OPEN ACCESS ARTICLE

This Article Free via Open Access: OA OA Abstract Full Text (PDF) OA All Versions of this Article: 146/4/1892    most recent pp.108.116285v1 Alert me when this article is cited Alert me if a correction is posted Services Email this article to a friend Related articles in Plant Physiol. Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Google Scholar Articles by Hennen-Bierwagen, T. A. Articles by Myers, A. M. PubMed PubMed Citation Articles by Hennen-Bierwagen, T. A. Articles by Myers, A. M. Agricola Articles by Hennen-Bierwagen, T. A. Articles by Myers, A. M.

BIOCHEMICAL PROCESSES AND MACROMOLECULAR STRUCTURES

Starch Biosynthetic Enzymes from Developing Maize Endosperm Associate in Multisubunit Complexes^1,[OA]

Department of Biochemistry, Biophysics, and Molecular Biology, Iowa State University, Ames, Iowa 50011 (T.A.H.-B., R.S.M., S.K., Q.G., M.G.J., A.M.M.); and Department of Molecular and Cellular Biology, College of Biological Science, University of Guelph, Guelph, Ontario, Canada N1G 2W1 (F.L., I.J.T., M.J.E.)

	ABSTRACT

TOP ABSTRACT RESULTS DISCUSSION MATERIALS AND METHODS LITERATURE CITED

 
Mutations affecting specific starch biosynthetic enzymes commonly have pleiotropic effects on other enzymes in the same metabolic pathway. Such genetic evidence indicates functional relationships between components of the starch biosynthetic system, including starch synthases (SSs), starch branching enzymes (BEs), and starch debranching enzymes; however, the molecular explanation for these functional interactions is not known. One possibility is that specific SSs, BEs, and/or starch debranching enzymes associate physically with each other in multisubunit complexes. To test this hypothesis, this study sought to identify stable associations between three separate SS polypeptides (SSI, SSIIa, and SSIII) and three separate BE polypeptides (BEI, BEIIa, and BEIIb) from maize (Zea mays) amyloplasts. Detection methods included in vivo protein-protein interaction tests in yeast (Saccharomyces cerevisiae) nuclei, immunoprecipitation, and affinity purification using recombinant proteins as the solid phase ligand. Eight different instances were detected of specific pairs of proteins associating either directly or indirectly in the same multisubunit complex, and direct, pairwise interactions were indicated by the in vivo test in yeast. In addition, SSIIa, SSIII, BEIIa, and BEIIb all comigrated in gel permeation chromatography in a high molecular mass form of approximately 600 kD, and SSIIa, BEIIa, and BEIIb also migrated in a second high molecular form, lacking SSIII, of approximately 300 kD. Monomer forms of all four proteins were also detected by gel permeation chromatography. The 600- and 300-kD complexes were stable at high salt concentration, suggesting that hydrophobic effects are involved in the association between subunits.

The chemical structures of amylopectin and glycogen are produced by the same classes of enzyme, specifically a glucan synthase that transfers Glc residues to a growing linear chain from a nucleotide sugar donor and a glucan branching enzyme that cleaves a linear chain at a glycoside bond and transfers one portion of it to a C-6 hydroxyl. A possible explanation for the different molecular architecture of amylopectin compared to glycogen is that the plant enzymes inherently have distinct substrate specificities that determine their sites of action. Amylopectin biosynthesis, however, likely requires coordination among these enzymes in addition to their individual substrate specificities. In contrast to glycogen biosynthesis, which typically requires only one glycogen synthase and one branching enzyme, multiple classes of starch synthase (SS) and starch branching enzyme (BE) are present in starch-producing tissues. Individual SSs and BEs are conserved to a large extent in the plant kingdom, and genetic analyses indicate that in many instances the functions of each class are nonredundant (Myers et al., 2000; Nakamura, 2002; Ball and Morell, 2003; James et al., 2003; Zhang et al., 2004, 2005; Delvalle et al., 2005; Fujita et al., 2007). Another distinguishing feature of amylopectin biosynthesis is that -(16) glucan hydrolase activity, i.e. starch debranching enzyme, is required for wild-type starch levels and for normal amylopectin structure (Ball et al., 1996; Myers et al., 2000; Ball and Morell, 2003). No such function has been implicated in glycogen biosynthesis. Little is known regarding how the activities of so many enzymes capable of covalently modifying glucan polymers are coordinated to achieve the crystallization competent structure of amylopectin.

Longstanding genetic evidence suggests that starch biosynthetic enzyme activities are coordinated as opposed to acting independently through stochastic mechanisms. A classical example is the maize (Zea mays) gene dull1 (du1) that codes for SSIII (Gao et al., 1998). The du1 gene was identified originally by mutant alleles that act as second-site genetic modifiers of a specific mutation of the gene sugary1 (su1; Mangelsdorf, 1947), which codes for the starch debranching enzyme isoform ISA1 (James et al., 1995). Neither the du1^– mutation nor the su1^– mutation have appreciable affects on kernel phenotype; however, when they are coupled in double homozygous mutants, there is a severe defect in starch biosynthesis that can be easily tracked genetically (Mangelsdorf, 1947). The fact that the mutant phenotypes are not additive suggests that SSIII and ISA1 function via a concerted mechanism as opposed to catalyzing independent steps in starch biosynthesis. Genetic evidence also indicates that numerous starch biosynthetic enzymes depend on other components of the pathway for their activity. For example, mutation of SSIII affects the enzymatic activity of both BEIIa and SSI, in the former instance reducing activity and in the latter causing an increase in enzyme activity observed in cell extracts (Boyer and Preiss, 1981; Singletary et al., 1997; Cao et al., 1999).

A hypothesis to explain these genetic interactions, as well as the organized architecture of amylopectin with regard to branch placement and frequency, is that the enzymes associate in one or more complexes. Such an arrangement would afford the possibility of coordinate and possibly reciprocal regulation and also might affect amylopectin architecture owing to spatial relationships between active sites. Consistent with this hypothesis, stable complexes between enzymes of the starch biosynthetic pathway were demonstrated in wheat (Triticum aestivum) endosperm extracts by immunoprecipitation (Tetlow et al., 2004). Specifically, BEIIb, BEI, and starch phosphorylase associated in a precipitable complex that requires Ser phosphorylation for stability. Further evidence from wheat endosperm is presented in the accompanying manuscript showing that trimeric complexes can be formed between SSI, SSII, and either of BEIIa or BEIIb (Tetlow et al., 2008).

