Multiple-step relayed correlation spectroscopy

Proc. Natl. Acad. Sci. USA Vol. 81, pp. 6286-6289, October 1984

Biochemistry

Multiple-step relayed correlation spectroscopy: Sequential resonance assignments in oligosaccharides (two-dimensional NMR spectroscopy/ovomucoid)

S. W. HOMANS, R. A. DWEK, D. L. FERNANDES, AND T. W. RADEMACHER Department of Biochemistry, University of Oxford, South Parks Road, Oxford, OX1 3QU, England

Communicated by R. R. Porter, June 25, 1984

these two-dimensional methods over their one-dimensional counterparts, it is in general not possible to obtain all of the necessary assignments due to the preponderance of tight coupling between protons within the unresolved envelope. This tight coupling results in poor spectral dispersion even with two-dimensional methods. Furthermore, since asparagine-linked oligosaccharides are branched and possess some degree of symmetry or repetitive structure, chemical shifts of proton resonances (other than H1) derived from the monosaccharide residues in a given oligosaccharide are often degenerate. Conventional COSY thus, in general, requires support from "subspectral editing" experiments to resolve ambiguities in resonance assignment, resulting in a time-consuming and uncertain assignment strategy. In this paper we report that the recently developed method of homonuclear relayed correlation spectroscopy (RECSY) (9-11), when extended with a series of coherence transfer steps, can aid the resonance assignment of the 1H NMR spectra of oligosaccharides. The method directly correlates the resolved anomeric and H2 protons with remote (H4, H5) protons through a linear network of couplings. In this manner it is possible to obtain resonance assignments by simple inspection of cross-peaks that appear in a well-resolved region of the two-dimensional spectrum. We have used this technique to assign the proton resonances in the naturally occurring pentasaccharide Manal-3(Manal-6)Manf314GlcNAcf31-4GlcNAc [M-(M)-M-Gn-Gn] (Fig. 1) isolated from hen ovomucoid. This pentasaccharide is the core unit of the chitobiosyl complex, high-mannose, and hybrid oligosaccharide classes found on glycoproteins and will therefore form the basis of a complete resonance assignment strategy for the larger oligosaccharide structures.

A general property of the high-resolution ABSTRACT proton NMR spectra of oligosaccharides is the appearance of low-field well-resolved resonances corresponding to the anomeric (H1) and H2 protons. The remaining skeletal protons resonate in the region 3-4 ppm, giving rise to an envelope of poorly resolved resonances. Assignments can be made from the H1 and H2 protons to their J-coupled neighbors (H2 and H3) within this main envelope by using 1H-1H correlated spectroscopy. However, the tight coupling (J 6 ) between further protons results in poor spectral dispersion with consequent assignment ambiguities. We describe here three-step two-dimensional relayed correlation spectroscopy and show how it can be used to correlate the resolved anomeric (H1) and H2 protons with remote (H4, H5) protons directly through a linear network of couplings using sequential magnetization transfer around the oligosaccharide rings. Resonance assignments are then obtained by inspection of cross-peaks that appear in wellresolved regions of the two-dimensional spectrum. This offers a general solution to the assignment problem in oligosaccharides and, importantly, these assignments will subsequently allow for the three-dimensional solution conformation to be determined by using one-dimensional and two-dimensional nuclear Overhauser experiments.

High-resolution 1H NMR is now firmly established as a powerful tool in the determination of both primary structure (1) and secondary structure (2-5) in a variety of biologically important oligosaccharides and their fragments. In particular, the determination of oligosaccharide conformations by NMR and their correlation with primary sequence represents an approach to the understanding of the mechanisms of biological processes in which oligosaccharides are implicitly involved. A major experimental difficulty in these NMR studies has been the unambiguous assignment of proton resonances. In oligosaccharide NMR spectra, the anomeric proton resonances, and occasionally H2 proton resonances, of the constituent monosaccharide residues are well resolved (1). The remaining resonances are found in a broad unresolved region between 3 and 4 ppm. Early NMR studies of complex oligosaccharides relied upon the assignment of the resolvable resonances by comparison with model compounds (6). Whilst these "reporter resonances" are in some cases sufficient to determine oligosaccharide primary sequence, conformational analyses using nuclear Overhauser effects (NOEs) require a complete set of unambiguous resonance assignments (2, 3). We have previously investigated several strategies to address the assignment problem in oligosaccharides. These have involved 'H-1H correlated spectroscopy (COSY) (7), double-quantum NMR spectroscopy (8), and the use of intraresidue NOE connectivities obtained from one- and two-dimensional NOE experiments (3). Despite the advantages of

