The Inhibitory Effects of a Rhamnogalacturonan Ι (RG-I) Domain from Ginseng Pectin on Galectin-3 and Its Structure-Activity Relationship*

Background: Structural elements in pectin that inhibit galectin-3, a β-galactoside-binding protein associated with cancer progression, are poorly defined. Results: Both backbone and side chains of pectin RG-I-4 were important for its anti-galectin-3 activity. Conclusion: High activity of RG-I-4 was due to cooperation between short β1,4-galactan side chains. Significance: The results are valuable for producing highly active pectin-based galectin-3 inhibitors. Pectin has been shown to inhibit the actions of galectin-3, a β-galactoside-binding protein associated with cancer progression. The structural features of pectin involved in this activity remain unclear. We investigated the effects of different ginseng pectins on galectin-3 action. The rhamnogalacturonan I-rich pectin fragment, RG-I-4, potently inhibited galectin-3-mediated hemagglutination, cancer cell adhesion and homotypic aggregation, and binding of galectin-3 to T-cells. RG-I-4 specifically bound to the carbohydrate recognition domain of galectin-3 with a dissociation constant of 22.2 nm, which was determined by surface plasmon resonance analysis. The structure-activity relationship of RG-I-4 was investigated by modifying the structure through various enzymatic and chemical methods followed by activity tests. The results showed that (a) galactan side chains were essential to the activity of RG-I-4, whereas arabinan side chains positively or negatively regulated the activity depending on their location within the RG-I-4 molecule. (b) The activity of galactan chain was proportional to its length up to 4 Gal residues and largely unchanged thereafter. (c) The majority of galactan side chains in RG-I-4 were short with low activities. (d) The high activity of RG-I-4 resulted from the cooperative action of these side chains. (e) The backbone of the molecule was very important to RG-I-4 activity, possibly by maintaining a structural conformation of the whole molecule. (f) The isolated backbone could bind galectin-3, which was insensitive to lactose treatment. The novel discovery that the side chains and backbone play distinct roles in regulating RG-I-4 activity is valuable for producing highly active pectin-based galectin-3 inhibitors.

Galectin-3 is the only chimeric member of the galectin family, which is characterized by a conserved carbohydrate recog-nition domain (CRD) 3 that specifically binds to ␤-galactosides (1-3). Galectin-3 has a C-terminal CRD connected to a long N-terminal proline-and glycine-rich domain (1). X-ray crystal structure analysis has revealed that galectin-3 CRD is composed of a five-stranded and a six-stranded ␤-sheet in a ␤-sandwich arrangement. The binding groove is long enough to hold a linear tetrasaccharide (4). Galectin-3 can oligomerize in the presence of multivalent carbohydrate ligands (5). This self-association property depends on the N-terminal domain (5). Oligomerization of galectin-3 is essential for some of its biological functions (6).
Galectin-3 is involved in various biological processes, such as cell-cell interactions, proliferation, differentiation, apoptosis, and mRNA splicing, and is associated with different pathologic conditions including cancer (1,3,7,8). Galectin-3 is highly expressed in a subset of metastatic cancer cells in comparison with benign or normal cells and is a potential diagnostic and/or prognostic marker for some types of cancers (9). The role of galectin-3 in cancer progression and metastasis has been extensively studied. Overexpression of galectin-3 protects cancer cells from apoptosis in the presence of ultraviolet radiation or antitumor drugs (10). Interaction of galectin-3 with the cancerassociated Thomsen-Friedenreich glycoantigen mediates the initial adhesion of cancer cells to the vascular wall and subsequent tumor cell homotypic aggregation at the site of primary attachment to the endothelium (11). Galectin-3 acts as a chemoattractant and induces endothelial cell migration and tube formation, which are two important steps in tumor angiogenesis (12). Galectin-3 may also facilitate immune evasion of tumor cells by binding to T-cell receptors, which leads to inhibition of T-cell activation and apoptosis of T-cells (13).
Because galectin-3 contributes to tumor progression, numerous studies have focused on the development of specific galectin-3 inhibitors including peptide antagonists (14), lactu-* This work was supported by National Natural Science Foundation of lose amines (15), and galactose-based inhibitors (16,17). In addition to these synthetic molecules, a natural product, pectin, has emerged as a good source for generating high affinity galectin-3 inhibitors with low toxicity. Several pectins and pectic polysaccharides, such as modified citrus pectin (MCP), swallow root pectin, Hemidesmus pectin, black cumin pectin, Andrographis pectin, and okra pectin have been reported to possess anti-galectin-3 activities (18 -22). Among these pectin-based inhibitors, MCP has attracted special attention. Studies from several groups have shown that MCP inhibits multiple steps of tumor metastasis via inhibition of galectin-3 including inhibition of cancer cell adhesion, homotypic aggregation, invasion, clonogenic survival, angiogenesis, sensitization of neoplastic cells to apoptosis induced by chemotherapeutic agents, and correction of the impaired function of tumor-infiltrating lymphocytes (20,(23)(24)(25)(26)(27)(28).
Although interaction between pectin and galectin-3 has been recognized for some time, the structural features of pectin that contribute to that interaction are poorly understood. One hypothesis that has been proposed in the literature is that the galactose (Gal) residues in pectin facilitate interaction with galectin-3 (19,23,24). However, there are many Gal-containing pectins in nature, and only a few have been shown to interact with galectin-3. Clearly, the presence of Gal residues alone is insufficient. Pectin has a very complex structure. It usually contains galacturonic acid (GalA), Gal, arabinose (Ara), and rhamnose (Rha) residues. The content and linkages of each residue differ between plants and can vary within different parts of the same plant. This does not mean that the residues are randomly linked. In fact, they are organized into distinct structural elements or domains, such as galactan/arabinogalactan (AG), homogalacturonan (HG), rhamnogalacturonan ⌱ (RG-I), rhamnogalacturonan II, and xylogalacturonan. Each element or domain differs considerably between species (29). Gunning et al. (30) reported that ␤1,4-galactan derived from potato pectin specifically recognized galectin-3. Our research group isolated ␤1,4-galactan fragments from MCP. We found that the chain termini and not the internal region regulated interactions with galectin-3 (31). Despite these findings, the structure-activity relationship remains far from clear due to the lack of structurally defined pectin fractions or fragments. In this regard, our research group isolated and characterized four HG-rich and four AG-rich pectins from ginseng and five RG-I-rich pectin fragments from endo-PG-treated ginseng pectin (32)(33)(34). In the present study, we examined the inhibitory effects of these pectins and fragments on galectin-3-mediated activities and identified one of the RG-I-rich pectin fragments as a potent inhibitor of galectin-3. Additional structure-activity studies demonstrated that, besides Gal residues, both the backbone and the side chains of this fragment were important for inhibition of galectin-3.
␤-Elimination-␤-Elimination was performed according to a published protocol (38). Briefly, 5 mg/ml RG-I-4, which was dissolved in 0.2 M sodium borate buffer (pH 7.3), was heated for 4 h at 120°C. The products were dialyzed and lyophilized.
De-esterification-De-esterification was performed based on the literature (32). Briefly, 10 mg/ml RG-I-4 was treated with 0.1 M NaOH at 4°C for 4 h followed by neutralization and desalting on a Sephadex G-25 column (2 ϫ 20 cm).