This study took a systematic approach to investigate the ability of SSs and BEs from developing maize endosperm to associate in multisubunit complexes. Initial studies using in vivo protein-protein interaction tests revealed many interactions between distinct enzymes of the pathway that were sufficiently stable to reconstitute activity of the GAL4 reporter protein in yeast (Saccharomyces cerevisiae) nuclei. Biochemical analyses using immunoprecipitation and affinity purification methods demonstrated directly that multiple components of the pathway are capable of forming stable complexes. Interactions were observed involving SSI, SSIIa, SSIII, BEI, BEIIa, and BEIIb. In addition, many of these proteins comigrate in high molecular mass peaks in gel permeation chromatography (GPC), centered at either 600 or 300 kD. These observations of specific associations in cell extracts support the hypothesis that starch biosynthetic enzymes are physically coordinated in vivo.

	RESULTS

TOP ABSTRACT RESULTS DISCUSSION MATERIALS AND METHODS LITERATURE CITED

 

In Vivo Protein-Protein Interaction Tests

Various maize endosperm starch biosynthetic enzymes were examined for their ability to form stable complexes using directed in vivo protein-protein interaction tests (Chien et al., 1991). Data are reported here for the SS isoforms SSI and SSIII, and the BE isoforms BEI, BEIIa, and BEIIb. With the exception of SSIII, the polypeptides tested were the full-length mature proteins lacking the plastid targeting sequence. For SSIII, four fragments were tested, comprising residues 1 to 181, residues 1 to 367, residues 366 to 648, or residues 760 to 1,438 of the full-length protein. All of these segments of SSIII are located in the long N-terminal extension, distinct from the catalytic structure formed by the C-terminal portion of the polypeptide (Gao et al., 1998). Each polypeptide was fused separately to both the activation domain and the DNA binding domain of the yeast transcriptional activator GAL4, referred to as GAL4-AD and GAL4-BD, respectively (Chien et al., 1991). Empty vectors expressing GAL4-AD or GAL4-BD, but lacking any exogenous cDNA fusion, were utilized as negative controls. Positive controls were provided by fusions including the SV40 large T antigen or mouse p53, which in combination yield a very strong signal in this system (Li and Fields, 1993).

Plasmids were cotransformed into yeast such that each protein pair was tested reciprocally with respect to which polypeptide was fused to GAL4-AD or GAL4-BD. In addition, each gene fusion plasmid was combined with the reciprocal negative control empty plasmid, and some were also tested with the SV40 and p53 positive controls. Results from the test matrix are presented in Table I, and sample data are shown in Figure 1 . The yeast host strain PJ69-4A contains as reporter genes the ADE2 and HIS3 prototrophic markers, as well as the Escherichia coli gene lacZ coding for β-galactosidase, in engineered forms that require reconstituted GAL4 activity for expression (James et al., 1996). All notations of positive signals in Table I indicate that the transformant grew on minimal medium lacking Ade and also generated obvious blue coloration in the filter lift assay that indicates cleavage of the substrate 5-bromo-4-chloro-3-indolyl-β-D-galactoside by β-galactosidase. The ADE2 reporter is known to be particularly stringent as a genetic marker in this system.

View larger version (72K): [in this window] [in a new window] [as a PowerPoint slide]   Figure 1. In vivo protein-protein interaction between SSIIIHDX and SSI. The indicated combinations of gene fusions containing either the GAL4 transcriptional activation domain (AD) or the GAL4 DNA binding domain (BD) were cotransformed into yeast reporter strain PJ69-4A. Single colony-purified transformants were patched on –Ade selective medium to test for activation of the ADE2 reporter gene or on minimal selective medium for filter lifts and β-galactoside assays after cell permeabilization to test for activation of the lacZ reporter gene. p53 and SV40 large T antigen (LTA) are well-documented positive controls for this system. SSIII in all instances represents the fragment SSIIIHDX comprising residues 760 to 1,438.

View larger version (7K): [in this window] [in a new window] [as a PowerPoint slide]   Figure 2. SSIII gene diagram. Numerals indicate the residue number of the SSIII full-length primary sequence. The black box indicates the region that is homologous to prokaryotic glycogen synthases and other -(14) glucan synthases and thus contains the domains responsible for catalytic activity. The gray box indicates SSIIIHD, a region of sequence conserved specifically in SSIII proteins in both monocots and dicots. Arrows indicate the region of SSIII primary sequence that bound to full-length SSI in yeast nuclei, and asterisks indicate the location of peptide sequences identified from a 60-kD polypeptide purified on a SSI-affinity column (see Fig. 7).

View larger version (44K): [in this window] [in a new window] [as a PowerPoint slide]   Figure 7. Mass spectrometry identification of affinity purified proteins. Amyloplast proteins were fractionated by affinity chromatography using S-protein agarose coupled to either SSI or BEIIa as the matrix. A, Sypro Ruby stain. M lanes contain molecular mass markers and E lanes contain proteins present in the pooled fraction from the indicated affinity column. Peptide sequences identified by mass spectrometry in specific eluted bands are indicated. Numbers in parentheses indicate the positions of those peptide fragments in the full-length protein. GenBank accession numbers to reference the positions of each peptide are specified in the text.

Polyclonal antisera raised against recombinant protein or synthetic peptides were developed as reagents for recognition of BEI, BEIIa, BEIIb, SSI, SSIIa, or SSIII from maize endosperm. Genetic tests to verify the specificity of each antiserum made use of mutations in the genes ae (Stinard et al., 1993), sbe2a (Blauth et al., 2001), sbe1 (Blauth et al., 2002), su2 (Zhang et al., 2004), and du1 (Gao et al., 1998), which code for BEIIb, BEIIa, BEI, SSIIa, or SSIII, respectively.

The serum designated BEIIa/b was raised using full-length recombinant BEIIa as the antigen. This serum detects two major protein bands in wild-type endosperm soluble extracts, with apparent molecular mass of approximately 90 kD or 75 kD (Fig. 3A , center). The protein generating the 90-kD signal is absent from sbe2a^– mutant endosperm extracts (Fig. 3A, center), and its apparent molecular mass matches the value predicted from the known cDNA sequence and mature amino terminus of BEIIa (Gao et al., 1997). Another antiserum, designated BEIIa, was raised using a BEIIa-specific synthetic peptide as the antigen. This serum also recognizes the 90-kD protein present in wild-type endosperm but fails to generate this signal in the sbe2a^– mutant (Fig. 3B, top right). These data demonstrate conclusively that both BEIIa/b and BEIIa recognize BEIIa in immunoblots.

