The Plant Cell Wall Polysaccharide Rhamnogalacturonan II Self-assembles into a Covalently Cross-linked Dimer*

The location of the 1:2 borate-diol ester cross-link in the dimer of the plant cell wall polysaccharide rhamnogalacturonan II (RG-II) has been determined. The ester cross-links the apiofuranosyl residue of the 2-O-methyl-d-xylose-containing side chains in each of the subunits of the dimer. The apiofuranosyl residue in each of the two aceric acid-containing side chains is not esterified. The site of borate esterification is identical in naturally occurring and inin vitro synthesized dimer. Pb2+, La3+, and Ca2+ increase dimer formationin vitro in a concentration- and pH-dependent manner. Pb2+ is the most effective cation. The dimer accounts for 55% of the RG-II when the monomer (0.5 mm) is treated for 5 min at pH 3.5 with boric acid (1 mm) and Pb2+ (0.5 mm); at pH 5 the rate of conversion is somewhat slower. Hg2+ does not increase the rate of dimer formation. A cation’s charge density and its ability to form a coordination complex with RG-II, in addition to steric factors, may regulate the rate and stability of dimer formation in vitro. Our data provide evidence that the structure of RG-II itself determines which apiofuranosyl residues are esterified with borate and that in the presence of boric acid and certain cations, two RG-II monomers self-assemble to form a dimer.

A single 1:2 borate-diol ester is believed to cross-link two RG-II molecules. 1 mol of the dimer contains 1 mol of boron (3,4,(12)(13)(14). We have proposed that two of the four 3Ј-linked apiofuranosyl (Apif) residues present in dRG-II-B are crosslinked by borate because the dimer contains approximately equimolar amounts of 3Ј-and 2,3,3Ј-linked Apif residues, whereas monomeric RG-II (mRG-II) contains only 3Ј-linked Apif residues (4,12). However, two different Apif-containing side chains are attached to the backbone of each RG-II monomer (chains A and B in Fig. 1). It is not known whether the borate cross-link involves the Apif residue in one of each of the two types of Apif-containing side chains or if the borate is attached to the same type of Apif-containing side chains in each monomer.
The ability to form dRG-II-B from mRG-II and boric acid in vitro provides a convenient model system to examine the mechanism of dimer formation (4). In addition, such studies are likely to provide information on the ability of dRG-II-B, which is present in fermented beverages such as wine, to form complexes with heavy metals (15). We have suggested that steric factors regulate dimer formation because only divalent cations with an ionic radius Ͼ1.1 Å (e.g. Pb 2ϩ , Sr 2ϩ , and Ba 2ϩ ) increase the rate of dRG-II-B formation significantly (4). However, the function of cations in dimer formation has not been established, nor is it known if the 1:2 borate-diol ester is located on the same glycosyl residue(s) in naturally occurring and in vitro synthesized dRG-II-B. We now report that the same two Apif residues are the sites of borate esterification in naturally occurring and in vitro synthesized dRG-II-B. The effects of selected mono-, di-, tri-, and tetravalent cations on dRG-II-B formation are described as are the abilities of Pb 2ϩ , Ca 2ϩ , and La 3ϩ to increase the rate of dimer formation from mRG-II and boric acid.

EXPERIMENTAL PROCEDURES
Isolation and Purification of RG-II-dRG-II-B was isolated from the cell walls of sugar beet tubers (3), potato tubers (16), bamboo shoots (17), and red wine (12) as described previously.
Formation of mRG-II and dRG-II-B-mRG-II was generated by treating dRG-II-B (50 mg) for 30 min at room temperature with 0.1 M HCl (10 ml). The solution was dialyzed (1,000 molecular weight cutoff) at 4°C against deionized water, and the RG-II was then converted to its sodium form by elution through a column (1 ϫ 5 cm) containing Chelex-100 (Na ϩ form, Bio-Rad). The eluant was then freeze dried.
dRG-II-B was generated in vitro by treating mRG-II (10 mg) for 4 days at room temperature with 50 mM potassium phthalate (10 ml), pH 3.5, containing 15 mM boric acid. dRG-II-B/Pb 2ϩ was generated by treating mRG-II (10 mg) for 24 h at room temperature with 50 mM potassium phthalate (5 ml), pH 3.5, containing 1 mM boric acid and 0.5 mM Pb(OAc) 2 . The solutions were dialyzed separately (1,000 molecular weight cutoff) against deionized water and freeze dried. The boron content of mRG-II, dRG-II-B, and dRG-II-B/Pb was determined by inductively coupled plasma atomic emission spectroscopy (4) and by inductively coupled plasma mass spectrometry (3).
