homophilic binding properties of galectin-3: involvement of the carbohydrate recognition domain

Journal ofNeurochemistryLippincott—Raven Publishers, Philadelphia© 1998 International Society for Neurochemistry

Homophilic Binding Properties of Galectin-3: Involvement ofthe Carbohydrate Recognition Domain

Stephan Kuklinski and Rainer Probstmeier

Department of Biochemistry, Institute for Animal Anatomy and Physiology, University of Bonn, Bonn, Germany

Abstract: Galectin-3, an animal lectin specific for fl-ga-lactosides, is composed of three different domains. TheN-terminal half of the molecule (N domain) consists of ashort N-terminal segment followed by glycine-, proline-,and tyrosirie-rich tandem repeats. The C-terminal domain(C domain) harbors the carbohydrate recognition domainhomologous to other members of the galectin family oflectins. Galectin-3 aggregates in solution, and participa-tion of the N domain of the molecule in this process hasalready been demonstrated. Using a solid-phase radioli-gand binding assay, which allows the direct analysis ofgalectin-3 self-association, here we provide evidence thatthe carbohydrate recognition domain of the lectin is in-volved in carbohydrate-dependent homophilic interac-tions: (a) Radiolabeled galectin-3 binds to immobilizedgalectin-3, and the addition of unlabeled galectin-3 insolution increases the rate of binding of radiolabeled lec-tin; (b) binding of radiolabeled galectin-3 to immobilizedgalectin-3 is inhibited by the C domain; (c) binding ofradiolabeled galectin-3 to immobilized galectin-3 or theC domain is inhibited by lactose but not by sucrose; and(d) the radiolabeled C domain does not bind to immobi-lized C domain. Taken together, these data suggest thatin addition to the N domain, the homophilic interac-tions of galectin-3 are mediated by the C domain. KeyWords: Galectin-3— Lectins—N-terminal domain—C-terminal domain.J. Neurochem. 70, 814—823 (1998).

Galectin-3, previously designated CBP-30, CBP-35,L-29, or Mac-2, is a member of the galectin family offl-galactoside-specific animal lectins (Barondes et al.,1994). According to their domain structure, galectinscan be subdivided into three subgroups: (a) prototype,(b) tandem-repeat type, and (c) chimera type (Kasaiand Hirabayashi, 1996). Prototype galectins with amolecular mass of ‘-~ 15 kDa contain one carbohydraterecognition domain (C domain), and some of themform noncovalently linked homodimers. Tandem-re-peat-type galectins contain two C domains in one poly-peptide chain. Chimera-type galectins contain one Cdomain in the C-terminal half of the molecule, whilethe N-terminal half shows no homology to other typesof galectins. In mammals, galectin-3 is the only chi-mera-type lectin so far known. A great part of the N-

terminal domain of galectin-3 is built up by repetitivesequences of nine amino acids rich in glycine, proline,and tyrosine and homologous to proteins of the hetero-geneous nuclear ribonucleoprotein particles. Depen-dent on the species, the number of these repetitivesequences varies, leading to molecular masses between26 and 30 kDa, although the apparent molecularmasses after sodium dodecyl sulfate—polyacrylamidegel electrophoresis (SDS-PAGE) are higher (Barondeset al., 1994). The very N-terminal part of galectin-3(up to amino acid 40) has been suggested to representa third individual domain as it shows homology to thetranscription activator serum response factor (Oda etal., 1991). Throughout this study we will refer to bothdomains together as an N-terminal domain (N do-main). The repetitive sequence within the N domainof galectin-3 is sensitive to treatment with collagenasesand thus allows the isolation of the C domain (gal-3C) (Hsu et al., 1992; Agrwal et al., 1993). Based onthe data in the literature, it can be concluded that theC and N domains of galectin-3 fold and function inde-pendently (Agrwal et al., 1993; Ochieng et al., 1993).The domain structure of galectin-3 is also reflectedby the composition of the conesponding gene, whichcontains six exons and five introns (Gritzmacher et al.,1992). Exon I encodes an untranslated region, while•exon II encodes an untranslated region, the proteintranslation initiation site, and the first six amino acidsincluding the initial methionine. The complete N-ter-minal domain of the molecule is encoded by exon III,and exons Tv—VT code for the C domain.

Galectin-3 is expressed in a number of different tis-sues and cell types (Wang et al., 1991). At the cellular

Received July 2, 1997; revised manuscript received September23, 1997; accepted September 23, 1997.

Address correspondence and reprint requests to Dr. R. Probstmeierat Department of Biochemistry, Institute of Animal Anatomy andPhysiology, Rheinische Friedrich-Wilhelms-Universitat Bonn, Kat-zenburgweg 9A, D-53115 Bonn, Germany.

Abbreviations used: ASF, asialofetuin; BSA, bovine serum albu-min; gal-3C, C domain of galectin-3; LPS, Iipopolysaccharide;MAG, myelin-associated glycoprotein; PBS, phosphate-buffered sa-line; SDS-PAGE, sodium dodecyl sulfate—polyacrylaniide gel elec-trophoresis; TBS, Tris-buffered saline; TBS/hem, TBS plus 10%hemoglobin and 2% BSA.