View larger version (54K): [in this window] [in a new window] [as a PowerPoint slide]   Figure 3. Antibody specificity tested by immunoblot analysis. Soluble endosperm extracts from maize lines of the indicated genotype were separated by SDS-PAGE and subjected to immunoblot analyses using the indicated antibodies at dilutions of 1/1,000 or 1/2,000. Abbreviations for mutations are as follows: ae–, ae-1; ae°, ae-B; sbe1°, sbe1::Mu; 2a°, sbe2a::Mu; su2°, su2-19791; du1°, du1-M3. WT, Wild-type inbred maize line W64A; M, lanes containing molecular mass standards. A, Characterization of BEI and BEIIa/b. B, Characterization of BEIIa, BEIIb, and SSII. C, Characterization of SSIIIHDP. In C, both lanes were loaded with 20 µg of amyloplast-enriched soluble protein fraction.

View larger version (18K): [in this window] [in a new window] [as a PowerPoint slide]   Figure 6. Affinity chromatography using amyloplast extracts. Amyloplast extracts in low salt buffer were incubated with S-protein agarose attached to the indicated recombinant maize proteins or S-protein agarose without any recombinant protein attached (indicated by –). After washing, the beads were boiled in SDS-PAGE sample buffer. Bound proteins were subjected to SDS-PAGE and immunoblot analyses using the indicated antisera. These SDS-PAGE gels were 4% to 12% acrylamide run using the NuPAGE buffer system. A, SSI affinity column. B. BEIIa affinity column. B2a indicates BEIIa.

The antipeptide serum SSIIa identified a protein with an apparent molecular mass of approximately 75 kD that is absent from the su2^– null mutant (Fig. 3B, bottom left). The apparent molecular mass of the signal matches the predicted value for mature SSIIa based on the cDNA sequence coding for the mature protein (Harn et al., 1998). These data indicate that SSIIa effectively identifies SSIIa.

The antipeptide serum SSIg recognized a protein with an apparent molecular mass of approximately 74 kD, as did the serum SSI that was raised against recombinant SSI (see Figs. 5A and 6B). This molecular mass is larger than the value predicted from the SSI cDNA; however, anomalous migration in SDS-PAGE is well documented for this particular protein (Mu et al., 1994; Knight et al., 1998). SSI maize mutants are not available, so genetic analysis could not be used to verify the specificity of the SSI and SSIg antisera.

View larger version (42K): [in this window] [in a new window] [as a PowerPoint slide]   Figure 5. Affinity chromatography using whole kernel extracts. Bound proteins were eluted in a step gradient of increasing KCl concentration, concentrated, and subjected to SDS-PAGE and immunoblot analysis with the indicated antisera. A, Affinity chromatography with immobilized BEIIa. B, Affinity chromatography with immobilized SSI. Open triangle, BEIIa; black triangle, BEIIb. Lane designations are the same for both segments. Lane M, Molecular mass markers. Lane 1, Total kernel extract prior to application to the column; lane 2, second low salt wash of the column after loading with proteins from the maize extracts; lane 3, fourth low salt wash of the column; lane 4, 0.2 M KCl elution; lane 5, 0.4 M KCl elution; lane 6, 0.6 M KCl elution; lane 7, 0.8 M KCl elution; and lane 8, 1.0 M KCl elution.

GPC Behavior of Maize Endosperm BEs and SSs

GPC was used to examine whether specific SSs and BEs fractionate in apparent high molecular mass complexes compared to the mass of the monomeric proteins. The starting fraction for this column was soluble lysate from partially purified amyloplast fractions obtained from either of two inbred lines, Oh43 or W64A, harvested 16 to 17 d after pollination (DAP). Column fractions were analyzed by immunoblot analyses using BEIIa/b, BEI, BEIIa, BEIIb, SSIIa, or SSIIIHDP as the probe. In standard phosphate buffered saline buffer (PBS; 50 mM sodium phosphate buffer, pH 7.5, 150 mM NaCl), BEIIb appears in a broad distribution across the column elution. There is an obvious low molecular mass peak containing the majority of BEIIb (Fig. 4 , row 1, fractions C10–C12). Minor amounts of BEIIb are also detectable in high molecular mass fractions extending all the way to the 670-kD marker, which is the permeability limit for this column matrix (Fig. 4, row 1). Quantitatively different results were obtained when the GPC fractionation was performed in the presence of PBS containing 1 M NaCl instead of the standard concentration of 0.15 M NaCl (i.e. high salt conditions). In this instance, the abundance of BEIIb in the high molecular mass fractions increased significantly and exhibited a peak in fractions B4 to B6 at the elution volume corresponding to approximately 600 kD (Fig. 4, row 2). Thus, high salt conditions appear to favor formation of a high molecular mass complex containing BEIIb.

View larger version (85K): [in this window] [in a new window] [as a PowerPoint slide]   Figure 4. Fractionation of BEs and SSs by GPC. Total amyloplast extracts (approximately 2 mg total protein, 0.5-mL loading volume) were applied to a Superdex 200 10/30 HR gel permeation column. Fractions from the column elution were subjected to SDS-PAGE and then immunoblot analyses. Labels on top of each column indicate the fraction, with earlier elution and larger molecular masses on the left. The GPC elution positions of molecular mass markers (Bio-Rad), as determined in independent column runs under identical conditions, are indicated by small vertical arrows. Horizontal arrows indicate molecular mass marker positions for SDS-PAGE. The antibody used for each immunoblot is indicated to the right of each segment, along with the maize inbred line used to prepare the amyloplasts and the salt condition.

The same analysis applied to SSIIa again revealed a high molecular mass fraction and a monomer fraction, in this instance with the great majority of the protein in the apparently larger form. In standard PBS, the high molecular mass peak is centered around fraction C1, corresponding to approximately 300 kD, and this well-defined peak tails off at fraction B3 (Fig. 4, row 5). This peak is also observed in high salt conditions, along with a second high molecular mass peak of SSIIa centered around fraction B5 at the elution volume corresponding to approximately 600 kD (Fig. 4, row 6). Thus, BEIIa, BEIIb, and SSIIa all appear to share the property that high salt concentration favors their inclusion in a multimeric complex of approximately 600 kD.

Fractionation of SSIII on the GPC column was monitored using SSIIIHDP. In standard PBS buffer, SSIII eluted in a discrete, high molecular mass peak corresponding to approximately 600 kD as judged by the elution position of the 670-kD molecular mass marker (Fig. 4, row 7, fractions B3–B6). Long exposures of these immunoblots reveal minor amounts of SSIII extending to fraction C2 (data not shown), corresponding to approximately 200 kD and approximating the monomer molecular mass. Essentially similar results were observed when the analysis was performed in 1 M NaCl (data not shown).

