iswell documented remains very limited. Althougholigosaccharide mass spectrometry, especially FABmass spec, seems very attractive on account of itsspeed and sensitivity, no method now known cangive more than limited partial structural informa-tion.

Of all the modern structural methods for com-plex carbohydrates, high field proton NMR yieldsthe most complete and detailed structural informa-tion. It is the only method which can, in princi-ple, give an ab initio structure without resort to anyother method. In practice, a complete structure de-termination by NMR is rarely the best approach toa completely new complex carbohydrate structureand generally other methods are used in conjunc-tion with NMR. The major weakness of NMR spec-troscopy as a method for structure determinationis its poor sensitivity (25 nanomoles to 10 micro-moles). On the other hand, since the experimentis non-destructive, it should always be consideredfirst after the isolation of a suitable sample. A sim-ple proton NMR spectrum is a modest experimentaleffort and will give immediate information on thepurity of the sample and perhaps some general in-formation on the structure. In the most favorablecases, the structure can be completely determinedby this simple experiment and in the-worst case onlytime is lost since the sample can generally be re-covered for subsequent analysis by methylation, en-zymatic degradation, periodate oxidation, or massspectrometry all of which are destructive. Perhapsthe greatest utility of the NMR method will be inproviding accurately known structures for calibra-tion of other more sensitive methods such as mono-


clonal antibody recognition, enzymatic degradation,mass spectrometry or some as yet undiscovered tech-nique.

III. The "Structural Reporter"Method in XH NMR

The proton NMR spectrum of a typical complexoligosaccharide in D2O solution shows some iso-lated resonances which have been called "struc-tural reporter resonances" in the pioneering workof Vliegenthart and coworkers who have assembleda large body of proton spectra relevant to glycopro-tein structure determination (7). As an illustrationof the method, the spectrum of the hexasaccharide,LND-1 from human milk (structure 1) is shown inFigure 1.

fucgal j3-(l—>3) glcNAc /?-(l—>3)gal j8-(l—»4)glc

fuc a-(l—>4

The resonances of the anomeric protons, thoseof methyl groups and of those of various protonswith distinctive chemical shifts have been correlatedwith known structures to yield a powerful methodfor use in conjunction with classical structural meth-ods which were outlined above. Since the method ofVliegenthart and coworkers is so simple and rapid,it should always be considered as the first step inthe identification and structure determination of anunknown complex carbphydrate. A water solublesample, such as a glycopeptide, oligosaccharide oroligosaccharide alditol, (100 nanomoles to 10 micro-moles) is exchanged with D2O and freeze dried inthe 5 mm NMR tube. High purity D2O is addedalong with a micromole of acetone as internal ref-erence and a simple 1-Dimensional proton NMRspectrum is recorded at a field of 300 to 500 MHz.Chemical shifts are reported in ppm below DSS us-ing 2.225 ppm for acetone as an internal reference(see Figure 1). While this chemical shift conventionis not completely accurate, it has proved to be veryreproducible among different laboratories facilitat-ing exact comparison of chemical shifts. It is oftenpossible to assign the structure of an oligosaccharideittfectly from the spectrum simply by comparison

to published data on similar or identical oligosac-charides. Milk oligosaccharides similar to LND-1have been studied by Dua and Bush (8) and by Duaet al. (9). In a typical spectrum (Figure 1) mostof the resonances crowded into the 3.5 to 4.0 ppmregion are those of methine protons with very sim-ilar chemical shifts. The "structural reporter reso-nances" in Figure 1 include those of the two over-lapping methyl groups of the 6-deoxy sugars, fucose,in the 1.27 ppm region. Resonances assigned to theamide methyl groups of acetamido sugars are near2.0 ppm and the resonance of H2 of /?-glucose isat 3.28 ppm. The chemical shifts of some of theseresonances may be distinctive of the position of theglycosidic linkage. Most of the structural reporterresonances appear in the low field region between 5.3and 4.0 ppm as shown in the expanded plot of Figure1 b. The resonances at lowest field are those of a-anomeric protons exemplified here by the two fucoseanomeric proton signals at 5.15 and 5.02 ppm. Theresonance of reducing terminal glucose a-anomer isat 5.23 ppm. The H5 resonance of fucose, which isextremely sensitive to linkage, is found at 4.88 ppmfor the 4-linked residue and at 4.35 ppm for thatwhich is 2-linked to galactose. The /?-anomeric res-onances are the 7 Hz doublets between 4.42 and 4.66ppm. A characteristic sharp resonance at 4.14 ppmis equatorial H4 of gal which is substituted at C3by glcNAc in the polylactosamine type of structure.This resonance overlaps with that of glcNAc H3 inthis compound. Although the chemical shifts of afew of the protons are slightly temperature depen-dent, the effect is sufficiently small that the temper-ature can be adjusted to move the resonance of theresidual HDO line so that it does not obscure anyresonances. Since a large sample was used in record-ing the spectrum of Figure 1, the HDO resonanaceat 4.8 ppm does not interfere with signals from thesample but residual HDO can pose a major problemwhen only a very small quantity of oligosaccharidesample is available for study.

In a pyranoside, it is found that the six-membered ring generally forms a chair of fixed con-formation providing a classification of protons as ax-ial or equatorial. Therefore the coupling patternsare characteristic of the stereochemistry of the car-bohydrate. For example, if H2 is axial, as it is forthe gluco and galacto stereochemistry, then a smallcoupling constant of 2-3 Hz is observed as a result

Vol. 10, No. 3/4 77

OP

3o

5.2 5 .0 4 .8 4 . 6 4 .4 4 . 2PPM

OQ

f u l l 1

o•8

3

•8

•§o

5 . 2 4 . 8 4 . 4 4 . 0 3 . 6 3 . 2 2 . 8 2 . 4 2 . 0 T . 6 1 . 2PPM

Figure 1. XH NMR spectrum of the human milk oligosaccharide, LND-I (structure 1) recorded at 300 MHzand 24 °C. la is full spectrum and lb is an expansion.

of the gauche conformation of HI and H2 follow-ing the Karplus relation. The chemical shifts andcoupling constants of the fucose a-anomeric proton

resonances in Figure 1 are typical. The fransdiaxialrelationship of HI and H2 in /?-anomers of sugarswith the gluco and galacto configuration leads to


larger coupling constants (7-9 Hz). Chemical shiftsbetween 4.4 and 4.8 ppm are typical of the anomericprotons of these ^-linked residues as shown in Fig-ure 1. Assignment of the anomeric configuration inthe case of mannose is made more difficult by theequatorial configuration of H2 for which the dihedralangle between HI and H2 is small for both a- and/?-anomers. Although interpretation of the chemi-cal shift and coupling constant of the structural re-porter resonances group is more subtle for mannoseresidues, Vliegenthart et al. (10) have shown thatdifferences in structural reporter resonances can beused for structure assignment if analogies can bedrawn to known structures and the interpretationsare well controlled.

The resonances of equatorial protons are oftenshifted downfield into the position of structural re-porter resonances and signals such as that of man-nose H2 and galactose H4 are often distinctive andprovide useful correlation with structure and link-age. Additional structural reporter resonances areobserved in deoxy sugars which have rigid geminallycoupled protons with distinctive chemical shifts aswell as a distinctive coupling pattern. For example,the sialic acids N-acetyl neuraminic acid (NANA)and N-glycolyl neuraminic acid (NGNA) are foundas 3-deoxy pyranosides and the characteristic chem-ical shifts of the resonances of the axial and equa-torial H-3 signals have been correlated with linkageby Vliegenthart and coworkers (7,10).