Preparation of Galacto-oligosaccharides
All galacto-oligosaccharides were prepared from potato galactan. Oligosaccharides A-E were prepared by enzymatic digestion. Briefly, 200 mg of potato galactan, which was dissolved in 20 ml of 50 mM sodium acetate buffer (pH 4.5), was incubated with 0.3 unit/ml endo-␤1,4-D-galactanase from A. niger at 30°C for 24 h. The digest was boiled for 5 min, centrifuged, loaded onto a Bio-Gel P-2 column (2 ϫ 90 cm), and eluted with distilled water at a flow rate of 0.15 ml/min. The eluate was collected at 4.5 ml/tube, analyzed by the phenolsulfuric acid assay (32), and pooled as shown in supplemental Fig. S1a.
Oligosaccharides F-M were prepared by partial acid hydrolysis. Briefly, 1 g of potato galactan was dissolved in 100 ml of 0.2 M trifluoroacetic acid and heated at 80°C for 4 h. The reaction was neutralized to pH 6 -7 with 10% aqueous ammonia, desalted on a Sephadex G-10 column (3 ϫ 20 cm), and passed through a small Q-Sepharose fast flow column (1 ϫ 12 cm). The non-binding material, referred to as P-oligo, was further separated on a Sephadex G-50 column (3 ϫ 90 cm) and eluted at 0.5 ml/min with 0.15 M NaCl. The eluate was collected at 3 ml/tube and pooled as shown in supplemental Fig. S1b.

Galectin-3-mediated Hemagglutination (G3H) Assay
The assay was performed as we described previously (31). Each well of a microtiter V plate contained 25 l of 1% bovine serum albumin (BSA) in 0.15 M NaCl, 25 l of 0.15 M NaCl (control) or the test samples in this solution, 25 l of 12.5 g/ml galectin-3, and 25 l of a 4% (v/v) chicken erythrocyte suspension. Agglutination proceeded for 90 min at room temperature. The minimum inhibitory concentration (MIC) of each test sample was determined by examining a series of dilutions. The results were the average of three independent experiments.

Galectin-3-mediated Cell Adhesion Assay
HT-29 cells (7 ϫ 10 4 ) in serum-free medium were added to each well of a 96-well plate, which was coated with galectin-3 or BSA (as background). Plates were incubated at 37°C for 1 h in the presence or absence of the test samples. Adherent cells were measured by the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide method (39). Each experiment was performed in triplicate and repeated at least three times. The half-inhibitory concentration (IC 50 ) for galectin-3-mediated cell adhesion, which was determined from the absorbance at 570 nm, was obtained for each test sample by creating a doseresponse curve with a nonlinear regression model in GraphPad Prism 5.0.

Asialofetuin-induced Homotypic Cell Aggregation Assay
MDA-MB-231 cells (0.6 ϫ 10 6 cells/ml in PBS) were mixed with buffer (control) or 15 g/ml ASF in the absence or presence of test samples in a 2-ml microcentrifuge tube and agitated for 1 h at 37°C. Cell aggregation was examined under a microscope and quantified as described previously (40). Each experiment was performed in triplicate and repeated three times.

Enzyme-linked Immunosorbent Assay (ELISA)
RG-I-4 and modified RG-I-4 were coated onto 96-well Nunc Immuno MaxiSorp microtiter plates at 4°C. The ␤1,4-galactan epitope was identified with monoclonal antibody LM5 according to a published method (41). Each experiment was repeated at least three times.