814

HOMOPHILIC BINDING OF GALECTIN-3 815

level, the protein has been found localized intracellu-larly (cytoplasm and nucleus) and extracellularly (onthe cell surface and in the extracellular matrix) (Wanget al., 1991). Like other galectins, galectin-3 does notcontain a secretion-specific signal peptide; however,an alternative secretion mechanism has been described(Lindstedt et al., 1993; Sato et al., 1993). Consideringthese different subcellular localizations, galectin-3 isnot supposed to display a distinct biological function.Thus, the molecule has been shown to participate inphenomena as different as cell adhesion (Sato andHughes, 1992) and pre-mRNA splicing (Dagher et al.,1995). In the developing and adult CNS, the expres-sion of galectin-3 is restricted toprojections of subpop-ulations of dorsal root ganglion neurons into the spinalcord (Regan et al., 1986), while under different patho-logical conditions, the lectin becomes expressed byactivated microglia, motoneurons and glioma cells(Probstmeier et al., 1996; Pesheva et al., 1997; S.Urschel et al., submitted for publication). We haverecently shown that different neural tissue-derived celladhesion molecules expressed at the cell surface or inthe extracellular matrix can serve as ligands for galec-tin-3 (Probstmeier et al., 1995). Furthermore, we havedemonstrated that binding of galectin-3 to the myelin-associated glycoprotein (MAG; Quarles, 1989) is car-bohydratedependent but notsaturable under the exper-imental conditions used, suggesting a self-associationof galectin-3 (Probstmeieret al., 1995). These data arein agreement with other observations on the binding ofgalectin-3 to IgE (Hsu et al., 1992) and laminin(Massa et al., 1993). Both reports provide evidencethat the di-/oligomerization of galectin-3 in solutionis dependent on the presence of the N domain as (a)unlabeled galectin-3 potentiates and unlabeled gal-3Cinhibits binding of radiolabeled galectin-3 to IgE and(b) galectin-3, but not gal-3C, causes hemagglutina-tion of erythrocytes. Although di-/oligomerization ofgalectin-3 occurs in solution (Hsu et al., 1992), thisprocess seems to be favored in the presence of “crys-tallization points” provided by galectin-3 moleculesbound to immobilized f3-galactoside carrying glyco-proteins (Massa et al., 1993).

The di-/oligomerization of galectin-3 described sofar is not dependent on the formation of S — S-linkedgalectin-3 dimers (Woo et al., 1991; Ochieng et al.,1993), as it also occurs in the presence of reducingagents. Covalently linked homodimers of galectin-3are formed at relatively high concentrations of the lec-tin via oxidation of cysteine 186 (Woo et al., 1991;Ochieng et al., 1993). By contrast, the molecularmechanism underlying the noncovalent di-! oligomer-ization of galectin-3 remains largely obscure. In thepresent study, we describe a solid-phase radioligandbinding assay that allows the direct analysis of homo-philic galectin-3 interactions. We demonstrate that theC domain of galectin-3 is involved in carbohydrate-dependent homophilic binding events as self-associa-

tion can be inhibited by lactose but not by irrelevantcarbohydrates.

MATERIALS AND METHODS

Purification and sources of proteins and otherreagents

Recombinant galectin-3 was expressed from plasmidprCBP35s in Escherichia coli and affinity purified on a lac-tose—agarose column as described (Agrwal et al., 1993;Probstmeier et al., 1995). Gal-3C was obtained after colla-genase treatment of galectin-3 and affinity purification onlactose—agarose as described (Agrwal et al., 1993). MAGwas immunoaffinity purified from adult mouse brain as de-scribed (Poltorak et al., 1987). Lipopolysaccharide (LPS)mixture from E. coli was from Difco Laboratories, and LPSfrom Salmonella minnesota R7 (Rd mutant) and S. minne-sota Re 595 (Re mutant) were from Sigma. Laminin andcollagenase D were from Boehringer Mannheim, and asialo-fetuin (ASF), lactose—agarose, latex beads (LB-30), andlectin from Erythrina cristagalli were from Sigma.

Radioactive labeling of galectin-3 and gal-3CProtein (100 to 200 ~sg)in 200—300 p3 solutions was

dialyzed overnight against 100 mM Na2HPO4/NaH2PO4,100 mM NaC1, and 1 mM /3-mercaptoethanol (pH 8.0),added to 1 mCi dried ‘

251-Bolton—Hunter reagent (Amer-sham; Bolton and Hunter, 1973), and incubated for 1 h at4°C.After addition of 50 i.tl of 1 M glycine and 1 mM /3-

mercaptoethanol (pH 8.0), protein solutions were incubatedfor a further 15 mm at 4°Cand dialyzed against phosphate-buffered saline (PBS; 50 mMNa

2HPO4/NaH2PO4, 100mMNaCI, pH 7.2) for 2 days at 4°Cwith four buffer changes.After dialysis, >90% of the total radioactivity could be pre-cipitated with trichioroacetic acid. The specific activity wasin the range of 0.2—1.0 x 106 cpmI~Lgof protein.

Solid-phase radioligand binding assayThe solid-phase radioligand-binding assay has been de-

scribed elsewhere (Probstmeier et al., 1995). Briefly, indi-vidual wells of 96-well Microtest III flexible assay plates(Falcon) were incubatedovernight at 4°Cwith different pro-teins (10—50 ~ig/ml in 100 mM NaHCO3, pH 8.1; 70 p3/well), 2md remaining free binding sites were blocked with1% bovine serum albumin (BSA) in 100 mM NaHCO5 (pH8.1). After washing with PBS containing 1% BSA (PBS-BSA), wells were incubated with radiolabeled proteins (di-luted in PBS-BSA) for 5 h at roomtemperature. Wells weresubsequentlywashed with PBS-BSA; thebottom of thewellswas cut off and bound radioactivity was counted directly.For SDS-PAGE analysis of the bound radioactive molecularspecies, individual wells were incubated for 15 mm withSDS—sample buffer (Laemmli, 1970). Under these condi-tions, 85—90% of the radioactivity bound could be removedfrom the plastic. Samples were boiled for 5 mm and sepa-rated by SDS-PAGE in 12% slab gels, and dried gels wereexposed to x-ray films (Kodak) at —70°C.