MATERIALS AND METHODS Oligosaccharide Preparation. Hen ovomucoid was obtained from Sigma (lot no. 128c-8045). The protein (2.5 g) was exhaustively dialyzed against distilled water at 4°C, lyophilized, and dried over phosphorus pentoxide under reduced pressure (7 days). It was then cryogenically dried under reduced pressure over activated charcoal (-196°C) for a further 7 days to remove protein-bound water. The sample was heated (100°C) under anhydrous 02-free argon at 1 atmosphere with 20 ml of freshly distilled anhydrous hydrazine for 8.5 hr. After removal of unreacted hydrazine the sample was N-acetylated with acetic anhydride. The N-acetylated mixture was applied to a 2.5 x 15 cm cellulose column equilibrated in 1-butanol/ethanol/water (4:1:1, vol/vol) and washed with 5 column volumes of equilibration solvent, and the oligosaccharide fraction was recovered by elution. The total yield of oligosaccharide was 0.67 g and the 1H NMR spectrum showed no contaminating peptide material. Abbreviations: COSY, 'H-'H correlated spectroscopy; RECSY, relayed correlation spectroscopy; NOE, nuclear Overhauser effect; Man or M, D-mannopyranose; GlcNAc or Gn, 2-acetamido-2deoxy-D-glucopyranose.

The publication costs of this article were defrayed in part by page charge payment. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. §1734 solely to indicate this fact. 6286

Biochem'istry: Homians et aL The purified oligosaccharides were reduced with NaB2H4 (Fluka). NaB3H4 (New England Nuclear; 500 mCi/mmol; 1 Ci = 37 GBq) was used to reduce a small fraction (2 mg) of the sample in parallel for use as a tracer. The neutral oligosaccharides were separated from the acidic components by paper electrophoresis in pyridine/acetate buffer, pH 5.4 (80 V/cm). The relative proportions of neutral (83%), monoacidic (14%), and diacidic (3%) sugars were found to be in agreement with those found by Yamashita et al. (12). The neutral fraction was then applied to a 1.5 cm x 200 cm Bio-Gel P-4 (-400 mesh) gel permeation column maintained at 55TC. The profile (Fig. 1) was found to be identical to that reported by Yamashita et al. (12). The oligosaccharide eluting at 7.12 glucose units was pooled and designated fraction N-8 after Yamashita et al. (12). Its structure was confirmed to be Manal3(Manal-6)Manfl-4GlcNAcf81-4GlcNAc by sequential exoglycosidase digestion. The sample was prepared for NMR studies by passage through a small column of Dowex AG50 X-12, AG3.4A, and Chelex 100 (Bio-Rad), followed by filtration through a 0.5-,tm-pore Teflon filter (Millipore). The oligosaccharide was deuterated by repetitive dissolution in 99.996% 2H20 (low in paramagnetic ions) (Aldrich) with intermediate flash evaporation, dissolved in 40 1L. of 99.996% 2H20, and deoxygenated by repeated evacuation and exchange with dry oxygen-free argon gas (prepared by catalytic removal of oxygen on BTS-catalyst) (Fluka). The sample (10 mM) was sealed anaerobically under dry argon within a 5-mm-bore precision-glass NMR tube (Wilmad). Two-Dimensional NMR Spectroscopy: COSY. Two-dimensional COSY experiments were conducted at 470 MHz according to Bax and Freeman (13). The basic pulse cycle was

90'(+x)-tj-0(0)acquisition (02, where k = +x, +y, -x, and -y and 6 = 900. For each t1 value 32 transients were collected. Coherence transfer echoes were selectively detected by the subtraction of alternate

Proc. NatL. Acad. Sci. USA 81 (1984)

6287

scans of the form

900-t1-0()-t2 and 9O0-tj-0(0 + 900)-t2, where 4 = 0° or 180°. All together, 1024 x 2048 spectra (sweep width = +900 Hz) were collected with a minimum evolution time (tj) of 5 ,us and an increment of 555 ,us to tj between spectra. Prior to Fourier transformation, the timedomain matrix was multiplied in both the tj and t2 dimensions with phase-shifted sine bell functions. The use of phase cycling to achieve quadrature phase detection in both dimensions allowed the transmitter frequency to be set to the Middle of the spectrum and thus reduced the size of the data matrix. Two-Dimensional NMR: RECSY. Two-dimensional 1H-1H multistep RECSY was performed at 470 MHz using the eight-pulse sequence

90.-ti-900 -

180 -Z 90', - 180 -D 90,, -V 180 -V

900,-t2.

