Identification of β-Glucosidase Aggregating Factor (BGAF) and Mapping of BGAF Binding Regions on Maize β-Glucosidase*

In certain maize genotypes (nulls), β-glucosidase does not enter the gel and therefore cannot be detected on zymograms. Such genotypes were initially thought to be homozygous for a null allele at the glu1 gene. We have shown that a β-glucosidase aggregating factor (BGAF) is responsible for the null phenotype, and it specifically interacts with maize β-glucosidases and forms large insoluble aggregates. To understand the mechanism of the β-glucosidase-BGAF interaction, we constructed chimeric enzymes by domain swapping between the maize β-glucosidase isozymes Glu1 and Gu2, to which BGAF binds, and the sorghum β-glucosidase (dhurrinase) isozyme Dhr1, to which BGAF does not bind. The results of binding assays with 12 different chimeric enzymes showed that an N-terminal region (Glu50-Val145) and an extreme C-terminal region (Phe466-Ala512) together form the BGAF binding site on the enzyme surface. In addition, we purified BGAF, determined its N-terminal sequence, amplified the BGAF cDNA by reverse transcriptase-polymerase chain reaction, expressed it inEscherichia coli, and showed that it encodes a protein whose binding and immunological properties are identical to the native BGAF isolated from maize tissues. A data base search revealed that BGAF is a member of the jasmonite-induced protein family. Interestingly, the deduced BGAF sequence contained an octapeptide sequence (G(P/R)WGGSGG) repeated twice. Each of these repeat units is postulated to be involved in forming a site for binding to maize β-glucosidases and thus provides a plausible explanation for the divalent function of BGAF predicted from binding assays.

␤-Glucosidase (␤-D-glucoside glucohydrolase, EC 3.2.1.21) occurs ubiquitously in all three (archaea, eubacteria, and eukarya) domains of living organisms. The enzyme catalyzes the hydrolysis of aryl and alkyl ␤-D-glucosides as well as glucosides with a carbohydrate moiety such as ␤-linked oligosaccharides (1). The occurrence and activity of ␤-glucosidase in maize is correlated with growth and certain desirable traits (2). The major function of maize ␤-glucosidase, however, may be in the defense of young plant parts against pathogens and herbivores by releasing toxic aglycones (e.g. hydroxamic acids) from their glucosides. Hydroxamic acids, derivatives of 1,4-benzoxazin-3-one, are believed to be the major defense compounds in maize, wheat, rye, and wild barley (3). The predominant hydroxamic acid glucoside in maize is 2-glucopyranosyl 4-hydroxy-7-methoxy-1,4-benzoxazin-3-one (DIMBOAGlc), whose aglycone DIMBOA is the primary defense chemical against aphids and the European corn borer (Ostrinia nubilalis). Several studies have shown a high correlation between DIMBOA content of maize genotypes and the level of resistance to or inhibitory affect on insects and pathogens (4 -8).
In certain maize genotypes (nulls), ␤-glucosidase occurs as part of large insoluble quaternary aggregates (9). The ␤-glucosidase zymograms of such genotypes are devoid of enzyme bands (10). These genotypes were initially thought to be homozygous for a null allele at the glu1 gene. However, biochemical and immunological data from our laboratory established that the so-called null genotypes have ␤-glucosidase activity when assayed in solution, and have a 60-kDa polypeptide reacting specifically with anti-␤-glucosidase sera on immunoblots (9). The enzyme is not detected on zymograms, because it occurs as large quaternary structures (Ͼ1.5 ϫ 10 6 Da), which fail to enter the gel. After dissociation of these structures by SDS, the enzyme can be detected on gels (11). We have recently shown that the null phenotype is due to a ␤-glucosidase aggregating factor (BGAF), 1 which specifically interacts with the enzyme, forming high molecular weight heterocomplexes (11).
We have identified BGAF as a protein belonging to the jasmonic acid-induced protein (JAp) family, and BGAF is solely responsible for ␤-glucosidase aggregation and insolubility, and thus, the apparent null phenotype. Jasmonic acid and salicylic acid are plant-signaling molecules that play an important role in induced disease resistance pathways. JAps are believed to function in some step of these pathways. Blocking the response to either of these signals can render plants more susceptible to pathogens (12)(13)(14)(15)(16)(17) and insects (18). Recently, it was shown that the jasmonic acid-dependent induced systemic resistance pathway and the salicylic acid-dependent systemic acquired resistance pathway are fully compatible, and they result in an additive effect on the level of induced protection (19).