To determine the coating efficiency, ‘251-labeled galectin-

3 or gal-3C was used as a tracer, mixed with unlabeledprotein at a ratio of 1:100 and coated as described previouslyat coating concentrations of 50 (for galectin-3) or 25 (forgal-3C) ~.tg/ml.Under these conditions, the coating efficien-cies of ‘251-labeled galectin-3 or gal-3C were comparable

J. Neurochem., Vol. 70. No. 2, 1998

816 S. KUKLINSKI AND R. PROBSTMEIER

FIG. 1. Binding of 1251-galectin-3 to laminin (LN), MAG, ASF,galectin-3 (GAL-3), or BSA as determined by solid-phase radioli-gand binding assay. Proteins were coated at a concentration of10 ~ig/ml[LN, MAG, ASF, and GAL-3(10)] or 50 pg/mI, andradiolabeled lectin (8.0 x io~cpm, 70 p1/well, which corre-sponds to 130 ng ofprotein) was incubated with the immobilizedproteins in the absence of carbohydrates or in the presence of10 mM lactose or sucrose. Values are given as means ±SD oftriplicate measurements from one representative of three inde-pendent experiments.

(~-~5%of the input material); i.e., comparable molaramounts of both molecules were immobilized.

Overlay assay with 1251-galectin-3Proteins separated by SDS-PAGE on 9% slab gels were

transferred onto nitrocellulose (Towbin et al., 1979) andvisualized by Ponceau S staining (Fahrig et al., 1987). Afterdestaining in PBS, nitrocellulose filters were blocked withTris-buffered saline (TBS; 50 mM Tris-HC1, 100 mM NaC1,pH 7.2), containing 10% hemoglobin and 2% BSA (TBS/hem) overnight at 4°C.1251-Galectin-3 (diluted in TBS/hem,1.0 x 106 cpm./ml) was incubated with the filter for 5 h atroom temperature. Filters were washed four times for 5 mmwith TBS/hem and four times for 5 mm with PBS-BSA, airdried, and analyzed by autoradiography.

Latex bead aggregation assayFifty microliters of a suspension of latex beads was

washed twice in PBS and incubated overnight at 4°Cwith1 ml ofASF- or BSA-containing solution (200 ~.tgofprotein!ml in PBS). Beads were then washed twice with PBS andincubated for 4 h with PBS containing 1% BSA. Beads weresubsequently washed twice in PBS and stored as a 10%suspension (vol/vol) in PBS at 4°C.

For the aggregation assay, 30 p3 of each bead suspensionwas resuspended in 500 p1 PBS containing 2 mM /3-mercap-toethanol and 1% BSA in the absence or presence of 20 mMlactose or sucrose. Galectin-3 or lectin from E. cristagalliwas added to a final concentration of 50 pg/ml, andthe latexbead suspensions were incubatedfor I h at roomtemperatureon arotatory shaker. The resulting aggregation patterns weresubsequently analyzed by light microscopy.

RESULTS

To analyze the homophilic binding properties of ga-lectin-3, we used recombinant lectin and its C domain

(gal-3C), which can be isolated after collagenase di-gestion of galectin-3 (Agrwal et al., 1993).

Binding of galectin-3 to galectin-3 is inhibitabieby lactose

To examine whether self-association of galectin-3can be analyzed directly by a solid-phase radioligandbinding assay, different glycoproteins (ASF, laminin,and MAG) and galectin-3 were immobilized and incu-bated with ‘251-galectin-3 in the absence of carbohy-drates or in the presence of 10 mM lactose or sucrose.As described previously (Probstmeier et al., 1995),1251-galectin-3 bound to ASF, laminin, and MAG in alactose-inhibitable manner (Fig. 1). Binding of 125I.galectin-3 to galectin-3 could be observed at coatingconcentrations of 50 but not 10 ~.tg!ml (Fig. 1). Athigher concentrations, no further increase in the bind-ing of ‘251-galectin-3 was observed (not shown). Nota-bly, the binding of galectin-3 to galectin-3 could beinhibited by lactose, suggesting the involvement of theC domain in homophilic interactions. To rule out thepossibility that the observed effect is due to (a) con-taminating protein or LPS that may be present in ourrecombinantgalectin-3 preparations or (b) an artificialglycosylation of bacteria-derived galectin-3, we per-formed the following control experiments.

1251-Galectin-3 was incubated with immobilizedMAG or galectin-3, and radioactive material bound tothese proteins was analyzed by SDS-PAGE and autora-diography. The radiolabeled galectin-3 preparationused in these experiments consisted solely of a proteinwith a molecular mass of 29 kDa (Fig. 2, lane 1)which was also the only component present in theradioactive material eluted from MAG and galectin-3(Fig. 2, lanes 2 and 3). Recently, it has been demon-strated that galectin-3 binds to LPS via its N-terminaldomain (Mey et al., 1996). As the recombinant galec-tin-3 used in this study is derived from E. coli, wefurther investigated the influence of different LPSpreparations on the self-association of galectin-3. Nei-ther a mixture of LPS from E. coli nor defined LPSpreparations devoid (Rd mutant) or not devoid (Remutant) of /3-galactosides altered the binding efficien-cies of homophilic galectin-3 interactions (Fig. 3).

