Purification, Characterization, Molecular Cloning, and Expression of Novel Members of Jacalin-related Lectins from Rhizomes of the True Fern Phlebodium aureum (L) J. Smith (Polypodiaceae)*

A lectin was purified from rhizomes of the fernPhlebodium aureum by affinity chromatography on mannose-Sepharose. The lectin, designated P. aureum lectin (PAL), is composed of two identical subunits of ∼15 kDa associated by noncovalent bonds. From a cDNA library and synthetic oligonucleotide probes based on a partial amino acid sequence, 5′- and 3′-rapid amplification of cDNA ends allowed the generation of two similar full-length cDNAs, termed PALa and PALb, each of which had an open reading frame of 438 bp encoding 146 amino acid residues. The two proteins share 88% sequence identity and showed structural similarity to jacalin-related lectins. PALa contained peptide sequences exactly matching those found in the isolated lectin. PALa and PALb were expressed in Escherichia coli using pET-22b(+) vector and purified by one-step affinity chromatography. Native and recombinant forms of PAL agglutinated rabbit erythrocytes and precipitated with yeast mannan, dextran, and the high mannose-containing glycoprotein invertase. The detailed carbohydrate-binding properties of the native and recombinant lectins were elucidated by agglutination inhibition assay, and native lectin was also studied by isothermal titration calorimetry. Based on the results of these assays, we conclude that this primitive vascular plant, like many higher plants, contains significant quantities of a mannose/glucose-binding protein in its storage tissue, whose binding specificity differs in detail from either legume mannose/glucose-binding lectins or monocot mannose-specific lectins. The identification of a jacalin-related lectin in a true fern reveals for the first time the widespread distribution and molecular evolution of this lectin family in the plant kingdom.

true ferns, have been largely overlooked in the study of lectins.
Lectins of known specificity serve as valuable reagents in glycobiological research. They can be employed for the detection and preliminary characterization of glycoconjugates on the surface of cells. Although many lectins belong to the same major specificity group of mannose-or mannose/glucose-binding lectins, their different reactivities toward more complex oligo-and polysaccharides render many of them specifically valuable for recognizing a particular type of saccharide structure and fuel the search for yet more novel lectins (3). Lectins are found in greatest quantity and are most readily purified from plant sources, especially storage tissues such as seeds, bark, bulbs, rhizomes, etc. Many lectins have been isolated and characterized from angiosperm subdivision of seed plants. On the basis of structural and evolutionary development, most of these plant lectins have been classified into seven families: legume lectins, chitin-binding proteins, type 2 ribosome-inactivating proteins, monocot mannose-binding lectins, amaranthins, curcurbitaceae phloem lectins, and jacalin-related lectins (JRLs) 1 (4). Each family has its own characteristic carbohydrate recognition domain. Jacalin, the prototype of JRLs, was isolated from seeds of jack fruit (Artocarpus integrifolia; Moraceae) (5). Subsequently, JRLs have also been isolated and characterized from various plant families of angiosperms such as Convolvulaceae (6), Asteraceae (7), Gramineae (8,9), Musaceae (10 -12), Fagaceae (13), and Mimosaceae (14). Although JRLs are widely distributed in higher plants, no information on JRLs outside of angiosperms is available except for a recently isolated lectin from the Japanese cycad (Cycas revoluta) of gymnosperm subdivision (15).
In an ab initio search for lectins in understudied groups of plants, we examined the large, fleshy, mesoterranian rhizomes of the tropical fern Phlebodium aureum for the presence of cell-agglutinating activity. We report herein the purification of P. aureum lectin (PAL), a mannose/glucose-specific lectin present in the rhizomes of this member of the Polypodiaceae family, as well as the cDNA cloning, expression, and characterization of this mannose/glucose-binding lectin and a closely related protein also having lectin activity. cDNA sequencing revealed that these fern lectins are novel members of the JRLs. This is the first report of the molecular cloning of JRLs from a lower plant, fern, which shows the structural and evolutionary relationship of JRLs in the plant kingdom. The two lectin genes were expressed in E. coli, and their physicochemical characterization is described and compared with the native lectin. This expression system should also be useful for mutagenesis studies to elucidate the structure-function relationship of JRLs.

MATERIALS AND METHODS
Rhizomes of Phlebodium aureum (L) J. Smith (also classified as Polypodium aureum L.) were collected from a specimen plant growing in the greenhouse at the Matthaei Botanical Gardens of the University of Michigan. Positive identification was provided by Dr. David Michener, collections curator of the botanical gardens.
Unless stated otherwise, saccharides and their derivatives and glycoproteins were purchased from Sigma. Ovine submaxillary mucin was a gift of Dr. R. N. Knibbs (University of Michigan). Mannose-Sepharose, prepared by divinyl sulfone coupling of mannose to Sepharose CL-4B (16), and yeast invertase-Sepharose, prepared using cyanogen bromideactivated Sepharose, were available from previous studies.