A similar approach has also been applied to in-terpretation of the *H NMR spectra of glycolipidsleading to valuable information on their carbohy-drate structures. Although they are not generallysoluble in water in a completely unaggregated form,Dabrowski and coworkers have shown that most gly-colipids give well resolved proton NMR spectra indeuterated dimethyl sulfoxide (DMSO) (11). In thecase of ionic glycolipids such as gangliosides, the ad-dition of 2% D2O improves solubility. The chemicalshifts are reported relative to internal tetramethylsilane (TMS) and they differ somewhat from thoseof similar carbohydrates in D2O (12). Although thestructural reporter approach is quite useful for cor-relation of the carbohydrate structure of glycolipidswith their 1H NMR spectra, chemical shift analogiesbetween spectra of glycolipids in DMSO and spectraof oligosaccharides and glycopeptides in D2O can-not be effectively drawn due not only to the differ-

ence in the chemical shift reference but also to otherperturbations in the chemical shifts (12). There isan extensive collection of reference spectra of glyco-lipids in DMSO solution to be found in the workof Dabrowski et al. (13) in gangliosides by Koernerand coworkers (14,15) and in blood group glycol-ipids by Hakomori and coworkers (16).

The method of structural reporter resonanceshas been applied not only to glycopeptides, oligosac-charides and glycolipids but also to bacterialpolysaccharides. Unfortunately the method isless effective for this system because of the muchgreater structural diversity of bacterial polysac-charides. Therefore an enormous library of spec-tra of reference structures is necessary for draw-ing the close chemical shift analogies required bythe structural reporter method. Some XH NMRdata have been reported for pneumococcal polysac-charides (17) and for E. coli polysaccharides (18)but substantially more 13C NMR data has been re-ported for high molecular weight polysaccharides.Gorin (19) has reviewed the general field and Jen-nings (20) has summarized the important work inbacterial polysaccharides. Expanded interest in the*H NMR spectroscopy of bacterial polysaccharideshas been stimulated by the recent demonstration ofcomplete proton assignments which greatly extendsthe capabilities of of proton NMR for the structuredetermination of bacterial polysaccharides (21,22).

The strength of structural reporter method is itssimplicity. Since only a simple 1-Dimensional pro-ton NMR spectrum is needed, it is the most sensi-tive method for very small samples. More sophis-ticated methods generally require more NMR ob-servation preventing acquisition of large numbers of

S ~- transients. The weakness of the method of struc-"**' tural reporters is that, in the absence of a sound

fundamental theory of proton chemical shift, spec-tra of identical or closely related carbohydrate struc-tures are required for reaching unambiguous con-clusions. Therefore it is the method of choice whendealing with members of a class which has been pre-viously studied in detail by proton NMR such asthe complex N-asparagine-linked glycopeptides orig-inally studied by Vliegenthart and coworkers (7,10).The high mannose N-linked glycopeptides have alsobeen extensively studied not only by Vliegenthartand coworkers (10) but also by Carver (23) and byAtkinson (24). Glycopeptides of the mucin type

Vol. 10, No. 3/4 79

having the galNAc-O- glycosidic linkage to serineor threonine are usually isolated by alkaline borohy-dride degradation which produces the oligosaccha-ride alditols which have also been extensively stud-ied by proton NMR (10,25). The oligosaccharides ofhuman milk, which are structurally related to gly-copeptides, have been studied by Dua et al. (8,9).

IV. Structure Determination byComplete Assignment of the*H NMR Spectrum

For interpretation of NMR spectra of oligosaccha-rides which are not identical to or closely relatedto known data, complete assignment of the methineresonances in the poorly resolved group of signals inthe 3.5 to 4.0 ppm region adds greatly to the struc-tural information over that provided by the methodof structural reporter resonances. It has been shownthat complete proton assignments in oligosaccha-rides in a size range up to 10 or 15 residues is infact generally practical with sufficient observationtime on 300-500 MHz NMR instruments (25,26).Composed mainly of linear chains of coupled spins,carbohydrates are especially suited to spin corre-lation methods such as difference decoupling or 2-dimensional spin correlation (COSY) and relatedtechniques for identification of all the protons of agiven sugar residue. The general approach is to as-sign an isolated structural reporter resonance, oftenan anomeric proton, then to correlate spins in a step-wise manner around the spin system of the ring. Forthe six-membered pyranoside ring, the stereochem-istry of the sugar can be identified from the values ofthe vicinal coupling constants. Although spin corre-lation can be done by 1-Dimensional difference de-coupling if only a few spin assignments are neededand sample quantity is very limited, difference de-coupling is a difficult experiment requiring carefulwork with a very stable spectrometer (9). In mostinstances, 2-dimensional methods are preferred dueto their more efficient use of the spectrometer re-sources for simultaneous determination of a largenumber of spin correlations.

Most of the 2-D NMR methods used in the struc-tural analysis of complex oligosaccharides difFer onlyin a few details from those used in the complete pro-ton assignments of proteins. Since the values of the

coupling constants are related to the stereochem-istry of the pyranoside ring, it is important to ob-serve the individual components of the multipletsin the crosspeaks for carbohydrates. Thus, specialattention should always be paid to the problem ofadequate digital resolution in 2-dimensional spectraof oligosaccharides. The difficulty of meeting thisrequirement is alleviated by the somewhat reducedchemical shift dispersion of carbohydrates comparedto proteins. Sugars have no aromatic protons andthe resonances of exchangeable amide protons arenot generally observed in D2O solution. The highestfield resonance is usually that of the methyl groupof a 6-deoxy hexose near 1.2 ppm and lowest fieldanomeric resonance in D2O is near 5.4 ppm. Therotational correlation times of most glycopeptides,oligosaccharides and glycolipids are such that 1 Hzsplittings are resolved. For polysaccharides incjud-ing those with high molecular weight, segmental mo-tion contributes to well resolved lines especially ifthe spectra are measured at elevated temperature.As in the case of proteins, the first reported spincorrelation experiments were magnitude COSY butthe advantages of phase sensitive experiments withpure absorptive lineshapes have been amply demon-strated in protein chemistry. Since most currentNMR spectrometers are capable of phased COSYwith a double quantum filter (DQF), application ofthis technique is becoming more common in complexoligosaccharides. Since in the magnitude COSY ex-periment, one calculates the absolute value ratherthan attempting to phase the spectrum, the lineshape is not pure absorptive and the dispersive con-tribution to the line shape causes excessive broad-ening and peak tailing. To improve this line shape,a sine bell apodization is generally used to form apseudo echo (27). The radical resolution enhance-ment is achieved at the price of suppressing muchof the original signal since sine apodization attenu-ates the early part of the FID as well as the end.If the data are encoded in such a way that thecorrect phases of a pure absorptive spectrum canbe calculated, then good lineshape can be coupledwith a data apodization which degrades the signalless severely. The two methods which are com-monly used for encoding the phases, time propor-tional phase increment (TPPI) (28) and the methodof States et al. (29) have been shown to be for-mally equivalent so the choice is dictated mainly by


instrumental effects (30). Although both methodsrequire acquisition of a raw data set twice as largeas that of a magnitude COSY, it has been shownthat no data are lost in phasing and the less radicalapodization leads to improved signal-to-noise ratios(30). Since there is no net magnetization transfer inCOSY, the multiplet components have alternatingsigns in phased spectra and inadequate digital res-olution can cause the overlapping multiplet compo-nents to cancel (31). The requirement for adequatedigital resolution in phased COSY spectra placesserious demands on the NMR computer for stor-age and processing of the large data matrices. Datasets of 2K x 2K points are in common use and theanticipated availability of more powerful comput-ers capable of processing larger matrices will be adecided advantage. Since it is impossible to storeand process such large data matrices on the prim-itive computers associated with NMR instrumentsmore than a few years old, many workers have be-gun to utilize a separate general purpose computerfor processing and storage after acquisition of theraw data using the computer on the NMR instru-ment. Although high speed data links capable oftransfer of 5 to 10 megabytes of data in a reasonabletime period are required, the advantage in computerspeed, data storage convenience and flexibility ofdata processing gained by using a generic multiusercomputer of the current generation is often worththis effort. Software for 2-dimensional NMR dataprocessing on such general purpose machines as theVAX and Sun computers is available from severalcommercial sources.

