Fax: +49 6221 565981
Cancer Cell Biology
Determination of structural and functional overlap/divergence of five proto-type galectins by analysis of the growth-regulatory interaction with ganglioside GM1in silico and in vitro on human neuroblastoma cells
Version of Record online: 2 NOV 2004
Copyright © 2004 Wiley-Liss, Inc.
International Journal of Cancer
Volume 114, Issue 1, pages 46–57, 10 March 2005
How to Cite
André, S., Kaltner, H., Lensch, M., Russwurm, R., Siebert, H.-C., Fallsehr, C., Tajkhorshid, E., Heck, A. J.R., von Knebel Doeberitz, M., Gabius, H.-J. and Kopitz, J. (2005), Determination of structural and functional overlap/divergence of five proto-type galectins by analysis of the growth-regulatory interaction with ganglioside GM1in silico and in vitro on human neuroblastoma cells. Int. J. Cancer, 114: 46–57. doi: 10.1002/ijc.20699
- Issue online: 18 JAN 2005
- Version of Record online: 2 NOV 2004
- Manuscript Accepted: 24 AUG 2004
- Manuscript Received: 2 JUN 2004
- Mizutani Foundation for Glycoscience (Tokyo, Japan)
The growth-regulatory interplay between ganglioside GM1 on human SK-N-MC neuroblastoma cells and an endogenous lectin provides a telling example for glycan (polysaccharide) functionality. Galectin-1 is the essential link between the sugar signal and the intracellular response. The emerging intrafamily complexity of galectins raises the question on defining extent of their structural and functional overlap/divergence. We address this problem for proto-type galectins in this system: ganglioside GM1 as ligand, neuroblastoma cells as target. Using the way human galectin-1 interacts with this complex natural ligand as template, we first defined equivalent positioning for distinct substitutions in the other tested proto-type galectins, e.g., Lys63 vs. Leu60/Gln72 in galectins-2 and -5. As predicted from our in silico work, the tested proto-type galectins have affinity for the pentasaccharide of ganglioside GM1. In contrast to solid-phase assays, cell surface presentation of the ganglioside did not support binding of galectin-5, revealing the first level of regulation. Next, a monomeric proto-type galectin (CG-14) can impair galectin-1-dependent negative growth control by competitively blocking access to the shared ligand without acting as effector. Thus, the quaternary structure of proto-type galectins is an efficient means to give rise to functional divergence. The identification of this second level of regulation is relevant for diagnostic monitoring. It might be exploited therapeutically by producing galectin variants tailored to interfere with galectin activities associated with the malignant phenotype. Moreover, the given strategy for comparative computational analysis of extended binding sites has implications for the rational design of galectin-type-specific ligands. © 2004 Wiley-Liss, Inc.
Our study focuses on an emerging class of endogenous growth regulators in a clinically relevant tumor model. For the orientation of the reader, we start with a brief primer of the concept. The cell surface is the obvious site for presentation of sensors for the cells' communication with the environment. Spatial accessibility, biochemical hardware to enable high-density coding and the translation of specific binding processes into signaling are essential means toward efficient information transfer. All three prerequisites are readily fulfilled by carbohydrate epitopes of cellular glycans (polysaccharides). In fact, their theoretical capacity for coding surpasses that of oligonucleotides and oligopeptides by orders of magnitude, and a complex enzymatic machinery of glycosyltransferases accounts for realization of an enormous structural diversity.1, 2, 3, 4, 5, 6, 7, 8 Spatially, the β-galactosides at antennae/branch ends of glycan chains are especially well separated from the membrane. They can in principle be easily engaged in biomolecular recognition. In this sense, the phenomenological mapping of disease-associated alterations in the glycomic profile is rather likely to acquire a functional dimension.3, 9, 10, 11, 12 As a general theme directing our studies, we thus aim to provide evidence for the concept to link distinct characteristics of tumor cell glycosylation with aspects of the malignant phenotype.
When interpreting oligosaccharides of glycan chains as code words, their message is expected to be biochemically decoded and then translated into cellular responses such as modulation of adhesion/migration or proliferation.13, 14 Laboratory applications of plant lectins extensively document the proof-of-principle versatility of proteins with distinct carbohydrate specificity in this respect.14, 15, 16 The detection of endogenous lectins and the fact that their expression matches that of enzymes involved in glycan assembly and remodeling in complexity strongly argue in favor of an elaborate in vivo system of protein(lectin)-glycan interactions.17, 18, 19, 20 In full accord with the assumed active role of diverse β-galactosides in functional glycomics, one particular family of endogenous lectins has evolved with specificity to this molecular category of targets, and this family is termed galectins.17, 21 Intriguingly, model studies with oligosaccharides representing branch-end epitopes of cell surface glycoconjugates and N-glycans harboring natural substitutions have already validated the predicted impact of structural and conformational features of the sugar ligand on affinity to galectins.22, 23, 24 Fitting the elaborate mechanisms to modify carbohydrate properties as ligands, two main factors on the side of galectins render effective fine-tuning and regulation likely: (i) the galectins' diversification in up to 14 different family members in mammals with subdivision into three groups (proto-, chimera-, and tandem-repeat types) and (ii) the observations from RT-PCR and immunohistochemical analyses (galectin fingerprinting) that a tumor (or nonmalignant) cell can often express more than one galectin type.25, 26, 27 These recent insights raise the pertinent question of defining the extent of structural and functional overlap/divergence among galectins. This issue characterizes the first main aim of our study.