Fluorescent Labeling of RG-I-4 and Galectin-3
RG-I-4 and galectin-3 were labeled with DTAF according to a published method (30). Free DTAF was removed on a Sephadex G-25 column.

Fluorescence Microscopy
GST-galectin-3-and GST-immobilized beads were prepared by incubating 1 ml of glutathione-Sepharose 4B beads (GE Healthcare) with either 20 mg of GST-galectin-3 or 10 mg of GST at 4°C for 4 h followed by extensive washing with PBS to remove unbound protein. An aliquot of the beads (5 l) was incubated with 50 l of 100 g/ml DTAF-RG-I-4 at room temperature for 1 h with agitation and then washed five times with PBS or PBS containing 200 mM lactose. The beads were transferred to glass microscope slides and viewed under an epifluorescence microscope (Olympus BX 51).

Flow Cytometry Assay
One hundred microliters of Jurkat cells (10 ϫ 10 6 cells/ml PBS) were incubated with 75 g/ml (i.e. 2.5 M) DTAF-galectin-3 in the absence or presence of 5 mg/ml lactose or RG-I-4 for 1 h at 25°C. After three washes with PBS, cells were measured on a FACScan flow cytometer (BD Biosciences). The results were representative of three independent experiments.

Surface Plasmon Resonance (SPR) Assay
The affinity of pectin for galectin-3 was measured with a BIAcore T100 instrument (GE Healthcare). Galectin-3 was coupled to an activated CM5 sensor chip by injecting 50 g/ml galectin-3, which was dissolved in 10 mM sodium acetate (pH 5.0), until an immobilized level of 8000 resonance units was obtained. The remaining active sites were blocked by injecting 1 M ethanolamine (pH 8.5). To determine the binding specificity, 0.3 mg/ml RG-I-4 or 1.65 mg/ml RG-I-4-RG was injected at 30 l/min for 90 s followed by PBS, 150 mM lactose, and a final injection of PBS. To determine the binding kinetics, five concentrations of each pectic sample (0.313, 0.625, 1.25, 2.5, and 5.0 M for RG-I-4 and potato galactan and 5, 10, 20, 40, and 80 M for MCP) were injected individually at a flow rate of 30 l/min with a single cycle kinetics program. Each concentration was injected for 90 s. Regeneration was performed with 150 mM lactose in PBS after all five concentrations had been injected. The control was obtained by injecting the same analyte solution on another channel that was immobilized with BSA. Data were analyzed with Biacore T100 evaluation software 2.0.1.

Other Analytical Methods
Sugar composition and methylation analyses were carried out according to methods that were described previously (32,34). One-dimensional 13 C nuclear magnetic resonance (NMR) and 1 H NMR spectra were obtained with a Bruker AV600 spectrometer (Germany) as described previously (32,34,42). Twodimensional correlation spectroscopy was recorded with a Bruker AV600 spectrometer at 298 K and 600 MHz with a spectral width of 3.15 KHz in two dimensions. High performance liquid chromatography (HPLC) was performed with a Shimadzu 10Avp HPLC system with the following columns: TSKgel G3000PW XL (7.8 ϫ 300 mm; Tosoh Corp.) eluted with 0.2 M NaCl at 0.5 ml/min, Sugar-PacK (6.5 ϫ 300 mm; Waters) eluted with 1 mM calcium-EDTA at 0.6 ml/min at 90°C, and Superose 6 10/300 GL (Amersham Biosciences) eluted with 0.15 M NaCl at 0.5 ml/min. Elution was monitored with a refractive index detector (RID-10A).
We further characterized the galactan portion of the side chains as galactan from other sources has been shown to bind galectin-3 (30,31). Methylation analysis showed that Gal residues were mainly terminal or 1,4-linkages at a 3:2 ratio, which is consistent with 13 C NMR results that RG-I-4 mainly contains ␤1,4-galactan side chains (34). The high ratio of terminal to 1,4-linked Gal suggested that the side chains of RG-I-4 might be short with some side chains containing only 1 Gal residue. This is in line with our previous estimation of 1.87 residues per chain (34). To evaluate the chain length, we used an antibody, LM5, which recognizes ␤1,4-galactan chains with at least four consecutive Gal residues (41). In this study, we classified ␤1,4-galactan side chains that contain Ն4 consecutive Gal residues as long chains and those containing Յ3 consecutive Gal residues as short chains. Intact RG-I-4 was recognized by LM5, indicating the presence of long side chains ( Table 2, RG-I-4-con). When 5% of Gal residues were removed by endo-␤1,4-D-galactanase, recognition of RG-I-4 by LM5 was reduced to 50% of the control ( Table 2, RG-I-4-endo-g). When 20% of the Gal residues were removed by ␤-galactosidase, recognition of RG-I-4 by LM5 was completely lost ( Table 2, RG-I-4-exo-g). These

TABLE 1 The inhibitory activities of ginseng pectins on galectin-3-mediated hemagglutination
Galectin-3-mediated hemagglutination was performed in the presence or absence of each test sample. The MIC for each sample is expressed as the mean Ϯ S.D.