In maize, cDNAs corresponding to the ␤-glucosidase genes (glu1 and glu2) have been cloned and sequenced (Refs. 20, 21; A. Esen and M. Shahid, direct submission GenBank accession number U25157). The putative protein products of these cDNAs, Glu1 and Glu2, show 90% sequence identity with each other. Additionally, a ␤-glucosidase cDNA (dhr1) from sorghum has been sequenced, which shares 70% sequence identity with Glu1 and Glu2 (22). Despite their high sequence identity, the maize and sorghum enzymes are functionally different with respect to BGAF binding. BGAF binds to both maize isozymes (Glu1 and Glu2) with high specificity but does not bind to their sorghum homolog Dhr1. Therefore, they provide an excellent system to study functional differences at nonconserved residues and elucidate the mechanism of BGAF-mediated enzyme aggregation and insolubility. The objective of the present study is to elucidate the mechanism of the ␤-glucosidase-BGAF interaction. To this end, we have generated a series of chimeras among Glu1, Glu2, and Dhr1 by domain swapping to identify the sites involved in BGAF binding. The binding properties of these chimeras enabled us to identify two separate and distinct polypeptide segments that together form a BGAF binding site on the surface of maize ␤-glucosidases. Finally, we have cloned and sequenced the BGAF cDNA and confirmed its identification by expressing it in Escherichia coli and demonstrating the activity of its recombinant protein product unequivocally in binding assays.
Purification of ␤-Glucosidases-The wild-type ␤-glucosidases and their chimeras (C-2, C-4, C-15, C-16, C-21, and C-22) were isolated from lysates of 800-ml cultures grown at 37°C and induced at 25°C. The enzymes were purified to near homogeneity using a combination of differential solubility fractionation by ammonium sulfate (AS) and hydrophobic interaction chromatography as described previously (25).
Purification of BGAF-Etiolated seedlings of 3-day-old maize inbred line H95 (null) was used as the source for purification of BGAF. BGAF was isolated in its free form from the pellet fraction of shoot homogenates that had been extracted four times with 50 mM NaAc (pH 5.0) buffer containing 30% AS (these conditions keep BGAF insoluble but removes ␤-glucosidase). The fourth pellet containing predominantly free BGAF was extracted with 50 mM NaAc buffer (pH 5.0). The extract was subjected to gel filtration on Sephacryl HR-200 (90 ϫ 1.6 cm), and the fractions containing BGAF were identified by enzyme-linked immu-nosorbent assay and pooled. Ammonium sulfate was added to the pooled fractions to a final concentration of 0.8 M and applied to a ToyoPearl-butyl 650 M hydrophobic interaction chromatography column equilibrated with 50 mM NaAc (pH 5.0) containing 0.8 M AS. The column was developed with a manual step gradient using 0.1 M increments from 0.8 to 0.1 M AS. BGAF-containing fractions were determined by enzyme-linked immunosorbent assay, pooled, and concentrated on a 10,000 cut-off spin column (Gelman Sciences). The purity of BGAF was checked by SDS-PAGE, and ϳ250 pmol of BGAF were subjected to N-terminal sequencing.
␤-Glucosidase-BGAF Binding Assays-The interaction between BGAF and ␤-glucosidase was measured in a binding assay by mixing purified BGAF at 10-fold molar excess with purified ␤-glucosidases or their chimeras. In the case of chimeras C-5, C-6, C-18, C-19, and C-26, crude expression extracts rather than purified enzymes were used as ␤-glucosidase source in binding assays. The BGAF-␤-glucosidase interaction is very specific and reminiscent of antigen-antibody interactions, therefore, crude expression extracts rather than purified enzyme could FIG. 1. A, sequence alignment showing the N-(1-214) and C-terminal (453-512) regions of maize (Glu1 and Glu2) and sorghum (Dhr1) ␤-glucosidases. The underlined peptides were used to design oligonucleotides to create the chimeric enzymes shown in B. The two regions (Glu 50 -Asn 127 and Phe 466 -Ala 512 ) that contain the residues forming a BGAF binding site are in green background. The residues that form the binding site are shown in red (invariant) and purple (variant), and those that are variant but equivalent or located outside the postulated boundary of the site are blue. See Fig. 4 for details. B, schematic representation of the 12 chimeric ␤-glucosidase constructs used to identify BGAF binding domains in maize (Glu1 and Glu2) and sorghum (Dhr1) ␤-glucosidases. The results from BGAF binding assays (ϩ for binding and Ϫ for not binding) are summarized under the column BGAF binding. The lengths of the domains in Glu1 or Glu2 replaced with their Dhr1 homolog by domain swapping are at the right of each chimera. be used as ligand source. The enzyme-BGAF mixes were incubated on ice for 1-2 h with occasional mixing. The reaction mixes were then electrophoresed on 8% native gels, and the gels were equilibrated with two changes of 50 mM citrate/100 mM phosphate buffer (pH 5.8) for 5 min each after electrophoresis. ␤-Glucosidase activity was detected by incubating the equilibrated gels in a 1 mM solution of the fluorogenic substrate 4-methylumbelliferyl-␤-D-glucoside (MUGlc) for 5-10 min. ␤-Glucosidase activity zones (bands) were visualized under UV light and documented using an AlphaImager 2000 documentation and analysis system (Alpha Innotech Corp., San Leandro, CA) (Fig. 2). The BGAF-Dhr1 binding assay was essentially the same as described above, except that mobility shifts of BGAF were analyzed by immunoblotting using anti-BGAF serum as probe instead of enzymatic activity, because Dhr1 does not hydrolyze MUGlc. Initially, purified BGAF was incubated in 2-fold incremental concentrations ranging from 20 to 2.5 g/ml with a fixed volume of a Dhr1 expression extract. Mixtures were electrophoresed on 8% native gels, blotted, and probed with anti-BGAF sera (Fig. 3). BGAF by itself and BGAF incubated with Glu1 served as negative and positive controls, respectively, in BGAF-Dhr1 binding assays.