In Figure 2 is shown an expanded section ofthe phased COSY spectrum of the milk hexasac-charide LND-1, (structure 1) whose 1-D spectrumwas discussed above (Figure 1). Both negative andpositive contour levels are plotted and the result-ing multiplet shapes are useful for the assignmentof overlapping crosspeaks. In Figure 2, the cross-peaks between the a-anomeric resonances and theircorrelated H2 are in the rows between 5.0 and 5.3ppm. The well resolved 3 Hz gauche couplings whichappear as anti-phase multiplet components in Fig-ure 2, cannot be generally discerned in magnitudeCOSY spectra. The larger coupling constants ofthe diaxial H1-H2 of the /?-anomeric proton reso-

nate seen in the rows between 4.42 and 4.66the corresponding H2 resonances on the

columns between 3.2 and 3.85 ppm. Assignment ofthe two overlapping /?-anomeric resonances at 4.66ppm in Figure 1 is readily made in the COSY asthe resonances of H2 of /?-gal3 and /?-glc are wellseparated at 3.61 and 3.28 ppm respectively. Thewell resolved multiplet structure and the alternat-ing signs of the antiphase crosspeaks are especiallyuseful in resolving the crosspeak at row 4.14 ppmand column 3.72 ppm which contains superimposedcrosspeaks between gal4 H4-H3 and between glc-NAc H3-H4. The multiplet shape is characteristicfor pyranosides since the size of the coupling con-stants is determined by the stereochemistry (transor gauche) of the protons which are coupled.

Since, when the phase of the crosspeaks in aCOSY spectrum are chosen to be absorptive, the di-agonal peaks are dispersive, the diagonal can inter-fere with crosspeaks between resonances with simi-lar chemical shifts which are close to the diagonal.To better visualize these crosspeaks, a double quan-tum filter (DQF) is often incorporated in the pulsesequence as it was in the DQF-COSY spectrum ofFigure 2. This sequence preferentially attenuatesthe single quantum resonances of the diagonal withrespect to the crosspeaks. In order to better visu-alize peaks with small coupling constants (0.5 to 1Hz) such as those between galactose H4 and H5, adelayed COSY will emphasize the small couplingsat the expense of some signal loss during the delay.

In spectra such as those of Figure 1 and 2 inwhich many resonances are crowded into the 3.5 to4.0 ppm region, one might anticipate that strongcoupling could introduce some difficulties in thescheme outlined for the assignment of the spins ofan individual ring by COSY even with 500 MHzspectrometers. Strong coupling, which arises whentwo coupled protons have similar chemical shifts,leads both to some distortion of the expected mul-tiplet shape and to COSY cross peaks which lieclose to the diagonal. The former effect may in-terfere with determination of the sugar stereochem-istry and the latter interferes with tracing the chainof spins within the sugar residue. If multiplet dis-tortion is not too severe it can be accurately inter-preted by spin simulation but additional experimen-tal methods are needed to complete the tracing ofthe spin connectivity. In these cases spin relay ex-periments or isotropic mixing techniques have beenshown to be especially valuable in assignments in

Vol. 10, No. 3/4 81

Figure 2. Expanded portion of the phase sensitive double quantum filtered COSY spectrum (300 MHz) ofLND-I (structure 1). Positive and negative contours are indicated by lines of different thickness.

oligosaccharides (12,32a). The most useful methodfor relay of coherence along the chain of spins is theisotropic mixing experiment in which net magnetiza-tion is transferred under spin locking. This exper-iment which is known as HOHAHA (homonuclearHartmann-Hahn) or also as TOCSY, can be doneeither in 1-Dimensional difference mode or as a 2-dimensional experiment with phase encoding (33).In this technique, crosspeaks are observed betweenresonances within a single spin system (eg. pyra-noside ring) which share common coupling partners.A complete spin system can thus be identified ifthere is at least one resonance in the spin system,such as the anomeric proton, which is well isolatedand which has a reasonably large coupling to itsneighboring spin. Figure 3 shows an example of thephased 2-D HOHAHA spectrum of a hexasaccharidealditol (12), structure 2.

Fuc(al-+2)Gal(/?lGalNac-ol

Fuc(al-*2)Gal(01->4)GlcNAcO31—>6)/

The spin systems of the glcNAc residue (H3 andH4) and of both fucosyl residues (H2 and H3) in theoligosaccharide show strong and intermediate cou-pled peaks which are difficult to interpret in theCOSY spectrum because of their proximity to thediagonal peaks. But in Figure 3, the rows contain-ing the anomeric proton resonances of fucose andGlcNAc show crosspeaks with the H2, H3 and H4which are far from diagonal. Since fucose has thegalacto configuration, the H4-H5 coupling is smalland the spin propagates only as far as H4 in this ex-


Ea- a

(O

4 ppm

Figure 3. Phase sensitive 2-dimensional Homonuclear Hartmann-Hahn (HOHAHA) spectrum (300 MHz) ofthe H hexasaccharide, structure 3 in pyridine solution from ref. (12). Chemical shifts are referenced to internalTMS.

Vol. 10, No. 3/4 83

periment. Since there is net magnetization transferin the HOHAHA experiment, the spectrum can bephased such that all peaks have a positive sign andthe effects of limited digital resolution are less seri-ous than in phased COSY. The technical implemen-tation of the HOHAHA sequence requires spin lock-ing at moderate power which is not always readilyavailable on older spectrometers. Bax and cowork-ers (34) have recommended the use of the transmit-ter "low-power" mode augmented by an outboardlinear amplifier while Ranee (35) has demonstratedschemes in which the decoupler can be used for spinlocking.

While a complete assignment of the resonancesof each spin system to individual pyranoside ringscan usually be achieved by the DQF COSY methodaugmented perhaps by HOHAHA, other methodshave been found useful for the solution of specificproblems in the assignment. The triple quantumfilter COSY (TQF COSY) selects only those spinsystems for which there are three mutually coupledspins. In the absence of strong coupling, the pyrano-sides commonly found in glycoproteins have only afew systems with three mutually coupled spins. Onesuch system in hexapyranosides is H5,H6 and H6'which often presents some difficulty in assignmentif the ring contains equatorial protons with smallcoupling constants which prevent transfer of coher-ence from the anomeric proton to H5 and H6. Dwekand coworkers have suggested this experiment forresolving the difficult and important problem of theassignment of H5,H6, H6' in oligomannosides in theN-linked glycopeptides (36). They have observedthat if any two spins in the three mutually coupledproton system are chemically equivalent (i.e. theyhave the same chemical shift) no TQF crosspeakwill be seen. Further difficulties in this experimentresult from strong coupling among two protons ofthe ring which may cause an artifactual crosspeak.Although this is a powerful experiment, the resultsmust be carefully interpreted.