Methylation and Carboxyl Reduction of Methylated RG-II-Separate solutions of mRG-II and dRG-II-B (ϳ5 mg) in water (5 l) were mixed with dimethyl sulfoxide (1 ml) and the mixtures purged with argon for 2 h at room temperature to remove air (18). Potassium methylsulfinylmethanide (200 l, 2.5 M in dimethyl sulfoxide) was then added, and the solutions were kept at room temperature for 2 h. The mixtures were frozen in ice, and then methyl iodide (220 l) was added dropwise. The solutions were kept for 2 h at room temperature to generate the methylated polysaccharide (prolonged exposure (Ͼ3 h) of the dimer to the methylating reagents resulted in partial cleavage of the borate ester). The solutions were then diluted with an equal volume of water and flushed with argon to remove the excess methyl iodide. The solutions were then dialyzed (1,000 molecular weight cutoff) for 24 h against deionized water and freeze dried. The methyl-esterified carboxyl groups of the methylated polysaccharide were carboxyl-reduced with Superdeuteride (1 ml, Aldrich) (19). The methylated, carboxyl-reduced polysaccharide was desalted by dialysis and then freeze dried.
Generation of the Rha-apiitol Derivatives by Partial Fragmentation of Methylated and Carboxyl-reduced RG-II-The methylated, carboxylreduced polysaccharide (ϳ 1 mg) was partially fragmented by treatment for 2 h at 70°C with 88% formic acid (300 l). This treatment hydrolyzes the 1:2 borate-diol ester thereby exposing the hydroxyls to which it was attached and also generates a mixture of partially methylated oligoglycoses. The formic acid was removed under a flow of nitrogen gas by coevaporation with isopropyl alcohol (2 ϫ 500 l). The partially methylated oligoglycoses that were generated were then converted to their corresponding partially methylated oligoglycosyl alditols by treatment for 4 h at room temperature with aq. 20% methanol (350 l) containing NaBD 4 (10 mg/ml). Excess NaBD 4 was destroyed by dropwise addition of glacial acetic acid, and the resulting solution was concentrated to dryness under a flow of nitrogen gas. The residue was treated with methanol containing 10% acetic acid (4 ϫ 500 l) and then with methanol (4 ϫ 500 l). The free hydroxyl groups of the methylated oligoglycosyl alditols were acetylated by treatment for 3 h at 120°C with acetic anhydride (100 l). Water (500 l) was added, and the acetic acid that formed was neutralized by the addition of solid sodium carbonate. Dichloromethane (1 ml) was added, and the organic and aqueous phases were separated by centrifugation. The organic phase was concentrated to dryness, and the residue was dissolved in acetone (20 l) before analysis by gas-liquid chromatography with mass spectrometry (GLC-MS).
Analytical Methods-GLC-MS was performed with a JEOL JMS-DX303HF mass spectrometer interfaced with a Hewlett-Packard 5890 gas chromatograph. GLC-chemical ionization MS (GLC-CI-MS) was performed with ammonia as the reagent gas and a source temperature of 150°C. GLC-electron impact ionization MS (GLC-EI-MS) was performed with an ionization current of 70 eV and an ion source temperature of 180°C. The partially methylated and O-acetylated oligoglycosyl alditols were separated using a DB-1 column (15 m ϫ 0.25 mm) as described (13). 11 B nuclear magnetic resonance spectroscopy ( 11 B NMR) was performed with a JEOL JNM-A600 spectrometer operating at 192 MHz and 25°C (3). Samples were analyzed in quartz NMR tubes without spinning or a field lock. Chemical shifts (␦) are reported in ppm relative to external boric acid (␦ 0 ppm).