Cloning and Sequencing of the BGAF cDNA-The N-terminal sequence of BGAF purified as described above was: (V/X)(I/E)(G/P) (N/L)YAPIGIGATV. The peptide APIGIGAT was used to design two (sense) degenerate primers (BGAF-6, CCNATHGGNATHGGNGCNAC; BGAF-7, GCNCCNATHGGNATHGGNGC). Messenger RNA was isolated from etiolated 2-day-old H95 shoots using oligo-dT-coated magnetic beads according to the vendor's protocol (Dynal). An oligo-dT primer was used for first-strand cDNA synthesis with avian myeloblastosis virus-reverse transcriptase (Promega). To amplify the BGAF cDNA, the primers BGAF-6 and BGAF-7 were individually paired with the oligo-dT primer in separate PCR reactions. A PCR product of 1 kb was obtained, gel purified, and cloned into the SmaI site of pBluescript II SK (Ϯ) for sequencing. A BLAST search of the maize EST data base "hit" three ESTs whose 3Ј-ends overlapped with the 5Ј-end of the BGAF cDNA. Of these, the longest one (459 bp, GenBank T70648) had the highest match (97%) and a 327-bp overlap with the 5Ј-end of the BGAF cDNA. A primer (BGAF-16, 5Ј-CAGCTCCTCATCTCAAGTGTG-3Ј) was derived from the extreme 5Ј-end of the EST T70648 and used with an extreme 3Ј-end primer (BGAF-12, 5Ј-CGATTCAAGTGCCAATCTCT-GGC-3Ј) to amplify, clone, and sequence the longest possible cDNA from H95 by RT-PCR.

RESULTS
BGAF Binding Assays-To examine the molecular basis of the BGAF-␤-glucosidase interaction and define the location of the putative binding site(s), chimeric ␤-glucosidases were constructed (Fig. 1B). The interaction between BGAF and ␤-glucosidase is very specific and reminiscent of an antigen-antibody interaction, therefore, binding assays could be performed with both purified and unpurified ligands with no effect on assay sensitivity and specificity (11). Consequently, 7 (Dhr1, C-3, C-5, C-6, C-18, C-19, and C-26) out of 16 ␤-glucosidases were used as crude bacterial cell lysates in binding assays.
