document

NATURE MEDICINE • VOLUME 7 • NUMBER 1 • JANUARY 2001 123

NEW TECHNOLOGY

In light of the increasing amount of genetic information that isbecoming available through sequencing efforts, it is now ap-parent that high-throughput systems that can provide a globalview of cell function and that lend themselves to automationare necessary to gain a comprehensive understanding of com-plex biological processes. To this end, proteomics, or ‘func-tional genomics’ has focused on the identification of a cell’sproteins, their post-translational modifications, and their in-teractions with one another. Together, an appreciation of a cell’s protein population can lead to a better understanding of how genotype elicits a phenotype. However, modifications to aprotein’s primary sequence and non-covalent protein–proteininteractions are but two mechanisms by which cells ‘fine-tune’a given protein’s activity; another equally important mecha-nism is non-covalent interactions between proteins andbiopolymeric saccharides at the cell surface and in the extracel-lular matrix. Such non-covalent interactions are critical for theproper functioning of many proteins, including growth factors,cytokines and morphogens.

Of the cellular- and extracellular-matrix polysaccharide con-stituents involved in the regulation of protein stability and func-tion, the heparin/heparan sulfate-like glycosaminoglycan(HLGAG) family is especially important. These polymeric saccha-rides are structurally characterized by a disaccharide repeat unit ofa uronic acid [either α-L-iduronate (I) or β-D-glucuronate (G)]linked 1→4 to a hexosamine (H). The chemical diversity ofHLGAGs arises not only from the identity of the uronic acid butalso from differential sulfation of individual disaccharide units.Four separate sites of potential sulfation exist, specifically, the 2-Oposition of the uronic acid (2S) and the 3-O (3S), 6-O (6S) and N(NS) positions of the glucosamine. Consistent with their structuralcomplexity, HLGAGs are known to play important regulatoryroles in diverse processes such as development1, morphogenesis2,cancer progression3,4, angiogenesis5 and anticoagulation6. Thepresence of diverse structural motifs that are specific for a cell typeand are tightly regulated—developmentally, temporally and spa-tially—allows HLGAGs to select for protein binding partners in ahighly selective manner. For instance, changes in HLGAG compo-sition and sequence are known to act as a developmental switch,carefully modulating the activity of mitogens both temporally andspatially7.

As HLGAGs play a fundamental role in the regulation of biolog-ical processes by binding to proteins and modulating their activ-ity1,2,8, identification of protein binding sites on cell-surfaceHLGAGs represents an important next step toward our under-standing of cell function beyond proteomics. In addition, it repre-sents an untapped opportunity to identify novel drug targets.Many of the proteins that bind to and are regulated by HLGAGsare clinically relevant molecules, including growth factors like fi-

broblast growth factor (FGF), angiogenesis inhibitors like endo-statin, and morphogens like WNT (ref. 9). For instance, the wide-spread use of mast-cell heparin (a highly sulfated HLGAG) fordecades as an anticoagulant is due primarily to the binding of aspecific HLGAG pentasaccharide sequence to antithrombin III(AT-III)10, that is, HNAc/S,6SGHNS,3S,6SI2SHNS,6S.

Despite the overwhelming success of heparin as an anticoagu-lant, clinical investigations into other protein binding sequencesin HLGAGs have been limited by the fact that these polysaccha-rides are only available in very small amounts and that no practi-cal methods exist for the isolation and direct sequencing ofspecific protein-binding oligosaccharides from HLGAGs at the cellsurface. Moreover, HLGAGs, unlike other biopolymers (for exam-ple, polynucleotides or polypeptides), cannot be amplified due totheir structural diversity and complex biosynthesis8. These issuespreclude the use of traditional analytical methods such as NMRand HPLC, and have impeded the discovery of HLGAG sequencesthat may have physiological and pharmacological significance.

Previous efforts to circumvent the problem of limited materialshave relied on heparin as a substitute for tissue-derived HLGAGs.Thus, to identify HLGAG sequences that bind to a particular pro-tein, the most common methodology involves column chro-matography of oligosaccharides derived from porcine intestinalmucosa heparin, the ‘universal’ HLGAG. Although this methodhas been successfully used to identify binding sequences for sev-eral proteins, including AT-III (ref. 10) and acidic fibroblast growthfactor11 (FGF-1), it is limited in that heparin contains a restrictednumber of sequences due to its high sulfate content, which biasesthe selection procedure. Thus, there is no opportunity to samplethe ‘real’ cell-surface HLGAG sequences that a protein encountersand binds to in vivo. To discover HLGAG sequences that bind withhigh affinity to a protein, identifying novel drug targets and ex-tending our understanding of biological phenomena in the post-genomic age, requires the development of sensitive methodologiesthat require only picomoles of material.