Purification of the Lectin-All procedures were conducted at 4°C. Pieces of rhizome from P. aureum were scraped to remove the soft, fuzzy layer and chopped into approximately 5-mm cubes. The light green chopped tissue (146 g fresh weight) was homogenized and extracted for 2-3 h with 600 ml of extraction buffer (PBS (10 mM sodium phosphate, 0.15 M NaCl, 0.135 mM CaCl 2 , 0.04% sodium azide, pH 7.2) containing 10 mM thiourea, 0.25 mM phenylmethylsulfonyl fluoride, and 1 g/liter ascorbic acid), with the addition of 10 g of poly(vinylpolypyrrolidone). The homogenate was squeezed through four layers of cheesecloth and centrifuged at 20,000 ϫ g for 20 min. To the supernatant solution was added solid ammonium sulfate to 10% saturation. After stirring overnight, any precipitate was removed by centrifugation, and the supernatant was filtered through coarse filter paper to remove a small amount of floating debris and was made 80% saturated with ammonium sulfate. After stirring for several hours followed by centrifugation, the precipitated protein was dissolved in approximately 60 ml of PBS, dialyzed against 2-3 2-liter changes of PBS, and clarified by centrifugation. This 10 -80% precipitated protein solution was applied onto a column (14 ϫ 2.5 cm; bed volume of 68 ml) of mannose-Sepharose, which had been equilibrated with PBS. The column was washed with PBS until the absorbance of the effluent at 280 nm decreased to a minimum value. The affinity-adsorbed lectin was desorbed with 0.1 M Me ␣-Man in PBS, dialyzed against PBS, and rechromatographed on the same column. Approximately 12 mg of purified lectin was obtained from 146 g of rhizome tissue.
Protein and Carbohydrate Estimations-Protein concentration was determined by a modified method of Lowry et al. (17), using bovine serum albumin as a standard. Total neutral sugar was determined colorimetrically by the phenol/sulfuric acid method (18). Specific sugars were identified on reverse-phase HPLC by the method of Fu and O'Neill (19).
Polyacrylamide Gel Electrophoresis in the Presence of SDS-SDS-PAGE was carried out on 0.75-mm slab gels in alkaline buffer system (Tris/glycine, pH 8.3) (20), using a mini-Protean II apparatus (Bio-Rad). BenchMark protein molecular mass standards used in SDS-PAGE were from Invitrogen.
Hemagglutination Assay-The hemagglutinating activity of the lectin was determined by a 2-fold serial dilution procedure using formaldehyde-treated (21) human and rabbit erythrocytes as described previously (22). The hemagglutination titer was defined as the reciprocal of the highest dilution exhibiting observable hemagglutination. Inhibition of agglutination by haptenic saccharides was assayed by serially diluting the solution of saccharide in the microtiter wells, followed by the addition of four agglutinin units of the lectin, followed by the addition of erythrocyte suspension after 30 min. The lowest concentration of saccharide that visibly decreased the extent of agglutination was defined as the minimum inhibitory concentration.
Molecular Mass and Subunit Structure-The molecular mass and subunit structure of purified P. aureum lectins were determined by gel filtration through a G2000-SWXL Progel-TSK column (30 ϫ 0.78 cm; Supelco, Bellefonte, PA) using a Beckman System Gold HPLC system as described previously (23) and by SDS-PAGE performed on samples with and without heating for 5 min in boiling water and in the presence or absence of 2-mercaptoethanol.
Amino Acid Composition Analysis and N-terminal Sequence Analysis-The amino acid composition and the N-terminal sequence of the purified lectin or peptides therefrom were analyzed by the University of Michigan Protein and Carbohydrate Structure Core Facility.
Quantitative Precipitation and Hapten Inhibition Assays-Quantitative precipitation assays were performed by a microprecipitation technique as described previously (22). Briefly, varying amounts of glycoproteins or polysaccharides, ranging from 0 to 100 g, were added to 50 g of purified PAL in a total volume of 250 l of PBS, pH 7.2. In some experiments, 20 g of lectin in a total volume of 120 l was used. After incubation at 37°C for 1 h, the reaction mixtures were stored at 4°C for 48 -72 h. The precipitates formed were centrifuged, washed two times with 400 l of ice-cold PBS, dissolved in 0.05 M NaOH, and evaluated for protein content by the Lowry method (17) using bovine serum albumin as a standard.
Hapten inhibition was performed in the same system with increasing amounts of potentially inhibitory saccharides added to the reaction mixture containing 50 g of the purified lectin and 8 g of precipitating polysaccharide in 250 l. The concentration of saccharide required for 50% inhibition was interpolated from corresponding inhibition curves.