When we tested intact Glu1 and Glu2 and a Glu2/Glu1 chimera (C-3) for BGAF binding, the gel-shift assay yielded positive binding results as evident from the formation of BGAF-␤-glucosidase complexes with reduced electrophoretic mobility on native PAGE gels. Thus, ␤-glucosidase activity zones (bands) shifted toward the cathodic end of the gel and a smear extending from the top of the resolving gel to the sample well in the stacking gel was present (Fig. 2, lanes 4, 6, and 20). The binding assays also demonstrate that when the C-terminal 47 amino acids from either Glu1 or Glu2 are replaced with the corresponding C-terminal 53 amino acids from Dhr1 (as is the case with C-2 and C-4, respectively) BGAF binding activity is almost completely lost (Fig. 2, lanes 8 and 10). However, when the C-terminal 23 amino acids of the Glu1/Dhr1 chimera C-2 are replaced by the C-terminal 17 amino acids of Glu1, which In the case of chimeras C-3, C-5, C-6, C-19, and C-26 crude expression extracts were used to mix with purified BGAF. Note that ␤-glucosidase activity zones (bands and smearing) detected with MUGlc are retarded in a region extending from the top of the resolving gel to the sample well in the stacking gel when BGAF binds wild-type ␤-glucosidases (lanes 4 and 6) and their chimeras (lanes 8, 10, 12, 14, 16, 18, 20, 22, and 24). In the case of C-5, there were two distinct ␤-glucosidase-BGAF complexes, whereas in others the complexes formed a smeared zone. yielded C-16, BGAF binding activity is mostly restored (Fig. 2,  lane 12). Not surprisingly, replacing the extreme C-terminal 17 amino acids of Glu1 by the 23 amino acids of Dhr1 (C-15) showed BGAF binding activity (Fig. 2, lane 14). The binding data from chimeras C-21 and C-22 show that BGAF binding becomes tighter when the disruptive region (from Dhr1) spanning amino acids Ser 466 -Leu 495 of C-16 is bisected, yielding C-21 and C-22 (Fig. 2, lanes 16 and 18). In the case of Dhr1/ Glu1 or Dhr1/Glu2 chimeras, in which the N-terminal Glu1 or Glu2 regions were replaced with the N-terminal Dhr1 regions varying from 127 (C-26) and 205 (C-19) to 461 amino acids long (C-18, data not shown), no BGAF binding to any of these three chimeras was observed (Fig. 2, lanes 26 and 28).
In contrast, a Dhr1/Glu1 chimera (C-5) and a Dhr1/Glu2 chimera (C-6) in which the extreme N-terminal 29-amino acidlong segments of Glu1 and Glu2 were replaced with their Dhr1 homologue bind BGAF, indicating that this segment is not involved in BGAF binding (Fig. 2, lanes 22 and 24). However, C-5, which had the highest electrophoretic mobility (Fig. 2, lane  21), produced predominantly two distinct bands after interaction with BGAF (Fig. 2, lane 22). The region spanning amino acids Glu 50 -Phe 205 was bisected through the construction of chimera C-26 in which the N-terminal Ser 1 -Asn 127 region of Glu1 was replaced with its Dhr1 homolog. Interestingly, C-26 did not have any BGAF binding activity (Fig. 2, lane 28) similar to C-19 (Fig. 2, lane 26), which establishes that the polypeptide segment spanning amino acids Glu 50 -Asn 127 must contain the other region(s) of Glu1 or Glu2 that is involved in forming the BGAF binding site. Binding assays show that BGAF does not bind to Dhr1 (Fig. 3, right panel) where the results were similar to those obtained with negative control BGAF by itself and BGAF plus E. coli lysate (data not shown). In contrast, the amount of BGAF detectable by immunoblotting decreased as the amount of BGAF reacting with the positive tester ␤-glucosidase (Glu1) increased, because the resulting complex was unable to enter the gel (Fig. 3, left panel). Furthermore, Dhr1 did not bind BGAF in coprecipitation assays (data not shown) performed as described previously (11).
Mapping BGAF Binding Regions-Our structural analysis relative to mapping was on the three-dimensional structure of Glu1 resolved by x-ray crystallography in collaboration with Bernard Henrissat's crystallography group in Marseilles, France (27). Our initial finding that two regions (28 and 17 amino acids long) within the C-terminal 47 amino acids (based on data with C-15 and C-16) were each essential, but not sufficient, for BGAF binding led to the analysis of these regions on the surface of the Glu1 three-dimensional structure. It appears that the extreme 17-amino acid-long C-terminal region alone makes a greater contribution to BGAF recognition and binding than the 30-amino acid-long C-terminal region preceding it (cf. Fig. 2, lanes 12 and 14). To identify other polypeptide regions that are involved in BGAF binding, we scanned structural elements and amino acids located on the surface in the direct vicinity of the C-terminal 17 amino acids. Analysis of the three-dimensional structure of Glu1 indicates that the N-terminal region maps proximally to the C-terminal region (Fig. 4). On this basis, chimeras C-5, C-6, C-19, and C-26 were tested for BGAF binding activity. Chimeras C-5 and C-6 had binding activity, whereas C-19 and C-26 did not. Collectively, these binding data indicate that the 77-amino acid-long region comprising amino acids Glu 50 -Asn 127 in the N-terminal half contain the other determinant(s) that is (are) involved in BGAF binding, and they map to the surface proximal to certain residues from the C-terminal 47-amino acid-long region. Both the binding data and the structural data corroborate the postulate that binding requires the formation of a site through folding by two distant regions of the primary structure.