Our previous work has defined a suitable tumor cell system, i.e., human SK-N-MC neuroblastoma cells, for analysis to contribute to the resolution of this problem. Due to the fact that neuroblastoma is a frequent extracranial solid tumor type in childhood, accounting for about 15% of pediatric cancer deaths, our project on endogenous growth regulators could spawn a clinical perspective. In detail, we have first shown that a distinct change in the glycomic profile, i.e., shift in the ganglioside population from higher sialylated forms to ganglioside GM1 due to upregulation of a cell surface ganglioside sialidase (neuraminidase), is the crucial control element to switch cell behavior from proliferation to differentiation.28, 29, 30 Next, we pinpointed galectin-1 as major receptor for the ganglioside's pentasaccharide chain.31 Sizable contributions of its GalNAc/Neu5Ac moieties to the enthalpic gain of binding can explain the notable selectivity of galectin-1 to pick this oligosaccharide from the wide panel of cell surface β-galactosides.32 We proposed galectin-1 to be a regulator of neuroblastoma growth. Verifying our hypothesis, carbohydrate-dependent binding to ganglioside GM1 by this homodimeric proto-type lectin impaired proliferation.33 Thus, we delineated a clear functional correlation between appearance of a distinct aspect of tumor cell glycosylation and its growth-regulatory functionality via an endogenous lectin. Interestingly, the p53-induced protein-1, another member of the galectin family, referred to as galectin-7, proved to be a functional homolog in this aspect.34 In contrast, galectin-3, which shares the target specificity to ganglioside GM1, failed to affect cell growth and in consequence competitively inhibited the activity of galectins-1 and -7.33, 34 This galectin is the only chimera-type galectin. It harbors a collagenase-sensitive stalk and a short N-terminal stretch with a phosphorylation site in addition to the carbohydrate recognition domain (CRD). In solution, it is predominantly monomeric but can oligomerize to pentamers to a small extent with increasing concentration.35, 36, 37, 38 The cell biological result emphasizes the importance of (i) ligand specificity, which in this case engenders direct competition for the same glycan epitope, and (ii) topological mode of CRD presentation for eliciting growth regulation.
In order to systematically answer the question given above with focus on proto-type proteins in a clinically relevant tumor system, we proceeded from the established basis in a stepwise manner. What follows in this paragraph is the rationale for our experiment route: we started by selecting galectins-2 and -5 and, notably, the two chicken galectins CG-14 and −16 which are monomeric (CG-14) or dimeric (CG-16), as the test panel. We then examined in silico how sequence variations among these galectins and also galectins-3, -4 and -7 affect ligand accommodation in this system. These computational data provide the first comparative mapping of the extended binding sites in complex with a physiological ligand. To set them in relation to experimental data on ligand affinity and growth regulation, we established recombinant production of the set of galectins and next performed binding studies in two settings: (i) a solid-phase assay using carrier-immobilized lysoganglioside GM1 and (ii) a cell-binding assay. The two assays naturally differ in the topological aspect of the ganglioside's presentation. Ligand selection on the cell surface was further probed by a ganglioside GM1-specific probe. Last but not least, functional assays gave the answer on the extent of functional overlap/divergence. They were flanked by competition assays focusing on cell growth. The major lesson from our in silico and in vitro study is the detection of functional divergence among proto-type galectins, a result relevant for rationally optimizing or interfering with clinically relevant functions of galectins.
Material and methods
Processing of sequence data
Amino acid sequences of human galectins-1 (accession no. P09382), -2 (P05162), -3 (P17931) and -7 (P47929), rat galectins-2 (Q9Z144), -4 (P38552) and -5 (P47967), mouse galectin-3 (P16110), as well as chicken galectins CG-14 (P07583) and CG-16 (P23668) were available from the Swiss-Prot database (http://www.expasy.org/sprot) and edited using the default text editor of Microsoft Windows (Redmond, WA, USA). Alignment of the sequences was established using the program Multalin (http://prodes.toulouse.inra.fr/multalin/multalin.html; version 5.4.1). The ClustalW algorithm, available on the website of the European Bioinformatics Institute (http://ebi.ac.uk/clustalw), was employed to calculate a tree cladogram illustrating putative phylogenic relationships among exon sequences encoding the homologous region of the CRD.