Prokaryotic organisms are also capable of proteinglycosylation (see, e.g., Messner and Sleytr, 1991).

FIG. 2. SDS-PAGE analysis in12% slab gels of 125l-galectin-3 used for binding (lane 1) andbound to immobilized MAG(lane 2) or galectin-3 (lane 3).Equal amounts of radioactivematerial/lane (—‘1.0 x 10~cpm)were loaded, and separatedprotein components were visu-alized by autoradiography. Theposition of galectin-3 (29 kDa)is indicated on the left.

.1. Neurochem., Vol. 70, No. 2, 1998


FIG. 3. Binding of 1251-galectin-3 to galectin-3 as determined bysolid-phase radioligand binding assay. Radiolabeled galectin-3(1.0 >< i0~cpm, 70 p1/well, which corresponds to 165 ng ofprotein) was incubated with immobilized galectin-3 at a coatingconcentration of 50 pg/mI in the absence of LPS or in the pres-ence of a mixture of [PS derived from E. co/i (LPS) and LPSfrom two S. minnesota mutants (Re or Rd) (see Materials andMethods). Values are given as means ±SD of triplicate mea-surements from one of two independent experiments. Numberson the x axis refer to concentrations (pg/mi) of LPS added.Background binding to BSA was in the range of 100 cpm.

To rule out the possibility that, in contrast to the non-glycosylated eukaryotic galectin-3, the recombinantgalectin-3 may be artificially glycosylated, galectin-3,ASF, and MAG were separated by SDS-PAGE, trans-ferred onto nitrocellulose, and incubated with ‘251-ga-lectin-3. All /3-galactoside-carrying glycoproteins ana-lyzed so far are still recognized by galectin-3 afterwestern blotting (see, e.g., Probstmeier et al., 1995).Under these experimental conditions, 125I-galectin-3bound to ASF and MAG but not to galectin-3 (Fig.4), suggesting that the recombinant galectin-3 carriesno /3-galactosidic or /3-galactose-like carbohydratestructures.

Unlabeled galectin-3 potentiates binding of‘251-galectin-3 to galectin-3

As unlabeled galectin-3 increases the binding effi-ciency of ‘25I-galectin-3 to MAG (Probstmeier et al.,1995), we were further interested in whether unlabeledgalectin-3 displays a similar effect on the homophilicgalectin-3 interactions as shown in Fig. 1. To examinethis question, immobilized galectin-3 was incubatedwith a constant amount of 1251-galectin-3 in the pres-ence of increasing amounts of unlabeled lectin (Fig.5A). The amount of 1251-galectin-3 bound to galectin-3 increased when unlabeled galectin-3 was added inup to a 10-fold excess and further decreased whenpresent in 20- or 50-fold excess. The absolute amountsof galectin-3 bound also increased, and no saturation ofbinding could be seen (Fig. 5B). Binding of galectin-3to galectin-3 could be specifically inhibited by lactose(Fig. 5A). In agreement with previous studies (Massaet al., 1993), binding of galectin-3 to laminin was

also potentiated in the presence of unlabeled galectin-3 (Fig. 5C and D).

The C domain of galectin-3 is directly involved inself-association of galectin-3

By the next set of experiments, we analyzed theinfluence of the C domain on the self-association ofgalectin-3. For this purpose, we used the isolated C-terminal half of galectin-3 (gal-3C), which harbors thelectin binding site of the molecule (Agrwal et al.,1993). When galectin-3 was immobilized and incu-bated with ‘251-galectin-3 either in the absence or inthe presence of unlabeled galectin-3 or gal-3C, strikingdifferences in the binding profiles were observed (Fig.6A). As already shown in Fig. 5A, binding of t25J..galectin-3 to immobilized galectin-3 increased at a 10-fold molar excess of unlabeled galectin-3. In the pres-ence of a 10-fold molar excess of unlabeled gal-3C,by contrast, binding of ‘251-galectin-3 to galectin-3 wasstrongly inhibited (Fig. 6A), suggesting that gal-3Ccan interact with galectin-3. Binding of 1251-gal-3C togalectin-3 and laminin showed only a weak tendency,if at all, to decrease in the presence of increasingamounts of unlabeled galectin-3 (Fig. 6B). Binding of‘251-galectin-3 to gal-3C was weaker than that of 1251.

galectin-3 to galectin-3 (see Fig. 1) when equimolarconcentrations were used for coating, but could be spe-cifically inhibited by lactose (Fig. 6C). These differ-ences are not due to different coating efficiencies (seeMaterials and Methods). At higher coating concentra-tions of gal-3C (40 ~tg/ml) and higher input valuesof ‘251-galectin-3, the amount of ‘251-galectin-3 boundincreased and signal/background (binding to BSA)ratios were in the range of 1.0:3.0 (not shown). Whenbinding of ‘251-gal-3C (1 ~.tg/well) to immobilized gal-3C (at a coating concentration 40 ,ug!mI) was ana-lyzed, no binding could be observed (signal/back-ground ratio 1.0:1.1—1.2).