The delays Ti and r2 were incremented sequentially with t1, to optimize transfer efficiency between protons in the alinked mannose residues 4 and 4', resulting in r1 = 60-120 ms and T2 = 20-75 ms. The value T3 was held constant at 20 ms. Phase cycling was used for 4 to eliminate axial peaks. No NOE cross-peaks in the two-dimensional spectrum were observed, reflecting the inefficiency of magnetization transfer by this mechanism for the compound under investigation. The experimental conditions and data processing were identical with those described above for COSY.

RESULTS Application of RECSY. As described by Eich et al. (9), the "mixing pulse," 6, of conventional two-dimensional COSY (900-1-1-t2) (Fig. 2A) is replaced in RECSY by the se-

FIG. 1. High-resolution gel permeation chromatogram of the neutral asparagine-linked oligosaccharides released from hen ovomucoid by hydrazinolysis. Tracer oligosaccharides were labeled by reduction with NaB3H4. The glucose oligomer positions are indicated by vertical arrows. The structure of the oligosaccharide eluting at 7.12 glucose units and used in this study is shown. Resolution enhancement of gel permeation data was achieved by deconvolution of the data by Fourier transformation.

6288

Biochemistry: Homans et aLPProc. NatL Acad. ScL USA 81

t2

ti

(A)

(B)

Pi

(C)

F]

ti

P2 'r

P2 Tl

P3

ti

P4

Q

I

I

:Kt, 1

Ir

P3

I

ti

(1984)

:Kt,

T2

1

1

:Lt,

1 1

I

I

I

P5 T2

:Lt, P 1 1

T3

'lm. 1 1

I

P7 T3 :H. N

t2

sequence used in COSY. P1 and P2 represent pulses, and t1 and t2 are the acquisition times in the W, sf FIG. 2. (A) Conventional two-pulse and w2 dimensions, respectively. (B) Pulse sequence used in the generation of single-step relayed correlation spectra. The rf pulse P2 in A is replaced by P2-1-P3-T-P4, which induces two consecutive coherence transfer steps. The delay T is a fixed interval, the length of which depends upon the J couplings of the spins involved in the relay (see text). (C) Eight-pulse sequence used to generate multistep relayed correlation spectra of the oligosaccharide shown in Fig. 1. The delays rl, T2, and T3 depend in a complex manner upon the neighboring J couplings of the five spins involved in the relay. The intervals Ktj, LDI, and Mtl are incremented sequentially with tj to optimize the efficiency of relayed magnetization transfer.

quence 90W-T-180W-T-90' (Fig. 2B), which can induce two consecutive coherence transfer processes. This can conveniently be illustrated by considering a two-spin '/2 system IS, with Hj = 2ifJsI-S. The effects of the pulse, sequence 90%t,-90 -r-180W-r-90' can be described in terms of the single spin and product spin operators (14). Transverse magnetization of spins I and S, created by the first pulse, evolves under the influence of Hz and Hj during the period tj and after the second pulse reaches the state 7rJ,Stl2IJS, ftIz + flstS, /2(k + Sx)

Sy)O (-I, cos fit, + Ix sin fit, - S, cos fQst, + Sx sin list,) cos 7Js5t1 - (+ 2IxS, cos ft,, + 2IS, sin flt, + 2SJI, -

(Iy +

cos

11stj

+

2SIy sin fQstj)sin uJisti,

[1]

generally possible for a typical monosaccharide due to the large variation in the J values around the ring. In a-D-mannose, for example, J12 = 1.9 Hz, J23 = 3.4 Hz, J34 = 9.4 Hz, and J45 = 10 Hz, leading to poor transfer efficiency. In practice, however, the transfer efficiency for any given residue can be optimized to give observable cross-peaks by incrementing the characteristic r values for each oligosaccharide type in concert with tj (Fig. 2C). 'H-'H Correlated NMR Spectruin of M-(M)-M-Gn-Gn. The conventional one-dimensional spectrum can be thought of as lying along the main diagonal W1 = W2 (Fig. 3). The offdiagonal peaks (cross-peaks) correlate J-coupled resonances. This is illustrated in Fig. 3, for mannose 4H1-4H2, and for mannose 4H2-4H3. However, attempts to obtain further correlations from within the main envelope lead to ambiguity. For example, at least three pairs of cross-peaks

where Q., and fQs are the Larmor precession frequencies of spins I and 5, respectively. The term -2IzSy sin fQt,, which describes antiphase magnetization of spin S, can now undergo relayed coherence transfer to a further spin M under the influence of HJ. Assuming JIM = 0, then

irJIs2T2IS, J-st,

-3.6 -3.8

' | ffi

7rJSM2T2SZMZ

-

_ , 4'H3

-4.0

~~4H3

-2IfS. sin kt, sin 7T (-2IS, cos IrJJs2T + S, cos TrJsM2T sin 1rJ152T + 2SMZ sin 1JSM2T sin lrJs52T)sin filt, sin 7rJ1stj. [2]