The presence of acetamido sugars in a structureis indicated by the amide methyl resonance near 2.0ppm but this signal does not generally indicate towhich residue the amide group belongs. The residuecontaining the amide group can be uniquely iden-tified by observation of the amide NH resonancewhich is coupled to the main ring spin system atH2 in the common 2-acetamido sugars galNAc and

glcNAc. Since in aqueous solution the amide pro-ton exchanges with solvent, it is generally neces-sary to identify the amide proton resonance withexperiments either in H2O solution or in an H2Oexchanged sample in non-protic solvents which donot promote exchange. For the case of experimentsin H2O solution, some water suppression method forimproving the dynamic range is usually necessary inorder to get reasonable sensitivity. Selective pulsesin 1-D difference decoupling spectroscopy are sim-ple and give a sensitivity equivalent to that of theD2O sample (25). The usual practice in 2-D spec-troscopy (COSY) involves water suppression withthe decoupler channel which, because of dynamicrange problems on most spectrometers, generallyleads to slightly degraded sensitivity. The capa-bilities of more modern spectrometers in generat-ing selective pulses may be exploited to improve thesensitivity of 2-D spectroscopy in H2O (37).

It is generally thought that the method of spincorrelation (eg. COSY) which is so crucial tothe successful application of the methods outlinedabove, cannot be applied to globular proteins overabout 20,000 Daltons as a result of the slow rota-tional correlation times which shorten T2 beyondthe point at which coupling correlations of 7 to10 Hz can be detected. One should therefore askwhether application of these methods to polysac-charides might encounter a similar limitation. Espe-cially if the polysaccharide adopts an extended rod-like conformation, quite low molecular weight poly-mers might exhibit long rotational correlation timespreventing the observation of spin correlation aboutthe pyranoside ring. Currently existing NMR dataon polysaccharides suggest that most of them aresufficiently flexible that this is not the case. Whilethe extent of segmental motion depends on the spe-cific glycosidic linkages of the polymer, the effectiverotational correlation times of most polysaccharidesare such that couplings of 1 to 3 Hz are resolved. Formore rigid polysaccharides, raising the probe tem-perature to 75°C may lengthen T2 and improve theresolution.

Once each sugar residue has been identified andits anomeric configuration determined, completionof the structure determination requires that the se-quence of the sugars and the linkage positions bedetermined. The vicinal proton coupling, which isso useful in the proton assignments of the individ-


ual pyranoside rings, is less valuable for relating therings to each other? Although the four-bond pro-ton couplings across the glycosidic linkage are largeenough to be useful in some aromatic glycosides,they are only about 0.2 Hz in disaccharides (38).Unfortunately such small coupling correlations aredifficult to detect in larger oligosaccharides and inpolysaccharides because of the longer rotational cor-relation times and short T2 associated with thesestructures. Proton nuclear Overhauser enhance-ment (NOE) which depends on proton proximity isa more effective method for determination of the se-quence of the sugar residues and in some cases theirlinkage positions as well. Although NOE is most of-ten observed between the anomeric proton and theproton connected to the carbon atom of the link-age (the aglycone proton), the effect depends on theconformation of the glycosidic linkage and some caremust be used in deduction of the oligosaccharidelinkage directly from proton NOE data. Several in-stances have been reported in blood group oligosac-charides and bacterial polysaccharides in which thelargest NOE between the anomeric proton and thatof the aglycone sugar is not to the aglycone protonitself, but to one adjacent to it. For the galNAca-(l—>3) gal linkage in the blood group A oligosac-charides, the NOE between galNAc HI and gal H4is much greater than to the aglycone proton, gal H3(25). This effect has been explained on the basisof the proposed conformation of the non-reducingterminal trisaccharide fragment of the blood groupA oligosaccharides (39). A similar effect observed inthe gal a-(l—*3) gal linkage of blood group B glycol-ipids presumably arises from a similar conformation(13). In a bacterial polysaccharide from Strepto-coccus sanguis 34 containing the sequence galNAca-(l—>3)rhamnose, the anomeric proton of galNAcshows the major NOE to the resonance of rha H2rather than to that of H3 (22). Although it is anequatorial proton adjacent to the linkage positionwhich shows the anomalously large NOE in eachof these examples, no simple rule can be extrap-olated from these observations since the man o>(1—>3)man sequence in N-asparagine glycopeptidesdoes not show this effect (40). The NOE results de-pend on the conformation of the glycosidic linkage,a topic which is not yet fully understood. There-fore, in any determination of the structure of a com-pletely new and unknown oligosaccharide system,

NOE data will reliably indicate the carbohydrate se-quence but there may be some ambiguities in linkageposition. Thus the NOE experiment may requiresome confirming data such as methylation analysis,enzymatic degradation, periodate oxidation or addi-tional NMR techniques, such as long range 13C-1Hcoupling correlation which will be discussed below.

NOE depends not only on proximities of pro-tons but also on the rotational correlation timeof the molecule, being positive for small molecules(hundreds of Daltons) and negative for large ones(thousands of Daltons). Although this effect canpresent some technical problems in the measure-ment of NOE in oligosaccharides, the extent of seg-mental motion in high molecular weight polysaccha-rides is usually such that substantial negative NOEare observed. By regulation of the probe temper-ature, conditions can generally be found for mostpolysaccharides such that NOE can be measured ei-ther by 1-D difference methods or by NOESY usingprocedures similar to those used for small proteins.When displayed in the phase sensitive mode, all the2-D NOESY cross peaks have the same sign. Thephases can be encoded either by the TPPI method(28) or by the method of States et al. (29) as de-scribed above for COSY.

We have alluded to the potential problem inmeasuring NOE for molecules of intermediate size.The rotational correlation times of oligosaccharideshaving 4 to 6 residues are such that these com-pounds generally show no NOE when studied at fieldstrengths of 300 to 500 MHz in D2O solution. Whilesubstantial negative NOE is generally observed forglycolipids in DMSO (41), NOE measurements onoligosaccharides and small glycopeptides may re-quire special measures. In our work, we have foundthat the rotational correlation time of oligosaccha-rides is a strong function of temperature. Figures4 and 5 show the 1-D difference NOE spectra for ahexasaccharide alditol (structure 2) at two differenttemperatures.

Since this oligosaccharide exhibits negative NOEat 5°C and positive NOE at 70°C , we conclude thatmeasurable effects can be measured quite generallyfor any size oligosaccharide by the 1-D differencemethod (12,42). When w is very close to rc, theeffects are modest (about 5 to 10 %) but can bemeasured with sufficient attention to control of theprobe temperature and to frequency stability of the

Vol. 10, No. 3/4 85

FucH1

2,3

Fuc"'H1

Gal3 Fuc Gal3

H3 H2 H2

Fuc'H2

.2,4

GalH2

/ / -—A'

w>.

Fuc FucH6 H6

PP«n

Figure 4. Difference NOE spectra (300 MHz) of the H hexasaccharide (structure 2) in D2O at 70°C showspositive enhancements (42).

spectrometer. Two-dimensional methods (NOESY) effects for resonances with similar chemical shifts,place lesser demands on spectrometer stability and The rotational correlation times of most polysac-suffer less from spectral overlap and cross saturation charides and of decasaccharides at 5 °C are sum-


H2 Gal3

Fuc2'3

H1

Fuc

NAc's

7=il //--^-T

-olGlcNAc

Fuc,H6s

1 ppm

Figure 5. Difference NOE spectra (300 MHz) of the H hexasaccharide (structure 2) in D2O at 5°C showsnegative enhancements (42).

ciently long that the NOESY experiment works well tures. Although it might appear that NOE measure-but for smaller oligomers, good NOESY data require ments in the rotating frame (ROESY) might offer amixed DMSO-water solvents and reduced tempera- solution to the experimental problem of 2-D NOE

Vol. 10, No. 3/4 87

measurement in oligosaccharides of medium size, ithas been pointed out that artifactual crosspeaks canarise when rotating frame NOE occurs to protonswhich are coupled and whose chemical shifts differby less than a few hundred Hz (43). This is preciselythe case for most oligosaccharides in which there areinvariably several coupled protons on the aglyconeresidue whose chemical shifts are quite similar.