Structures of Pectin Related to Galectin-3 Inhibition
data clearly demonstrate the correlation between removal of Gal residues and loss of antibody recognition. Furthermore, our results suggest that less than 20% of Gal residues are linked in long chains. Endo-␤1,4-D-galactanase, which has a long chain substrate (␤1,4-galactan), removed only a small percentage (5%) of Gal residues in RG-I-4 ( Table 2, RG-I-4endo-g), supporting the concept that most of the Gal residues are in short chains. However, under the same conditions, potato galactan, which contains long chains, was extensively hydrolyzed into oligosaccharides (supplemental Fig. S1a). Collectively, our previous data and the data presented here predict the structure for RG-I-4 shown in Fig. 2.
Galacto-oligosaccharides-Galacto-oligosaccharides were prepared from potato galactan, which contains mainly ␤1,4galactan (30). Oligosaccharides A-E were prepared via enzymatic digestion (supplemental Fig. S1a), and oligosaccharides F-M were obtained by partial acid hydrolysis (supplemental Fig. S1b). Oligosaccharides A-D were eluted as single and symmetrical peaks on a Sugar-PacK column at the positions of monomer, dimer, trimer, and tetramer, respectively (Fig. 3a). Structural analysis by 1 H NMR and two-dimensional 1 H NMR spectra (supplemental Figs. S2 and S3) confirmed that oligosaccharides A-D were Gal, Gal␤1,4-Gal, Gal␤1,4-Gal␤1,4-Gal, and Gal␤1,4-Gal␤1,4-Gal␤1,4-Gal, respectively. The assignment of each signal is listed in supplemental Table S1. Oligo E was a mixture with an average molecular mass of 1 kDa, which was equivalent to 6.1 sugar residues (6.1-mer) (Fig. 3b). Enzymatic digestion gave rise to few oligosaccharides that were longer than a tetramer. Longer oligosaccharides were prepared by partial acid hydrolysis. After removing the acidic portion with an anion exchange column, the neutral oligosaccharide portion, referred to as P-oligo, was obtained. P-oligo was composed of 96.7% Gal and 3.3% Ara residues and displayed a broad peak on a TSK-gel G3000 column (Fig. 3b). Linkage analysis by methylation showed that almost all of the Gal residues were in the form of 1,4-and terminal linkages, indicating that P-oligo consists of linear ␤1,4-galacto-oligosaccharides. The subfractions, oligosaccharides F-M, had similar sugar compositions and displayed narrower peaks on the TSK-gel G3000 column ( Fig. 3b). Their molecular masses ranged from 1.2 (oligosaccharide F) to 10.6 kDa (oligosaccharide M), which was equivalent to 7.3-65.3 Gal residues per oligosaccharide.

RG-I-4, a Ginseng Pectic Domain, Was a Potent Inhibitor in the G3H Assay
The G3H assay is widely used for evaluating potential galectin-3 inhibitors by measuring the MIC of sample for galectin-3-mediated hemagglutination (19). All ginseng pectins prepared from the native ginseng polysaccharides and the RG-I fragments prepared from endo-PG-digested ginseng pectin were examined (Table 1). Among the intact ginseng pectins, WGPA-1-HG exhibited the strongest inhibition of G3H with an MIC of 45 Ϯ 3.0 g/ml followed by WGPA-1-RG, which had an MIC of 91 Ϯ 3.8 g/ml. The weakest ginseng pectin was WGPA-4-HG, which had an MIC of 323 Ϯ 13.9 g/ml. In comparison, lactose, a standard galectin-3 inhibitor, showed an MIC of 10 Ϯ 1.3 g/ml. These results indicated that all intact ginseng pectins including HG-rich and AG-rich pectins could inhibit galectin-3 but in a relatively weak manner. Among the RG-I-rich fragments, the MICs of RG-I-2 and RG-I-3B were 60 Ϯ 3.8 and 120 Ϯ 6.3 g/ml, respectively, similar to the intact ginseng pectins and the commercially purchased potato RG-I fragments (MIC ϭ 181 Ϯ 6.3 g/ml) and larch wood AG (MIC ϭ 181 Ϯ 6.3 g/ml). In contrast, the MIC of RG-I-4 was 0.25 Ϯ 0.02 g/ml, which was significantly lower than other ginseng pectins. In fact, it was the lowest among all of the pectic samples examined including MCP (MIC ϭ 0.6 Ϯ 0.05 g/ml) and potato galactan (MIC ϭ 9.0 Ϯ 1.1 g/ml), which are two well established galectin-3 ligands. Thus, we identified an RG-I-rich fragment that is a potent inhibitor of galectin-3.

RG-I-4 Inhibited Galectin-3-mediated Cell Adhesion
RG-I-4-mediated inhibition of galectin-3 was further investigated at the cellular level. Galectin-3-mediated cell adhesion plays an important role in cancer progression and metastasis. Here, the effect of RG-I-4 on galectin-3-mediated HT-29 cell adhesion was examined and compared with that of lactose,  a The contents of Gal and Ara are shown relative to the control, which was set as 100. b Antibody response was determined by ELISA and is shown relative to the control, which was set as 100. c The -fold activity is shown relative to the control, which was set as 1. 1, increased; 2, decreased.
Remarkably, RG-I-4 showed 15-fold stronger inhibition than MCP with an IC 50 of 0.02 g/ml (Fig. 4d). In contrast, sucrose did not inhibit galectin-3-mediated cell adhesion at concentrations up to 800 g/ml (data not shown). The inhibition exerted by these samples was in an order similar to that observed in the G3H assay, which suggests that RG-I-4 is a potent inhibitor of galectin-3.