ASF-coated latex beads aggregate in presence ofgalectin-3

As previously demonstrated, galectin-3 agglutinateserythrocytes in a lactose-dependent manner (Frigeriand Liu, 1992). This hemagglutination may resultfrom a trans bridging of either (a) glycostructures ex-pressed on opposed cell surfaces (galectin-3 dimers

FIG. 4. Binding of 1251-galectin-3 togalectin-3 (lane 1), ASF (lane 2), andMAG (lane 3) as determined by overlayassay. Proteins (20 pg/lane) wereseparated by SDS-PAGE in 9% slabgels, transferred onto nitrocellulose,and incubated with 1251-galectin-3(106cpm/ml). Radioactive material boundwas visualized by autoradiography.Apparent molecular masses (kDa) aregiven at the left margin.

J. Neurochem., Vol. 70, No. 2, 1998


FIG. 5. Binding of 1251-galectin-3 to galectin-3 (A and B) or laminin (C and D) as determined by solid-phase radioligand binding assay.Proteins were immobilized at a concentration of 50 (galectin-3) and 10 (laminin) pg/mI. A fixed amount of 1251-galectin-3/weIl (8>< 1o~cpm, which corresponds to 130 ng of protein) was added in the presence of increasing amounts of unlabeled galectin-3 (GAL-3)without (circles) or with (triangles) the inclusion of 10 mM lactose in4he solution. The data were plotted as follows: A and C: cpmbound vs. -fold excess of unlabeled galectin-3; B and D: total lectin bound (ng) vs. total lectin added (pg). Values are given as means±SD of triplicate measurements. Background binding to BSA was in the range of 100 cpm.

or multimers would be a prerequisite for this kind ofinteraction) or (b) aputative cell surface receptor (rec-ognized by a carbohydrate-independent cell bindingsite on galectin-3) and a glycostructure (recognizedby the carbohydrate recognition domain of galectin-3). Tn the latter case, a di-!oligomerization of galectin-3 molecules would not be necessary. To distinguishbetween these two possibilities, latex beads werecoated with ASF or BSA and incubated with galectin-3 or lectin from E. cristagalli (recognizing terminal/3-galactosides; Sharon, 1993) in the absence or in thepresence of 20 mM lactose or sucrose. To avoid the

formation of covalently linked galectin-3 dimers (Wooet al., 1991), the assay was carried out in the presenceof 2 mM /3-mercaptoethanol.

ASF-coated beads aggregated in the presence of ga-lectin-3 and lectin from E. cristagalli (Fig. 7A andD). The aggregation could be inhibited in the presenceof 20 mM lactose (Fig. 7B and E) but not in thepresence of 20 mM sucrose (Fig. 7C and F). ASF-coated beads did not aggregate in the absence of addi-tives (Fig. 7G), and BSA-coated beads did not aggre-gate in the presence of galectin-3 or lectin from E.cristagalli (Fig. 7H and I).



These data clearly demonstrate that galectin-3 formsdi- or multimers in solution that are able to mediatecell aggregation solely via a trans bridging of glyco-structures expressed on opposed cell surfaces.

DISCUSSION

FIG. 6. Characterization of the gal-3C-mediated homophilicbinding as determined by solid-phase radioligand binding assay.A: Galectin-3 (GAL-3) or BSA was immobilized and incubatedwith 1251-galectin-3 (8 x i0~cpm/well) in the absence or pres-ence of a 10-fold excess of unlabeled galectin-3 (lOx GAL-3)or gal-3C (lox GAL-3C). B: Galectin-3 (GAL-3), laminin (LN),or BSA was immobilized and incubated with 1251-gal-3C (8 x 1 O~cpm/well) in the absence or presence of a 5- or 10-fold excessof unlabeled galectin-3. C: Gal-3C or BSA was immobilized andincubated with 125l-galectin-3 (4x 1 Q4 cpm/well) in the absenceof carbohydrates or in the presence of 10 mM lactose or sucrose.In all experiments, proteins were immobilized at a concentrationof 50 (galectin-3) and 25 (gal-3C) pg/mi. Values are given asmeans ±SD of triplicate measurements.

In the present study, we have shown for the firsttime a direct interaction of galectin-3 molecules usinga solid-phase radioligand binding assay. Our studiesexclude the possibility that protein or LPS contami-nants and artificial glycosylation of galectin-3 may ac-count for the observed homophilic interactions. Onlythis type of assay allowed the demonstration of a lac-tose-dependent self-association of galectin-3, sug-gesting that the C domain of the molecule is involvedin this process. Thus, two different homophilic bindingmechanisms of galectin-3 can now be distinguished:(a) a carbohydrate-independent mechanism mediatedby the interaction of N domains (Hsu et al., 1992;Massa et al., 1993; Mehul et al., 1994) and (b) acarbohydrate-dependent mechanism mediated by theC domain (present study).