Isothermal Titration Calorimetry-Isothermal titration calorimetry was carried out, and results were calculated as previously described (11) using lectin concentrations of approximately 4 mg/ml, equivalent to a subunit concentration of approximately 0.25 mM. Titrations generally were conducted by 25 additions of 5 l of ligand solutions having concentrations between 1 and 50 mM, depending on the expected range of binding constant.
RNA Isolation and cDNA Cloning-For RNA isolation, chopped rhizome tissue was immediately ground to a powder with a pestle under liquid nitrogen. Total cellular RNA was isolated with Concert Plant RNA reagent (Invitrogen), and subsequently poly(A) ϩ RNA was isolated with the Micro-FastTrack 2.0 kit (Invitrogen). Using this protocol, 1 g of poly(A) ϩ RNA per 10 g of rhizome was isolated.
An adapter ligated cDNA library was constructed with the Marathon cDNA amplification kit (Clontech, Palo Alto, CA). Two degenerate forward primers (PALF1, CARGTNGTNTAYGGNAAYGGNACNAC-NAAR; PALF2, GCNAAYGGNCARACNAARGARATHGAYGTN) were designed from the amino acid sequence VNGLQVVYGNGTTKLHGX-ANGQTKEIDV of a cyanogen bromide cleavage fragment of fern lectin for rapid amplification of cDNA ends (RACE). 3Ј-RACE was conducted with a combination of primers, adapter primer 1 (Invitrogen) and PALF1, and Platinum Pfx DNA polymerase (Invitrogen) as follows. DNA was denatured at 94°C for 3 min, followed by three-step cycles (40 cycles) (92°C for 0.5 min, 50°C for 0.5 min, and 68°C for 1 min), and further extended at 65°C for 15 min. This amplified DNA fragment was subsequently amplified with adapter primer 2 (Invitrogen) and PALF2. The amplified 0.5-kbp fragment was cloned using Zero Blunt TOPO PCR cloning kit (Invitrogen). Inserted DNA was sequenced with T7 and SP6 primers by the DNA Sequencing Core Facility of the University of Michigan, and two similar but different genes (termed PALa and PALb) including poly(A) ϩ were obtained. Two specific reverse primers for each gene (PALaR1, GCCTAGTAAAGCGACCGACATGGCTACAAGAGCG-CTAC; PALbR1, GACATAGAGGCCGAGGCGATCCAAACGGTCTCC) were designed, and 5Ј-RACE was conducted with adapter primer 1 and PALaR1/PALbR1, respectively.
Construction, Expression, and Purification of Recombinant Fern Lectin-The full-length coding sequence PCR products of PALa and PALb incorporating NdeI and XhoI sites into its forward and reverse primers, respectively, were cloned into pCR-Blunt II-TOPO vector (Invitrogen) and subsequently cloned into expression vector pET-22b(ϩ) (Novagen) to generate carboxyl-terminally His 6 -tagged proteins, yielding pET-PALa and pET-PALb, respectively. The Nova Blue (DE3) strain of E. coli harboring expression plasmid pET-PALa and pET-PALb was precultured in 5 ml of LB medium containing 50 g/ml ampicillin at 37°C for 3 h and was added to 1 liter of medium. After the optical density at 600 nm reached 0.4 -0.6, 1 ml of 1 M isopropyl-D-thiogalactoside was added to the medium, and the cells were further cultured at 25°C for 6 h. The induced cells were collected by centrifugation, resuspended in a lysis buffer (PBS, containing 10 mM 2-mercaptoethanol, 1% Nonidet P-40, and 1 mM phenylmethylsulfonyl fluoride), and sonicated. The insoluble fraction was removed by centrifugation at 10,000 ϫ g for 30 min at 4°C. Recombinant PALa (rPALa) and recombinant PALb (rPALb) were purified from the soluble fraction by absorption on a mannose-Sepharose 4B column and elution by 0.2 M Me ␣-mannose.
Effect of Temperature and pH-To examine their thermostability, lectin solutions (0.1 mg/ml in PBS) were incubated for various periods at 40, 50, 55, 60, or 70°C. After 10 l of lectin solution was cooled on ice, its hemagglutination activity was assayed as described above. pH stability of the lectins were determined in the following buffers: 0. Sequence Data Processing-Multiple sequence alignment was performed by the ClustalW program (25). Homologous sequences were searched for by the FASTA program. A phylogenetic tree was constructed by the neighbor-joining algorithm based on an evolutionary distance matrix constructed by Kimura's method (26) The degrees of confidence for internal lineages in phylogenetic trees were calculated by the bootstrap procedure.