Isolation and Identification of the BGAF cDNA-The BGAF cDNA was cloned and sequenced as described above and reported (GenBank accession number AF232008). The longest BGAF cDNA isolated from H95 is 1118 bp long and includes a 918-bp coding sequence and a 43-bp 5Ј-and a 157-bp 3Ј-untranslated region. Fig. 5 shows the deduced primary structure of the 306-amino acid-long putative BGAF precursor, which contains two octapeptide (G(P/R)WGGSGG) repeats that are separated by 40 amino acids. These two repeats are postulated to play an essential role in forming the sites involved in binding to ␤-glucosidase (see below). The experimentally determined N-terminal sequence VISNKAPIGI of the mature protein starts 38 amino acids after the first methionine in the precursor. Thus, the BGAF precursor has a 38-amino acid-long presequence (i.e. signal peptide), which leaves a mature protein that is 268 amino acids long after the cleavage of the signal peptide.
Expression of the BGAF cDNA in E. coli-Two BGAF cDNA expression constructs (one coding for the putative BGAF precursor and the other for the mature protein) were prepared and expressed in E. coli. The mature protein encoding BGAF cDNA was expressed in moderate amount, and it yielded a protein whose electrophoretic mobility was faster than BGAF isolated from plant tissue (Fig. 6A, lanes 3 and 4). Yet, this smaller (27.5 kDa) recombinant BGAF protein was still able to bind and aggregate ␤-glucosidase (Fig. 6B, lane 4). In contrast, the pu-

FIG. 4. Three-dimensional structure of the maize ␤-glucosidase isozyme Glu1 showing a surface patch formed by amino acids that occur in two different regions (N-terminal Glu 50 -Asn 127 and C-terminal Phe 466 -Ala 512 ).
Within the patch (postulated BGAF binding site), the residues shown in red are invariant in Glu1, Glu2, and Dhr1, whereas those in purple are unique to Glu1 (and Glu2) and different in Dhr1. The Glu1 and Glu2-specific residues (purple) are postulated to be essential for BGAF binding, whereas one or more of the homologous Dhr1-specific sites are disruptive for BGAF binding. The variant sites shown in light blue may not be essential for binding because of their location or functional equivalence or variation between Glu1 and Glu2. The arrowhead on the left points to the active site, which is away from the BGAF binding site. tative precursor protein encoding BGAF cDNA yielded a protein (306-amino acid-long protein, calculated molecular mass 31.8 kDa) whose electrophoretic mobility was similar to that of BGAF isolated from plant tissue (Fig. 6A, lanes 5-7). The immunoblot analysis of the protein products of both BGAF cDNAs showed that they had the same immunoreactivity toward anti-BGAF serum as did BGAF extracted from maize shoots (Fig. 6A, lanes 3-7). The immunoblotting data clearly indicated that the presumptive BGAF cDNAs encode polypeptides that have the same immunoreactivity with the anti-native BGAF serum as the native BGAF. More unequivocal evidence that the putative BGAF cDNAs encode BGAF came from functional assays. The gel-shift assay in Fig. 6B (lanes 4 and 6) shows that BGAF expressed in E. coli binds ␤-glucosidase and retards its electrophoretic mobility in a manner identical to that obtained with native BGAF isolated from maize.

DISCUSSION
The ␤-glucosidase null phenotype previously reported in maize is due to a specific interaction between the enzyme and an aggregating factor or BGAF (11). The specificity of the interaction is proven by the fact that BGAF does not bind ␤-glucosidases from fungi (Trichoderma and Aspergillus) and other plant sources (e.g. almond, black cherry, sorghum, rice, and oats). We have cloned and sequenced the cDNA encoding BGAF (GenBank accession number AF232008) and identified BGAF as a member of the JAp family (16,17). It is possible that BGAF is also related functionally to small heat shock proteins, which (28 -30) bind nonnative proteins, preventing their aggregation and maintaining them in a state competent for ATP-dependent refolding by other chaperones.