Molecular modeling of galectin-ligand complexes
The coordinates of the topological relationship between human galectin-1 and the bound-state low-energy conformer of the pentasaccharide of ganglioside GM1 had been determined previously.32 This structure was used as template for modeling the interaction of the homologous CRDs with the ligand's low-energy conformer, which is preferentially present in solution and selected by galectin-1. CRD topologies were calculated by extensive homology modeling. These structures were superimposed over the template together with the carbohydrate ligand in its low-energy conformation, using especially the strictly conserved Trp residue and the neighboring Gly moiety as common point of reference, in visual molecular dynamics (VMD) with customized tcl scripts as described.39 For internal control of the validity of the results of the homology modeling procedure, we compared the datasets obtained from our computations with those of available crystal structures for human galectin-2 (1HLC), galectin-3 (1A3K), galectin-7 (1BKZ) and CG-16 (1QMJ). This comparison served to validate the relevance of the results of calculations for cases in which no crystal structure is listed in the databank. A color code is introduced to the illustration of the modeling results to allow the reader to spot substitutions in equivalent positions readily.
Cloning and recombinant production of the galectins
cDNAs for human and rat galectin-2 and for rat galectin-5 were cloned from total RNA of the human colon carcinoma line HT-29, rat duodenum and rat kidney, respectively, with primer sets designed on the basis of published cDNA sequences.40, 41, 42 Because CG-14 and CG-16 were both present in embryonic kidney,43 cDNAs for CG-14 and CG-16 were cloned from total kidney RNA either with suitable primer sets by RT-PCR in the case of CG-1444 or by consecutive primer-directed RT-PCR and then 3′-RACE-PCR to complete the terminal sequence section for CG-16 (Swiss-Prot accession no. AY553270). Recombinant expression with the system combinations of pQE-60 (Qiagen, Hilden, Germany)/E. coli strain M15[pREP4] or pUC540 (KanR)/E. coli strain HB 101 was performed in TB or 2YT media (Roth, Karlsruhe, Germany) at 30°C or 37°C using final concentrations of isopropyl-β-D-thiogalactoside of 25–500 μM. Galectins-1, -3 and -7 were produced as described previously.33, 34
Purification, labeling and activity controls of galectins
The galectins were purified to homogeneity by affinity chromatography on lactosylated Sepharose 4B, obtained by divinyl sulfone activation, as a crucial step.45 Elution included stabilization of the lectin by iodoacetamide treatment to prevent loss of carbohydrate-binding activity by oxidation.46 Purity controls were routinely performed by 1- and 2-dimensional gel electrophoresis and gel filtration. Mass spectrometric analysis of the aggregation status of rat galectin-5 was carried out with an LC-T nanoelectrospray ionization orthogonal time-of-flight mass spectrometer (Micromass, Manchester, UK) operating in the positive ion mode with 10 pmol of protein sample in 1 μl using a solution of 50 mM ammonium acetate at pH 6.8 to retain noncovalent interactions in contrast to a mixture of acetonitrile:water (1:1) with 0.1% formic acid to establish denaturing conditions.34 Analytical gel filtration with 100 μg aliquots was run on a prepacked Superose 12 HR 10/30 column (24-ml bed volume) connected to a high performance liquid chromatography (HPLC) system (Hitachi-Merck, Darmstadt, Germany) with 50 mM PBS (pH 7.2) without/with 100 mM lactose, its presence required to block galectin-matrix interactions and hereby preclude size-independent retardation, at a flow rate of 0.7 ml/min, as described.37 Hemagglutination assays with glutaraldehyde-fixed, trypsin-treated rabbit erythrocytes, biotinylation of galectins under conditions to maintain activity and assessment of extent of their labeling by a proteomics protocol were performed as described.47, 48, 49 Iodo beads (Pierce, Bonn, Germany) and carrier-free Na 125I (Amersham Biosciences, Freiburg, Germany) facilitated radioiodination of galectins in the presence of 100 mM lactose to protect the carbohydrate-binding sites from chemical modification.34
Solid-phase and cell-binding assays
Preparations of the sphingosine N-alkyl(sulfosuccinimidyl)ester derivative of the lysoganglioside GM1 obtained from purified ganglioside, its covalent coupling to carbohydrate-free BSA, adsorption of the resulting neoglycoprotein to the surface of microtiter plate wells from solution (20 μg/ml in 20 mM PBS, pH 7.2) for 12 hr at 4°C and the solid-phase assays using biotinylated galectins in solution with parallel controls to determine the extent of carbohydrate-dependent binding using a mixture of 75 mM lactose and 0.5 mg asialofetuin/ml as glycoinhibitors followed the previously described protocols.34, 50 For internal comparison of galectin activity toward N-glycans, the glycoproteins serum amyloid P component with its single biantennary complex-type N-glycan at Asn32 per subunit in the pentamer and asialofetuin with its three predominantly (74%) triantennary N-glycans were also assayed as surface-immobilized ligands.48 Human neuroblastoma cells (strain SK-N-MC) were routinely cultured in DMEM containing 10% FCS (PAA Laboratories, Cölbe, Germany) and antibiotics. Cells were seeded in 96-well plates and routinely grown for 5 days to reach confluence with a density of about 105 cells/well in serum-supplemented medium. The medium was changed to serum-free DMEM with 25 mM HEPES (pH 7.4) and 0.01% BSA, cell culture continued for 16 hr and radioiodinated galectin was added in the absence of any inhibitor or in the presence of 25 μg cholera toxin B-subunit (Sigma, Munich, Germany)/ml, label-free galectins or a mixture of 150 mM lactose and 0.5 mg asialofetuin/ml, as described.31, 34 When the effects of galectin-1 binding on endocytosis were tested, the cultures were cooled to 12°C or pretreated (30 min) with 100 μM vinblastine prior to adding radiolabeled galectin-1.