RESULTS
Hemagglutinating Activity-Crude extracts from P. aureum rhizomes weakly agglutinated formaldehyde-stabilized rabbit erythrocytes (titer ϭ 32-64) but not sheep nor any type of human erythrocytes. After purification as described below, the lectin agglutinated rabbit erythrocytes at a minimum concentration of 1 g/ml (titer at 4 mg/ml ϭ ϳ4,000), whereas sheep or any type of human erythrocytes required approximately 150 g/ml for agglutination (titer at 4 mg/ml ϭ ϳ64). Lectin that was dialyzed extensively against metal-free, EDTA-containing buffer and assayed in the same buffer had identical hemagglutination titer against rabbit erythrocytes, indicating that it has no requirement for metal ions.
Purification of Native Fern Lectin-Because the hemagglutination activity of crude extracts of P. aureum was inhibited by D-mannose, mannose-Sepharose was used as an affinity absorbent for isolation of the lectin. After elution of nonabsorbed protein from the mannose-Sepharose column, 0.1 M Me ␣-Man eluted a sharp band of protein, which in a second affinity chromatography was totally bound and eluted in a similar manner (Fig. 1). The lectin was also bound to and eluted from a column of immobilized yeast invertase under the same conditions. As shown in Fig. 2A, upon SDS-PAGE at pH 8.3 with unheated samples, the lectin preparation obtained from affinity chromatography on mannose-Sepharose or invertase-Sepharose showed a major band at approximately 38 kDa. The low, broad band appearing before and after elution of the sharp lectin peak (Fig. 1A) appeared to contain several bands of nearly equal intensity. Upon rechromatography of the eluted lectin fraction (Fig. 1B), this contaminating broad band was allowed to wash off before commencing elution of the lectin with a haptenic sugar, yielding a slightly more purified material. When lectin samples were boiled in SDS with 2-mercaptoethanol, a single band of approximately 15 kDa was observed ( Fig. 2B), suggesting that the native structure is a dimer of this monomer. The earlier and later contaminating material gave a single band at slightly lower mass. We have not investigated the nature of this apparent contaminant further.
Molecular Mass and Subunit Structure-The molecular mass was also estimated by size exclusion chromatography on a silica-based matrix. The purified lectin migrated as a single, nearly symmetrical band of approximately 31 kDa, based on standardization with known proteins (data not shown). Together with the SDS-PAGE analysis, these results indicate that the lectin exists as a dimer of approximately 15-kDa subunits that requires boiling in SDS to dissociate completely.
Amino Acid Composition and N-terminal Amino Acid Sequence-The amino acid composition of the purified fern lectin (data not shown) indicated that it contains a single residue each of methionine and histidine and the typically large amounts of aspartic acid/asparagine, serine, and glycine observed in many other lectins. Attempts to sequence the Nterminal region were not successful; however, after cyanogen bromide cleavage, a large peptide having the N-terminal sequence VNGLQVVYGNGTTKLHGXANGQTKEIDV was detected. Cleavage with Achromobacter protease I yielded a peptide with the partial sequence LGPWGGSGGDSFDD-GSDNGG. Carbohydrate Analysis-No periodate-Schiff staining bands were observed on SDS-polyacrylamide gels of the purified native lectin, although a small amount of neutral sugar (approximately 0.8 -1.5 hexose units/subunit) was detected in some preparations at high concentration by the phenol/sulfuric acid assay. Analysis of such PAL preparations by the method of Fu and O'Neill (19) indicated primarily mannose to be present. Although the partial sequence shown above contains a putative glycosylation sequence (NGT), the asparagine residue must be largely or totally nonglycosylated to be detected by automated sequencing. Most likely, the mannose arises from trace amounts of endogenous mannan or high mannose glycoprotein contaminants, which remain associated with the lectin during the purification procedures, but the lectin itself is not glycosylated.
Inhibition of Agglutination-The ability of a number of mono-and oligosaccharides to inhibit agglutination of formaldehyde-treated rabbit erythrocytes is shown in Table I. It is evident that branched mannose oligosaccharides are the best inhibitors, whereas glucose-containing disaccharides were much less effective. This result indicates that PAL is a mannose/glucose-binding lectin.
Quantitative Precipitation and Precipitation Inhibition-In qualitative precipitin tests in capillary tubes, mannans from Saccharomyces cerevisae and Saccharomyces rouxii and dextran B-1355-S from Leuconostoc mesenteroides (27) gave visible precipitation, whereas many other polysaccharides, glycoproteins, and neoglycoproteins did not. Quantitative precipitin assays (Fig. 3) confirmed that the two yeast mannans precipitated to a significant extent with an equal weight of lectin; dextran B-1355-S at 3-4-fold higher concentration by weight also strongly precipitated the lectin. Two other dextrans from L. mesenteroides, B-742-S and B-742-L (27), partially precipitated the lectin at much higher concentrations. Among glycoproteins, only yeast invertase, a high mannose glycoprotein, partially precipitated the lectin. Soybean agglutinin, which also contains a high mannose glycan, was inactive. Precipitation of yeast mannan was inhibited 50% by Man 3 at 3.1 mM, mannose at 19 mM, Me ␣-Man at 80 mM, or Me ␤-Man at 130 mM (data not shown). These results are in qualitative agreement with the inhibition of agglutination but show that the precipitation of mannan involves a considerably stronger interaction than does rabbit erythrocyte agglutination, since higher concentrations of inhibitory sugars are required.