Carbohydrate-dependent homophilic binding ofgalectin-3 is mediated by N—C domain interaction

To explain the molecular mechanism of the carbohy-drate-dependent homophilic binding of galectin-3,three main possibilities can be considered: The twohomophilic binding sites are located (a) in the N do-main, (b) in the C domain, or (c) one in the C andthe other in the N domain. It is not likely that the twohomophilic binding sites are localized in the N domainbecause, to our present knowledge, the C and N do-mains fold and function independently (Agrwal et al.,1993; Ochieng et al., 1993) and isolated gal-3C iscapable of binding galectin-3 (the present study). It isalso not likely that the two binding sites are localizedwithin the C domain as, (a) in contrast to the effect ofthe whole molecule, binding of galectin-3 to differentglycoproteins or to galectin-3 is inhibited and not en-hanced in the presence of the isolated C domain and(b) gal-3C does not bind to immobilized gal-3C (Hsuet al., 1992; present study). Thus, the most plausibleexplanation for the carbohydrate-mediated self-associ-ation of galectin-3 seems to be that one binding siteis localized in the C and the other in the N domainof the molecule. This statement is supported by ourobservations on the binding of radiolabeled gal-3C tolaminin in the presence of increasing concentrationsof unlabeled galectin-3 (Fig. 6B). Binding of ‘251-gal-3C to laminin is only weakly inhibited, if at all, inthe presence of increasing concentrations of unlabeledgalectin-3. This can be interpreted in the way that eachunlabeled galectin-3 molecule bound to laminin pre-vents the binding of gal-3C molecules to the polylacto-samine side chains of laminin, on the one hand, butits N domain is still disposable for interaction with



FIG. 7. Aggregation assay using ASF- and BSA-coated latex beads. Latex beads coated withASF (A—G) or BSA (H and I) were incubated for4 h at room temperature without lectins (G) orin the presence of lectin from E. cristagalli (A-C, H) or galectin-3 (D—F, I) in the absence ofcarbohydrates (A, D, G—l) or in the presence of20 mM lactose (B and E) or sucrose (C and F).

free gal-3C molecules, on the other. Therefore, thenumber of binding sites for gal-3C should not be re-duced over a wide range of unlabeled galectin-3 con-centrations added.

Carbohydrates bound to C domain preventcarbohydrate-dependent homophilic interactions

Our experiments have shown that gal-3C interfereswith the binding of galectin-3 to galectin-3 (Fig. 6A)and that gal-3C interacts with galectin-3 (Fig. 6B andC). Furthermore, direct binding of radiolabeled galec-tin-3 to gal-3C is specifically inhibitable by lactose(Fig. 6C). Taken together, these data suggest that (a)one homophilic binding site must be localized withinthe C domain (otherwise gal-3C would not prevent thebinding of galectin-3 to galectin-3) and (b) binding

of carbohydrates to the C domain prevents its interac-tion with the N domain (as demonstrated by the lac-tose-dependent binding of radiolabeled galectin-3 togal-3C).

The N domain is involved in carbohydrate-independent homophilic interactions of galectin-3

Several pieces of evidence demonstrate the partici-pation of the N domain in homophilic galectin-3 inter-actions: (a) Binding of radiolabeled galectin-3 to IgEis potentiated by unlabeled galectin-3 and inhibitedby gal-3C (Hsu et al., 1992). These data suggest aninteraction between N domains but do not, however,exclude interactions between N and C domains. (b)Chemical cross-linking experiments have demon-strated that galectin-3 aggregates in solution (Hsu et



al., 1992; Mehul et al., 1994). However, no evidencefor the formation of galectin-3 aggregates in solutioncould be provided by circular dichroism (Massa et al.,1993) or gel filtration experiments (Hsu et al., 1992;Massa et al., 1993). The N domain, by contrast, aggre-gates much more efficiently than the whole moleculeas determined by chemical cross-linking and circulardichroism studies (Mehul et al., 1994).

As the isolated N domain strongly aggregates insolution to form not only dimeric but also oligomericstructures (Mehul et al., 1994), at least two bindingsites involved in homophilic interactions must be local-ized in this domain. The weak self-aggregation of thewhole molecule is most likely due to conformationalchanges or a sterical hindrance caused by the C do-main. Differences in the melting temperatures of theisolated N domain and the intact molecule further sug-gest that the absence of the C domain increases thestability of the N domain (Agrwal et al., 1993). Bind-ing studies using the isolated N domain should helpto understand the homophilic binding properties of ga-lectin-3 in more detail.

Homophilic interactions of galectin-3 and theirparticipation in cell-to-cell adhesion

In principle, galectin-3 can be anchored to the cellsurface either (a) via its C domain bound to /3-galac-tosidic carbohydrates expressed on glycoproteins orglycolipids or (b) via another putative cell bindingsite. Summarizing the data from the literature, it seemslikely that both possibilities are realized: (a) For mac-rophages, the cell surface expression of galectin-3 cor-relates with the expression of high-affinity glycopro-tein ligands for this lectin, and in cross-linking experi-ments, lectin and ligand have been found to co-precipitate (Sato and Hughes, 1992). (b) Treatmentof mast cells with lactose results in the detachment ofa great part of cell surface-expressed galectin-3, while10% of all galectin-3 molecules still remain attachedto the cell surface after this kind of “elution” (Frigeriand Liu, 1992), arguing for an additional lactose-inde-pendent attachment mechanism. (c) Recombinant ga-lectin-3 expressed on the surface of Sf9 insect cellscannot be detached by lactose treatment, and thesecells aggregate upon addition of ASF (Inohara andRaz, 1995), suggesting a carbohydrate-independent at-tachment mechanism.