The development of mass spectrometric (MS) approaches for theanalysis of proteins has propelled forward the entire field of pro-teomics by allowing routine analysis of minute amounts of pro-tein, levels that can readily be isolated from in vivo sources. SuchMS methodologies have extended protein analysis beyond meredisplay of proteins to identification and quantification of proteinsin a cell and discovery of protein–protein interactions critical forcellular communication. Similar analytical techniques must be de-veloped for the analysis of protein–polysaccharide interactions.

We recently developed a highly sensitive matrix-assistedlaser desorption ionization mass spectrometric (MALDI-MS) ap-proach to analyze picomole amounts of HLGAG oligosaccha-rides12. We reasoned that a combination of our MALDI-MSmethod to detect HLGAGs and immobilization of an HLGAG-

Direct isolation and sequencing of specific protein-bindingglycosaminoglycans

NISHLA KEISER1, GANESH VENKATARAMAN1,2, ZACHARY SHRIVER1 & RAM SASISEKHARAN1

1Division of Bioengineering & Environmental Health, Center for Biomedical Engineering, 2Harvard-MIT Division of Health Sciences and Technology, Massachusetts Institute of Technology,

Cambridge, Massachusetts 02139, USA Correspondence should be addressed to R.S.; email: [email protected]

©20

01 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://m

edic

ine.

nat

ure

.co

m© 2001 Nature Publishing Group http://medicine.nature.com

124 NATURE MEDICINE • VOLUME 7 • NUMBER 1 • JANUARY 2001

NEW TECHNOLOGY

binding protein on a surface should make it possible to performaffinity isolation of tissue-derived HLGAG oligosaccharides ona MS surface. Here, we demonstrate how a surface non-covalentaffinity mass spectrometry (SNA-MS) strategy allows us to cou-ple several rigorous analytical methodologies into a single pro-cedure. Using well-established HLGAG-binding proteins as amodel system, we show how SNA-MS can be used to directlyisolate, enrich and sequence HLGAG oligosaccharides thatbind to a specific protein. Moreover, we establish that thismethod can be extended to cell-surface HLGAGs, demonstrat-ing that—for the first time—biologically relevant HLGAGoligosaccharides can be isolated and directly sequenced.

Isolation, selection and sequencing strategyThe overall strategy for SNA-MS involves protein immobiliza-tion on a thin hydrophobic film or on a metal surface, and sub-sequent addition of an aqueous mixture of oligosaccharidesthat is allowed to bind to the protein (Fig. 1). Following a low-salt wash to remove the non-specific binders (for example, 0.2M NaCl), a synthetic basic peptide, (RG)19R, is added in matrixsolution to chelate specific binders, and the oligosaccharidesare detected as (RG)19R peptide–oligosaccharide complexesusing MALDI-MS (ref. 13). Using enzymatic or chemical meth-ods, HLGAG oligosaccharides can also be subjected to depoly-merization before chelation with (RG)19R, enabling their directsequencing. SNA-MS therefore combines protein immobiliza-tion, saccharide selection and sequencing into a single proce-dure wherein high-affinity binders can be rapidly identifiedand sequenced.

In general, proteins that bind to HLGAGs are in the15,000–60,000-Dalton range; with proteins of this size theHLGAG binding site can at most accommodate a tetra- tohexasaccharide. This ‘minimal-binding size’ of HLGAGs is ob-served in crystal structures of several protein–HLGAG com-plexes, including FGF-1 and FGF-2 (refs. 14–16), AT-III (ref.17), and foot-and-mouth disease virus18 (FMDV). Therefore, weused penta- or hexasaccharide libraries in the development ofSNA-MS. However, this technique can be easily extended tolonger saccharide fragments should such fragments representminimal binders.

SNA-MS of antithrombin-IIIAs a first step in the development of SNA-MS, we focused onthe AT-III system for a number of reasons. First, the high-affin-ity HLGAG binding site for AT-III is well established (that is,HNAc/S,6SGHNS,3S,6SI2SHNS,6S), providing an opportunity to develophighly specific binding, washing and elution conditions forSNA-MS. Second, synthetic compounds are available that pos-sess the intact AT-III site or low-affinity variations that lack onesubstituent essential for high-affinity binding. Thus, with thisrepertoire of compounds, we are able to address issues of selec-tivity. Finally, AT-III, like many mammalian proteins, is heav-ily glycosylated, representing a challenge for immobilizationchemistries that can be widely extended to many systems.