In addition to the value of the NOE for struc-ture determination, the technique has been a majorsource of experimental information for determiningthe conformation of complex carbohydrates. SinceNOE depend on distances between protons, it is pos-sible in principle to determine inter-proton distancesdirectly from NOE data. In practice it is gener-ally impossible to measure a sufficient number ofindependent NOE data to rigorously determine anoligosaccharide conformation. However, when theNOE data are combined with the geometric con-straints dictated by the known covalent structureand the conformation of the pyranoside rings, a de-tailed conformation can often be deduced. Thereis a growing body of literature in which variouscomputational methods of molecular modeling arecombined with proton NOE data to deduce confor-mations of such complex carbohydrates as the N-asparagine glycopeptides (40,44) and blood groupoligosaccharides (39,45).

Essentially all of the structural studies of com-plex oligosaccharides, glycopeptides and of bacterialpolysaccharides by proton NMR, spectroscopy haveutilized D2O as the solvent. For glycolipids, whichaggregate into micelles in D2O solution, deuterateddimethyl sulfoxide has invariably been the solventof choice for structural studies and a substantial lit-erature exists both for gangliosides and for neutralglycolipids. It is clear from comparison of the XHspectra of similar oligosaccharides in DMSO andD2O that there are real differences in the relativeproton chemical shifts in these two solvent systemswhich does not result simply from the differencein chemical shift referencing. In recent studies onblood group oligosaccharides which are soluble inDMSO, we have completed proton assignments ofthe spectra by the same methods as those used inD2O solutions and we have extended this to the useof deuterated pyridine as the solvent (12). In Fig-ure 6 are compared the spectra of a blood groupA tetrasaccharide (structure 3) in D2O and in pyri-

dine showing even greater differences in the chemicalshifts in this solvent.

Fuc(alGal(/?l->3)GalNAc-ol

While the anomeric proton of fucose resonatesdownfield from that of galNAc in the aqueous sys-tem, the galNAc anomeric is the lowest field reso-nance in the spectrum in pyridine solution. Also thecommon rule that the anomeric resonances are themost downfield is not followed for the spectrum inpyridine solution, the resonance assigned to galNAcH2 being downfield of that of the anomeric pro-ton of /?-gal. The substantial influence of the sol-vent on the chemical shifts can be used to advan-tage in structural studies on complex carbohydratesfor which unfavorable signal overlapping and strongcoupling of the spectra in D2O solution causes diffi-culties in the tracing of the spin systems by couplingcorrelation and in identification of the sequence andcarbohydrate linkage by NOE (12). If a lE NMRstudy in D2O fails to yield data adequate to fullyidentify a new oligosaccharide structure, the exper-iment can be repeated in pyridine solution withoutcompromising the sample. This solvent can be evap-orated for subsequent studies in another solvent.

It might be suspected that the remarkable de-pendence of relative chemical shifts on solvent couldarise from a difference in the oligosaccharide confor-mation. But at least for the blood group A and Holigosaccharides, we have evidence based on mea-surements of the NOE in different solvents that theconformations of these oligosaccharides are similarin water, DMSO and in pyridine (45). We proposeinstead that the chemical shift differences depend ondifferent local magnetic susceptibility. This hypoth-esis is consistent with the observation that pyridine,with its highly anisotropic aromatic ring shows thegreatest effect as a result of the "ring current" ef-fects. Although a more detailed explanation of thesolvent perturbations of the chemical shifts mightprovide a valuable correlation with oligosaccharidestructure, we cannot yet provide any simple ratio-nalization of these effects.


GalNAc-olH2

FucH6

GalNAcHI

2 . 0 1 . 8 1 . 6 t . H 1 . 2 P P MS . S S.M 5 . 2 S . O 4 . 8

S . B 5 . 4 S . Z S . O 4 . 8 1 . 6 M . I 1 . 2 4 . 0 3 . 8 3 . 6 2 . 2 2 . 0 1 . 8 J . 6 l . f l . Z 1 . 0

Figure 6. Spectrum of the blood group A tetrasaccharide (structure 3) at 300 MHz and 60°C in D2O solutionreferenced to DSS (lower trace) and in pyridine solution referenced to TMS (upper trace) showing changes inrelative chemical shifts (12).

V. Prospects for Carbon NMR

While * H NMR spectroscopy has been the most im-portant source of structural information for complexcarbohydrates, 13C NMR spectroscopy has enor-mous potential for carbohydrates as a result of itschemical shift dispersion which is much greater thanthat of proton NMR. In contrast to the rathercrowded proton spectrum, the 13C NMR spectrumof a pentasaccharide has few overlapping lines evenwhen recorded at low field in an iron magnet spec-trometer at 25 MHz. In higher field superconduct-ing spectrometers individual lines are observed foressentially all the carbons of an oligosaccharide ofmodest size (see Figure 7). In the proton-decoupledcarbon spectrum of a milk pentasaccharide LNF-II

(structure 4) measured at 75 MHz, the anomericregion between 95 and 105 ppm contains resolvedsignals for the a-and /?-anomers of reducing termi-nal glucose and one line for each of the other fourresidues (46).

gal /?-(lfuc a-

*3)glcNAc •3)gal

The resonances of the two methyl carbons, onefrom C6 of fu cose and one from the amide methyl ofGlcNAc are at 16.5 and 23 ppm respectively. Theresonances at 80 and 83 ppm are assigned to glcC4 and to gal4 C3, the positions of glycosylation.

Vol. 10, No. 3/4

A J i /V u v78.0 77.0 76.0 75.0 74.0 73.0 72.0 71.0 70.0 69.0 68.0

PPM

I112.0 104.0 96 .0 88 .0 80 .0 72 .0 6 4 . 0 56 .0 48.0 40 .0 32 .0 24 .0 16.0

PPM

Figure 7. 13C NMR spectrum (75 MHz) of the milk tetrasaccharide LNF-II (structure 4). 7a is full spectrumand 7b is an expansion. •


These downfield shifts are characteristic of glyco-sidic substitution. The resonance of glcNAc C2 isat 57 ppm, an upfield position which is characteristicof the acetamido substituted carbons. The methy-lene carbons of the C6 are all resolved and locatedbetween 61 and 63 ppm. In addition to these rela-tively isolated and well resolved resonances, the ex-panded plot of Figure 7b between 66 and 78 ppmcontains 19 resolved lines in a region in which 21resonances are expected (46). In spite of the im-proved chemical shift dispersion of the 13C spectrumof Figure 7 over that of the XH spectrum of a simi-lar oligosaccharide (Figure 1), carbon spectroscopyhas not found wide use in structure determination ofglycopeptides and related oligosaccharides. One ofthe major barriers to wider exploitation of 13C NMRspectroscopy in complex carbohydrates has been itsrather poor sensitivity. Therefore, published dataon 13C spectroscopy of glycopeptides from higheranimals, which have been available only in very lim-ited sample sizes, have been modest. In contrast,polysaccharides such as those from yeast and bacte-rial cell surfaces can often be isolated in 100 mgquantities. Therefore, the availability of Fouriertransform 13C NMR spectrometers soon led to anextensive literature on the 13C spectra of polysac-charides. In a review of this field, Gorin (19) has de-scribed the correlation of the 13C spectra with struc-ture of the polysaccharide repeating unit. For mostpolysaccharides, line widths of only a few Hz wereobserved and the 13C NMR data have made signif-icant contributions to the overall understanding ofthe structure of bacterial polysaccharides (20).