Although most of the experimental evidence sug-gesting an interaction between N domains of intactgalectin-3 molecules is indirect, this possibility offersa straightforward interpretation for the ability of galec-tin-3 to induce cell aggregation. Hemagglutination ex-periments in the presenceof reducing agents have dem-onstrated that cell aggregation occurs independently ofthe formation of covalently linked galectin-3 dimers(Hsu et al., 1992; present study). Assuming an N—Ndomain association of two galectin-3 molecules, theirbinding to carbohydrate ligands exposed on the cell

FIG. 8. A hypothetical model on the implication of the homo-philic interaction ofgalectin-3 in cell-to-cell adhesion. A: A galec-tin-3 dimer bridges two neighboring cell surfaces via noncovalentassociation of the N domains. B: N—C domain interactions alonedo not lead to cell bridging unless a cell binding site other thanthe lectin binding site (C) exists. D: N—C domain interactions inaddition to N—N domain interactions may allow a greater flexibil-ity and variability of lectin-bridged cellular complexes, for exam-ple, by an increased spacer length (cf. A and D).

surface of two neighboring cells via their C domainswould lead to a bridging of these cells (Fig. 8A). Thepossibility of an N—C domain interaction, as demon-strated in our study, will never lead to cell bridging(Fig. 8B) unless one postulates a cell binding site ongalectin-3 other than the carbohydrate binding site(Fig. 8C). We would like to speculate that N—C do-main interactions may additionally occur to create agreater flexibility and variability of the cellular bridgesformed (Fig. 8D).

.1. Neurochem., Vol. 70, No. 2. 1998


The functional implication of the lactose-mediatedhomophilic binding of galectin-3 must notbe restrictedto cell adhesion phenomena. (a) In HeLa cells, galec-tin-3 seems to be involved in the assembly of splico-somes in the nucleus (Dagher et al., 1995). Release ofgalectin-3 from nuclear extracts after lactose treatmentleads to an inhibition of in vitro splicing, although nonuclear glycoproteins carrying /3-galactosidic carbohy-drate structures have so far been identified (Dagher etal., 1995). One could argue that in nuclear extracts,galectin-3 appears to be bound to protein compo-nent(s) in a multimeric form and that the lactose-de-pendent release of the molecule is, at least partially,due to a disruption of homophilic binding by the mech-anismdescribed by us. (b) For T cells, bcl-2 has beenshown to bind to immobilized galectin-3 in a lactose-dependent manner (Yang et al., 1996). However, adirect interaction between the two molecules has notyet been demonstrated. (c) In nuclear extracts of thehuman tumor cell line HL6O, galectin-3 has been foundassociated with a glucose binding lectin, and evidencefor the disruption of this association by lactose hasbeen provided (Seve et al., 1993). Unfortunately, theauthors were not able to provide direct evidence thattheglucose binding lectin does not carry /3-galactosidiccarbohydrate structures. If the two lectins, as sug-gested, bind to each other solely via their protein parts,this would provide another confirmation of our findingthat protein-based interactions of galectin-3 can be per-turbed via a carbohydrate binding-dependent mecha-nism. Toward this possibility are observations basedon differential scanning calorimetry studies demonstra-ting that upon lactose binding, galectin-3 undergoesconformational changes (Agrwal et al., 1993).

Acknowledgment: This work was supported by theDeutsche Forschungsgemeinschaft (PR 278/2-1 and PR278/2-2). The authors thank C. Heimann for excellent tech-nical assistance and P. Pesheva for critical reading of themanuscript.

REFERENCES

Agrwal N., Sun Q., Wang S.-Y., and Wang J. L. (1993) Carbohy-drate-binding protein 35. I. Properties of the recombinant poly-peptide and the individuality of the domains. J. Biol. Chem.268, 14932—14939.

Barondes S. H., Cooper D. N. W., Gitt M. A., and Leffler H. (1994)Galectins. Structure and function of a large family of animallectins. J. Biol. Chem. 269, 20807—20810.

Bolton A. E. and Hunter W. M. (1973) The labelling of proteins tohigh specific radioactivity by conjugation to a

251-containingacylating agent. Biochem. J. 133, 529—536.

Dagher S.F., Wang J. L., and Patterson R. J. (1995) Identificationof galectin-3 as a factor in pre-mRNA splicing. Proc. NatI.Acad. Sci. USA 92, 1213—1217.

Fahrig T., Landa C., Pesheva P., Kuhn K., and Schachner M. (1987)Characterization of binding properties of the myelin-associatedglycoprotein to extracellular matrix constituents. EMBO J. 6,2875 —2883.

Frigeri L. G. and Liu F.-T. (1992) Surface expression of functional

IgE binding protein, an endogenous lectin, on mast cells andmacrophages. J. Immunol. 148, 861—867.

Gritzmacher C. A., Robertson M. W., and Fu-Tong L. (1988) IgE-binding protein. Subcellular location and gene expression inmany murine tissues and cells. J. Immunol. 141, 2801—2806.

Gritzmacher C. A., MehI V. S., and Liu F.-T. (1992) Genomic clon-ing ofthe gene for an IgE-binding lectin reveals unusual utiliza-tion of 5’ untranslated regions. Biochemistry 31, 9533—9538.

Ho M. and Springer T. A. (1982) Mac-2, a novel 32,000 Mr mousemacrophage subpopulation specific antigen defined by mono-clonal antibodies. J. immunol. 128, 1221—1228.

Hsu D. K., Zuberi R. I., and Liu F. T. (1992) Biochemical and bio-physical characterization of human recombinant IgE-bindingprotein, an S-type animal lectin. J. Biol. Chem. 267, 14167—14 174.

Inohara H. and Raz A. (1995) Functional evidence that cell surfacegalectin-3 mediates homotypic cell adhesion. Cancer Res. 55,3267—3271.