Investigation of a variety of surfaces and immobilizationschemes were completed to bind AT-III in a rapid and repro-ducible manner. We found that the optimal scheme in-volved biotinylation of the protein followed by its additionto avidin, which had been adsorbed on a thin hydrophobicfilm. Penta 1 (see Table 1), containing an intact AT-III pen-

Laser

Protein

MALDI surface

Selection

Biotin

Minimal bindingHLGAGoligosaccharides

Mass Spectrogram

Avidin

Fig. 1 Schematic of surface non-covalent association mass spectrometry(SNA-MS). Avidin is adsorbed on a surface suitable for MALDI-MS analysis(stainless steel plate or hydrophobic film). Biotinylated protein is then im-mobilized on the avidin. Saccharides (shown as disaccharide repeat unitsof three for hexasaccharides) are applied to the protein spot and washedwith salt to select for high-affinity binders (in green). After application of(RG)19R in caffeic acid matrix, the sample is directly analyzed by MALDI-MS in a one-step procedure. Saccharides are detected as non-covalentcomplexes with (RG)19R; the mass of the saccharide can be determined bysubtraction of the (M+H)+ value of the peptide from the observed (M+H)+

value of the 1:1 peptide–saccharide complex.

Table 1 In the oligosaccharide abbreviations, ∆U refers to a C4-C5 unsat-urated uronic acid, I is α-L-iduronate, G is β-D-glucuronate, and H is hex-osamine. The 2S, 3S, 6S and NS subscripts indicate sulfation at therespective O or N positions, whereas OMe represents methylation of theanomeric hydroxyl group of the glucosamine. Hexa 1 refers to two separatehexasaccharides used in this study either derived from (a) heparinase di-gestion of heparin or (b) nitrous acid degradation of the same followed bysodium borohydride reduction. The former has a mass of 1732.5 whereasthe latter has a mass of 1655.4; despite their slight chemical difference,both have been shown to bind with the same affinity to growth factors, forexample, FGF. All oligosaccharides are detected as non-covalent complexeswith a basic peptide (RG)19R using this procedure, such that the theoreticalm/z value of (M+H)+ for a 1:1 complex between the saccharide and thepeptide is also presented. The theoretical (M+H)+ value of (RG)19R is 4226.8.

HLGAG oligosaccharides used in this study and theircorresponding theoretical masses

Saccharide Sequence Mass m/z ComplexHexa 1 (a) ∆U2SHNS,6SI2SHNS,6SI2SHNS,6S (a) 1732.5 (a) 5959.3

(b) I2SHNS,6SI2SHNS,6SI2SMan6S (b) 1655.4 (b) 5882.2

Hexa 2 ∆U2SHNS,6SIHNAc,6SGHNS,3S,6S 1614.3 5841.1

Hexa 3 ∆U2SHNS,6SI2SHNS,6SGHNS,6S 1652.4 5879.2

Synthetic HNS,6SGHNS,3S,6SI2SHNS,6S,OMe 1508.2 5735.0Penta 1

Synthetic HNS,6SGHNS,6SI2SHNS,6S,OMe 1428.1 5654.9Penta 2

Tetra 1 ∆U2SHNS,6SGHNS,6S 1074.9 5301.7

Di 1 ∆U2SHNS,6S 577.5 4804.3

©20

01 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://m

edic

ine.

nat

ure

.co


Fig. 3 Selection of HLGAG oligosaccharide sequences that bind withhigh affinity to FGF-2. All saccharides are observed as 1:1 non-covalentcomplexes with the basic peptide (RG)19R. The indicated masses are nor-malized by subtraction of the (RG)19R contribution, whereas the x-axisrepresents the observed mass of the complex. All masses are within 1.5units of the theoretical mass for each saccharide. a, Addition of 5 pmolHexa 1 (I2SHNS,6SI2SHNS,6SI2SMan6S) and Hexa 2 to FGF-2, followed by a saltwash to remove low-affinity binders results in a signal for Hexa 1 only.Previous studies have shown that Hexa 1 binds FGF-2 with nanomolaraffinity, independent of whether the saccharide is derived from hepari-nase cleavage (∆U2SHNS,6SI2SHNS,6SI2SHNS,6S, Table 1) or from nitrous acid scis-sion (I2SHNS,6SI2SHNS,6SI2SMan6S, Table 1). b, A hexasaccharide mixture

derived from a partial digestion of heparin before selection on FGF-2.Consistent with previous observations, several hexasaccharide species areidentified in the mixture including nona-, octa- and heptasulfated hexas-accharides (1731.7, 1652.1 and 1573.2) and hepta- and hexasulfated,singly acetylated oligosaccharides (1613.6 and 1534.4). c, Selection ofFGF-2 binders from the hexasaccharide mixture. After selection with FGF-2, only the octa- and nonasulfated hexasaccharides (Hexa 3, 1652.2 andHexa 1, 1732.0) are observed. d, Application of heparinase I to saccha-rides selected by FGF-2 resulted in the disappearance of both hexasaccha-rides. In addition, there was the appearance of a trisulfated disaccharide(Di 1, 578.4) and a pentasulfated tetrasaccharide (Tetra 1, 1075.5). *, anH2SO4 adduct that arises from matrix contamination.