A Chemical Procedure to Locate the Borate-esterified Apiosyl
Residues in dRG-II-B-Two of the four Apif residues in dRG-II-B have been shown, by glycosyl linkage composition analysis, to be the probable sites of borate esterification (4,12,13). There are three ways the borate ester can cross-link two RG-II monomers. The borate could cross-link an Apif residue of side chain A to B or to another A, or two B side chains may be cross-linked. We describe in the following paragraphs experiments designed to determine which apiosyl residues are crosslinked by the borate ester using a chemical method (see Fig. 2) originally developed to sequence complex carbohydrates (19).
The 11 B NMR spectrum of methylated dRG-II-B in dimethyl sulfoxide contains a signal at ␦ 8.8 which corresponds to a 1:2 borate-diol ester (2)(3)(4) and establishes that the cross-link had not been hydrolyzed during the methylation reaction. Borateesterified Apif residues are not methylated, whereas unesterified Apif residues are methylated at O-2 and O-3 ( Fig. 2, step i). The methylated, carboxyl-reduced RG-II dimer was then partially fragmented by treatment with formic acid (Fig. 2, step ii). This treatment hydrolyzes the 1:2 borate-diol ester thereby exposing the hydroxyls to which it was attached and also generates a mixture of partially methylated oligoglycoses including the partially methylated disaccharide Rhap-(133Ј)-Api. The Rhap residue linked to the Apif residue from side chain A has no O-methyl groups (Fig. 2, step i) whereas the Rhap residue linked to Apif residue from side chain B has O-methyl groups at O-2 and O-4 ( Fig. 2, step i). This provides a way to identify the side chain from which each Rhap-(133Ј)-Api originated. The methylated oligosaccharides were converted, by reduction with NaBD 4 , to their corresponding partially methylated oligoglycosyl alditols (Fig. 2, step iii) and the free hydroxyl groups then O-acetylated (Fig. 2, step iv).
The  3C) and thus originated from an Apif residue that was not esterified with borate. The absence of the ion at m/z 308 (aldJ 1 ) indicates that the Rhap residue is not methylated at O-3 and thus must have been originally substituted at O-3 (19). No evidence was obtained by GLC-EI-MS for the presence of the Rhap-(133i)-apiitol derivative that would originate from a borate-esterified Apif residue of side chain B. The apiitol of such a derivative would be acetylated at O-2 and O-3 and thus differs from derivative 3, which is methylated at O-2 and O-3 (see Fig. 2). These results provide strong evidence that in naturally occurring dRG-II-B the Apif residue of the aceric acid-containing side chain (side chain B in Fig. 1) is not esterified with borate.
Two Rhap-(133Ј)-apiitol derivatives were generated from methylated, carboxyl-reduced mRG-II which were shown by GLC-CI-MS to have [M ϩ NH 4 ] ϩ ions at m/z 499 (3) and 555 (2). The derivative (3) Fig. 1). The results establish that neither of the two Apif residues in mRG-II is esterified with borate. Indeed, mRG-II contains no borate. It is important that these results demonstrate that our chemical procedure does distinguish between borate-esterified and unesterified apiosyl residues.
The disaccharide ␣-L-Rhap-(135)-KdopA is attached to C-3 of one of the RG-II backbone GalpA residues (20). The Kdo residue has cis hydroxyl groups at C-7 and C-8 which are a potential site for borate esterification. The following experiments were performed to determine if the Kdo is borate-esterified. The partially methylated and O-acetylated monoglycosyl alditol derivative of Rhap-(135)-Kdo was generated by partial acid hydrolysis of methylated, carboxyl-reduced mRG-II and from naturally occurring dRG-II-B. The GLC-CI mass spectrum of the Rhap-(135)-Kdo'ol derivative contained a [M ϩ NH 4 ] ϩ ion at m/z 603 (data not shown) irrespective of whether it was generated from mRG-II or dRG-II-B. The mass of this ion establishes that Rhap-(135)-Kdo'ol did not originate from borate-esterified Rhap-(135)-Kdo. No evidence was obtained by GLC-CI-MS for a Rhap-(135)-Kdo'ol derivative with a [M ϩ NH 4 ] ϩ ion at m/z 659, the mass of the ion expected for a monoglycosyl 3-deoxyoctitol originating from borate-esterified Rhap-(135)-Kdo. Thus, we conclude that the Kdo residue in RG-II is unlikely to be cross-linked by a borate ester.