The BGAF-␤-glucosidase system represents a highly specific protein-protein interaction, providing insights into the molecular mechanism of the interaction based on binding data from Glu1/Dhr1 and Glu2/Dhr1 chimeric enzymes. The BGAF binding assays with chimeras clearly show that replacement of the C-terminal 47 amino acids of Glu1 with its Dhr1 homolog abolishes binding (Fig. 2, lane 8), indicating that this region is essential for BGAF recognition and binding. The results of binding assays with chimeras C-2, C-4, C-16, C-15, C-21, and C-22 (Fig. 2, lanes 8, 10, 12, 14, 16, and 18) suggest that both the C-terminal 28-amino acid-long (Phe 466 -Lys 493 in C-16) and the extreme 17-amino acid-long (Lys 496 -Ala 512 in C-15) regions of Glu1 are individually capable of restoring BGAF binding, albeit not to the same extent. Thus, the Glu1 region Phe 466 -Lys 493 is compatible with the Dhr1 region Gln 492 -Asn 514 as is the Dhr1 region Ser 462 -Arg 489 with the Glu1 region Lys 496 -Ala 512 . This result is not surprising, because sites for most protein-protein interaction interfaces are confined to a surface patch that is usually composed of more than one stretch of the same polypeptide chain, which incrementally and combinatorially contribute to the interaction with the ligand. Within the region Phe 466 -Lys 493 , Glu1 and Glu2 differ from Dhr1 by eight amino acid substitutions where a given site in Dhr1 is occupied by a different and nonequivalent (except K493R) residue from that in Glu1 and Glu2 (Fig. 1A). Of these, F466S/ Y466S, A467S, Y473F, and K493R can be ruled out for involvement in BGAF binding, because the first three are buried in the active site cavity of the enzyme and the K493R substitution is likely to be equivalent. This narrows down the candidate sites for contribution to BGAF recognition and binding to N483G, T485E, Y487T, and E490R. These four amino acid substitutions separating Glu1 and Glu2 from Dhr1 are likely to have significant effects on BGAF recognition and binding, because they change the bulkiness, hydrophilicity, or charge of the side chain. Similarly, within the region Lys 496 -Ala 512 , Glu1 and Glu2 differ from Dhr1 by six amino acid substitutions (K496Q, T500G, K502A, P504K, S505N, and K506N), an internal dipeptide (VE) addition and a terminal tetrapeptide addition (GQLN), each of which alone or in combination with others may affect BGAF recognition and binding (Fig. 1A). In short, both C-terminal regions (Phe 466 -Lys 493 and Lys 496 -Ala 512 ) are individually capable of complementing an N-terminal region (Glu 50 -Val 145 , see below) to form a functional BGAF binding site. The finding that the BGAF binding site on Glu1 and Glu2 is made up of more than one stretch of polypeptide is also supported by the finding that chimeras C-18 (data not shown) and C-19 do not bind BGAF (Fig. 2, lane 26). Although the data indicate that the BGAF binding site is a surface patch that includes certain amino acids from the C-terminal (e.g. C-2, C-4) region, they alone are not sufficient to evoke BGAF binding. Moreover, BGAF showed no binding activity toward inactive, denatured Glu1 extracted from inclusion bodies (data not shown), indicating that the tertiary structure of the correctly folded enzyme is essential to form a functional binding site.
The finding that chimera C-19 did not bind BGAF suggested that the N-terminal 205 amino acids contain the other region(s) that is (are) involved in BGAF binding. Analysis of the three- FIG. 5. Primary structure of the putative BGAF precursor protein. The two octapeptide repeat regions postulated to be involved in binding to maize ␤-glucosidases are shown in underlined and italicized. The hydrophilicity plot (not shown) predicts that these regions reside on the surface. The N-terminal sequence (in boldface) determined by sequencing a purified BGAF preparation isolated from the "null" maize inbred H95 follows a 38-amino acid-long signal peptide (underlined). The peptide sequence APIGIGAT used for designing oligonucleotide sequences to amplify the original BGAF cDNA is underlined.

FIG. 6. Expression of the BGAF cDNA in E. coli.