NEW TECHNOLOGY

tasaccharide binding sequence, was used to verify that AT-IIIwas immobilized on the surface and was able to bind saccha-rides. Penta 1 binding to AT-III could be observed with waterwashes and salt washes of up to 0.5 M (Fig. 2a), consistentwith Penta 1 being a strong binder to AT-III and AT-III re-maining intact during the immobilization procedure.Selectivity was confirmed by the fact that Penta 1 bound toAT-III under conditions where a different pentasaccharide(Penta 2)—identical to Penta 1 save for the absence of the es-sential 3-O sulfate—did not bind. To ensure that this wasthe result of specific binding, introduction of a mixture ofHexa 1, Hexa 2 and Penta 1 to immobilized AT-III followedby a 0.2 M salt wash resulted in a signal only for Penta 1 (Fig.2b and c). Interestingly, under these washing conditions,there was no signal from Hexa 2, which contains a partiallyintact AT-III binding site, again suggesting that, under ourselection conditions, only sequences with a full binding sitewill be selected as specific binders. In addition, even ahighly charged saccharide like Hexa 1 is not retained after a0.2M salt wash, despite the fact that it contains a largernumber of sulfates (3 sulfates per disaccharide unit). Thus,SNA-MS is able to differentiate between specific and non-specific ionic interactions, even allowing the isolation andmass analysis of specific binders within a complex saccha-ride mixture.

Isolation of specific oligosaccharide binders to FGF-2To ensure the binding and wash conditions developed withthe AT-III system could be extended, we chose to investigatethe FGF system using its prototypic member, FGF-2. FGF-2 isknown to bind HLGAGs avidly, and this binding at the cellsurface is critical for proper FGF-2 function. As with the AT-IIIsystem, the optimal procedure used 30 pmol (~ 0.5 µg) of pro-tein and 5–20 pmol of saccharide. Initial experiments in-volved the use of a purified oligosaccharide (Hexa 1 of Table1) that is known to bind with high affinity to FGF-2. Wefound that Hexa 1 binds to FGF-2 and can be detected, evenwith a salt wash of 0.5M NaCl, consistent with the knownaffinity of Hexa 1 for FGF-2. Heat-denatured FGF-2 did notbind Hexa 1, further indicating that the protein–saccharideinteraction is specific under these conditions. In addition,when an equimolar mixture of Hexa 1 and Hexa 2 (a low-affinity binder) was applied to FGF-2 and washed with 0.2 MNaCl to eliminate nonspecific binding, only Hexa 1 was ob-served (Fig. 3a). Together, these results point to the fact that,under the conditions of the experiment, immobilized FGF-2retains the same binding specificity as FGF-2 in solution. Inconjunction with the AT-III results of the first example, itseems that a 0.2 M NaCl wash can be broadly applied to re-move low-affinity binders from HLGAG binding proteins.

As with the AT-III system, we determined with FGF-2

a b c

Fig. 2 Binding of HLGAG oligosaccharides to AT-III. AT-III was biotiny-lated and immobilized on an avidin-coated plate. All saccharides are ob-served as 1:1 non-covalent complexes with the basic peptide (RG)19R. Theindicated masses are normalized by subtraction of the (RG)19R contributionto the 1:1 complex, while the x-axis represents the observed mass of thecomplex. All observed masses were within 3.2 units of the theoretical massof Penta 1 (1508.2) and Hexa 1 (1732.4). This resolution is sufficient for

unambiguous saccharide identification. a, Penta 1 (20 pmols) was addedto AT-III, washed with 0.2M NaCl and allowed to dry. The major peak(1507.9) represents Penta 1. Penta 1 was then mixed with an equimolaramount of Hexa 1 and Hexa 2 and spotted on AT-III. b, A water wash indi-cated that there was some non-specific binding as demonstrated by thepresence of Hexa 1 (1730.9). c, Washing with a solution of 0.2 M NaCleliminated non-specific binding, leaving only Penta 1 (1506.9).

a b c d

©20

01 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://m

edic

ine.

nat

ure

.co



NEW TECHNOLOGY

whether specific binders could be selected from a more com-plex mixture. In this case we used a size-fractionated hexasac-charide pool derived from incomplete heparinase-I digestionof porcine intestinal mucosa heparin19. The hexasaccharidepool was spotted on immobilized FGF-2. At least five uniquestructures were detected in the hexasaccharide mixture beforeselection (Fig. 3b). Upon a 0.2 M salt wash, only two struc-tures, the octa- and nonasulfated hexasaccharides (Hexa 3and Hexa 1 of Table 1, respectively), remained (Fig. 3c, seebelow for sequence information).