The Location of the Borate Ester Is Identical in Naturally Occurring and in Vitro Synthesized dRG-II-B-The mechanism of dRG-II-B formation in plants is not known (4,5). Thus the location of the 1:2 borate-diol ester may differ in naturally occurring and in in vitro synthesized dRG-II-B. We performed an experiment to determine whether the location of the borate ester is the same in naturally occurring and in vitro synthesized dRG-II-Bs.
Dimeric RG-II-B isolated from sugar beet, potato, bamboo shoots, and red wine, dRG-II-B/Pb synthesized from mRG-II, boric acid and Pb 2ϩ , and dRG-II-B synthesized from mRG-II and boric acid were methylated and carboxyl reduced. The Rhap-(133Ј)-apiitol derivatives were generated (see Fig. 2 (Table I).
Additional evidence that our chemical procedure is suitable for determining the presence and location of borate-esterified Apif residues in dRG-II-B was obtained by the analysis of a presumptive mRG-II generated from wine dRG-II-B. Rha-(133Ј)-apiitol derivative 1 (see Fig. 3A) was generated from the methylated and carboxyl-reduced wine mRG-II, which suggested the presence of a borate-esterified apiosyl residue. Indeed, the dimer was shown, by SEC, to account for ϳ20% of the wine mRG-II (see Table I). Taken together our results provide evidence that the borate ester cross-links the Apif residues of the A side chains irrespective of whether the dRG-II-B is obtained from a natural source, synthesized from mRG-II in the presence of boric acid, or synthesized from mRG-II in the presence of boric acid and a divalent cation.
Di-and Trivalent Cations Increase the Rate of Formation of dRG-II-B-We have provided evidence that the location of the 1:2 borate-diol ester is identical in naturally occurring and in vitro synthesized dRG-II-Bs. Thus, the mechanism of dimer formation can be analyzed using in vitro synthesis of dRG-II-B. In a previous study we showed that only divalent cations with an ionic radius of Ͼ1.1 Å increased dimer formation in vitro (4). We now provide evidence that the ionic radius of the cation is only one of several factors that regulate dimer formation in vitro.
Divalent cations (Sr 2ϩ , Pb 2ϩ , and Ba 2ϩ ) with an ionic radius Ͼ1.10 Å and trivalent cations (Eu 3ϩ , Pr 3ϩ , La 3ϩ , and Ce 3ϩ ) with an ionic radius Ͼ0.90 Å significantly increased the amount of dimer formed in 24 h, whereas Ca 2ϩ and Cd 2ϩ (ionic radius of 0.99 and 0.95 Å, respectively) caused only a small increase in the amount of dimer formed (Table II). Somewhat unexpectedly, Hg 2ϩ , which has an ionic radius of 1.10 Å, has no discernible effect on the amount of dimer that formed (Table  II). This may result from the fact that Hg 2ϩ typically does not form stable coordination complexes with oxygen-donor ligands such as RG-II (21). Those cations that do increase the amount of dimer formed all have an affinity for oxygen-donor ligands (21). These results provide additional evidence that the cationdependent increase in the rate of dimer formation in vitro most likely involves the formation of an RG-II-cation coordination complex. Thus, charge, ionic radius, and ligand-donor-atom selection are all factors that determine whether a particular cation will increase the amount of dRG-II-B formed in vitro.