Although it might appear that complete assign-ment of all the well resolved lines of the 13C spec-trum of a complex carbohydrate could be easier thanthe assignment of a 1H sp'ectrum, with its arrayof overlapping multiplets, a truly reliable assign-ment of the carbon spectrum of an oligosaccharidepresents some special problems. As a result of thelow natural abundance of 13C, there is no simpleanalog of the vicinal coupling of protons which pro-vides a rigorous assignment of the resonances of thesugar ring. Therefore other approaches to the as-signment of 13C resonances have been employed inthe studies on bacterial polysaccharides as well as inthe limited number of studies on glycopeptides andglycolipids. Many of the assignments were made byanalogies of the chemical shifts to the assigned res-

onances of the constituent monosaccharides and tothose of simple oligosaccharides following a build-up scheme which incorporates the effect of a-and/?-substituents in an empirical manner. This ap-proach, which has been called the glycosylation shiftmethod, has been applied in the most systematicway to the mannose oligosaccharides of the typeseen in N-linked glycopeptides (47,48). This em-pirical method suffers from the lack of a fundamen-tal theory of chemical shift and has no scheme toaccount for conformational perturbations which re-sult in long range effects. Therefore it is not veryreliable and has led to many errors in the assign-ments of the 13C NMR spectra of carbohydrates(46). Among other techniques which have been em-ployed to assist in assignment of the carbon reso-nances, specific labeling with 13C is most reliablebut its utility for assignments in complex carbo-hydrates isolated from natural sources is extremelylimited (49). The "attached proton test" (APT) andrelated DEPT pulse sequences involving transfer ofpolarization from 1H to 13C are very useful in or-ganic chemistry for distinguishing among methine,methylene and methyl carbons, each of which is di-rectly coupled to a different number of protons (50).For carbohydrates, most of whose carbons are of themethine type, this experiment is less useful servingmainly for rigorous identification of C6 in hexapyra-nosides (46). A third method which has been usedfor carbon assignments in complex carbohydrates isthe deuterium isotope shift (DIS) which depends onthe difference between the chemical shift of carbonatoms connected to OH and those connected to OD.The differential of a few hertz in 13C chemical shiftbetween carbohydrates in D2O and in H2O solutioncan be measured either in a coaxial tube scheme(51) or with two separate carefully referenced spec-tra, one measured in H2O and one measured in D2O(47). Since the deuterium isotope shift depends onwhether a 13C is a-or /?-to a hydroxyl group, itshould be possible to distinguish not only the signalsassigned to glycosidically linked carbons, but alsothose adjacent. Unfortunately differences in mag-netic susceptibility and concentration dependence ofthe chemical shift result in some technical difficul-ties in making the highly accurate chemical shiftmeasurements required by this experiment and theempirical interpretation requires some caution (52).The above methods have been used to derive com-

Vol. 10, No. 3/4 91

plete carbon assignments for complex carbohydratesystems including milk oligosaccharides (46), man-nosidosis oligosaccharides (53) blood group oligosac-charides (54) gangliosides (55) and for glycopeptides(56). In a review Bock et al. (57) have compiled ta-bles of the chemical shifts of many oligosaccharidesand Dill et al., (58) have reviewed the literature inglycopeptides. Although most of these resonanceassignments are likely to be correct, a few of themmay have been interchanged in these studies.

Several recent technical developments in 13CNMR spectroscopy may alleviate the difficulties inderiving reliable 13C assignments as well as con-tribute powerful new tools for structure determi-nation of complex carbohydrates. First the recentprogress in obtaining complete proton assignmentsfor complex carbohydrates can be exploited in as-signment of the carbon resonances by means of XH-13C chemical shift correlation. Although this corre-lation can be done in principle by established 13C de-tected heteronuclear COSY techniques which havebeen widely used in organic chemistry, the low sensi-tivity of this method presents a major obstacle to itsapplication for higher molecular weight complex car-bohydrates. Recent hardware developments makepossible proton detected 2-D heteronuclear multi-ple quantum coherence spectra which gives a gainin sensitivity sufficient to make 13C-1H correlationspectroscopy of complex oligosaccharides a real pos-sibility. The experiment requires a decoupler oper-ating at the 13C frequency and a dual tuned probe,preferably one with the proton detection coil in-side the carbon coil for improved sensitivity (34).Through special pulse sequences one can suppressthe 1H signals from protons connected to 12C andthe experiment requires as little as a micromole ofsample for a modern 500 MHz spectrometer. Anextended version of this reverse detected COSY, inwhich the spin is relayed from 13C to *H to vicinallycoupled 1H further extends its power in derivingcomplete assignment of the carbon spectrum fromthe assigned proton spectrum (59). Although thereare few reports in which this approach has been usedfor assignment of a large unknown complex carbo-hydrate, it appears that the added resolution gainedby spreading the congested proton spectrum in thecarbon dimension may extend considerably the highresolution NMR methods for complicated structures(60a).

Once the carbon spectrum has been completelyassigned, an unambiguous determination of the gly-cosidic linkage position can be obtained from the3-bond C-H coupling correlation. Detection of longrange coupling between either the anomeric carbonand the aglycone proton or between the anomericproton and the aglycone carbon would serve to iden-tify the linkage much more reliably than can bedone by proton NOE. Although these coupling con-stants are known to depend on geometry, unam-biguous determination of the linkage requires onlythat either one of these couplings be greater thanabout 5 Hz permitting detection in long-range 13C-*H correlation spectra. It has been proposed thatthis correlation can be observed either by the selec-tive INEPT method with 13C detection (61) or bythe 2-dimensional heteronuclear multiple quantumcoherence method with proton detection of the 3-bond correlation (59,62). It should be noted thatthe values of these long range carbon-proton cou-pling constants across the glycosidic linkage couldbe very useful in determining the conformation ofthe glycosidic linkage. The long range 3JCJJ be-tween the anomeric proton and the aglycone car-bon is related to the glycosidic dihedral angle <f>and that between the anomeric carbon and the pro-ton attached to the aglycone carbon is related tor]> by a Karplus relation. While the correct pa-rameterization of this Karplus relation is not yetunderstood, detailed studies of these coupling con-stants for model systems could contribute a valuabletool for investigation of the conformation of complexoligosaccharides.

VI. NMR Spectroscopy of OtherNuclei: 31P, 17O, 15N

Few nuclei other than XH and 13C have been ex-ploited in structural studies of complex carbohy-drates. 17O has very low natural abundance andsince it is quadrupolar it does not give high reso-lution spectra. The nitrogen nucleus of acetamidosugars is a candidate for study but the natural abun-dance of 15N is low and since no interesting struc-tural information is expected few data have been re-ported. Although study by NMR of other carbohy-drate substituents such as sulfate would be of greatinterest, detection of the sulfur nucleus by NMR isimpractical and only the effect of sulfation on *H


aa.

oo

ino

I Mf I IM| I I I I [I til l III IJITII j 11 i 111 llirrnTJlM I l l l l l | l l l l l IMI| 11 Hill I l|l I II MM I |M II MM I Jl II M III I ]M N Ml III5 . " 4 5 . 2 5 . 0 4 . 8 4 . 6 4 . 4 4 . 2 4 . 0 3 . 8 3 . 6 3 . 4

ppm

Figure 8. Two dimensional ^ P ^ H long range coupling correlated spectrum of the capsular polysaccharidefrom Streptococcus sangus 34, structure .5.

or 13C chemical shift has been used. Phosphoryla-tion of complex oligosaccharides is relatively com-mon and the sensitivity of 31P makes this nucleusparticularly suitable for high resolution NMR. Thechemical shift of 31P in phosphates is quite sensitiveto the state of ionization of the phosphate so phos-phomonoesters and diesters can generally be distin-guished from inorganic phosphate as a result of theirdifferent pKa (63). Since 31P in phosphate estersshows reasonably large vicinal and geminal couplingconstants to protons and to carbon atoms (5-7 Hz),long range coupling correlation can be used to as-sign proton or carbon resonances which are coupledto 31P (64). This approach can be used in the as-signment of the position of phosphate ester substi-tution in phosphorylated carbohydrates and in bac-terial poly saccharides containing internal phospho-diester linkages (65,66). Figure 8 shows the longrange 31P-1H 2-D correlated spectrum and the 1H

NMR spectrum of the capsular polysaccharide of thebacterium, Streptococcus Sanguis

This bacterial polysaccharide is composed of arepeating linear hexasaccharide linked with phos-phodiester bonds as shown in structure 5.