Cell growth assays
Cells were seeded at an initial density of 104 cells/well and cultured for 16 hr to allow cell attachment. Then culture continued for 48 hr in serum-supplemented medium containing galectins at the standard concentration of 125 μg/ml, which has been determined to be strongly inhibitory for cell growth in the cases of galectins-1 and -7, followed by quantitation of cell numbers by application of reagents of a commercial kit (CellTiter 96; Promega, Mannheim, Germany). Controls to ascertain inhibition of activity by presence of glycoinhibitors and assays to assess potency of other galectins to block galectin-1 activity using a mixture, e.g., galectin-1 and CG-14, followed routine procedures.33, 34
Results and discussion
Sequence comparison as a measure to infer evolutionary relationship
Galectins share reactivity to β-galactosides and the jelly-roll-like folding pattern.17 Occurrence of a set of invariable amino acids at sites crucial for ligand contact underlies the common selectivity. The way the standard sequence signature can vary within the group of proto-type galectins and relative to a tandem-repeat-type and the chimera-type galectin is illustrated by a detailed sequence alignment (Fig. 1). The basic requirement for presence of the indolyl moiety (W68 in human galectin-1, placed at position 87 in the alignment of Fig. 1) engaged in stacking and C-H/π-interactions with the B-face of galactose is readily apparent by its strict conservation. It also becomes clear that considerable sequence diversification has taken place, an argument for functional divergence.
Further processing of these data was performed to infer the degree of relationship based on sequence similarity. The obtained result indicates that galectins-1 and -7, which were both proven to be growth regulators for neuroblastoma cells,33, 34 are rather widely separated in the cladogram (Fig. 2). Remarkably, analyzing the same galectin type from two mammalian species delineates effects less pronounced than comparing different galectins in a species, and the chicken galectins have notable similarity to human galectin-1. Looking at these rather equal characteristics, functional divergence might then be solely based on differences in the quaternary structure. However, a serious caveat is to be raised: the given reasoning might lead to an error of judgement, because sequence alignments are generally not weighted with respect to the structural role of variable positions. Of course, it can be taken for granted that the binding of the core part of a β-galactoside calls for a constant set of interactions, as chemically defined by thorough mapping with deoxy- and fluoro- derivatives of lactose.51 But how a galectin accomplishes reaching its documented specificity toward particular complex β-galactoside-containing ligands on the cell surface or in the extracellular matrix, such as laminin, carcinoembryonic antigen, certain integrins or, as outlined above, ganglioside GM1 in this cell model, is an open question.31, 52 Most likely, an extension of the binding site and the recruitment of further contact points on the complex glycan of the main cellular target beyond the disaccharide core should be operative.
The ensuing concept to explain the remarkable selectivity of endogenous lectins to decode sugar-based information had recently been verified for P- and E-selectins. L-Fucose is the primary site, with galactose, sialic acid (neuraminic acid) and suitably positioned tyrosine sulfates serving as discriminatory auxiliary contacts.53 P-Type lectins, too, engage more than the central sugar unit (the mannose-6-phosphate residue in this case) in the binding process, the binding site encompassing a total of three mannose moieties.54 Notably, the first study on a galectin dissecting its interaction with a complex ligand in structural terms in solution by nuclear magnetic resonance (NMR) spectroscopy and molecular modeling was equally instrumental to highlight the importance of amino acids in the vicinity of the primary contact site.32 Like the core recognition structure which is bound in its low-energy syn-conformation,55, 56 human galectin-1 selects a low-energy conformation of the complete pentasaccharide to bring the GalNAc/Neu5Ac residues in ganglioside GM1 in enthalpically favorable contact with Arg48, His52, Lys63 and Glu71.32 Examining these positions in Figure 1 confronts us with sequence deviations. This result teaches the lesson that the presented sequence alignment will not be sufficient to answer questions on fine-specificity, when one substitution can matter. Owing to the availability of a detailed view on how human galectin-1 accommodates the key-like low-energy conformation fitting of ganglioside GM1 like a lock, we were now able to sort out common or disparate denominators in the galectins. To do this, we scrutinized the binding sites for this ligand by a sophisticated computational approach, using the experimentally based data on galectin-1 as template. In a broader context, testing this procedure thoroughly will have merit to predict fine-specificity features in other cases as well. Hereby, we are establishing a tool for rational design of selective reagents.51
Comparative computational analysis of galectin-ganglioside GM1 interaction
Before we started the detailed monitoring, we performed homology modeling of galectin structures. We set out from bovine galectin-1 (1SLT) and compared the results obtained to coordinates of the crystal structures as far as available, i.e., for human galectins-2, -3 and -7 and chicken liver galectin CG-16. The remarkable ease of reconciling the datasets argues in favor of the validity of results from the computational process. Therefore, we proceeded to complete modeling also for those cases where no crystal structure is known. The way the ganglioside's low-energy conformation makes contacts to the individual binding sites will translate into predictions on ligand affinity. As shown in Figure 3, which uses a color code to allow rapid orientation for spotting structurally equivalent positions, galectins-1 and -7 closely resemble each other, although they are widely separated in the cladogram (Fig. 2). A substitution from His52 to Thr56 (blue moiety in upper part) had no major bearing on total affinity in the binding assays.34 The distance from His52 to Trp68 is 11.6 Å in human galectin-1, whereas Thr56 is separated from Trp68 by 16.7 Å. As a common denominator, the distance between Trp68 and Arg48 varies only within the narrow range between 9.1 and 10.1 Å. This feature, in combination with the depicted spatial display of the amino acids, will have to be kept in mind for the design of galectin-type-specific ligands (carbohydrates or glycomimetics) and of peptides substituting a galectin for therapeutic purposes.57, 58, 59