Titration Calorimetric Determination of Carbohydrate Binding-Thermodynamic parameters for the binding of various mannose and glucose-containing saccharides are shown in Table II. These results also agree with the previous data, confirming that the branched mannose trisaccharide and pentasaccharides are the best ligands, with the additional mannose units strongly enhancing the binding.
Molecular Cloning of Fern Lectin-3Ј-RACE with the adapter primers and the degenerate primers that were designed from the cyanogen bromide fragment yielded a 0.5-kbp product. Of eight clones sequenced, seven were identical (PALa), whereas one clone contained a similar but apparently different sequence (PALb). Cloning and sequencing of 5Ј-RACE products generated two full-length nucleotide sequences including polyadenylation. PALa contains a 24-bp 5Ј-untranslated region, followed by a 438-bp open reading frame encoding 146 amino acid residues and 272 bp 3Ј-untranslated region, whereas PALb contains a 5-bp 5Ј-untranslated region, followed  by a 438-bp open reading frame encoding 146 amino acid residues and 275-bp 3Ј-untranslated region (Fig. 4). Neither contain an adenylation signal sequence. Since the isolated lectin contains only one methionyl residue, the N-terminal methionine of mature lectin has been removed, resulting in N-terminal serine. The calculated molecular masses of PALa (14,854 Da) and PALb (14,895 Da) without N-terminal Met are in good agreement with the molecular mass of native fern lectin (nPAL) (14,903 Da) determined by matrix-assisted laser desorption ionization time-of-flight mass spectrometry. Since nPAL appears to be blocked at its N terminus, the slightly higher molecular mass of nPAL than the calculated molecular mass of PALa is probably due to the presence of a blocking group, such as an N-acetyl moiety (M ϭ 42 kDa) present on the native lectin. The deduced amino acid sequence of PALa contains the same sequences as those found in the native fern lectin, but PALb contains a slightly different sequence (Fig. 4), suggesting that the PALa gene encodes the native PAL. No signal sequence could be discerned in the deduced amino acid sequences of either PAL gene, indicating their syntheses on free polysomes. PALa and PALb have two and one potential N-linked glycosylation site (NX(S/T)), respectively in the sequences.

Construction, Isolation, and Characterization of Recombinant Fern Lectin-To express recombinant PALa (rPALa) and
PALb (rPALb), the entire open reading frames of each were cloned into expression vector pET-22b(ϩ) and introduced into E. coli strain Nova Blue (DE3) cells. Active recombinant lectins expressed as carboxyl-terminal His 6 -tagged fusion proteins were purified from E. coli extract by a single chromatographic step on a mannose-Sepharose 4B column. The yields of rPALa and rPALb purified by affinity chromatography were each approximately 4 mg from 1 liter of culture.
The N-terminal amino acid sequences of rPALa (H 2 N-SSAG-SEVAKLGPWGGSGGDS) and rPALb (H 2 N-SSASSEVAKLG-PWGGSGGDS), determined by a gas phase protein sequencer without pretreatment, indicated that the initial methionine of these proteins was also removed in the bacterial expression system but that no N-terminal blocking occurred.
SDS-PAGE analysis of the recombinant lectins was compared with that of native lectin ( Fig. 5; cf. Fig. 2). The two recombinant lectins gave virtually the same pattern of bands as did the native lectins in both unheated and heated samples, except for the absence of minor contaminating or isolectin bands in the recombinant samples. After boiling for 5 min either in the presence or in the absence of 2-mercaptoethanol in SDS sample buffer, the three forms each gave a single band at about 14 kDa (nPAL) or 14.5 kDa (rPALs), respectively. The slightly greater apparent size of the recombinant subunits is accounted for by the presence of the His 6 tags. Likewise, size exclusion chromatography of intact recombinant lectins in solution gave elution profiles indistinguishable from that of the native lectin, indicating each to have a molecular mass of ϳ31 kDa.