The presence of BEI was also monitored throughout column elutions under both high and low salt GPC fractionation conditions. In both instances, the only form of BEI detected was a low molecular mass form, indicative of the monomer (Fig. 4, row 8; data not shown). Thus, under these GPC conditions, BEI behaved differently than BEIIa, BEIIb, SSIIa, or SSIII in the regard that it was not detected in the high molecular mass fractions.

In summary, the GPC data suggest the existence of at least two distinct high molecular complexes that contain multiple starch biosynthetic enzymes. SSIIa fractionates in a well-resolved peak of approximately 300 kD (Fig. 4, rows 5 and 6), designated here as "C300." The same fractions also contain BEIIa and BEIIb (Fig. 4, rows 1–4). A second apparent complex, referred to as "C600," is defined by migration of the SSIII protein in a peak at about 600 kD (Fig. 4, row 7). This same high molecular fraction peak also contains SSIIa, BEIIa, and BEIIb (Fig. 4, rows 2, 4, and 6). Formation of C600 appears to be favored by high salt conditions, represented here by 1 M NaCl. Further proteomic analysis will be required to define the nature of the complex(es) located in the C300 or C600 chromatography peaks. The specific elution behaviors of BEIIa, BEIIb, SSIIa, and SSIII, however, suggest that each of these proteins may be present in the same multisubunit complex. Whether or not SSI might also comigrate with the other starch biosynthetic enzymes could not be determined because the available antibodies were not effective in these conditions (data not shown).

Protein Interactions Detected Using Immobilized Recombinant Enzymes

Interacting Proteins from Whole Cell Extracts
Affinity purification with recombinant proteins immobilized on agarose beads was used as an additional method of detecting interactions among starch biosynthetic enzymes. An advantage of this technique is that the high concentration of the ligand in the solid phase can favor formation of complexes between proteins with relatively low binding affinity. The immobilized proteins, either BEIIa or SSI, were shown to be enzymatically active after binding to the matrix by biochemical assay (data not shown), indicating that natively folded protein is available for binding to potential partners in the cell extracts. Whole-cell soluble extracts from maize kernels were first passed through a column containing bound bovine serum albumin (BSA) to minimize nonspecific protein-protein binding interactions. The extracts were then applied to the immobilized affinity ligands in low salt conditions, specifically 0.1 M KCl. The columns were washed in low salt buffer until no further protein elution could be detected, and bound proteins were then eluted from the matrix in a step gradient of increasing salt concentration from 0.2 M to 1.0 M KCl. The eluted proteins were concentrated, separated by SDS-PAGE, and characterized by immunoblot analyses.

Recombinant BEIIa was immobilized to S-protein agarose by means of an S-tag sequence fused at the amino terminus, and this matrix was used to isolate proteins from whole cell extracts of developing maize kernels. Immunoblot analysis using SSI revealed that the SS bound to the BEIIa column and eluted in a specific peak in KCl concentrations of 0.2 M to 0.6 M (Fig. 5A ). This signal was not detected when the same analysis was applied using S-agarose beads without any maize protein attached (data not shown).

The reciprocal experiment was performed in which recombinant maize SSI was immobilized on S-protein agarose. Immunoblot analysis using BEIIa/b revealed two proteins that bind to the SSI matrix and migrate in SDS-PAGE at the apparent molecular masses known to correspond to BEIIa and BEIIb (Fig. 5B). Again, the bound protein eluted in a specific peak of salt concentration from 0.2 M to 0.6 M KCl. Thus, in both orientations in this affinity purification procedure, BEIIa and SSI were found to associate. In addition, an association between BEIIb and SSI was observed using the SSI affinity matrix.

Interacting Proteins from Amyloplast Extracts
Affinity purification with recombinant SSI or BEIIa as the immobilized ligand was also performed using maize amyloplast extracts. In these analyses, after application of the extract to the column, the beads were washed extensively in 0.15 M KCl to remove nonspecific interactions then boiled in SDS-PAGE loading buffer. Thus, all proteins attached to the affinity matrix were released in a single step, as opposed to the salt gradient elution employed for the whole kernel extracts. S-protein agarose beads without any bound recombinant protein were used as the negative control. The bound proteins were separated by SDS-PAGE and analyzed by immunoblot analyses using antisera that recognize SSI, SSIIa, BEIIb, BEIIa, or BEI.

Analysis of the proteins collected on the SSI affinity column verified the results obtained using whole cell extracts with regard to purification of BEIIb (Fig. 6A ) and BEIIa (data not shown). In addition, a signal for SSIIa was detected in the bound fraction by SSIIa serum (Fig. 6A). The BEIIa affinity column retained both SSI and SSIIa (Fig. 6B), but no signal was observed in the eluate using BEIIb as the immunoblot probe (data not shown). The affinity data taken together indicate the existence in these cell extracts of stable multisubunit complexes comprising SSI/SSIIa/BEIIa and SSI/SSIIa/BEIIb.

Identification of Affinity Purified Proteins by Mass Spectrometry
Mass spectrometry determination of amino acid sequences from affinity purified proteins was used as a complement to immunological identification. In these experiments, amyloplast extracts were applied directly to S-agarose columns bearing SSI or BEIIa as the affinity ligand. Samples were applied in buffer containing 0.15 M NaCl, and the columns were washed extensively in the same buffer. Bound proteins were eluted with a step gradient of increasing KCl concentration, and the fractions from 0.6 M to 1.0 M from each column were pooled, desalted, and concentrated. These proteins were then separated by SDS-PAGE and visualized by staining with Sypro Ruby.

The SSI affinity matrix bound two major protein bands, one migrating at an apparent molecular mass of approximately 70 kD and the other at approximately 60 kD (Fig. 7 ). Each band was excised from the agarose gel, digested with trypsin, and analyzed by electrospray tandem mass spectrometry. Two peptide sequences present in the 70-kD band matched precisely to sequences within maize BEIIa (GenBank accession no. AAB67316). Thus, BEIIa was again observed to bind to the SSI affinity column, in this instance eluting at a salt concentration that overlaps one in which it was identified immunologically. One of these two peptides is also identical in sequence to a segment of BEIIb (GenBank accession no. AAC33764), raising the possibility that BEIIb may also be present in this 70-kD band.

The 60-kD band from the SSI affinity column yielded two peptides that match precisely to sequences within maize SSIII (GenBank accession no. AAC14014). These data indicate that SSIII either binds directly to SSI or is part of a complex that contains SSI. The two peptides observed are located within the SSIIIHD portion of this class of SS (Fig. 2). The location of the two peptides, approximately 50 kD apart in primary sequence, considered together with the 60-kD molecular mass of the SSIII fragment retained on the SSI affinity column, indicates that the purified fragment comprises most or all of the SSIIIHD region and little else from the SSIII primary sequence.