Substitutions with experimentally so far unknown impact in the binding pocket concern Lys63, with Leu60 and Gln72 taking its place in human galectin-2 (and also the rat protein) and rat galectin-5, and His52, where these two lectins present Glu moieties. Looking at galectins-3 and -4 which bind GM1,34 Asn or Trp take the place of His52 (not shown), without an impairment of affinity.34 Of special note, the two chicken galectins present no substitution relative to human galectin-1 in the spatial profile of relevant amino acids for the molecular rendezvous with the pentasaccharide. Consequently, their binding to the ganglioside is likely, and a difference in quaternary structure will make them attractive tools to relate ability for cross-linking to biological activity. When purified from the intestine or liver and subjected to gel filtration at physiological ionic strength, CG-14 is monomeric; CG-16, in contrast, is dimeric and forms stoichiometric complexes with the triantennary N-glycans of asialofetuin.60, 61, 62 To be independent of natural sources, we proceeded to facilitate recombinant production of the galectins in the test panel. The availability of respective vectors will also enable any mutational engineering to alter quaternary structure or to rationally optimize ligand properties in the future. Here, the purified proteins afforded to experimentally determine binding properties and relate the insights from computational calculations to the proteins' activities. In the first step, we determined the quaternary structures of the recombinant products, a process which also constitutes a purity control.
Solution structures of galectins in the test panel
Hemagglutination of glutaraldehyde-fixed and trypsin-treated erythrocytes is a common assay to measure sugar-binding and cross-linking activities of lectins. When appropriate ligands are presented on the cell surface, agglutination is an indicator for oligomer formation. Each reaction was completely abolished by lactose but not a nonspecific sugar, ascertaining carbohydrate dependence. The chicken galectins were widely separated in their activities by a factor of 60-fold for the minimal lectin quantity to yield a positive result, as already noted for these 2 lectins from natural sources.63 Human and rat galectins-2 were less reactive than human galectin-1. Interestingly, their retention times and that of human galectin-7 in gel filtration were also longer than for galectin-1, albeit clearly not reaching the range for a monomer, so that shape differences among dimeric proto-type galectins in solution become apparent. Controls in the presence of lactose excluded retention by carbohydrate-dependent interaction to the matrix, as shown previously to be the case for galectin-3.37 No evidence for dissociation or formation of higher aggregates could be obtained in gel filtration analysis, as was also previously seen in a study on human galectin-1 by small angle neutron scattering.64 Regarding rat galectin-5, we could confirm its weak but positive reaction as agglutinin and its behavior as monomer in gel filtration, which had been reported previously.41 To further analyze its aggregation status regardless of shape, we added the highly sensitive method of mass determination by nanoelectrospray ionization mass spectrometry under conditions not harmful for noncovalent interactions.34, 65 In addition to recording spectra of samples exposed to acidic conditions, we applied a pseudophysiological milieu as well, to look at the aggregation status. Previous experimental series with galectins-1, -3 and -7 had ascertained the suitability of this approach to detect galectin dimerization.34
As illustrated in Figure 4, the galectin-5 preparation is pure. Analyzed in a denaturing (acidic) environment multiple ionization occurs (Fig. 4a), while this pattern is restricted under “native” conditions (Fig. 4b). The determined molecular masses of 16,046.75 ± 2.33 Da and 16,048.27 ± 0.11 Da from the spectra in Figure 4a and b, respectively, are in accord with galectin-5 being devoid of the N-terminal methionine residue. In this form, the galectin has a calculated mass of 16,048.23 Da. A minor peak at 15,976.3 ± 0.2 Da corresponds to a galectin-5 after the additional loss of alanine from the N-terminus (calculated mass: 15,977.15 Da). The spectrum under pseudophysiological conditions, which maintain dimerization of galectins-1 and -7 monitored previously,34 raises no evidence for any aggregation of monomeric galectin-5 at a concentration of 10 μM (Fig. 4b). The concern that sample processing will be harmful to an aggregate is further allayed by running an analysis in the presence of a carbohydrate ligand. As similarly seen with galectin-7,34 occurrence of peaks of lectin-ligand complexes add to the evidence for stability of noncovalent complexes under these conditions (not shown). However, we cannot exclude the possibility that galectin-5, as galectin-3,36 might form an aggregate when exposed to a multivalent ligand or a cell surface. This process could explain the weak but significant hemagglutination activity. In solution, we thus deal with two (at least preferentially) monomeric proto-type galectins in our test panel, i.e., CG-14 and rat galectin-5. Although database mining for evidence for a human galectin-5 was negative (Lensch et al., unpublished), rat galectin-5 together with CG-14 are valuable tools for the further experiments, i.e., binding and functional assays. To exclude any variations of experimental conditions the following binding studies were performed in parallel for our test panel with the same set of reagents or cell batches.