The minimum concentration for hemagglutination activity of rPALa and rPALb against formaldehyde-treated rabbit erythrocytes was estimated to be 0.6 g/ml, which is not significantly different from the native fern lectin (1 g/ml). Sugar-binding specificities (estimated by inhibition of hemagglutination) of recombinant fern lectins, especially rPALa, were very similar to the native lectin (Table I)  each case. Most ␣and ␤-diglucosides were weakly inhibitory, but it is noteworthy that maltose (Glc␣1-4Glc), laminaribiose (Glc␤1-3Glc), and gentiobiose (Glc␤1-6Glc) were very weak or noninhibitory at the maximum concentration tested. This pattern of specificity distinguishes these lectins from the mannose/ glucose-binding banana lectin, a member of the JRLs (11,12).
rPALa generated quantitative precipitation curves with the same yeast mannans and dextran 1355-S as did native PAL (Fig. 3). Interestingly, rPALa also gave a modest precipitin curve with rabbit liver glycogen, whereas native PAL gave no appreciable precipitation. rPALb precipitated neither dextran 1355-S nor glycogen (Fig. 3B). Native and recombinant fern lectins showed similar stability to pH and temperature. No appreciable change was observed in the hemagglutinating activity of lectins preincubated in a pH range from 4 to 9, and they retained more than half of their activity at pH 2.0, 3.0, and 10.0. In contrast, the lectins are less stable to heat; all forms retained their activity up to 50°C but lost half of their activity within 15 min at 60°C.

DISCUSSION
The so-called "primitive" vascular plants (ferns and fern allies) have only occasionally been investigated for the presence of lectins, despite the widespread occurrence of high levels of lectins in storage tissues (seeds, bark, rhizomes, bulbs, etc.) of flowering plants and the carpophores of fungi and the almost universal occurrence of low levels of lectins and related pro- teins in all types of plant tissues, animals, and microorganisms. Mellor and co-workers (28) purified a hexameric 120-kDa lectin from the Azolla-Anabaena symbiosis and, at lower levels, from nonsymbiotic Azolla (Salviniaceae) plants. This lectin was inhibited by galactose and more weakly by GalNAc but not by lactose. A brief report by Vasheka et al. (29) showed the presence of agglutinating activity toward rat erythrocytes in the rhizomes of three species of Dryopteris (Polypodiaceae) ferns; in one of the species, an extract of the fronds also had high activity. These activities were not purified; nor was any carbohydrate specificity established. Both of these genera are in the class of true ferns, Filicinae. To our knowledge, no other reports of lectins in true ferns or fern allies have been made. This report thus represent the most detailed and extensive investigation of a lectin from the plant division Pteridophyta to date.
Agglutination inhibition, precipitation, and calorimetric titration data all indicate that PAL is a mannose/glucose-binding lectin, although its affinity for glucose is considerably weaker than typical mannose/glucose-binding legume lectins. Mannose-binding legume lectins typically have affinities for glucose of 20 -50% that of mannose, whether comparing the free sugars or the Me ␣-glycosides (2). In the case of PAL, however, affinity for Me ␣-Glc is ϳ10% or less that of Me ␣-Man (Tables I and II). The very weak interaction of PAL with maltose is also in sharp contrast to legume Man/Glc-binding lectins (2). Monocot mannose-binding lectins, however, exhibit no detectable binding to glucose-containing mono-or oligosaccharides (3,30). PAL may thus be considered to be a new class or subclass of mannosebinding lectins exhibiting weak but measurable activity toward glucose structures.
By definition, a lectin is a sugar-binding protein or glycoprotein of nonimmune origin that agglutinates cells and/or precipitates glycoconjugates (1). In order to form detectable precipitate, however, both the lectin and the glycoconjugate must be multivalent and must be mixed in an appropriate stoichiometric ratio. Furthermore, in order to form the cross-linked aggregates necessary for precipitation, at least one of the components (lectin or polymeric ligand) must possess three or more binding sites; otherwise, only linear, nonprecipitating aggregates can form. The strong precipitation of yeast mannan is readily understandable. Likewise, yeast invertase contains 9 -10 N-linked glycan structures/polypeptide, each containing 26 -54 mannose residues/residue of N-acetylglucosamine (31,32), which provides a high density of branched mannose structures for cross-linking. On the other hand, soybean agglutinin contains one Man 9 GlcNAc 2 structure per subunit of a tetrameric protein and is readily precipitated by concanavalin A (2). It fails to precipitate PAL, however, suggesting that this lectin does not bind as tightly, either because of differing fine specificity or poorer accessibility of the binding sites than in the case of concanavalin A. Bovine ribonuclease B possesses a single Man 6 GlcNAc 2 structure per 15-kDa monomer (33) and thus would not be expected to precipitate with the lectin in any case. Although glucose and its oligosaccharides are very poor ligands, highly branched glucan structures might provide sufficient binding interactions to precipitate the lectin. The highly branched dextran B-1355-S, which contains almost equal amounts of ␣1,6and ␣1,3-linked glucose units and a small amount of ␣1,4 linkages (27), is an especially good precipitant, and those designated B-742-L and -S, which contain about 20% ␣1,4 linkages and 0% (L) or 26% (S) ␣1,3 linkages, are moderate precipitants. Dextrans B-1208 and B-512 (clinical dextran), almost exclusively (95%) ␣1,6-linked, are essentially linear structures. Hence they are not precipitated by the lectin, again in sharp contrast to legume Man/Glc-binding lectins such as concanavalin A. Isolichenan and elsinan, which also failed to precipitate the lectin, are linear ␣-glucans of maltose and maltotriose units linked by ␣1,6 or ␣1,3 bonds, respectively (11). Interestingly, the banana lectin, a mannose/glucose-binding member of the jacalin-related lectin family, is precipitated by elsinan by virtue of reacting with its internal ␣1,3-glucan structures but is unreactive with isolichenan (11). As discussed below, PAL is also related to the jacalin family of lectins.