The BEIIa column purified four visible protein bands that migrated at apparent molecular mass between approximately 70 kD and 80 kD. These four bands were excised as a single group and analyzed by mass spectrometry. Among these proteins, one peptide sequence was identified that matches precisely to a sequence within maize SSI (GenBank accession no. AAB99957; Fig. 7). Although a single peptide is not sufficient by itself to conclude that the protein is present, taken together with the immunological identifications, these data provide strong evidence that SSI is bound onto the BEIIa affinity matrix. The identities of the other three proteins bound to BEIIa could not be determined by the mass spectrometry analysis, probably owing to the low level of material present.

Further Analysis of SSIIIHD

The SSIII and BEIIa polypeptides purified on the SSI affinity column and identified by mass spectrometry reproducibly migrated in SDS-PAGE at apparent molecular masses less than that of the full-length protein. This is especially obvious for SSIII, because the purified polypeptide had an apparent molecular mass of 60 kD, whereas the full-length protein is approximately 180 kD. To test the authenticity of the SSI-SSIII interaction, immunoblot analysis was used to examine whether the 60-kD SSIIIHD fragment exists in the cell extract starting material applied to the SSI-affinity column. Use of the SSIIIHDP IgG fraction allowed identification of intact SSIII or any subfragments that contain SSIIIHD. In addition to apparent full-length SSIII, a 60-kD signal was clearly observed in total amyloplast extracts (Fig. 3C) and total endosperm cell extracts (data not shown). Thus, the fragment identified by peptide sequence is readily available in cell extracts.

In GPC analysis, the 60-kD SSIIIHD fragment was detected exclusively in the high molecular mass fraction C600 that also contains full-length SSIII, and none was observed in the fractions containing the SSIII monomeric form (Fig. 4, row 7). The 60-kD fragment, therefore, is associated with a high molecular mass complex. At this point, it is not possible to distinguish whether some portion of SSIII is fragmented in the C600 complex in vivo, or if limited proteolysis occurs after cell lysis. Protease inhibitor cocktail was included in all extraction buffers so as to minimize post-lysis degradation; however, whether or not the inhibition was completely effective cannot be known.

Further validation of the interaction between SSIIIHD and SSI is provided by the results of the in vivo protein-protein interaction tests. As described previously, a fragment made up for the most part of the complete SSIIIHD sequence tested strongly positive for interaction with SSI in the two-hybrid test (Fig. 1; Table I).

Coimmunoprecipitation of Multiple SSs and BEs

Immunoprecipitation from Kernel Extracts
Coimmunoprecipitation was used as another approach to analyze potential binding partners of maize BEIIa, BEIIb, SSI, SSIIa, and SSIII. Protein-A Sepharose beads preincubated with BEIIa/b serum effectively precipitated BEIIa from whole kernel extracts (Fig. 8A ). Little if any BEIIb was found in these eluted fractions. Thus, BEIIa/b appears to be more effective at precipitating BEIIa than BEIIb, even though in immunoblot analysis the antiserum recognizes both proteins in total extracts (Figs. 3 and 8A). The same procedure using preimmune serum did not result in any detectable BEIIa or BEIIb in the precipitate (Fig. 8A). Thus, binding of any coprecipitating proteins to the immunoaffinity matrix depends on the presence of antibodies that recognize BEIIa. Probing the precipitate fractions with SSI antiserum revealed an immunoblot signal of the size expected for the SSI protein (Fig. 8A). The preimmune serum did not precipitate SSI (Fig. 8A), again demonstrating that the apparent coprecipitation depends on antibodies that bind BEIIa.

View larger version (37K): [in this window] [in a new window] [as a PowerPoint slide]   Figure 8. Coimmunoprecipitation of starch biosynthetic enzymes. A, Precipitation from total kernel soluble extracts using BEIIa/b serum. The supernatant after immunoprecipitation and the eluted proteins were fractionated by SDS-PAGE and characterized by immunoblot analysis using the indicated antiserum. S lanes indicate the supernatant and P lanes indicate the precipitated proteins. B, Precipitation from total kernel soluble extracts using SS-specific antibodies. Details are as in A. C, Precipitation from amyloplast extracts using SSIg. After extensive washing, the immunoprecipitated beads were boiled in SDS-PAGE loading buffer, and bound proteins were separated by 10% SDS-PAGE, transblotted, and immunodetected using the antisera indicated. These SDS-PAGE gels included 10% acrylamide. D, Precipitation from amyloplast extracts using SSIIa (S2). Details are as in C. –, Mock precipitation without any serum exposed to the protein-A Sepharose.

The same approach was used to identify binding partners for SSIII. An antiserum termed DU1N, which was raised against residues 1 to 648 of SSIII (Cao et al., 1999), was preincubated with protein A-Sepharose. Those beads were then incubated with kernel extract, and the eluate was probed for the presence of BEIIa, BEIIb, and SSI. The results showed that SSI was precipitated by the DU1N antiserum (Fig. 8B). No signal was detected for BEIIa or BEIIb in the DU1N precipitate using BEIIa/b as the immunoblot probe (data not shown). This last result is contrast to the interactions between SSIII and the BEs observed by the in vivo interaction test, which might be explained by the capabilities of the particular antiserum used for immunoprecipitation.

Immunoprecipitation from Amyloplast Extracts
Immunoprecipitation was also performed using amyloplast extract as the starting material, with either SSIg or SSIIa as the precipitating serum. In these instances, the precipitates were boiled in SDS-PAGE sample buffer to release the precipitating antibodies and any bound proteins. The proteins were then separated by SDS-PAGE and analyzed by immunoblotting with various antisera. As expected, immunoprecipitation with SSIg effectively collected SSI from the amyloplast extracts (Fig. 8C). BEIIa and BEIIb were observed in the SSI coimmunoprecipitate (Fig. 8C), as had been seen using whole cell extracts (Fig. 8B). In addition, SSIIa and BEI were detected in the SSI coimmunoprecipitate by immunoblot analyses using SSIIa or BEIg serum (Fig. 8C). Control precipitations using protein-A Sepharose without any bound antibodies did not yield any signals in the immunoblot analysis (data not shown).

Coimmunoprecipitation using SSIIa as the precipitating agent was also performed. As expected, the SSIIa serum effectively precipitated SSIIa (data not shown). In addition, SSI and BEIIb were also detected as coprecipitated proteins in the immunoblot analysis (Fig. 8D).