-4.2 E 0L

o

-4.4 - 4.6

The third term in Eq. 2 is converted into observable antiphase M magnetization by the last pulse

-4.8

I

x+ M. ) lr/2(S -

2SM, sin 7TJSM2T sin 7rJ/s2T sin flt, sin fTJisti -2SzMv sin dJSM2T sin rrJ152T sin fkQt, sin 7rJs5tj. [3] As pointed out by Eich et al. (9), multiple coherence transfer steps may be performed. In the four-step transfer (three-step relay) experiment described below, the transfer function sin tJSM2T sin irJ,52r in Eq. 3 becomes modified for an ISMQX spin system, to sin 7rJs52T1 sin ITJsM2Ti sin ITJSM2T2 sin 7TJMQ2r2 sin IrJMQ2T3 sin frJQX2T3, assuming all couplings except neighboring couplings are zero. Therefore, for maximal transfer efficiency, this function requires that the delays T1, T2, and T3 are matched to the J values of the residues of interest (e.g., Ti has to be matched to Jjs and JSM). This is not

'4'H1'

4'H2 '

4H1,,e.__

4H2 '

5.2

4.2 4.0

-5.0 L

5.0 4.8 4.6 4.4

3.8

rz Z3

3.6

W1, ppm

FIG. 3. Contour plot of the conventional 'H-'H correlated twodimensional NMR spectrum (3.6-5.2 ppm) of the oligosaccharide shown in Fig. 1. The off-diagonal peaks (cross-peaks) correlate Jcoupled protons, which lie along the main diagonal. The broken lines show the correlations between mannose 4H1-4H2, 4'H1-4'H2, 4H2-4H3, and 4'H2-4'H3. The pair of broken lines at W2 = 3.87 ppm illustrates the ambiguity in the assignment of mannose 4H4 and 4'H4.

Proc. Nati. Acad. Sci USA 81 (1984)

Biochemistry: Homans et aL

I--

-I9

IU---

-

-

-

0 -1~ ~~~

. ..

S.

qs

ws-

Table 1.

-3.8

4GIcNAc derived from hen ovomucoid

pentasaccharide

E

-4.4 H H ol~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ -4.6 -4.8

go

I

a

-5.0

H3 H5 H4

H2

4H1l

_

_

I

5.2

5.0 4.8

4.4 4.2 oi, PPm

4.6

4.0

3.8

H6a H5 H4 H3 H2 4.23 -t 3.79 3.84 3.75 3.55 3.75 4.64 2 3.63 3.177 3.76 4.24 4.78 3 3.74 3.79 3.62 3.87 4.05 5.10 4 3.70§ 3.75 3.62 3.86 3.95 4.90 4' 2,2-dimethyl-2-silapen*Chemical shifts are given relative to sodium 5 tane-4-sulfonate (indirectly to acetone, = 2.225 ppm at 250C). Shifts are accurate to +0.01 ppm. tThe residue reduced by NaB2H4. tApproximate value due to mixing of the eigenstates of H3 and H4. §Tentative assignment due to weak cross-peak.

(Fig. 1)

*

- 5.2

3.6

FIG. 4. Contour plot of the two-dimensional multistep relayed correlation spectrum (3.6-5.2 ppm) of the oligosaccharide of Fig. 1. As in the conventional COSY spectrum of Fig. 3, the broken lines correlate mannose 4H1 with 412. In addition, mannose 4H1 can be correlated directly with mannose 4H3 as shown. The correlations with mannose 4H4 and 4H5 are determined in a similar manner. The correlation between mannose 4H2 and 4H6a is also illustrated.

could correlate with the resonance position of mannose 4H3 = W2 = 3.87 ppm. The assignment strategy would at clearly be simpler if correlations were directly obtained from mannose 4H1 sequentially around the ring. Such correlations are never seen in COSY spectra, due to the vanishingly small couplings of remote protons, resulting in zero crosspeak intensity. They can, however, be directly observed by using multistep RECSY, since, as shown above, nonzero transfer efficiency is obtainable even with zero coupling bewI

tween remote protons.