Estimates of relative binding (or dissociation) constants for mono-and oligosaccharides by inhibition of erythrocyte agglutination, direct calorimetric titration, and inhibition of mannan precipitation are in good agreement, although absolute values vary with the assay system used. To the limited extent that precipitation inhibition assays have been carried out, inhibition constants are considerably higher (e.g. 3 versus 0.1 mM for Man 3 and 80 versus 1-1.6 mM for Me ␣-Man). Inhibition of rabbit cell agglutination also requires somewhat higher concentrations than indicated by dissociation constants determined calorimetrically. These quantitative differences reflect the fact that the lectin's affinity for mannan is quite strong and involves multiple interactions that are difficult to reverse or inhibit, as compared with the rabbit erythrocyte cell surfaces or binding of monovalent ligands in solution.
It is clear from Tables I and II that the most favorable structural element for binding is the Man␣1,3Man structure. Man␣1,4Man is a much poorer inhibitor than the other dimannosides, as is the case also with the corresponding diglucosides, where maltose requires at least 50 mM, whereas the other three glucosidic linkages show somewhat stronger inhibition. Man␣1,2Man, whose linkage involves the axial hydroxyl group of the reducing mannose unit and thus has a considerably different conformation than the other disaccharides, exhibits poorer inhibition or binding than the ␣1,3-disaccharide or the monosaccharide, and also appears from the calorimetric data (Table II) to have considerably different entropy and enthalpy contributions to binding. Man␣1,6Man shows enhanced binding over the monomannoside, although this structure in Man 3 has a small variable effect on ␣1,3-linked disaccharide binding, depending on the form of the lectin and assay system used. A second ␣1,3-linked moiety in the Man 5 structure further enhances the binding by almost the same factor as the first ␣1,3-linked residue. On the other hand, the O-methyl glycoside moiety has little effect in the case of the disaccharides as seen in Table II. Also, as noted in the agglutination inhibition studies, mannose may actually be a slightly better inhibitor than Me ␣-Man, and the bulkier p-nitrophenyl aglycone appears to interfere significantly with binding when in the ␣but not the ␤-anomeric position.
The complete amino acid sequences of two closely related gene products were deduced from clones of two full-length cDNAs, termed PALa and PALb, obtained by 5Ј,3Ј-RACE procedure using primers designed from the amino acid sequence of a cyanogen bromide cleavage fragment. The subunits of these proteins were composed of 146 amino acid residues each, with 88% sequence identity (Fig. 4).
The mannose-binding site of Heltuba consists of five residues (Gly 18 , Gly 135 , Asp 136 , Val 137 , and Asp 139 ) linked to O-3, O-4, O-5, and O-6 of mannose by creating a network of eight hydrogen bonds as shown by x-ray crystallographic studies (24). In addition, Met 92 mediates hydrophobic interaction with the pyranose ring of mannose. Four of the six amino acid residues are conserved in the PAL polypeptide (Gly 18 , Gly 132 , Asp 133 , Asp 136 ), although the replacement of Met 92 and Val 137 of Heltuba with Gly 89 and Arg 134 of PAL was observed (Fig. 6). The difference in sugar binding specificity between Heltuba and fern lectin might be caused by the replacement of these key residues. Man␣1-2Man and Man␣1-3Man are equally good inhibitors for the hemagglutination activity of Heltuba. On the other hand, among the ␣-linked mannose disaccharides, Man␣1-3Man is the best inhibitor of PAL, with the ␣1,2 and ␣1,6 being approximately one-fifth as active, whereas Man␣1-4Man is essentially a noninhibitor.