RG-I-4 Inhibited Galectin-3-related Cancer Cell Aggregation
Homotypic cell aggregation in the blood stream contributes to tumor cell embolization and is associated with increased metastasis (11,43). MCP has been shown to inhibit ASF-induced cell aggregation by disrupting interactions between galectin-3 on the cell surface and glycol ligands on ASF (18,40). We examined whether RG-I-4 inhibits ASF-induced cell aggregation (Fig. 5a). The inhibitory activity of RG-I-4 was compared with lactose, potato galactan, MCP, and sucrose. As shown in Fig. 5b, all of the tested samples except sucrose (data not shown) showed inhibitory effects. However, the effective doses for each compound were markedly different. Lactose and potato galac-tan did not inhibit aggregation until 10 g/ml. In contrast, RG-I-4 and MCP markedly inhibited aggregation at concentrations as low as 0.05 and 0.1 g/ml, respectively. These data indicated that RG-I-4 was a potent inhibitor of ASF-induced cell aggregation.

RG-I-4 Inhibited the Binding of Galectin-3 to T-cells
Galectin-3 binds to T-cells, resulting in inhibition of T-cell activation and induction of apoptosis. These events may contribute to the escape of tumor cells from the immune system (13). Here, we used flow cytometry to determine whether RG-I-4 inhibits binding of fluorescently labeled galectin-3 (DTAF-galectin-3) to Jurkat cells, a human T-lymphocyte line. In the absence of any inhibitors (Fig. 6a), a high density of DTAF-galectin-3 bound to cells as indicated by the high fluorescence intensity. In the presence of 5 mg/ml lactose (Fig. 6b), the fluorescence intensity decreased dramatically. In the presence of 5 mg/ml RG-I-4 (Fig. 6c), the fluorescence intensity was further decreased, indicating that RG-I-4 exerted stronger inhibition than lactose. Thus, RG-I-4 potently inhibited binding of galectin-3 to T-cells.

RG-I-4 Bound to the CRD of Galectin-3
To validate direct binding between RG-I-4 and galectin-3, we labeled RG-I-4 with a fluorescent dye, DTAF, and incubated  DTAF-RG-I-4 with galectin-3-coupled beads. After incubation, the beads were washed with PBS or 200 mM lactose and viewed under a fluorescence microscope. As shown in Fig. 7, DTAF-RG-I-4 did not bind to the control beads ( Fig. 7a) but bound to the beads with immobilized galectin-3 (Fig. 7b). Bound DTAF-RG-I-4 was completely washed off by lactose (Fig. 7c). These results demonstrate that RG-I-4 binds specifically to the CRD of galectin-3.
Binding of RG-I-4 to galectin-3 was examined further by SPR analysis with a Biacore T100 instrument. Galectin-3 was immobilized on a CM5 sensor chip, and RG-I-4 in PBS was allowed to flow over the chip. As shown in Fig. 8a, the response increased by ϳ150 resonance units after injection of 0.3 mg/ml RG-I-4 for 90 s (arrow A). The response remained largely unchanged when PBS was injected (arrow B), indicating that RG-I-4 was tightly associated with galectin-3. After the injection of 150 mM lactose (arrow C) followed by PBS (arrow D), the response returned to base line, indicating that RG-I-4 was completely washed off by lactose. This provided further evidence that binding of RG-I-4 was dependent on the CRD of galectin-3.

RG-I-4 Bound Galectin-3 with High Affinity
The binding kinetics of RG-I-4 with galectin-3 were determined by SPR and compared with those of MCP and potato galactan. Five concentrations were tested for each pectic polysaccharide with a "single cycle kinetics" program and analyzed with Biacore T100 evaluation software 2.0.1 (Fig. 8, b and c). The dissociation constants (K D ) derived from the association and dissociation rate constants were 22.2 nM for RG-I-4 (i.e. 1.33 g/ml based on its molecular mass of 60 kDa), 2590 nM for potato galactan (i.e. 38.9 g/ml based on its molecular mass of 272 kDa), and 143 nM for MCP (i.e. 61.5 g/ml based on its molecular mass of 30 kDa). These data indicate that the binding affinity of RG-I-4 for galectin-3 was significantly higher than that of MCP and potato galactan, which is consistent with our data showing potent inhibition of galectin-3-mediated effects by RG-I-4.

Ara Residues Regulate the Activity of RG-I-4
According to our previous work, Ara residues in RG-I-4 exist as 5-substituted or terminal residues within the ␣1,5-arabinan/ AG-I side chains (34). To study the roles played by Ara, we treated RG-I-4 with endo-␣1,5-L-arabinanase or ␣-L-arabinofuranosidase, yielding RG-I-4-endo-a and RG-I-4-exo-a, respectively. In comparison with buffer-treated RG-I-4 (RG-I-4-   (Table 2), respectively. Endo-␣1,5-Larabinanase treatment increased the G3H activity of RG-I-4 by 7.5-fold ( Table 2, RG-I-4-endo-a) in comparison with the control, suggesting that ␣1,5-L-arabinan chains are not required but may restrict the activity of RG-I-4. In contrast, ␣-L-arabinofuranosidase, which cleaves the terminal Ara residues in both ␣1,5-L-arabinan and AG-I chains, decreased the G3H activity of RG-I-4 by 2-fold ( Table 2, RG-I-4-exo-a). Together, these data indicate that Ara residues can positively or negatively regulate the activity of RG-I-4 depending on their locations. The ␣1,5-L-arabinan side chains impaired the activity, whereas the terminal Ara residues in the AG-I side chains promoted the G3H activity of RG-I-4.