Solid-phase and cell binding assays
The neoglycoprotein with the lysoganglioside as ligand part was adsorbed to the surface of microtiter plate wells, hereby establishing a model for cell surface presentation. An advantage of this model is the uniform presence of only one class of ligand. But it should also be mentioned that distinct features on the cell membrane such as local clustering can hereby not be mimicked. As documented previously for galectins-1, -3, -7 and the N-terminal domain of tandem-repeat-type galectin-4,34 binding of the biotinylated galectins was saturable and inhibitable by glycosubstances (0.5 mg asialofetuin/ml and 75 mM lactose). Examples of binding curves are shown in Figure 5. As further controls against the influence of carbohydrate-independent binding via protein/lipid parts of this neoglycoprotein, we performed parallel assays with the glycoproteins' serum amyloid P component and asialofetuin with the same inhibition protocol. These series ascertained binding activity of the labeled lectins with one exception. CG-16 failed to bind to the lysoganglioside and reacted considerably weaker with the glycoproteins than expected based on its cross-linking activity with asialofetuin and the microcalorimetrically determined affinity.61, 66 In this case, an impact of labeling (direct impairment or indirect effects, e.g., by further lowering its isoelectric point (pI) value, which already is the most acidic among the test panel) cannot be excluded. When the binding data were algebraically transformed, the obtained Scatchard plots were linear, as evidence for presence of a single class of binding sites and absence of cooperativity (Fig. 5). The calculated dissociation constants were in the range of those determined previously for galectins-1, -3, and -7 (and also galectin-4's N-terminal domain with 1.41 ± 0.34 μM), CG-14 showing an increased affinity (Table I). Evidently, changes in the architecture of the extended binding sites, which are shown in Figure 3, will not cause dramatic changes in affinity under these experimental conditions. These results should not yet be automatically extrapolated to the physiological situation with a lectin interacting with cell surface ligands. In fact, the following caveats warrant to be acknowledged and should be addressed: (i) presentation of the carbohydrate ligand in this model and on the cell surface can differ with implications for affinity and (ii) any chemical modification can influence binding properties so that further measurements, especially competition and functional assays, are necessary. To address these issues experimentally in a stepwise manner, we performed cell binding studies.
|Type of lectin||Solid-phase assay KD (μM)||Cell assay||Activity as growth regulator|
|KD (μM)||Bmax (106 molecules/cell)|
|Human galectin-1||1.62 ± 0.551||0.76 ± 0.181||2.23 ± 0.291||++3|
|Human galectin-2||1.09 ± 0.88||0.63 ± 0.22||2.07 ± 0.41||++|
|Rat galectin-2||1.71 ± 1.20||0.47 ± 0.19||1.92 ± 0.35||++|
|Murine galectin-3||2.85 ± 0.641||0.94 ± 0.251||2.70 ± 0.372||−3|
|Rat galectin-5||1.45 ± 1.10||n.d.||n.d.||−|
|Human galectin-7||2.29 ± 1.271||0.88 ± 0.291||2.22 ± 0.341||++1|
|CG-14||0.19 ± 0.09||0.51 ± 0.27||2.22 ± 0.39||−|
|CG-16||n.d.||0.56 ± 0.32||1.93 ± 0.31||+|
Binding of the galectins was saturable and inhibitable by presence of inhibitors of β-galactoside-dependent interaction (0.5 mg asialofetuin/ml and 150 mM lactose), as measured for galectins-1, -3 and -7 and as reported previously.33, 34 The inherent control by using the cholera toxin B-subunit as specific blocking reagent ascertained that access to ganglioside GM1 is responsible for a major contribution to cell surface binding of the tested galectins (Fig. 6). The calculation of the dissociation constants for cell surface binding was based on binding curves, which are exemplarily shown in Figure 6. Scatchard plots yielded rather similar data, when compared to the previously obtained result set (Table I). In interspecies comparison, galectins-2 of rat and human origin behaved indistinguishably. CG-16 proved active in this system and the blocking effect of the cholera toxin B-subunit indicated interaction with ganglioside GM1 (Fig. 6c). Also, the number of binding sites at saturation was rather similar, implying comparable selection of binding sites. The only exception was galectin-5, which failed to bind to the cell surface. To exclude that radioiodination had impaired galectin-5's binding activity toward ganglioside GM1, which had been seen in the solid-phase assay, we performed an inhibition study with label-free galectin using the well-characterized binding of galectin-131, 33, 34 as standard. This assay was run for the complete panel beyond galectin-5 in order to strengthen the evidence for common target specificity on the neuroblastoma cell surface. After all, analysis of binding of sialylated β-galactosides with increasing chain length and of 90K/MAC-2 BP had delineated distinct ligand preferences for galectins-1 and -3.22, 67 If a galectin—similar to the cholera toxin B-subunit—will home in on ganglioside GM1, it will reduce the extent of binding of galectin-1. The rather close range of measured dissociation constant (KD)-values predicts efficient competition. Its extent was determined after strict coincubation and also after coincubation preceded by a 1-hr preincubation step with galectin-1.