dRG-II-B Formation from mRG-II and Boric Acid Is Rapid in the Presence of Selected Di-and Trivalent
Cations-We showed previously that certain divalent cations increase the rate of dRG-II-B formation in vitro, although the rate achieved was relatively slow (4). We now show that dRG-II-B is formed within minutes from mRG-II and boric acid at pH 3.5 when in the presence of an appropriate divalent or trivalent cation.
Pb 2ϩ and La 3ϩ both induce a rapid, concentration-dependent increase in the rate of dRG-II-B formation at pH 3.5 (Fig. 4, A  and B). Within 5 min in the presence of 0.5 mM Pb 2ϩ , ϳ55% of   the monomer is converted to the dimer, and Ͼ90% conversion occurs within 1 h (Fig. 4A). La 3ϩ is less effective than Pb 2ϩ because only 60% of the monomer is converted to the dimer in 1 h, and the conversion is ϳ80% after 6 h (Fig. 4B). Ca 2ϩ is considerably less effective than both Pb 2ϩ and La 3ϩ (Fig. 4C). The rate of dimer formation in the presence of Pb 2ϩ and La 3ϩ is somewhat slower at pH 5 (Fig. 4, D and E). Ca 2ϩ , even at high concentration (50 mM), is again considerably less effective than both Pb 2ϩ and La 3ϩ at pH 5 (Fig. 4F). Dimer formation is barely detectable within 6 h at pH 3.5 or 5 in the absence of added cations under the conditions of our experiments (Fig. 4,  A and D), confirming that dimer formation in vitro is pH-and cation-dependent (4).

The Location of the Borate Ester Cross-link in dRG-II-B-We
have provided evidence that a single 1:2 borate-diol ester in dRG-II-B cross-links the 3Ј-linked Apif residues of the 2-O-MeXyl-containing side chains of the two mRG-II subunits but does not cross-link the 3Ј-linked Apif residues of the aceric acid-containing side chains. The location of the B ester is the same in dRG-II-B isolated from natural sources and synthesized in vitro. However, 1:2 borate-apiose esters can exist in either of two diastereomeric forms (bis(␤-D-Apif)-(R)-2,3:2,3 and -(S)-2,3:2,3-borate). Indeed, both diastereoisomers are formed when methyl-␤-D-apioside is reacted with borate. 2 It is not known if naturally occurring and in vitro synthesized dRG-II-B contain the same diastereoisomer. Nevertheless, we conclude that, irrespective of the diastereoisomer formed, the structure of RG-II itself determines which Apif residues are esterified with borate.
We have reported previously that the maximum rate of dRG-II-B formation in vitro occurs between pH 3 and 4 and in the presence of selected divalent cations (4). In contrast, the 1:2 borate-diol esters of methyl-␤-D-apioside form only above pH 5.2 even in the presence of cations. 2 Thus, the pH-and cationdependent formation of Apif 1:2 borate-diol esters in dRG-II-B is determined by the structural characteristics of RG-II. We propose that the charge density of side chain A, which contains three uronosyl residues (see Fig. 1), may explain, in part, the requirement for divalent cations in dimer formation. Additional factors, including the conformations of both the 2-O-MeXyl-and aceric acid-containing side chains, are likely to contribute to the specific location of the borate ester.
dRG-II-B Formation in Vitro and in Muro May Be Promoted by Different Divalent Cations-The rate of cation-dependent dimer formation in vitro and the rate of dimer formation in suspension-cultured Chenopodium album cells are comparable. 3 However, those cations that promote dimer formation in vitro (see Table II) are unlikely to be present at concentrations sufficiently high to promote dimer formation in muro (22,23). In contrast, calcium is present in plant cell walls at mM concentrations (24), although most (Ͼ95%) of this calcium is bound, and the "free" calcium content of the wall is typically Ͻ5 mM. Furthermore, calcium ions have been reported to stabilize the borate ester cross-link in muro. 3 Nevertheless, low concen-trations of Ca 2ϩ (0.5 mM) do not promote rapid dimer formation in vitro under the conditions of our experiments, although higher concentrations (Ͼ5 mM) are somewhat effective (see Fig.  4, C and F). We suggest that the roles of divalent cations in regulating borate ester cross-linking of RG-II in vitro and in muro are not the same because soluble and wall-bound mRG-II differ.