␤1,4-Galactan Side Chains Played Important Roles in the Activity of RG-I-4
To study the roles played by Gal residues, we treated RG-I-4 with either ␤-galactosidase or endo-␤1,4-D-galactanase, yielding RG-I-4-exo-g and RG-I-4-endo-g, respectively. In comparison with the control, the Gal content in RG-I-4-exo-g and RG-I-4-endo-g was reduced by 20 and 5%, respectively ( Table  2). The activities of RG-I-4-exo-g and RG-I-4-endo-g in the G3H assay were decreased by 37.5-and 9.3-fold (Table 2), respectively, indicative of the important roles played by Gal residues. Because Gal residues in RG-I-4 are mainly located in the ␤1,4-galactan side chains, we concluded that these side chains are the functional structural elements.

The Contributions of the Length of ␤1,4-Galactan Side Chains to the Activity of RG-I-4
Structural analysis indicated that RG-I-4 contained mainly short chains of Յ3 Gal residues and only a small percentage of side chains with Ն4 Gal residues. When the long chains were partially removed, the activity decreased by 9.3-fold (Table 2) as in the case of RG-I-4-endo-g, which exhibited a 50% reduction in antibody recognition. When the long chains were completely removed, the activity decreased by 37.5-fold (Table 2) as in the case of RG-I-4-exo-g, which lost antibody recognition. These data demonstrate the important contributions that long chains make to RG-I-4 activities. However, even when the long chains were completely removed as in the case of RG-I-4-exo-g, activity still remained relatively high (MIC of 9.4 Ϯ 0.5 g/ml) and was equivalent to the activity of potato galactan (MIC of 9.0 Ϯ 1.1 g/ml). Thus, the short chains also contribute significantly to RG-I-4 activity.
To understand the correlation between chain length and activity, we prepared a series of ␤1,4-galacto-oligosaccharides from potato galactan and examined their activities in the G3H assay (Fig. 3c). The MIC decreased significantly for the tetramer versus monomer and changed little thereafter (Fig. 3c). The MIC of the tetramer was slightly lower than lactose (Fig. 3c,  asterisk). These data indicate that oligosaccharides that were longer than or equivalent to a tetramer had similar activities that were slightly better than lactose. The oligosaccharides that were shorter than a tetramer had 13-(trimer), 48.7-(dimer), and 185-fold (monomer) lower activities than lactose. These data suggest that a chain of 4 Gal residues meets the basic need for maximum activity. Based on these data, we deduce that a long side chain (greater than or equal to a tetramer) in RG-I-4 has slightly better activity than lactose, whereas a short chain (less than or equal to a trimer) has lower activity than lactose.

Multiple ␤1,4-Galactan Chains Determined the High Activity of RG-I-4
As shown above, the functional elements of RG-I-4 were ␤1,4-galactan side chains. However, RG-I-4 had 40 -1000-fold higher activity than a single ␤1,4-galactan chain. Why did RG-I-4 have such high activity? Structural analysis showed that an RG-I-4 molecule has 55 side chains, a feature distinct from other ginseng RG-I samples of low activity (e.g. RG-I-2 had 1.5 side chains, and RG-I-3B had 4 side chains; Table 3). To determine whether this feature is related to the extraordinary activity, we fragmented RG-I-4 by ␤-elimination, which cleaves pectic polysaccharides at the methyl-esterified GalA residues on the backbone (38). We hypothesized that if the cooperative action of the side chains contributes to the activity of RG-I-4 then fragments with fewer side chains would have much lower activity. As shown in Fig. 9, RG-I-4 was cleaved into fragments, which are referred to as RG-I-4␤. Each RG-I-4␤ fragment has an estimated molecular mass of 11 kDa and about 10 side chains. The 13 C NMR spectrum of RG-I-4␤ was similar to that of RG-I-4 except for slight differences corresponding to the GalA and Rha residues (Fig. 1). These data indicate that ␤-elimination occurred only on some GalA residues and did not alter the side chains as reported previously (38). The G3H assay showed that the activity of RG-I-4␤ was decreased by 8000-fold in comparison with untreated RG-I-4 (Table 2). Clearly, the number of side chains played a major role in the activity of RG-I-4. Based on the predicted molecular mass of 11 kDa, the MIC of RG-I-4␤ (2000 g/ml) was calculated to be 180 M, which is between the MIC of the dimer (366 M) and trimer (99

M).
This result is consistent with our conclusion that the majority of side chains in RG-I-4 are short.

The Backbone of RG-I-4 Regulated Its Activity
Two common modifications that occur naturally to the GalA residues of the RG-I backbone are methyl esterification and acetylation. Because these modifications were shown to often affect the activities of other pectins, we investigated whether they affect the activity of RG-I-4. We de-esterified RG-I-4 (RG-I-4-de) according to the published method (32). This treatment successfully removed methyl and acetyl groups as indicated by the disappearance of 13 C NMR signals at 53.4 ppm, which represents the methyl carbon of the methyl ester groups of GalA, and 22.0 ppm, which represents the methyl carbon of the acetyl groups in O-2 or O-3 of GalA (Fig. 1). The G3H assay showed that the inhibitory activity of the de-esterified RG-I-4 was decreased by 50-fold in comparison with untreated RG-I-4 ( Table 2). This experiment provided further evidence that the backbone of RG-I-4 plays important roles.