JRLs have been classified as galactose-specific (gJRLs) and mannose-specific (mJRLs) according to their sugar specificities, which relate to a specific structural difference. gJRLs including Jacalin and M. pomifera lectin are built up of cleaved protomers consisting of a ␤-chain (20 amino acids) and an ␣-chain (133 amino acids). Proteolytic processing of the proproteins generates an N-terminal Gly, whose amino group mediates a hydrogen bond with O-4 of galactose, which is responsible for the galactose-binding specificity of jacalin and M. pomifera lectin. In contrast, mJRLs, including all other JRLs, consist of uncleaved protomers of about 150 amino acid residues. Another major difference between gJRLs and mJRLs is their biosynthesis, processing, and localization. Jacalin (gJRL) is synthesized on the endoplasmic reticulum as a preproprotein and is targeted in the vacuolar compartment after a complex series of processing steps. In contrast, C. sepium lectin (mJRL) is synthesized and localized in the cytoplasm without processing, due to the absence of a signal peptide. The mature protein corresponds to the entire open reading frame. A phylogenetic tree constructed based on the amino acid sequences of  1-154); CCA-C, C. crenata lectin (residues 155-309); Pk-1, P. platycephala lectin (residues 1-147); Pk-2, P. platycephala lectin (residues 150 -293); Pk-3, P. platycephala lectin (residues 296 -442). Barley is a jacalin-like protein from Hordeum vulgare (residues 148 -306).
18 jacalin-related lectin carbohydrate recognition domains (JRL-CRDs) from 15 lectins showed their evolutionary relationships (Fig. 7). The cluster of JRLs is in good agreement with the taxnomic classification of angiosperms, because JRLs from dicots are separated from JRLs from monocots (O. sativa lectin, Barley, and M. acuminata lectin). In this tree, JRLs from monocots are evolutionarily closer to PALs from fern. Two tandemly repeated JRL-CRDs from C. crenata and three tandemly repeated JRL-CRDs from P. platycephala might have been duplicated after the divergence of these plant families. The only other mJRL from a nonangiosperm is that recently isolated and partially characterized from a gymnosperm, Japanese cycad (C. revoluta), but only a partial amino acid sequence is available (15). The identification of mJRLs in a nonspermotophyte, fern, reveals that JRLs are not restricted to Spermatophyta but rather are widely distributed in the plant kingdom. In contrast, gJRLs, which were believed to be a major subgroup of JRLs, have been found exclusively in the dicot Moraceae. It appears that gJRLs were evolved from mJRLs by the insertion of signal peptide and vacuolar targeting signal after the divergence of Moraceae from other flowering plants.
In other plant families, the molecular evolution of this group of lectins can also be observed. The subunits of JRLs from C. crenata and P. platycephala consist of two or three tandemly repeated jacalin-related lectin domains, indicating that these JRLs have evolved by gene duplication and/or exon shuffling. In addition, chimeric JRLs have been identified in Brassica napus. Therefore, it is likely that an ancestral protein of JRLs has evolved to play diverse roles in each plant family. A relatively small number of invariant amino acid residues among JRLs and different sugar binding specificity support this hypothesis.
Although the physiological function of the JRLs is not known, there are several possibilities. mJRLs have been thought to be stress-responsive proteins. A JRL from rice is identical to a salt and drought stress-inducible salT gene product, and this lectin is expressed only after induction by either jasmonate or NaCl. Myrosinase-binding proteins from B. napus, which contain one or two jacalin-related lectin domains in the sequences, are also stress-inducible. mJRLs are also assumed to be self-defense proteins because of their binding affinity to mannose, which is a relatively scarce sugar in plants but is widely distributed in microorganisms, insects, and animals.
The expression system of PALa and PALb containing carboxyl-terminal His 6 tags in E. coli was constructed, and their sugar-binding specificity and physicochemical characteristics were compared with those of the native lectin isolated from fern rhizomes. The studies indicated that rPALa and rPALb resemble the native lectin in most respects. It is not surprising that rPALb had similar sugar binding specificity to rPALa, because they share high (88%) sequence identity and conserve all key amino acids among JRLs (Figs. 4 and 6). rPALa gave precipitin curves with the highly branched ␣-mannan from yeast and ␣-D-glucan (dextran B-1355-S) as expected (Fig. 3B). Interestingly, rPALa also precipitated moderately with rabbit liver glycogen, whereas native PAL did not. The carboxyl-terminal His 6 tag and/or the free N terminus of rPALa might account for the difference in binding affinity with glycogen. Although rPALb gave a precipitin curve with mannan, in contrast to rPALa, it precipitated neither with dextran nor with glycogen, indicating that even small differences in noncritical residues can affect the lectin's reactivity with polysaccharides.
The identification of a JRL in a true fern further expands the widespread distribution of JRLs in the plant kingdom. Because fern lectin is probably the closest known relative of the common ancestor of JRLs, the finding of JRLs from ferns might help to explain the original function of JRLs in plants. Although the specificity of PAL as detailed here does not suggest any unique utility of this particular lectin at this time, its distinct character is nevertheless of interest in understanding lectin binding interactions. Recombinantly expressed fern lectins will be useful for such investigations by site-directed mutagenesis. Their applications in biological and medical research by virtue of their specificities toward mannose and oligomannosides and their ease of preparation as recombinant expression proteins from E. coli by one-step affinity chromatography might also be significant.