Kasai K. and Hirabayashi J. (1996) Galectins: a family of animallectins that decipher glycocodes. .1. Biochem. (Tokyo) 119, 1—8.

Laemmli U. K. (1970) Cleavage of structural proteins during assem-bly of the head of the bacteriophage T4. Nature 227, 680—685.

Lindstedt R., Apodaca G., Barondes S. H., Mostov K. E., and LefflerH. (1993) Apical secretion of a cytosolic protein by Madin—Darby canine kidney cells. J. Biol. Chem. 268, 11750—11757.

Liu F.-T., Hsu D. K., Zuberi R. I., Hill P. N., Shenhav A., KuwabaraI., and Chen S-S. (1996) Modulation of functional propertiesof galectin-3 by monoclonal antibodies binding to thenon-lectindomains. Biochemistry 35, 6073—6079.

Massa S. M., Cooper D. N. W., Leffier H., and Barondes S. H.(1993) L-29, an endogenous lectin, binds to glycoconjugateligands with positive cooperativity. Biochemistry 32, 260—267.

Mehul B., Bawumia S., Martin S. R., and Hughes R. C. (1994)Structure of baby hamster kidney carbohydrate binding proteinCBP3O, an S-type animal lectin. J. Biol. Chem. 269, 18250—18258.

Messner P. and Sleytr U. W. (1991) Bacterial surface layer glyco-protein. Glycobiology 1, 545—551.

Mey A., Leffler H., Hamam Z., Normier G., and Revillard J.-P.(1996) The animal lectin galectin-3 interacts with bacterial lipo-polysaccharides via two independent sites. J. Immunol. 156,1572—1577.

Ochieng J., Platt D., Tait L., Hogan V., Raz T., Carmi P., and RazA. (1993) Structure—function relationship of a recombinanthuman galactoside-binding protein. Biochemistry 32, 4455—4460.

Oda Y., Leffler H., Sakakura Y., Kasai K., and Barondes S. H.• (1991) Human breast carcinoma eDNA sequence encoding a

galactoside-binding lectin homologous to mouse Mac-2 antigen.Gene 99, 279—283.

Pesheva P., Urschel S., Frei K., and Probstmeier R. (1997) Murinemicroglial cells express functionally active galectin-3 in vitro.J. Neurosci. Res. (in press).

Poltorak M., Sadoul R., Keilhauer G., Landa C., Fahrig T., andSchachner M. (1987) The myelin-associated glycoprotein(MAG), a member of the L2/HNK-l family of neural celladhesion molecules, is involved in neuron—oligodendrocyte andoligodendrocyte—oligodendrocyte interaction. J. Cell Biol. 105,1893—1899.

Probstmeier R., Montag D., and Schachner M. (1995) Galectin-3, a/3-galactoside-bindinganimal lectin, binds to neural recognitionmolecules. J. Neurochem. 64, 2465—2472.

Probstmeier R., Angelov D. N., and Neiss W. F. (1996) Galectin-3is expressed by subpopulations of microglial and neuronal cellsin the facial nucleus after axotomy. Neuroforum 2, 96.

Quarles R. H. (1989) Glycoproteins of myelin and myelin-formingcells, in Neurobiology of Glycoproteins (Margolis R. U. andMargolis R. K., eds), pp. 243—275. Plenum Press, New York.

Regan L. J., Dodd J., Barondes S. H., and JessellT. M. (1986) Selec-



five expression of endogenous lactose-binding lectins and lac-toseries glycoconjugates in subsets of rat sensory neurons. Proc.NaIl. Acad. Sci. USA 83, 2248—2252.

Sato S. and Hughes R. C. (1992) Binding specificity of a babyhamster kidney lectin for H type I and II chains, polylactos-amine glycans, and appropriately glycosylated forms of lamininand fibronectin. J. Biol. Chem. 267, 6983—6990.

Sato S., Burdett I., and Hughes R. C. (1993) Secretion of the babyhamster kidney 30-kDa galactose-binding lectin from polarizedand nonpolarized cells: a pathway independent of the endo-plasmatic reticulum—Golgi complex. Exp. Cell Res. 207, 8—18.

Seve A. P., Felin M., Doyennette-Moyne M.-A., Sahraoui T., AuberyM., and Hubert J. (1993) Evidence for a lactose-mediated asso-ciation between two nuclear carbohydrate-binding proteins.Glycobiology 3, 23—30.

Sharon N. (1993) Lectin—carbohydrate complexes of plants andanimals: an atomic view. Trends Biochem. Sci. 18, 221—226.

Towbin H., Staehelin T., and Gordon J. (1979) Electrophoretic trans-fer ofproteins from polyacrylamide gels to nitrocellulose sheets:procedure and some applications. Proc. Natl. Acad. Sd. USA76, 4350—4354.

Wang J. L., Laing J. G., and Anderson R. L. (1991) Lectins in thecell nucleus. Glycobiology 1, 243—252.

Woo H.-J., Lotz M. M., Jung J. U., and Mercurio A. M. (1991)Carbohydrate-binding protein 35 (Mac-2), a laminin-bindinglectin, forms functional dimers using cysteine 186. J. Biol.Chem. 266, 18419—18422.

Yang R.-Y., Hsu D. K., and Liu F.-T. (1996) Expression of galectin-3 mediates T-cell growth and apoptosis. Proc. NatI. Acad. Sci.USA 93, 6737—6742.


homophilic binding properties of galectin-3: involvement of the carbohydrate recognition domain

Documents