(l(l

—»6) a-D-GalpNAc (1-+4)/?-D-Glcp (l-^6)/8-D-Gal/+6)/3-D-GalpNAc (l-+3)a-D-Galp -(-1- ]r

5

The 1H spectrum was fully assigned using meth-ods such as COSY, HOHAHA and NOE which havebeen described above (22). The position of the phos-phodiester linkage was then determined by the vici-nal 31P-XH coupling correlation between phosphorusnucleus and the protons of the carbohydrate at thelinkage positions, a-gal HI and a-galNAc H6.

Vol. 10, No. 3/4 93

VII. High Resolution NMR ofIntact Glycoproteins

One of the major advantages of structural analy-sis by NMR spectroscopy, that no chemical treat-ment of an oligosaccharide or glycopeptide is re-quired, not only simplifies the experimental proce-dure but also allows easy recovery of the samplefor further study by other methods. A valuable ex-tension of this idea would be the extraction of in-formation about the structure of the carbohydratesidechains directly from the NMR spectrum of an in-tact glycoprotein. While aH NMR experiments onintact polysaccharides and glycolipids are routine,1H NMR spectroscopy of glycoproteins has beensomewhat limited for two reasons. First, the long ro-tational correlation times of globular proteins largerthan about 20,000 Dal tons shortens T2 beyond thepoint at which spin correlation can be used in theproton assignment. Second, the resonances of manyof the protons of the peptide overlap those of thesugars. For example the resonances of the a-and/?-protons of the amino acids fall in the 4.8 to 4.0ppm region and cover many anomeric proton reso-nances. For proteins in which amino acids are muchmore abundant than carbohydrate residues, it is im-possible to distinguish the sugars so complete protonassignments for glycoproteins have been limited tosome special types of glycoproteins that are inter-nally flexible and have high sugar content and somerepetitive structure. Our laboratory has carried outcomplete proton assignments of several antifreezeglycoproteins, important components of the bloodof polar fish. This class of glycoproteins has a sim-ple repeating sequence and the completely assignedproton NMR spectra have been used in conforma-tional modeling studies (67,68). A somewhat sim-ilar situation occurs in mucin glycoproteins whichare structurally related to antifreeze glycoproteinsand have been studied by proton NMR spectroscopy(69).

The improved chemical shift dispersion of 13CNMR spectroscopy provides greater possibilities forexperiments on intact glycoproteins. Dill and Aller-hand (70) have pointed out that there is a window inthe carbon resonances of the peptide residues abovethe aromatic resonances (160-110 ppm) and belowthe resonances of the a-carbons at about 70 ppm.Since the resonances of all the anomeric carbons of

the sugar fall in between 110 and 90 ppm they can bedetected even in the presence of a substantial excessof peptide signal. Some of the remaining carbon res-onances fall into this window region permitting theextraction of information on the carbohydrate struc-ture. In spite of the poor sensitivity of direct 13Cdetection, valuable studies on intact glycoproteinshave been reported including some on globular pro-teins (71) and the 13C spectrum of antifreeze glyco-protein has been completely assigned (72). Severalextensive studies of mucins, including those withblood group active carbohydrate sidechains, haveexplored the dynamics of the carbohydrate showingthat it is generally more flexible than the peptidecore (73,74).

Carbon-13 detection, which was used in all thestudies described above, requires 10-30 micromolesof sample in the NMR probe which translates into100 mg of a 30,000 Dal ton protein and presents amajor obstacle to these studies. Even if a glyco-protein can be obtained in such prodigious quan-tities, it may not form well behaved aqueous solu-tions at such high concentration. The prospect ofovercoming the sensitivity problem is offered by themethod of 1H detected 13C spectroscopy which canprovide 2-dimensional C-H correlated spectra on afew micromoles of sample. This technique, whichwas described above in connection with the problemof assignment of the 13C spectra of oligosaccharides,has the effect of spreading out the proton spectrumin the carbon dimension. Thus the anomeric pro-ton resonances which occur in an intact glycopro-tein between 4.4 and 5.4 ppm, would appear be-tween 95 and 105 ppm in the 13C axis, a region inwhich peptide resonances are absent. It is possiblethat inverse detected 2-dimensional 13C-1H correla-tion spectroscopy could provide detailed structuraldata on a few milligrams of an intact glycoprotein.

Acknowledgments

I wish to thank Perseveranda Cagas for the *Hdata on LND-I and C. Abeygunawardanafor the XH-31P correlation data on the S.s. 34 polysaccharide.Research supported by NSF grant DMB 8517421and by NIH grant GM-31449.


VIII. References1Murayama, J., Tomita, M., and Hamada, A. J.Biochem. (Tokyo), 91 , 1829-1836 (1982).

2Kornfeld, R. and Kornfeld, S. Ann. Rev.Biochem., 54, 631-664 (1985).

3Harford, J. and Ashwell, G. Methods in Enzy-mol. 109, 232-246 (1985).

4Kornfeld, S. Federation J. 1, 462-468 (1987).5Feeney, R. E., Burcham, T. S., and Yeh, Y.

Annu. Rev. Biophys. Biophys. Chem. 15, 59-78(1986).

6Karlsson, K. A. Chem Phys Lipids 42, 153-172(1986).

7Vliegenthart, J. F. G., Dorland, L., and van-Halbeek, H. Advan. Carbohyd. Chem. Biochem.,41, 209-374 (1983).

8Dua, V. K. and Bush, C. A. Analyt. Biochem.133, 1-10. (4), 386-397 (1983).

9Dua, V. K., Goso, K., Dube, V. E., and Bush,C. A. / . Chromatogr.

328, 259-269 (1985).10Vliegenthart, J. F. G., vanHalbeek, H., and

Dorland, L. Pure Appl. Chem. 53, 45-77 (1981)."Hanfland, P., Egge, H., Dabrowski, U., Kuhn,

S., Roelke, D., and Dabrowski, J. Biochemistry 20,5310-5319 (1981).

12Rao, B. N. N. and Bush, C. A. Carbohyd. Res.180, 111-128 (1988).

"Dabrowski, J., Egge, H., and Hanfland, P.Pure Appl Chem. 56, 807-819 (1984).

14Yu, R. K., Koerner, T. A., Scarsdale, JN, andPrestegard, J. H. Chem. Phys. Lipids 42, 27-48(1986).

15 Koerner, T. A. W., Prestegard, J. H., and Yu,R. K. Methods Enzymoi, 138, 38-59 (1987).

16 Clausen, H., Levery, S. B., Nudelman, E. D.,Stroud, M., Salyan, M. E., and Hakomori, S. J. Biol.Chem., 262, 14228-34 (1987).

17Jones, C. Carbohyd. Res., 139, 75-83 (1985).18Kenne, L. Stud. Org. Chem., 26, 199-210

(1986).19Gorin, P. A. J. Advan. Carbohyd. Chem.