Sequencing of protein-bound oligosaccharidesTo test whether we could derive structural information for FGFhigh-affinity binders on the surface, we subjected Hexa 1 andHexa 3 to enzymatic depolymerization by heparinase I directlyafter selection (Fig. 3d). Importantly, enzymatic depolymeriza-tion appeared to proceed efficiently even in the presence ofFGF-2, indicating that it is possible to derive sequence informa-tion from bound oligosaccharides. The nonasulfated hexasac-charide was reduced to a single trisulfated disaccharide,indicating that this saccharide is a repeat of the disaccharideunit I2SHNS,6S (Fig. 3d and Di 1, Table 1). Digestion of the octasul-fated hexasaccharide yielded a trisulfated disaccharide (Di 1,Table 1) and a pentasulfated tetrasaccharide (Tetra 1, Table 1).Heparinase III cleavage, which resulted in the disappearance ofTetra 1, confirmed that this tetrasaccharide contains an unsul-fated uronic acid in the reducing end-disaccharide unit. Otherenzymatic and chemical constraints can also be used to derivestructural information, including the exoenzymes that have

previously been shown to be effective in sequence determina-tion with MALDI-MS (ref. 12). Thus, through judicious applica-tion of enzymatic constraints, we find that the sequence ofHexa 1 is ∆U2SHNS,6SI2SHNS,6SI2SHNS,6S and the sequence of the Hexa3 is ∆U2SHNS,6SI2SHNS,6SGHNS,6S. Our sequencing assignments wereconfirmed by isolation and sequencing of Hexa 1 and 3. In thiscase, compositional analysis indicated that Hexa 3 consists oftwo trisulfated disaccharides (∆U2SHNS,6S) and one disulfated dis-accharide (∆UHNS,6S). Both the short and long heparinase I di-gests yielded a trisulfated disaccharide (∆U2SHNS,6S) and apentasulfated tetrasaccharide, which must be ∆U2SHNS,6SGHNS,6S

in accordance with the known substrate specificity of hepari-nase I. Heparinase III digestion resulted in the formation of ahexasulfated tetrasaccharide (∆U2SHNS,6SI2SHNS,6S) and a disulfateddisaccharide (∆UHNS,6S). Together, these results allow us to con-verge on a sequence of ∆U2SHNS,6SI2SHNS,6SGHNS,6S for Hexa 3.

The sequences determined for hexa 1 and 3 are consistentwith the known sequence specificity of FGF-2 in that the twobinding oligosaccharides are found to differ only in their re-ducing end-disaccharide unit. As observed in the crystal struc-ture of an FGF-2–HLGAG hexasaccharide complex14, thenon-reducing end tetrasaccharide ∆U2SHNS,6SI2SHNS,6S is essentialfor high-affinity binding.

SNA-MS analysis of smooth muscle cell-derived HLGAGsHeparan sulfate at the cell surface of vascular smooth musclecells (SMCs) is known to contain high-affinity FGF-2 bindingsites20. To demonstrate that we indeed can extend our studiesto cell-derived HLGAGs, we sought to identify FGF-binding

MethodsProtein preparation and immobilization. Recombinant FGF-2was prepared as described21. Wild-type AT-III was a gift from R.Rosenberg. AT-III was incubated overnight with excess porcinemucosal heparin (Celsus Lab, Cincinnati, Ohio), then biotinylatedwith EZ-link sulfo-NHS biotin (Pierce, Rockford, Illinois). Excess bi-otin was removed by spin column with a molecular weight cutoff(MWCO) of 10,000 (Millipore). Canon NP Type E transparencyfilm was taped to the MALDI sample plate and used as a proteinimmobilization surface22. AT-III was immobilized by first drying 4 µg neutravidin (Pierce) on the film surface, then adding 30pmols biotinylated AT-III to the neutravidin spot. Heparin was re-moved by washing 10 times with 1 M NaCl and 10 times withwater. FGF-2 was immobilized by spotting 1 µl of a 30 pmol/µlaqueous solution on the film and air-drying. For some experi-ments, FGF-2 was denatured by heating at 80 °C for 30 min.

Saccharide binding, selection and analysis. Penta 1 and 2 weregifts from R. Rosenberg. Other saccharides were derived from apartial digest of porcine mucosal heparin by heparinase I as re-ported19. The hexasaccharide fraction was obtained by size exclu-sion chromatography on Biogel P-6 and lyophilized to dryness.Saccharides were bound to immobilized proteins by spotting 1µlof aqueous solution of 5 to 20 pmol/µl on the protein spot for atleast 5 min. Unbound saccharides were removed by washingwith water 15 times. For selection experiments, the spot waswashed 10 times with NaCl concentrations ranging from 0.2 to 1 M, followed by 10 water washes. Caffeic acid matrix in 50%acetonitrile (saturated solution) with 2 pmol/µl (RG)19R wasadded to the spot before MALDI analysis. All saccharides were de-

tected as non-covalent complexes with (RG)19R using MALDI pa-rameters as described12.

Saccharide digestion by heparinase I or III. Saccharides se-lected for FGF-2 binding were digested with heparinases I or III byspotting 8 ng of enzyme in water after selection was completed.The spot was kept wet for the desired digestion time by addingwater as necessary. Caffeic acid matrix with 2 pmol/µl (RG)19Rwas added to the spot for MALDI analysis.