Contour Plot of the Two-Dimensional Three-Step Relayed Coherence Transfer Experiment. In comparison with conventional COSY (Fig. 3), three new cross-peaks octur at w02 = 5.1 ppm; these new cross-peaks correlate mannose 4H1 directly with mannose 4H3 (8 = 3.87 ppm), mannose 4H4 (6 = 3.62.ppm), and mannose 4H5 (8 = 3.79) (Fig. 4). Although at first sight this extra connectivity information might be considered to increase assignment ambiguity, it is important to realize that assignments may reliably be determined sequentially around the ring by application of one-step, two-step, and three-step relayed coherence transfers, respectively. We have chosen to illustrate the result of the three-step sequence in Fig. 4. By correlating protons within the same network of spins, it is possible to obtain unambiguous proton resonance assignments for M-(M)-M-Gn-Gn. These are shown in Table 1.

DISCUSSION Several important points emerge from the data of Fig. 4. It is notable that magnetization transfer occurs with good efficiency only from the anomeric proton to remote protons in the a-mannosyl residues (4, 4') in M-(M)-M-Gn-Gn. This residue specificity is a direct result of the dependence of the transfer efficiency upon all of the J couplings in the ring (see above). Since each different oligosaccharide residue has a characteristic "I-signature," the relayed experiment can be performed with selectivity on any given monosaccharide residue type by the correct choice of the T values. This clearly extends even to the level of the a and anomers of a given monosaccharide-i.e., transfer efficiency is very low for a

Manai-3(Manai-6)Manpi-4GlcNAc,1iChemical shift,* ppm

Residue

-af

ia -4.2

/

'H NMR resonance assignments for the reduced

-3.6

-4.0

__4H2N

6289

H1

P-linked mannose 3 as shown in Fig. 4. In cases in which H2 protons are well resolved (e.g., mannosyl H2s), sufficient magnetization is transferred into the hydroxymethyl grouping to allow identification of one of the H6 protons (Fig. 4). The relay experiment can be applied to oligosaccharides of greater complexity or even to oligosaccharide mixtures; This capability arises from the inherent property of high-resolution 'H NMR spectra of oligosaccharides that the anomeric protons are always at low field and well resolved, allowing correlations to be made in a well-resolved region of the twodimensional spectrum. In conclusion, we have presented an assignment strategy for oligosaccharides that depends upon sequential magnetization transfer around the ring. Although optimization of transfer efficiency for mtiftistep transfer requ'es prior knowledge of all neighboring couplings in the ring, in practice these can be predicted with accuracy since it is found that monosaccharide ring geometry in oligosaccharides is essentially fixed and J couplings approximate those of the isolated monosaccharides. Assignments can be made sequentially and unambiguously, and this should be of value in the conformational analysis of oligosaccharides. This research has been funded by grants from the Medical Ressearch Council, the Science and Engineering Rdtedtch Council (Oxford Enzyme Group), and Monsanto. 1. Vliegenthart, J. F. G., Van Halbeek, H. & Dorland, L. (1981) Pure Appl. Chem. 53, 45-77. 2. Homans, S. W., Dwek, R. A., Fernandes, D. L. & Rademacher, T. W. (1982) FEBS Lett. 150, 503-506. 3. Homans, S. W., Dwek, R. A., Fernandes, D. L. & Rademacher, T. W. (1983) FEBS Lett. 164, 231-235. 4. Brisson, J.-R. & Carver, J. P. (1983) Biochemistry 22, 36713680. 5. Brisson, J.-R. & Carver, J. P. (1983) Biochemistry 22, 36803686. 6. Dorland, L., Haverkamp, J., Schut, B. L., Vliegenthart, J. F. G., Spik, G., Strecker, G., Fournet, B. & Montreuil, J. (1977) FEBS Lett. 77, 15-20. 7. Homans, S. W., Dwek, R. A., Fernandes, D. L. & Rademacher, T. W. (1983) Biochim. Biophys. Acta 760, 256-261. 8. Homans, S. W., Dwek, R. A., Fernandes, D. L. & Rademacher, T. W. (1984) Biochim. Biophys. Acta 798, 78-83. 9. Eich, G., Bodenhausen,- G. & Ernst, R. R. (1982) J. Am. Chem. Soc. 104, 3731-3732. 10. Wagner, G. (1983) J. Magn. Reson. 5S 151-156. 328-332. 11. King, G. & Wright, P. E. (1983) J. Magn. Reson. 54, 12. Yamashita, K., Kamerling, J. P. & Kobata, A. (1982) J. Biol. Chem. 257, 12809-12814. 13. Bax, A. & Freeman, R. (1981) J. Magn. Reson. 44, 542-561. 14. Sorensen" 0. W., Eich, G. W., Levitt, M. H., Bodenhausen, G. & Ernst, R. R. (1983) Prog. NMR Spectrosc. 16, 163-192.