The soluble mRG-II used to form the dimer in vitro and wall-bound RG-II are not identical. Soluble mRG-II has a backbone that contains between 7 and 15 1,4-linked ␣-D-galacturonosyl residues (12,25), whereas the wall-bound RG-II backbone is believed to be covalently inserted within a homogalacturonan chain (1). Calcium ions may interact with both homogalacturonan and RG-II and thereby promote cross-link formation in muro. Moreover, wall-bound mRG-II molecules may be structurally constrained in a manner that favors borate ester cross-link formation. In contrast, borate ester formation in vitro is dependent on the direct interaction of di-and trivalent cations with RG-II itself. This interaction is determined, in part, by steric factors, because only divalent cations with ionic radii Ͼ1.10 Å and trivalent cations with ionic radii Ͼ0.95 Å promote dimer formation in vitro (Table II). Di-and trivalent cations may also stabilize the borate ester cross-link because treating naturally occurring dRG-II-B with EDTA results in the slow but discernible formation of mRG-II. 3 Such results are consistent with a previous report showing that calcium ions form coordination complexes with and stabilize the 1:2 boratediol esters of glucaric acid (26).
dRG-II-B Is Formed by the Self-assembly of Two mRG-II Molecules-dRG-II-B formation in muro must result from either a spontaneous self-assembly or an enzymically catalyzed process. Our results do not preclude that borate ester formation is enzymically catalyzed in muro. However, we have shown that in the presence of boric acid and certain cations, two RG-II monomers rapidly self-assemble to form a dimer and that the structure of RG-II itself may determine the location of the borate ester.
The ability of plant cell wall polysaccharides to self-assemble into ordered structures has become the subject of considerable debate (27). For example, the parallel arrangement of 1,4linked ␤-D-glucan chains in naturally occurring cellulose is believed to result in large part from the organization of the membrane-bound cellulose synthases because the spontaneous assembly of parallel glucan chains is entropically unfavorable (27). The formation of ordered structures is a characteristic of homogalacturonan because this polysaccharide spontaneously forms gels in the presence of calcium. Cellulose formation and the calcium-dependent gelation of homogalacturonan result from noncovalent interchain bonding (27). Moreover, a glycosyl residue in one glucan or galacturonan chain may interact with any of the glycosyl residues in a second chain. In contrast, RG-II dimer self-assembly requires the formation of a covalent borate ester cross-link between the Apif residue of the same side chain (Fig. 1, side chain A) in each mRG-II subunit. The specificity and cation dependence of this cross-linking suggest that there are precise structural requirements for dRG-II-B formation, and this may explain why the structure of RG-II is highly conserved in higher plants (1).
Borate Ester Cross-linking of RG-II May Alter the Mechanical Properties of the Plant Cell Wall-We have provided evidence that a single borate ester cross-links two mRG-II molecules and that the location of the ester is the same in dRG-II-B isolated from different plants. The results of numerous studies suggest that the boron requirement and wall pectin content are correlated in many plants (28,29) and that boron is required to maintain the mechanical properties of the primary wall (6, 8 -11, 30, 31). For example, borate ester cross-linking of RG-II results in a rapid decrease in the wall pore size of suspensioncultured plant cells. 3 Moreover, ester cross-link formation is required to prevent the walls from rupturing when the cells enter the stationary phase of their growth. 3 Thus, the mechanical properties of the cell wall may be determined in large part by the macromolecular pectin network that most likely forms when RG-II is cross-linked by a borate ester. Nevertheless, additional roles for borate ester cross-linking of RG-II cannot be excluded.
In summary, we have shown that a single borate ester crosslinks the Apif residue of the 2-O-MeXyl-containing side chain of each mRG-II subunit in naturally occurring and in vitro synthesized dRG-II-B. Dimer formation in vitro is pH-dependent and occurs within minutes in the presence of certain di-and trivalent cations. RG-II is, to the best of our knowledge, the first example of a plant cell wall pectic polysaccharide that self-assembles to form structurally identical dimers.