Isolation, purification and selection of FGF binders from SMCheparan sulfate. Bovine aortic smooth muscle cells were grownto confluence in T75 flasks. Cells were washed twice with PBS andthen 200 nM heparinase I or III was added for 1 h. The super-natant was heated to 50 °C for 10 min to inactivate heparinase,and the sample was filtered. To remove polynucleotide contami-nation, the samples were treated with DNase and RNase (Roche)at room temperature overnight. Heparan sulfate was isolated bybinding to a DEAE filter (Millipore), washing away unbound ma-terial, and eluting with 10 mM sodium phosphate 1 M NaCl pH6.0. The material was then concentrated using a 3,000 MWCOmembrane (Millipore) and buffer exchanged into water. The re-tentate was lyophilized and reconstituted in water.Compositional analysis was performed on each sample as de-scribed12,23. Heparinase II (100 nM) was added, allowed to reactat room temperature, and aliquots were taken at 5, 10, 20 and 30min post-addition. The reaction mixture (1 µl) was spotted onimmobilized FGF. After drying, the sample was washed, 2pmol/µl of the (RG)19R in matrix was added, and the sample wasanalyzed as outlined above.

©20

01 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://m

edic

ine.

nat

ure

.co



NEW TECHNOLOGY

oligosaccharides at the cell surface of SMCs. To this end, SMCswere treated with either heparinase I or heparinase III to releaseheparan-sulfate oligosaccharide fragments from the cell sur-face. This procedure resulted in the isolation and purificationof approximately 100 ng of cell-surface HLGAGs (Fig. 4a).Consistent with the known substrate specificity of heparinasesI and III, the composition of released fragments is different ineach case. Heparinase I treatment of SMC heparan sulfate pro-teoglycans released oligosaccharide fragments that are less sul-fated (Fig. 4b and d), whereas heparinase III treatment releasedmore highly sulfated fragments (Fig. 4c and d). In both cases,fragments were then treated with heparinase II to reduce themto minimal binding size before being subjected to the surfaceaffinity assay. At given time points in the heparinase-II diges-tion, fragments were spotted on FGF-2 and binders were se-lected. A single hexasaccharide (Hexa 1) was identified as ahigh-affinity binder for FGF-2 only from the heparinase-III–de-rived cell-surface HLGAGs (Fig. 4e). Importantly, this oligosac-charide was not found in heparinase-I–derived cell-surfaceHLGAGs. This observation is consistent with the known actionof heparinase I, which readily cleaves this hexasaccharide andother highly sulfated regions, as is evidenced in the composi-tion data (Fig. 4b and d). Moreover, we observe that, in the SMCHLGAG preparations, no Hexa 3 is present—a binder that isfound in the heparin-derived hexasaccharide pool.

In summary, we describe a powerful methodology to isolate,enrich and sequence tissue-derived HLGAGs that bind to spe-cific proteins. The generalized procedure involves immobiliza-tion of a protein on a hydrophobic surface either directly orthrough a biotin-avidin system. HLGAGs with high affinity forthe protein can be selected by applying a saccharide mixture tothe immobilized protein (1:6 to 1:1 saccharide:protein ratio),followed by a 0.2 M salt wash to remove non-specific binders.Bound saccharides can then be directly analyzed by mass spec-trometry. The methodology outlined here has a number of crit-ical advantages over available methods. First, the analysis canbe performed with picomoles of material for both the proteinand HLGAG oligosaccharides. Second, it is possible to derive se-

quence information from the bound HLGAG oligosaccharidesdirectly on target. Third, this method makes it feasible to useHLGAGs isolated from the cell surface, rather than highly sul-fated heparin or the ‘universal’ HLGAG from mast cells.

The broad applicability of SNA-MS methodology makes itpossible to examine how dynamic changes in cell-surfaceHLGAG composition and sequence influences ability of cells toalter responses to signaling molecules that impinge on the cel-lular phenotypic states. Just as mass spectrometry has advancedthe realm of proteomics, providing a ready methodology toprobe cell function and discover novel drug targets, SNA-MSshould provide a foundation for elucidating key polysaccha-ride–protein interactions.

AcknowledgmentsThis study was supported in part by funds from the Burroughs WellcomeFoundation, Arnold and Mabel Beckman Foundation, National Institutes ofHealth [R01HL59966], National Science Foundation Fellowship andWhitaker Health Sciences Fund Fellowship. We thank C.-P. Kwan for recombi-nant FGF-2, D. Berry for help with cell culture, K. Pojasek for generation ofthe hexasaccharide library and V. Sasisekharan for critical reading of themanuscript.

1. Lin, X. & Perrimon, N. Dally cooperates with Drosophila Frizzled 2 to transduceWingless signalling. Nature 400, 281–284 (1999).

2. Tsuda, M. et al. The cell-surface proteoglycan Dally regulates Wingless signalling inDrosophila. Nature 400, 276–280 (1999).