	DISCUSSION

TOP ABSTRACT RESULTS DISCUSSION MATERIALS AND METHODS LITERATURE CITED

 

Multisubunit Complexes Containing Starch Biosynthetic Enzymes

The results of this study indicate that many of the starch biosynthetic enzymes in maize endosperm are capable of associating with each other, either directly or indirectly, in multisubunit complexes. The pairs of proteins that were observed by at least one analytical method to be present in the same complex are SSI/SSIIa, SSI/SSIII, SSI/BEI, SSI/BEIIa, SSI/BEIIb, SSIIa/BEIIa, SSIIa/BEIIb, and SSIII/BEIIa. Three different methods were employed to test these interactions, and many of the pairs were detected in multiple analyses (Table III ). The most frequently observed association was SSI with BEIIa, which was detected in both test orientations by immunoprecipitation and affinity purification and in one orientation of the in vivo protein-protein interaction test. The association between SSI and SSIII was also observed by all three methods. Evidence for the SSI/SSIIa and SSI/BEIIb-containing complexes was obtained from both immunoprecipitation and affinity purification. The remaining associations were detected only by one analytical method, and further work is in progress to gather additional support that these proteins are in fact components of multisubunit complexes. In addition to the pairwise association data, GPC analyses directly demonstrated that some population of the SSIIa, SSIII, BEIIa, and BEIIb proteins assemble into one or more high molecular mass complexes. Taken together, this body of evidence demonstrates definitively that maize endosperm SSs and BEs are capable of forming stable multisubunit complexes.

The GPC analyses shown in this study suggest the existence of at least two distinct high molecular mass complexes. SSIIa, BEIIa, and BEIIb all migrate in GPC in the same high molecular mass fractions of approximately 300 kD, in the C300 peak (Fig. 4). The presence of SSI in C300 could not be examined in this study for technical reasons, although given the pairwise association data, there is a strong possibility that SSI will also be found in the same fractions. SSIII is not present in C300, instead appearing only at a higher molecular mass in the C600 peak (Fig. 4). Further purification and direct protein characterization will be required to define the nature of each complex, including whether one or more particular species is present in each of the C300 and C600 peaks. The data are consistent with the possibility that some combinations of SSIIa, SSI, BEIIa, and BEIIb associate to form C300, and this complex is joined by SSIII to form C600. In wheat endosperm, a 260-kD heteromeric complex has been observed in cross-linking studies between SSI, SSII, and either BEIIa or BEIIb (Tetlow et al., 2008), in agreement with the elution profile described here for C300. Also of note is the fact that the relative distribution of each enzyme between its monomer form and high molecular mass form(s) varies. SSIII and SSIIa, for example, are predominantly found in a high molecular mass form as compared to the monomer (Fig. 4, rows 5–7). To the contrary, the great majority of BEIIa is present as a monomer (Fig. 4, row 3). BEIIb is apparently broadly distributed among different quaternary forms (Fig. 4, rows 1 and 2), and BEI is exclusively present as a monomer under these conditions.

Hydrophobic forces are likely to be important for association between components of the apparent complexes. Both the C600 and C300 complexes are stable in a chemical environment that energetically favors hydrophobic interactions (i.e. 1 M NaCl), and, furthermore, these conditions shift the distribution of BEIIa, BEIIb, and SSIIa toward C600 (Fig. 4, rows 2, 4, and 6). This observation does not obviously correspond with the fact that increasing salt concentration was used to break interactions and thus elute proteins from affinity columns. A potential explanation for the apparent anomaly is that both ionic interactions and hydrophobic effects are involved in associations within the complexes. The quaternary structures detected by GPC are preformed in the cell extracts, whereas new complexes with the column-bound ligand must form during affinity purification. Thus, different associations may be detected by the two methods, and ionic bonds and hydrophobic effects may vary in significance in each instance.

Hydrophobic interactions in quaternary structure might also be unexpected for proteins well known to function independently as enzymes, which is the case for all of the proteins examined in this study. Surface hydrophobicity interactions are known for enzymes with the same structural fold as the BEs, specifically the broadly conserved -(14) glucan hydrolase family of -amylases. These enzymes catalyze the same reaction as the BEs, with the distinction that in the latter instance, the cleaved glucan chain is transferred to a C-6 hydroxyl group, whereas the -amylases utilize a free hydroxyl. Association of the glucan chain substrate with -amylases is mediated in part by hydrophobic interactions on the surface of the enzyme, at noncatalytic sites (Bozonnet et al., 2007, and refs. therein). Structural interactions of the same type might be involved in assembly of the BEs, and by extension the SSs, into the complexes present in the C300 and C600 fractions.

BEI stands out from the other five enzymes tested by its appearance exclusively as a monomer in the GPC analysis. One physical interaction involving BEI was detected, specifically its coimmunoprecipitation with SSI (Fig. 8C). Again, different assembly states may be observed by immunoprecipitation and GPC. In analyses of wheat amyloplast proteins, BEI was detected as a component of a high molecular mass complex also containing starch phosphorylase and BEIIb (Tetlow et al., 2004). Further experimental analyses will be required to resolve the protein-binding capabilities of BEI in maize.

Specificity of Interaction

The observed specificity of interactions between the SSs and BEs rules out the trivial explanation that the proteins may simply stick to each other in cell extracts through random or glucan-mediated association. In particular, the data reveal interactions between SSI and both BEIIa and BEIIb; however, no direct association between the two BEs was detected in the immunoprecipitation or affinity purification analyses. Thus, BEIIa and BEIIb from the cell extracts are both competent to join complexes, but they do not do so with each other. Furthermore, a structurally and functionally similar enzyme, BEI, was not observed to be present in any of the complexes demonstrated by GPC (Fig. 4, row 8). This degree of specificity argues against a general adhesion among the proteins as the explanation for the observed associations. This certainly applies for all of the complexes observed with BEIIa or BEIIb and by extension is likely the case for the SSs as well.

Further evidence for specificity in starch biosynthetic enzyme associations is provided from the analysis of wheat endosperm presented in the accompanying article (Tetlow et al., 2008). Associations between SSI, SSIIa, and BEIIa or BEIIb were observed in tissue harvested at mid-developmental stages but not at earlier stages, even though the enzymes are present and glucan synthesis is proceeding. Thus, some aspect of cell physiology affects whether or not the complexes can form, indicative of in vivo functions. Observation of complex formation in yeast nuclei, in particular between SSI and BEIIa or SSI and SSIII, also supports direct interactions between the binding partners, because in this instance there are no other maize factors present. Finally, pretreatment of the wheat amyloplast extracts with amyloglucosidase to remove glucan polymers had no effect on the observed immunoprecipitations (Tetlow et al., 2008), contrary to what would be predicted if the SS and BE enzymes were associated nonspecifically through a common polymeric substrate.