The Isolated Backbone of RG-I-4 Bound Galectin-3 in a CRD-independent Manner
Next, we isolated the backbone of RG-I-4, which is referred to as RG-I-4-RG, and investigated whether it interacts with galectin-3. We speculated that RG-I-4-RG would not have antigalectin-3 activity as it contained very few Gal residues. To our surprise, RG-I-4-RG inhibited activity in the G3H assay, although higher concentrations were required relative to intact RG-I-4 (Table 2). To understand its activity further, we measured interaction of RG-I-4-RG with galectin-3 by SPR. As shown in Fig. 8a, the response increased by ϳ100 resonance units after injection of 1.65 mg/ml RG-I-4-RG for 90 s (arrow A). The response dropped slightly when PBS was injected (arrow B), indicating that RG-I-4-RG was tightly associated with galectin-3. However, bound RG-I-4-RG could not be washed off by 150 mM lactose (arrow C), indicating that binding was independent of the CRD of galectin-3 (i.e. the binding was nonspecific). We questioned whether this nonspecific interaction facilitates binding of the whole RG-I-4 molecule to galectin-3. Based on the fact that the bond between galectin-3 and RG-I-4 could be completely broken by lactose (Fig. 8a), we con-cluded that binding between galectin-3 and RG-I-4 did not involve nonspecific interactions.

DISCUSSION
Inhibitors of galectin-3 that are derived from edible natural products are of important value because of their low toxicities. Pectin-derived inhibitors have been shown to inhibit various galectin-3-mediated events during tumor development and metastasis (18 -24, 27, 28). However, the structural elements of pectin that mediate inhibition of galectin-3 are poorly defined. In this study, we address this issue using a series of ginseng pectins and fragments that were prepared and characterized by our group (32)(33)(34).

The Inhibitory Activities and Binding Properties of RG-I-4 to Galectin-3
All of the fractions that were isolated from intact ginseng pectin were evaluated for their abilities to inhibit galectin-3 by the G3H assay. All of the fractions exhibited weak inhibition with MIC values ranging from 45 to 323 g/ml, although the Gal contents ranged from 3.5 to 56.2% (32,33). These results suggested that the activity of the pectin was not solely determined by its Gal content. Subsequently, we examined RG-I fragments isolated from endo-PG-treated ginseng pectin and found that one of the RG-I fragments, RG-I-4, strongly inhibited galectin-3-mediated hemagglutination. Its effect was 40 times stronger than lactose and stronger than all reported inhibitors derived from pectic sources. RG-I-4 also inhibited galectin-3-mediated cell adhesion, homotypic cell aggregation, and galectin-3 binding to T-cells.
Binding of RG-I-4 to galectin-3 was investigated using microscopy and SPR. Both methods demonstrated that RG-I-4 specifically bound to the CRD of galectin-3. The affinity of RG-I-4 for galectin-3 was higher than for other pectic polysaccharides, consistent with its high inhibitory activity of galectin-3.
It is interesting that intact ginseng pectin has low activity, whereas modification (endo-PG digestion in this case) generated highly active components such as RG-I-4. This phenomenon was also observed in the case of citrus pectin, which is activated by pH modification, although the parental material is inactive. The active elements might be concealed in the parental material by other sugars, linkages, or the overall conformation of the molecules.