A, SDS-PAGE gel (12%, anode at bottom) blot of E. coli lysates probed with anti-BGAF serum. Lanes 1 and 2, total and soluble protein, respectively, from nonrecombinant pET21a containing E. coli cells (negative control); 3 and 4, total and soluble protein, respectively, from cells containing recombinant pET21a with the mature protein coding BGAF cDNA; 5 and 6, total and soluble protein, respectively, from cells containing recombinant pET21a with the precursor protein coding BGAF cDNA; 7, shoot extract from maize inbred H95 serving as a positive control. Note that the mature BGAF protein (lanes 3 and 4) produced in E. coli is smaller in size (nonglycosylated) than that produced in the maize plant (lane 7). B, BGAF binding to ␤-glucosidase detected on a native-PAGE gel (8%, anode at bottom) after incubation with MUGlc. Lane 1, lysate of E. coli cells containing nonrecombinant pET21a (negative control); 2, Glu1 (no BGAF); 3, expression extract from the mature protein coding BGAF clone (no ␤-glucosidase); 4, Glu1 ϩ lysate from the mature protein coding BGAF clone; 5, lysate from the putative precursor protein encoding BGAF clone; 6, same as lane 5, but mixed and incubated with Glu1; 7, Glu1 ϩ lysate from cells containing nonrecombinant pET21a (no BGAF). Note that only the lanes containing recombinant mature (lane 4) and putative precursor BGAF (lane 6) show retarded ␤-glucosidase activity zones indicating the presence of a functional BGAF. dimensional structure of Glu1 (27) indicated that several segments from the N-terminal 205-amino acid-long region are located on the surface proximal to the C terminus (Fig. 4). The BGAF binding data from C-5, C-6, C-19, and C-26 together bracketed the region spanning Glu 50 -Asn 127 as the other region contributing to the structure of the BGAF binding site. Again, within this region, Glu1 and Glu2 differ from Dhr1 by 13 amino acid substitutions. Of these, four sites (I72V, N75D, K81A, and T82A; highlighted in purple in Figs. 1A and 4) are likely to make a major contribution to the formation of the BGAF binding site, because they all cluster within a surface patch (Fig. 4). The patch includes four amino acids (N483G, T485E, Y487T, and E490R; purple in Figs. 1A and 4) from region Phe 466 -Lys 493 and 2 (K496Q and T500G; purple in Figs. 1A and 4) of six from region Lys 496 -Ala 512 from the extreme 47-amino acid-long Cterminal region. Four sites (K502A, P504K, S505N, and K506N) at which Glu1 and Glu2 differ from Dhr1 are not shown in Fig. 4, because they are within the last 11-amino acid-long free coil region at the extreme C terminus that could not be resolved in the crystal structure. Thus, the binding assays have identified two discontinuous segments (Glu 50 -Asn 127 near the N terminus and Phe 466 -Ala 512 at the C terminus) that are brought together on the surface of the ␤-glucosidase tertiary structure to form a functional BGAF binding site. This provides a plausible explanation for the lack of BGAF binding to Dhr1 and certain Glu1/Dhr1 or Glu2/Dhr1 chimeras (C-2, C-4, C-18, C-19, and C-26) in which local structural changes due to amino acid substitutions at one or more of the 11 plus postulated sites disrupt binding.
The N-terminal sequence data from a purified BGAF preparation and the three maize ESTs whose 3Ј-end overlapped with the 5Ј-end of the cloned BGAF cDNA were key to the isolation of a cDNA (1118 bp) with a full-length coding sequence from the maize inbred H95. The fact that the experimentally determined N-terminal sequence (VISNKAPIGIGATV) starts 38 amino acids after the first methionine in the deduced primary structure of the putative BGAF precursor suggests that the cDNA contains a full-length (918 bp) coding sequence (Fig. 5). Thus, the BGAF precursor is 306 amino acids long and has a 38-amino acid-long signal peptide. Thus, our experimentally determined N-terminal sequence belongs to a 268-amino acid-long mature BGAF protein whose calculated molecular mass and that of the one expressed in E. coli are similar (27.5 kDa) but is smaller than expected (ϳ35 kDa), suggesting that the native BGAF isolated from maize is post-translationally modified (e.g. glycosylated). Interestingly, BLAST searches indicated that BGAF shared significant amino acid identity (58 to 61%) with three small heat-shock proteins (GenBank AF021258, U43496, and U43497) from barley that function in the systemic acquired resistance response and with a similar protein (GenBank U32427 (31)) from wheat (47% identity), which is also involved in systemic acquired resistance.
Immunoblots of E. coli lysates, in which the precursor BGAF and mature BGAF protein coding clones were expressed, provided unequivocal identification of the isolated BGAF cDNA. The blots showed that an immunoreactive band with the same electrophoretic mobility and molecular size as BGAF isolated from plant extracts in the case of the precursor (Fig. 6A, lanes  5 and 6) and a smaller than expected polypeptide in the case of the recombinant mature protein (Fig. 6A, lanes 3 and 4). In addition, functional assays confirmed the presence of BGAF in E. coli expression extracts, which tested positive for ␤-glucosidase aggregating activity in gel-shift assays (Fig. 6B, lanes  3-6).