3. Vlodavsky, I. et al. Mammalian heparanase: Gene cloning, expression and function intumor progression and metastasis. Nature Med. 5, 793–802 (1999).

4. Hulett, M.D. et al. Cloning of mammalian heparanase, an important enzyme intumor invasion and metastasis. Nature Med. 5, 803–809 (1999).

5. Sasisekharan, R., Moses, M.A., Nugent, M.A., Cooney, C.L. & Langer, R. Heparinaseinhibits neovascularization. Proc. Natl. Acad. Sci. USA 91, 1524–1528 (1994).

6. Petitou, M. et al. Synthesis of thrombin-inhibiting heparin mimetics without side ef-fectsΩ. Nature 398, 417–422 (1999).

7. Perrimon, N. & Bernfield, M. Specificities of heparan sulphate proteoglycans in de-velopmental processes. Nature 404, 725–728 (2000).

8. Conrad, H.E. Heparin-Binding Proteins (Academic Press, San Diego, California, 1998).9. Tumova, S., Woods, A. & Couchman, J.R. Heparan sulfate proteoglycans on the cell

surface: versatile coordinators of cellular functions. Int. J. Biochem. Cell. Biol. 32,269–288 (2000).

10. Atha, D.H., Stephens, A.W., Rimon, A. & Rosenberg, R.D. Sequence variation in he-parin octasaccharides with high affinity for antithrombin III. Biochemistry 23,5801–5812 (1984).

SMC HLGAG HepI HepIII Isolation/Purification Hep II to generate minimal binding oligosaccharides

∆U2SHNS,6S

∆U2SHNS

∆UHNS

∆UHNAc,6S

∆U2SHNS,6S

∆U2SHNS

∆UHNS,6S

∆UHNS

Sacch. code Hep

I Hep III

D ∆U2S-HNS,6S 5 16 9 ∆U2S-HNS 22 30 5 ∆U-HNS,6S 6 8 C ∆U2S-HNAc,6S 0 0 1 ∆U-HNS 31 20 8 ∆U2S-HNAc 0 0 4 ∆U-HNAc,6S 8 7 0 ∆U-HNAc 28 19

Fig. 4 Identification of a FGF-2 high-affinity binder in the heparan sulfate oligosaccharides from cell surface-derived proteoglycans from bovine aortic smooth muscle cells. a, Outline of the strategy employed to identifybinders. Compositional analysis using capillary electrophoresis of heparinase III- or heparinase I-generatedSMC heparan sulfate indicated that the samples had different compositions, consistent with the known substrate specificities of the enzymes. b, Heparinase III cleaves the undersulfated regions of HLGAGs leavingbehind oligosaccharides primarily comprised of tri-, di-, and monosulfated disaccharides. c, Conversely, heparinase I cleaves the highly sulfated regions of HLGAGs leaving behind oligosaccharides primarily comprised of di-, mono-,and nonsulfated disaccharides. d, Table summarizing the compositional data of (b)and (c). The amounts of each disaccharide are presented as a percentage of total disaccharide. e, With SMC heparan sulfate derived from heparinase III digestion, a single hexasaccharide binder is observed, viz., Hexa 1.As with the other samples, Hexa 1 is observed as a 1:1 non-covalent complex with the basic peptide (RG)19R.The indicated mass is normalized by subtraction of the (RG)19R contribution and is within 1 D of the theoretical mass of Hexa 1. The x-axis represents the observed mass of the peptide–saccharide complex.

a b c d

e

©20

01 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://m

edic

ine.

nat

ure

.co



NEW TECHNOLOGY

11. Kreuger, J., Prydz, K., Pettersson, R.F., Lindahl, U. & Salmivirta, M. Characterizationof fibroblast growth factor 1 binding heparan sulfate domain. Glycobiology 9,723–729 (1999).

12. Venkataraman, G., Shriver, Z., Raman, R. & Sasisekharan, R. Sequencing complexpolysaccharides. Science 286, 537–542 (1999).

13. Rhomberg, A.J., Ernst, S., Sasisekharan, R. & Biemann, K. Mass spectrometric andcapillary electrophoretic investigation of the enzymatic degradation of heparin-likeglycosaminoglycans. Proc. Natl. Acad. Sci. USA 95, 4176–4181 (1998).

14. Faham, S., Hileman, R.E., Fromm, J.R., Linhardt, R.J. & Rees, D.C. Heparin structureand interactions with basic fibroblast growth factor. Science 271, 1116–1120 (1996).

15. DiGabriele, A.D. et al. Structure of a heparin-linked biologically active dimer of fi-broblast growth factor. Nature 393, 812–817 (1998).

16. Faham, S., Linhardt, R.J. & Rees, D.C. Diversity does make a difference: Fibroblastgrowth factor-heparin interactions. Curr. Opin. Struct. Biol. 8, 578–586 (1998).

17. Jin, L. et al. The anticoagulant activation of antithrombin by heparin. Proc. Natl.Acad. Sci. USA 94, 14683–14688 (1997).