Potential Functions

The long amino terminal extension of SSIII beyond the catalytic domain could serve as a scaffolding protein that brings together multiple components of the starch biosynthetic pathway. Part of this region is highly conserved among the chlorophytes (Ral et al., 2004, and refs. therein), specifically the region referred to as SSIIIHD, and all known plant SSIII isoforms contain a further extension toward the amino terminus beyond this central conserved region (Fig. 2). The data presented here indicate that the amino terminal domain specific to maize endosperm SSIII, as well as the central SSIIIHD region, each bind to different components of the starch biosynthetic pathway (Fig. 2). Specifically, in vivo protein-protein interaction test data indicate that the amino terminal portion of SSIII in the first 181 residues can bind to BEIIa, whereas the SSIIIHD domain contained within residues 760 to 1,438 binds to SSI (Fig. 1; Table I). There may be two distinct BEIIa binding domains, because residues 366 to 648 of SSIII also bind to BEIIa in yeast nuclei. Likewise, there may be two domains of SSIII that bind to SSI, because in addition to the SSI-SSIIIHD interaction, the amino terminal fragment from residues 1 to 367 also interacts with SSI in yeast (Fig. 2). Whether these multiple binding interactions of SSIII are physiologically significant remains to be determined; however, further analyses to investigate functions of the long, conserved amino terminal extension of the SSIII class of enzymes are warranted.

Suggested potential functions for the complexes include substrate channeling and/or effects on the crystallization of the linear chains of amylopectin to form crystalline lamellae. An intriguing consideration about starch biosynthesis is that the products of SSs are the substrates of BEs, and the products of BEs are the substrates of SSs. Thus, as opposed to random stochastic interactions between enzyme and substrate, biosynthesis of starch may occur while the growing polymer interacts simultaneously or near simultaneously with multiple biosynthetic enzymes. This consideration is complicated further by the presence of multiple classes of each enzyme, many of which are required for specific structural aspects of amylopectin. For example, mutations of SSI, SSII, or SSIII all cause specific and nonoverlapping structural defects in amylopectin (Zhang et al., 2004, 2005; Delvalle et al., 2005; Fujita et al., 2007). Association of the enzymes in complexes could provide an environment for ordered construction of the glucan polymer destined to crystallize and form starch granules. One possibility is that the space for crystallization is restricted by the presence of the enzymes in complexes. A second potential function is that organization of the enzymes in complexes could have direct effects on substrate binding and thus specific architectural consequences on amylopectin structure. For example, the position of a branch point might be determined by the spatial locations of an SS and BE active site relative to each other. These two proposed in vivo functions are not mutually exclusive.

Genetic data indicating regulatory interactions between SSs and BEs provide strong evidence to support the hypothesis that starch biosynthetic enzyme complexes are physiologically significant. For example, mutations in SSIII are known to result in increased SSI activity (Singletary et al., 1997; Cao et al., 1999), and SSI and SSIII are shown here to be capable of associating in a multisubunit complex. The simplest explanation for the combined genetic and biochemical observations is that the two proteins function together in vivo in a complex and that this association confers a negative regulatory effect on SSI in wild-type kernels. These genetic indications of in vivo function are indirect, however, and numerous further experiments will be required to test for physiological significance of the observed complexes. Both in vivo crosslinking and in vivo BRET/FRET analyses can determine whether specific starch biosynthetic enzymes are located, in cells, within atomic distances of each other. Purification of the complexes present in the C300 and C600 peaks to homogeneity will be required to define the interactions. Effects of specific interactions on both SS and BE enzymatic activity can also be examined after purification. Further purification could also allow biophysical characterization to examine hydrophobicity effects on complex formation. Eventually, genetic modifications that prevent specific quaternary associations could be useful to further investigate potential physiological functions of any in vivo complex.

	MATERIALS AND METHODS

TOP ABSTRACT RESULTS DISCUSSION MATERIALS AND METHODS LITERATURE CITED

 

Plant Materials

Maize (Zea mays) plants were field grown in the summer at Iowa State University and at Guelph University. Kernels from self-pollinated ears were collected at specified days after pollination, quickly frozen in liquid nitrogen, and stored at –80°C until use. Kernels of the maize inbred lines W64A and Oh43 were used to prepare endosperm whole cell soluble extracts or amyloplast extracts. All mutant alleles were backcrossed into the W64A inbred line. The following mutant alleles were utilized: ae1 (Maize Genetic Stock Center no. 517B), an uncharacterized mutation in the gene coding for BEIIb; su2-19791, an uncharacterized spontaneous mutation known to condition lack of SSII protein (Zhang et al., 2004); sbe1::Mu, a Mutator transposon insertion in the gene coding for BEI (Blauth et al., 2002); sbe2a::Mu, a Mutator transposon insertion in the gene coding for BEIIa (Blauth et al., 2001); ae-B, a 882-bp deletion in the gene coding for BEIIb that removes all of exon 9 (GenBank accession no. AF072725; Fisher et al., 1996; M. Yandeau-Nelson and M. Guiltinan, personal communication); and du1-M3, a Mutator insertion in the first exon of the gene coding for SSIII (M.G. James, unpublished data).

Recombinant Plasmids

Plasmids used to express hybrid proteins containing either the DNA binding domain or the transcriptional activation domain of yeast (Saccharomyces cerevisiae) GAL4, respectively, were constructed from the vectors pGAD-C or pGBD-C (James et al., 1996). In all instances, the GAL4 fragment was located at the amino terminus of the fusion protein. The maize portion of the fusion gene in plasmid pBEIIb-AD or pBEIIb-BD started at codon 62 of the full-length BEIIb open reading frame and terminated at the native stop codon at position 801. The source of the BEIIb cDNA sequence was plasmid pET23d-MBEII (Guan et al., 1994). The maize portion of the fusion gene in plasmids pBEIIa-AD or pBEIIa-BD began at the codon known to correspond to the mature amino terminus of BEIIa protein as it exists in maize amyloplasts (Gao et al., 1997) and extended to the native termination codon following residue 794 of the mature protein. The source of the BEIIa cDNA sequence was a plasmid provided by Dr. M. Guiltinan (Pennsylvania State University). The maize portion of the fusion gene in plasmid pSSI-AD and pSSI-BD started at codon 40 of the full length SSI open reading frame and extended to the native stop codon at position 584. Four different portions of the amino terminus of SSIII were tested, and these amino acid spans are specified in the "Results" section. The source of the SSIII cDNA sequence was plasmid pMG10-6 (Gao et al., 1998). Positive controls for the in vivo protein interaction assay were residues 84 to 708 of the SV40 large T antigen fused to the Gal4p transcriptional activation domain, and residues 72 to 390 of murine p53 fused to the Gal4p DNA binding domain (Li and Fields, 1993). The vectors used to express the positive control proteins were pAD-GAL4-2.1 for the T antigen fragment and pBD-GAL4 for the p53 fragment (Stratagene).