Biochem., 38, 13-104 (1981).20Jennings,. H. J. Advan. Carbohyd. Chem

Biochem., 41 , 155-208 (1983).21Romanowska, E., Gamian, A., and Dabrowski,

J. Eur. J. Biochem., 161, 557-564 (1986).

22McIntire, F. C , Bush, C. A., Wu, S. S., Li, S.C., Li, Y. T., McNeil, M., Tjoa, S. S. and Fennessey,P. V. Carbohyd. Res., 166, 133-143 (1987).

^Carver, J. P. and Grey, A. A. Biochemistry 20,6607-6616 (1981).

24Trimble, R. B. and Atkinson, P. H. / . Biol.Chem. 261, 9815.24 (1986).

2SDua, V. K., Rao, B. N., Wu, S. S., Dube, V.E., and Bush, C. A. / . Biol Chem. 261, 1599-608(1986).

26Homans, S. W., Dwek,R., A, and Rademacher, T. W. Biochemistry 26,6571-6578 (1987).

27Bax, A. "Two-dimensional Nuclear MagneticResonance in Liquids" D. Reidel, Dordrect, Holland,1982. j

^Marion, D. and Wuthrich, K. Biochem. Bio-phys. Res. Comm. 113,967-974(1983).

"^States, D. J., Haberkorn, R. A., and Ruben,D. J. (1982) / . Magn. Reson. 48, 286-292.

^Keeler, J. and Neuhaus, D. J. Magn. Reson.63, 454-72 (1985).

31Widmer, H. and Wuthrich, K. J. Magn. Res.74, 316-336 (1987).

32Homans, S. W., Dwek, R. A., Boyd, J., Soffe,N., and Rademacher, T. W. Proc. Nail Acad. Sci.,tf&4 84, 1202-1205 (1987).

32aDabrowski, J., Dabrowski, U., Bermel, W.,Kordowicz, W. and Hanfland, P. Biochemistry 27,5149-5155 (1988).

33 Bax, A. and Lerner, L. Science (Washington)232, 960-967 (1986).

^Summers, M. F., Marzilli, L. G., and Bax, A.J. Am. Chem. Soc. 108, 4285-4894 (1986) .

35Ranee, M. J. Magn. Res. 74, 557-564 (1987).^Homans, S. W., Dwek, R. A., Boyd, J., Mah-

moudian, M., Richards, W. G. and Rademacher, T.W. Biochemistry 25,,6342-6350 (1986).

37Skelnar, V., Tschudin, R., and Bax, A. J.Magn. Res. 75, 352-357 (1987).

^Batta , G. and Liptak, A. / . Am. Chem. Soc.106, 248-250 (1984).

^Bush, C A., Yan, Z. -Y., and Rao, B. N. N.,/ . Am. Chem. Soc. 108, 6168-6173 (1986).

^Homans, S. W., Dwek, R. A. and Rademacher,T. W., Biochemistry 26, 6553-6560 (1987).

41Koerner, T. A. W., Prestegard, J. C , Demou,P. C., and Yu, R. K. Biochemistry 22, 2687-2690(1983).

Vol. 10, No. 3/4

42Rao, B. N. N., Dua, V. K. and Bush, C. A.Biopolymers 24, 2207-2229 (1985).

43Bax, A. / . Magn. Resonance 77, 134-147(1988).

44Cumming, D. A. and Carver, J. P. Biochem-istry 26, 6664-6676 (1987).

45Yan, Z. -Y., Rao, B. N. N. and Bush, C. A. J.Am. Chem. Soc. 109, 7663-7669 (1987).

46Bush, C. A., Panitch, M. M., Dua, V. K., andRohr, T. E. Anal. Biochem. 145, 124-136 (1985).

47Allerhand, A. and Berman, E. / . Am. Chem.Soc. 106, 2400-2412 (1984).

48Allerhand, A. and Berman, E. / . Am. Chem.Soc. 106, 2412-2420 (1984).

49Barker, R. and Serianni, A. S. Ace. Chem.Res. 19, 307-313 (1986).

^Doddrell, D. M. and Pegg, D. T. J. Am.Chem. Soc. 102, 6388-6390 (1980).

51Pfeffer, P., Valentine, K., and Parrish, F. J.Am. Chem. Soc. 101, 1265-1274 (1979).

52Blumberg, K. and Bush, C. A. Analyt.Biochem. 119, 397-406 (1982).

^Matsuura, F., Nunez, H. A., Grabowski, G. A.,and Sweeley, C. C. Archives Biochem. and Biophys.207, 337-352 (1981).

54Shashkov, A. S., Arbatsky, N. P., Derevit-skaya, V. A., and Kochetkov, N. K. Carbohyd. Res.72, 218-221 (1979).

55Sillerud, L. O., Prestegard, J. H., Yu, R. K.,Schafer, D. E., and Konigs berg, W. H. Biochemistry17, 2619-2628 (1978).

^Prohaska, R., Koerner, T. A. W., Armitage, I.M., and Furthmayr, H. J. Biol. Chem. 256, 5781-5791 (1981).

57Bock, K., Pedersen, C. and Pedersen, H. Adv.Carbohyd. Chem. Biochem. 42, 193-225 (1984).

58Dill, K., Berman, E., and Pavia, A. A. Advan.Carbohyd. Chem.

and Biochem. 43, 1-49 (1985).59Lerner, L. and Bax, A. Carbohyd. Res. 166,

35-46 (1987).^Dabrowski, J., Hauck, M., Romanowska, E.

and Gamian, A. Biochem. Biophys. Res. Commun.146, 1283-1285 (1987).

^ T s u i , F. P., Egan, W., Summers, M. F., Byrd,R. A., Schneerson, R. and Robbins, J. B. Carbohyd.Res. 173, 65-74 (1988).

61 Bax, A., Egan, W-, and Kovac, P. J. Carbohyd.Chem. 3,593-611 (1984).

62 Byrd, R. A., Egan, W., Summers, M. F., andBax, A. Carbohyd. Res. 166, 47-58 (1987).

^Gorenstein, D. G. Ann. Rev. Biophys. Bioen-gin. 10, 355-386 (1981).

64Jennings, H. J., Rosell, K. . G., and Carlo, D.J. Canad. J. Chem. 58, 1069-1074 (1980).

^Hellerqvist, C. G., Sundell, H., and Gettins, P.Proc. Natl. Acad. Sci., USA 84, 51-55 (1987).

^Mclntire, F. C , Crosby, L. K., Vatter, A. E.,Cisar, J. O., McNeil, M. R., Bush, C. A., Tjoa, S. S.,and Fennessey, P. V. / . Bacterioi 170, 2229-2235(1988).

67Bush, C. A. and Feeney, R. E. Int. J. Pept.Protein Res. 28, 386-397 (1986).

68 Rao, B. N. N. and Bush, C. A. Biopolymers26, 1227-1244 (1987).

^Gerken, T. A. Arch. Biochem. Biophys. 247,239-253 (1986).

70Dill, K. and Allerhand, A. J. Biol. Chem.254, 4524-4531 (1979).

71Berman, E., Walters, D. E., and Allerhand, A./ . Biol. Chem. 256, 3852-3857 (1981).

72Berman, E., Allerhand, A., and DeVries, A. L.J. Biol. Chem. 255,4407-4410(1980).

73Gerken, T. A. and Dearborn, D. G. Biochem-istry 23, 1485-1497 (1984).

74Gerken, T. A. and Jentoft, N. Biochemistry26, 4689-4699 (1987).

H igh R esolution N M R in the D eterm ination of S tructure in C … · 2010-04-16 ·...

Documents

Transcript of H igh R esolution N M R in the D eterm ination of S tructure in C … · 2010-04-16 ·...