Correlation of the Structure and Activity of RG-I-4
The Effects of the Side Chains of RG-I-4 on Binding to Galectin-3-The results of the G3H assay showed that hydrolyzing Ara residues either slightly impaired or enhanced the activity depending on the location of the residues. Thus, Ara residues regulate RG-I-4 activity. In contrast, hydrolysis of Gal residues markedly decreased RG-I-4 activity, suggesting that Gal residues play a pivotal role. Because Gal residues in RG-I-4 are located in the galactan side chains, we concluded that these side chains are the functional structural elements. This is consistent with previous reports that showed that galactans from potato pectin (30) and MCP (31) interact with galectin-3. It has been reported that potato galactan consists of long ␤1,4-galactan chains with an average chain length of ϳ22 residues (30). We confirmed the presence of long chains by isolating galacto-oligosaccharides as long as 63.5 residues from potato galactan (Fig. 3b). MCP galactan also consists of long ␤1,4-galactan chains with ϳ110 Gal residues (18 kDa) according to our previous study (31). By analogy, we initially speculated that RG-I-4 activity is regulated by its long galactan side chains. However, structural analysis indicated that RG-I-4 contained mainly short chains of Յ3 Gal residues and only a small percentage of side chains with Ն4 Gal residues. Furthermore, when the long chains (Ն4 Gal residues) were completely hydrolyzed, activity was considerably reduced but still high. Thus, although the long chains contribute to its activity, the short chains play major roles in regulating the activity of RG-I-4 in contrast to potato and MCP galactans.
The above discovery led us to further investigate the relationship between chain length and RG-I-4 activity. Because of the difficulty of isolating individual side chains in reasonable quantities, we prepared various ␤1,4-galacto-oligosaccharides from potato galactan. We found that the activity was significantly increased for the tetramer versus the monomer with only a slight change in activity for long chains up to a 63.5-mer. Based on these data, we deduced that the activity of a single chain in RG-I-4 is proportional to its length up to 4 Gal residues. Furthermore, all long chains with Ն4 Gal residues had similar activities.
It has been proposed that the CRD of galectin-3 possesses an extended binding site that can accommodate 4 sugar residues. Our data showing that the activity increased as the chain lengthened from a monomer to a tetramer of Gal residues fit well with this model. Thus, we provide experimental evidence supporting the proposed mechanisms of binding. The remaining question is whether galectin-3 recognizes the internal or terminal tetrasaccharide within a galactan chain. According to our studies on MCP galactan that showed that the activities are proportional to the number of chain termini (31), we propose that galectin-3 recognizes terminal tetrasaccharides. In the case of poly(LacNAc), galectin-3 appears to primarily recognize internal LacNAc units as stated by the authors (44).
The Cooperation of the Side Chains-It is intriguing that RG-I-4 had significantly higher activity than other galactans. Further studies demonstrated that the high activity of RG-I-4 resulted from the cooperative actions of the side chains. This conclusion is based on four lines of evidence. First, RG-I-4 is a highly branched molecule with galactan side chains, providing the opportunity for multivalent interactions. Second, an individual side chain has significantly lower activity than the whole molecule. Third, cleaving the backbone yielded fragments with fewer side chains that had significantly lower activity. Finally, modification of the backbone, which might sterically affect the orientation of the side chains, significantly decreased the activity.
How does cooperation between side chains affect the activity of pectin? In contrast to many carbohydrate-binding proteins, galectin-3 has one binding site. However, upon binding to multivalent ligands, galectin-3 forms multimers and acquires multiple binding sites (5). As a result, the avidity of a multivalent ligand for galectin-3 is dramatically increased because there are multiple binding sites on galectin-3 and its ligand. Multivalency of the ligand is more prominent in the case of galectin-3 than for other carbohydrate-binding proteins. Thus, it is not surprising that the multivalency and cooperativity of the RG-I-4 side chains become dominating factors determining pectin activity.
The mode of cooperative action explains the relatively low activity of MCP galactans (31). The activity of MCP galactans (MIC ϭ 50 g/ml) falls in the range of galacto-oligosaccharides. As demonstrated by our previous study, MCP galactans are linear chains (31), suggesting that they are monovalent and unable to induce multimerization of galectin-3. The lack of multivalency in both MCP galactan and galectin-3 is associated with low affinity. In the case of potato galactan, the situation is more complicated. The activity of potato galactan is similar to its long oligosaccharide derivatives. Although there are multiple side chains, they do not appear to work cooperatively. The location and/or orientation of the side chains might not be suitable for binding and inducing formation of a galectin-3 oligomer. An interesting future goal is to find out which pectin structures can induce oligomerization.
The Roles of the Backbone-The significant decrease in activity after cleavage and modification of the backbone clearly demonstrated the involvement of the backbone in the activity of the whole molecule. The backbone maintains a proper structural conformation so that the side chains can work cooperatively. Although nonspecific interaction between galectin-3 and the isolated backbone was observed, such nonspecific interaction was not observed between galectin-3 and the intact molecule. The backbone in RG-I-4 has a high branching frequency of 0.67; i.e. there are two side chains for every three disaccharide (␣1,2-Rha-␣1,4-GalA-) units (34). Such high branching frequency might completely conceal the active groups of the backbone.
How do we understand the interaction between galectin-3 and the isolated backbone? In this study, the G3H and SPR experiments were performed at pH 7.2, a physiological condition. At this pH, galectin-3 was positively charged (30), whereas the isolated backbone RG-I-4-RG, which contains 49% GalA, was negatively charged. Thus, the association between galectin-3 and the isolated backbone might be based on ionic interaction. Protein-pectin interactions based on ionic charge were observed for galectin-3 and other pectic polysaccharides such as potato RG-I and citrus homogalacturonic acid (30). However, our preliminary experiments revealed that DTAF-labeled RG-I-4-RG could not be eluted from galectin-3-immobilized beads by a high concentration of NaCl up to 4 M (data not shown). Thus, ionic interaction might not be an appropriate explanation. There might be secondary carbohydrate recognition sites within the galectin-3 molecule. Investigations to solve this issue are going on in our research group.
More Evidence Derived from Comparative Analyses of RG-I-2, RG-I-3B, and RG-I-4-RG-I-2, RG-I-3B, and RG-I-4 are RG-I-type ginseng pectins. RG-I-2 and RG-I-3B showed much weaker activity than RG-I-4. Table 3 lists the structural data for these pectins and is based on data from a previous publication (34) and from analyses in the current study. We compared the structures and found that (a) the content of Gal and the ratio of Gal/Ara in RG-I-4 (19.5% and 2.12, respectively) were about twice those in RG-I-3B (13.7% and 1.15) and RG-I-2 (12.4% and 0.86). (b) All Gal residues in RG-I-4 resided in the AG-I-type side chains, whereas Gal residues in RG-I-2 and RG-I-3B resided in either AG-I-or AG-II-type (i.e. ␤1,3,6-D-galactan or arabino-␤1,3,6-D-galactan) side chains. (c) 55 side chains were linked to C-4 of Rha in the RG-I-4 molecule, which was dramatically higher than RG-I-3B and RG-I-2 (4 and 1.5, respectively). These data suggest that high Gal content, the Gal/Ara ratio, the AG-I-type side chains, and the total number of side chains on the backbone play important roles. Thus, the information gained by comparing these pectins was consistent with the experimental results obtained in this study.