Plasmid pHC16 was used for recombinant expression of maize BEIIa in Escherichia coli. pHC16 was built in the expression vector pET-29a(+) (Novagen) and contains the complete coding sequence of mature BEIIa, specifically amino acids 21 to 847 as specified by the available cDNA sequence (GenBank accession no. U65948). Amino acid 21 is known to be the mature amino acid terminus of BEIIa after cleavage of the targeting peptide during the import into the plastid (Fisher et al., 1993). The BEIIa coding region is fused at the N terminus to the 15-amino acid S-tag coding sequence, and the gene fusion is expressed from the bacteriophage T7 promoter.

Plasmid pHC18 was used for recombinant expression of maize SSI in E. coli. Also built in pET-29a(+), this plasmid contains the complete coding sequence of mature SSI, specifically amino acids 39 to 670 according to the full-length cDNA sequence (GenBank accession no. AAB99957). Amino acid 39 was directly identified as the mature N terminus by amino acid sequence of the purified protein (Knight et al., 1998). The recombinant SSI protein is fused to the S-tag sequence at its amino terminus.

Plasmid pTB-3829 was used to express SSIII residues 770 to 1,225 fused to glutathione S-transferase (GST) at the amino terminus. A fragment of the SSIII cDNA sequence was produced by PCR from plasmid pMG10-6 (Gao et al., 1998). The downstream oligonucleotide primer included a stop codon following SSIII codon 1,225, and the upstream primer included a six-amino acid coding sequence providing a thrombin cleavage site immediately upstream of SSIII codon 770. The primers also contained the sequences required for in vitro recombination into plasmid vector pDONR221, using BP clonase of the Gateway cloning system (Invitrogen). This recombination generated the plasmid pTB-3828, which was confirmed by restriction mapping and determining the DNA sequence of the complete SSIII coding region as well as parts of the vector. In the second step, a segment of pTB-3828 was inserted into pDEST15 (Invitrogen) using the LR clonase of the Gateway system to produce the plasmid pTB-3829. The relevant expression construct within pTB-3829 contains the following elements: (1) the phage T7 promoter, inducible with isopropylthio-β-galactoside (IPTG); (2) a ribosome binding site sequence and a start codon; (3) a coding region for 223 amino acids of Schistosoma japonicum GST; (4) a linker sequence from the acceptor plasmid; (5) codons specifying the designed thrombin cleavage site; (6) SSIII codons 770 to 1,225; (7) a stop codon; and (8) a transcriptional termination sequence.

In Vivo Protein-Protein Interaction Tests

Standard methods and growth media were used for transformation and culturing of yeast (Ausubel et al., 1989). The yeast strain PJ69-4A was used as the reporter for in vivo protein-protein interactions (James et al., 1996). The genotype of PJ69-4A is MATa ura3 his3 leu2 trp1 ade2 gal4 gal80 GAL2-ADE2 met2::GAL7-lacZ. Pairs of recombinant plasmids built in either the pGAD-C or pGBD-C vectors (James et al., 1996) were cotransformed into PJ69-4A. Multiple independent transformants were selected that contained both the LEU2 and TRP1 markers of the two plasmids, respectively, and these clones were single colony purified. Cells were maintained on selective minimal medium lacking Leu and Trp, and then spread very lightly onto selective minimal test medium lacking Ade, Leu, and Trp. Growth of the colonies on the –Ade medium indicated the two proteins being tested interact in the yeast nucleus to regenerate a GAL4 activity. Qualitative assay of β-galactosidase activity in permeabilized yeast colonies on filter lifts was as described (Ausubel et al., 1989).

Recombinant Protein Expression and Purification

Recombinant BEIIa and SSI were expressed in E. coli and purified as follows. Host strain BL21DE3 transformed with expression plasmid pHC16 or pHC18 was grown in Luria-Bertani medium containing 30 µg/mL kanamycin to mid log phase in 400-mL cultures in 2-L flasks. Expression was induced by the addition of 1 mM IPTG, and cultures were shaken at room temperature for an additional 16 to 20 h. Cell pellets were washed and suspended in 10 mL sonication buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 5 mM dithiothreitol [DTT], 0.1% Triton X-100, 1 mM phenylmethylsulfonyl fluoride [PMSF]) and lysed by treatment with 100 µg/mL lysozyme followed by sonication. Lysates were centrifuged at 20,000 rpm for 20 min and the supernatant was collected. Recombinant SSI or BEIIa in the lysates was bound to S-protein-agarose through the high affinity S-tag:S-protein interaction according to the manufacturer's protocol (Novagen catalog no. 69704-4). To purify BEIIa for use as antigen, biotinylated thrombin was used to cleave the fusion protein and release it from the matrix, and streptavidin-agarose was used to remove the protease, as described by the manufacturer (Novagen catalog no. 69022-3).

Recombinant SSIIIHD was expressed in E. coli and purified as follows. pTB-3829 was transformed into host strain RosettaDE3pLysS (Novagen). Expression cultures in Luria-Bertani medium supplemented with 50 µg/mL carbenicillin (Sigma; 1 L in 2-L flasks) were inoculated from overnight precultures at a dilution of 1/100 and shaken vigorously for 2 h at 37°C. IPTG was then added to a final concentration of 0.5 mM, and the cultures were grown for an additional 4 h at 30°C. Cells were collected by centrifugation, washed in water, suspended in 20 mL of binding/wash buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.1% Triton X-100, 5 mM DTT) supplemented before use with 1 mM PMSF and 0.1 mg/mL lysozyme (Sigma), then incubated at 30°C for 15 min. The cell suspension was sonicated on ice with six bursts of 20 s each, then centrifuged at 17,000 rpm at 4°C for 20 min. The supernatant was filtered through a 0.45-µm nitrocellulose filter, and additional DTT was added to make the final concentration 10 mM. The GST-SSIIIHD fusion protein was purified from this lysate using glutathione-Sepharose 4B (GE Healthcare) according to the manufacturer's protocol. Th