Indeed, competition was effective for the galectins which bound to the cell surface, as shown in Fig. 7a–c. When the galectins used as competitors were included 1 hr later than galectin-1, the binding of the latter was not completely abolished. In order to check the possibility that galectin-1 enhances endocytosis, leaving less molecules of the ligand available, we conducted the same experiments at reduced temperature and in the presence of vinblastine, an inhibitor of endocytosis. Since neither reduced temperature nor presence of the inhibitor affected residual galectin-1 binding after the addition of competitor, galectin-1 is unlikely to enhance endocytosis (data not shown). Galectin-5 failed to interfere with galectin-1 binding (Fig. 7d). Because galectin-5 was not chemically modified in this assay, we can conclude that monomeric galectin-5—in contrast to monomeric CG-14—shows no evidence for measurable recognition of the ganglioside on the cell surface. Despite similar, albeit weak, activity in hemagglutination for both galectins as measure for aggregation on a cell surface, the binding characteristics in this system are markedly different. Of note, proto-type CG-14, with its close similarity to galectin-1, is a suitable tool to put the hypothesis to the test as to whether cell binding without cross-linking is sufficient to elicit the biological response. Also, blocking of galectin-1 binding by CG-14 could mean that presence of a monomeric proto-type galectin is a way to interfere with distinct aspects of functionality of galectin-1. The monitoring of growth of cells exposed to galectins will provide the respective answer.
Cell growth assays
In preliminary assays, we ascertained the reactivity of the currently tested cell batches as control for constant responsiveness relative to our previous studies. Having verified this essential prerequisite, as documented exemplarily in Fig. 8a and b, we set forth to examine the growth-regulatory effects of cell binding for the test panel. The following results were obtained: (i) galectin-2 (human and rat) acted as functional homolog of human galectin-1, (ii) CG-16 was likewise an effector but to a reduced extent and (iii) the monomeric galectin-5 and CG-14 failed to affect growth (Figs. 8 and 9; Table I). Because cell growth was unaffected by galectins in the presence of glycoinhibitors to prevent carbohydrate-dependent cell binding (not shown), this process was essential for growth regulation. Presence of the glycoinhibitors did not affect cell growth (not shown). Apparently, cell binding and a dimeric structure are necessary to lead to signaling for growth inhibition. In contrast to functional differences among galectins-1 and -7 in in vitro assays with blood and immune cells and in wound healing68, 69 galectins-1, -2, and -7 shared functionality in this cell model. To add further evidence to this conclusion, the results on CG-14 and galectin-5, shown in Figures 7–9, led us to the following experiment: to test both galectins as inhibitors of galectin-1's activity as negative growth regulator. The model we developed predicts that CG-14 should interfere with galectin-1's activity on the level of ligand binding, hereby blocking galectin-1-dependent negative growth regulation. In contrast, galectin-5 should be without functional effect. The microphotograph in Figure 8h and the data in Figure 10 illustrate the validity of the concept. We have added a parallel series with CG-16 in Figure 10 to underscore its activity as effector in comparison to CG-14.
In summary, we thus detected divergence in the group of proto-type galectins on the structural and functional levels. The first lesson emerging from this study is the requirement to monitor complex natural glycans with respect to their ligand properties. For this purpose, structurally well-characterized glycoproteins will be suited as well as glycolipids to unveil intrafamily regulation of affinity among galectins. These modifications can concern substitutions in the chain (e.g., introduction of bisecting GlcNAc) or in terminal β-galactosides (e.g., α1,2-fucosylation or α1,3-substitutions) as well as changes in the number of glycan antennae and degree of clustering.23, 24, 70, 71, 72, 73 Our studies follow the hypothesis that distinct glycan modifications act as regulators of lectin activity. Moreover, our current results underline the importance to take results from assays with a single target to the test by binding studies with cell surfaces. Whereas galectin-2, irrespective of its origin from rat or human tissue, and CG-16 are homologs in both respects, galectin-5 failed to bind ganglioside GM1 (and other ligand site(s)) on the cell surface. Once the major ligand(s) for a cell-type-selective galectin effect have also been defined in other tumor cell systems, corresponding studies can follow the guideline and strategy presented in this report. Hereby, it is then possible to factor cell surface presentation of a glycan epitope into the assessment as ligand and to determine why distinct aspects of tumor cell glycosylation matter for establishing the malignant phenotype, as has been shown in this system. With this knowledge acquired, their expression profiles can be manipulated on the level of the responsible glycosyltransferases or glycosidases, merging work with galectins, as we are currently pursuing.