18. Fry, E.E. et al. The structure and function of a foot-and-mouth disease virus-oligosac-charide receptor complex. EMBO J. 18, 543–554 (1999).

19. Pervin, A., Gallo, C., Jandik, K.A., Han, X.J. & Linhardt, R.J. Preparation and structuralcharacterization of large heparin-derived oligosaccharides. Glycobiology 5, 83–95(1995).

20. Sperinde, G.V. & Nugent, M.A. Heparan sulfate proteoglycans control intracellularprocessing of bFGF in vascular smooth muscle cells. Biochemistry 37, 13153–13164(1998).

21. Padera, R., Venkataraman, G., Berry, D., Godavarti, R. & Sasisekharan, R. FGF-2/fi-broblast growth factor receptor/heparin-like glycosaminoglycan interactions: Acompensation model for FGF-2 signaling. FASEB J. 13, 1677–1687 (1999).

22. Wang, H., Tseng, K. & Lebrilla, C.B. A general method for producing bioaffinityMALDI probes. Anal. Chem. 71, 2014–2020 (1999).

23. Brickman, Y.G. et al. Structural comparison of fibroblast growth factor-specific he-paran sulfates derived from a growing or differentiating neuroepithelial cell line.Glycobiology 8, 463–471(1998).

ON THE MARKET

THERMAL CYCLERS

Optimize experiments at the touch of ascreen, says Techne, using its Touchgenegradient thermal cycler. The device of-fers a gradient range of 30 °C and ‘Quad’circuit technology, where four indepen-dent temperature sensors control eightpeltier units. The graphical display showsthe sample temperature profile in real timewhile the program is running, includingthe upper and lower temperature limits ofthe gradient. Special gradient features thatenable users to replicate experimental con-ditions include a gradient calculator func-tion that displays the actual temperatureof each column, and active ramping andfourfold control that ensures that eachsample is at the target temperature for thesame hold time. Other design features in-clude multiple interchangeable block for-mats, memory cards and a maximumramp rate of greater than 3 °C/s.Tel. (+44) (0) 1223-832401Fax (+44) (0) 1223-836838www.techneuk.co.uk

ONLINE SHOPPING

BioSupplies, a provider of e-commercesolutions for the life sciences, haslaunched PurePeptides. PurePeptides is anInternet-based tool that provides re-searchers with a convenient way to orderpeptides. The service enables users to builda peptide or protein online and select froman extensive range of modifications andoptions. PurePeptides then enables the

user to view a list of potential manufactur-ers, check their prices and order the prod-uct online. PurePeptides complements thecompany’s launch of PureDNA.Tel. (+1) 215-387-1689Fax (+1) 215-895-3101Website: www.biosupplies.com

SIMPLIFYING SEQUENCING SETUP

Beckman Coulter’s new kit — the CEQDTCS Quick Start Kit — simplifies dye ter-minator cycle sequencing reaction setup.The kit features a master mix of compo-nents that reduces the required number ofpipetting steps from ten to four, producingsignificant time savings, according to thecompany. At the same time, the kit useslarger transfer volumes to reduce pipettingerrors. The kit is designed to work with theCEQ 2000XL DNA sequencing system. Forcomplete automation of reaction setup,the kit can be used with the Biomek FX andBiomek 2000 automated liquid handlingsystems, also from Beckman Coulter.Tel. (+1) 714-871-4848Fax (+1) 714-773-6611www.beckmancoulter.com

LIQUID HANDLING SOLUTIONS

Matrix Technologies has introduced thePlateMate Plus, a high-thoughput sys-tem with interchangeable pipettingheads. Each platform offers the ability towork with 96-, 384- and/or 1,536-wellplates using either a 96- or 384-tip con-figured head for liquid processing. In ad-dition to precise reagent addition,mother/daughter plate replication andplate pooling applications, the PlateMatePlus can also accomplish serial dilutionsand microplate reformatting between 96-,384- and 1,536-well plates. Availablewith removable front-loading universalstackers and ControlMate Windows-based software.Tel (+1) 800-345-0206www.matrixtechcorp.com

NOT A CASE OF ‘ALL HANDS TOTHE PUMP’

The HandSaver microemulsion pumpfrom Stoelting automates the labora-tory task of mixing adjuvant/antigenemulsions by forcing the fluid back andforth through a double-hub needle.With the HandSaver device, simplymount the syringes on the pump andreturn when thorough emulsions areachieved, says the company.Tel. (+1) 630-860-9700Fax (+1) 630-860-9775www.stoeltingco.com

Touchgene gradient thermal cycler.

Beckman’s new kit simplfies reactionsetup for DNA sequencing.

MatrixPlateMate Plus.

HandSaver microemulsion device takes theeffort out of mixing adjuvant antigenemulsions.

©20

01 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://m

edic

ine.

nat

ure

.co


Documents

document