A most intriguing feature in the primary structure of BGAF was the presence of two octapeptide repeats (G(P/R)WGGSGG), which occur 40 amino acids apart (Fig. 5) and whose variants are also present in three barley JAps as GPWGG(N/S)GG and in one wheat JAp as GPWG(K/G)(I/P)(S/C)G. We postulate that each of these octapeptides either individually forms a ␤-glucosidase binding site or makes a major contribution to it. The hydrophilicity plot predicts that both octapeptide repeats reside on the surface of BGAF and, thus, would be available for interaction with maize ␤-glucosidases. Our previous finding that the BGAF-␤-glucosidase aggregates can grow to a size in excess of 1.5 ϫ 10 6 Da (11) suggests that both BGAF and ␤-glucosidase must be bivalent. If BGAF were monovalent, it could only bind one ␤-glucosidase dimer, resulting in a soluble quaternary association of discrete size (ϳ180 -190 kDa), which is not observed. Densitometric analysis of BGAF and ␤-glucosidase monomer intensities after cosedimentation suggests a stoichiometry of about 2 molecules of ␤-glucosidase (two monomers of a homodimer) to 1 molecule of BGAF (monomer), consistent with the bivalency of both molecules. Based on the binding assays and the finding of two octapeptide repeats in the primary structure of BGAF, we propose that one BGAF molecule or monomer binds as a divalent ligand to two ␤-glucosidase dimers, linking ␤-glucosidase dimers in a linear chain in which monomeric BGAF with two binding sites and dimeric ␤-glucosidase with two BGAF binding sites alternate (Fig. 7). In fact, the ␤-glucosidase-BGAF interaction is similar to antigen-antibody interactions, having an equivalence point (11). When soluble ␤-glucosidase and BGAF are present in the correct ratios, optimal precipitation occurs. In the region of either ␤-glucosidase or BGAF excess, only small complexes are formed. There are other examples of ␤-glucosidase aggregation and ␤-glucosidase binding proteins in plants. ␤-Glucosidases from flax and oat occur in high molecular mass forms ranging from 245 to 1200 kDa (32)(33)(34). Additionally, Falk and Rask (35) reported two myrosinase (␤-thioglucosidase)-binding proteins (50 and 52 kDa) from rapeseed.
It appears that binding of BGAF to ␤-glucosidase does not affect enzyme activity and kinetic parameters, suggesting that BGAF binding neither sterically blocks the active site nor changes the conformation to alter enzyme activity. This suggestion is corroborated by the finding that the postulated BGAF binding site (formed by residues in domains Glu 50 -Asn 127 and Phe 466 -Ala 512 ) on ␤-glucosidase is away from the active site (Fig. 4). One plausible function of BGAF-␤-glucosidase interaction may be that BGAF plays a protective role for ␤-glucosidase, shielding the enzyme from endogenous proteases or proteases in the secretions of invading pests. Additionally, the BGAF-␤-glucosidase interaction would keep active ␤-glucosidase at the wound site, preventing the enzyme from diffusing to other parts of the plant where it has been shown to elicit deleterious effects (36). FIG. 7. A model for ␤-glucosidase-BGAF interaction leading to ␤-glucosidase aggregation and insolubility. The model is based on BGAF-␤-glucosidase binding assays, which have identified one BGAF binding site per ␤-glucosidase monomer, two per homodimer. Additionally, BGAF in its uncomplexed form exists as a monomer and thus, must minimally be bivalent. The two octapeptide repeats found in the primary structure of BGAF ( Fig. 5; G(P/R)WGGSGG) is postulated to be involved in ␤-glucosidase recognition and binding, and they support the model along with stoichiometric data, which show a ratio of 2 molecules of ␤-glucosidase (dimer) per one molecule of BGAF.
In conclusion, we have shown that maize BGAF is a member of the JAp family that specifically interacts with ␤-glucosidase. Based on corroboratory binding and structural data we have identified two different regions in the primary structure of ␤-glucosidase, which form a BGAF binding site on the surface of the enzyme. We have also isolated the BGAF cDNA, deduced the sequence of its protein product, identified two octapeptide repeats in the sequence, and postulated that they form two binding sites each of which binds a monomeric unit of the ␤-glucosidase homodimer. We have confirmed the identity of the BGAF cDNA further by expressing it in E. coli and demonstrating that its recombinant protein products (mature and precursor BGAF) are functionally and immunologically identical to the native BGAF isolated from maize. The finding that BGAF shares significant amino acid identity (58 -61%) with three defense proteins (GenBank AF021258, U43496, and U43497) from barley involved in systemic acquired resistance and the fact that ␤-glucosidase also plays a role in defense suggest that the specific interaction between BGAF and ␤-glucosidase has physiological relevance. Future studies will focus on the precise identification of specific amino acids within the binding sites of BGAF and ␤-glucosidase and defining their roles using site-directed mutagenesis and x-ray crystallography, as well as understanding the physiological function of the BGAF-␤-glucosidase interaction.