The second lesson enhances our understanding of the functionality of the complex galectin network in tumor cells.74 Differences in quaternary structure, in this case the formation of monomeric and dimeric modules for cross-linking,75 translate into divergent effects on the cellular level, i.e., triggering of growth regulation and its abrogation in this cell model. Thus, our data broaden the basis to refer to galectin-1 as a potent growth regulator in human neuroblastoma. Hereby, these data add dimeric galectins to the recently compiled panel of intracellular proteins relevant for cell cycle regulation and apoptosis in this tumor type, e.g., the Id2 protein and survivin.76 The availability of vector collection for galectins will be instrumental to taking the next step in our concept. Aside from ensuring a supply of test material, it enables the design of variants, deliberately altering the fine-specificity and the quaternary structure. That efforts in this area are promising has been recently demonstrated. Targeting the intracellular interaction of galectin-1 with oncogenic H-Ras, a custom-made galectin-1 mutant (i.e., L11A) was designed, and it acted as a dominant negative effector for active H-Ras.77 This work and the perspective raised by our report intimate that options for new treatment modalities might arise from rational exploitation of galectin functionality.
Having defined cell surface β-galactosides as code words for cell communication and delineated the intrafamily diversity of galectins, the next issue is to take stock of the extent of structural and functional overlap/divergence among galectins. Here, we focused on proto-type galectins, including cross-species comparison. With the coordinates of the extended binding site of galectin-1 for ganglioside GM1 and the potency of regulation of neuroblastoma cell growth through the interaction at hand, we first defined distinct substitutions distinguishing combining sites for this glycan chain, e.g., between galectin-1 and galectins-2 and -5. This knowledge-based computational procedure will find further application to design galectin-type-specific reagents and galectin-mimetic peptides. Regulation of galectin activity on cell growth was detected on two levels, i.e., binding to the cell surface and triggering of growth inhibition. Galectin-5, which showed affinity to the carrier-immobilized lysoganglioside in a solid-phase assay, failed to bind to ganglioside GM1 on the cell surface, and monomeric CG-14 did not induce growth reduction. Because it blocked galectin-1's activity competitively, the quaternary structure of proto-type galectins is a key factor for this type of functional divergence. This result is proposed to have potential therapeutic relevance, i.e., in cases where expression of a cross-linking galectin, for example galectin-1 in glioblastoma or colon cancer,74, 78, 79 is associated with the malignant phenotype.
We thank Drs. S. Namirha and S. Goldmann for helpful discussion and S. Himmelsbach, B. Hofer and L. Mantel for excellent technical assistance.
- 1Glycosyltransferases involved in N- and O-glycan biosynthesis. In: GabiusHJ, GabiusS, eds. Glycosciences: status and perspectives. Weinheim-London: Chapman & Hall, 1997. 79–113., .
- 2The information-storing potential of the sugar code. In: GabiusHJ, GabiusS, eds. Glycosciences: status and perspectives. Weinheim-London: Chapman & Hall, 1997. 1–14..
- 19Handbook of animal lectins. New York: J. Wiley & Sons, Inc. 2000. 468 p..
- 20Gabius HJ, ed. Special issue: animal lectins. Biochim Biophys Acta 2002; 1572: 163–434.
- 23Structure-activity profiles of complex biantennary N-glycans with core fucosylation and with/without additional α2,3/α2,6-sialylation: synthesis of neoglycoproteins and their properties in lectin assays, cell binding, and organ uptake. J Med Chem 2002; 45: 478–91., , , , , , , , .
- 55NMR-based, molecular dynamics- and random walk molecular mechanics-supported study of conformational aspects of a carbohydrate ligand (Galβ1,2Galβ1,R) for an animal galectin in the free and in the bound state. Biochem Biophys Res Commun 1996; 219: 205–12., , , , , , , , .
- 70Neoglycoproteins with the synthetic complex biantennary nonasaccharide or its α2,3/α2,6-sialylated derivatives: their preparation, assessment of their ligand properties for purified lectins, for tumor cells in vitro and in tissue sections, and their biodistribution in tumor-bearing mice. Bioconjugate Chem 1997; 8: 845–55., , , , , , .
- 73Effects of polyvalency of glycotopes and natural modifications of human blood group ABH/Lewis sugars at the Glaβ1-terminated core saccharides on the binding of domain-I of recombinant tandem-repeat-type galectin-4 from rat gastrointestinal tract (G4-N). Biochimie 2004; 86: 317–26., , , , , , .