Differential Characteristics and Subcellular Localization of Two Starch-branching Enzyme Isoforms Encoded by a Single Gene inPhaseolus vulgaris L.*

Starch-branching enzymes (SBE) have a dominant role for amylopectin structure as they define chain length and frequency of branch points. We have previously shown that one of the SBE isoforms of kidney bean (Phaseolus vulgaris L.), designated PvSBE2, has a molecular mass (82 kDa) significantly smaller than those reported for isologous SBEs from pea (SBEI), maize (BEIIb), and rice (RBE3). Additionally, in contrast to the dual location of the pea SBEI in both the soluble and starch granule fractions, PvSBE2 was found only in the soluble fraction during seed development. Analysis of a pvsbe2 cDNA suggested that PvSBE2 is generated from a larger precursor with a putative plastid targeting sequence of 156 residues. Here we describe the occurrence of a larger 100-kDa form (LF-PvSBE2) of PvSBE2 found both in the soluble and starch granule fractions of the developing seeds. The determined N-terminal sequence, VKSSHDSD, of LF-PvSBE2 corresponded to a peptide sequence located 111 amino acids upstream from the N terminus of purified PvSBE2, suggesting that LF-PvSBE2 and PvSBE2 are products of the same gene. Analysis of the products by 5′-RACE (rapid amplification of cDNA ends) and reverse transcription PCR indicated that the two transcripts for pre-LF-PvSBE2 and pre-PvSBE2 are generated by alternative splicing. Recombinant LF-PvSBE2 (rLF-PvSBE2) was purified from Escherichia coli and the kinetic properties were compared with those of recombinant PvSBE2 (rPvSBE2). rLF-PvSBE2 had much higher affinity for amylopectin (K m = 4.4 mg/ml) than rPvSBE2 (18.4 mg/ml), whereas theV max of rLF-PvSBE2 (135 units/mg) for this substrate was much lower than that of rPvSBE2 (561 units/mg). These results suggest that the N-terminal extension of LF-PvSBE2 plays a critical role for localization in starch granules by altering its enzymatic properties.

Starch-branching enzyme (SBE) 1 (1,4-␣-D-glucan: 1,4-␣-Dglucan 6-␣-D-(1,4-␣-D-glucan)-transferase, EC 2.4.1.18) catalyzes the cleavage of ␣-1,4-linkages and the subsequent transfer of ␣-1,4 glucan to form an ␣-1,6 branch point in amylopectin (3). SBE is a member of the ␣-amylase family of enzymes, characterized by four highly conserved regions and a central (␤/␣) 8 barrel domain (4). Apart from the barrel domain, SBEs show considerable structural variation in the length and amino acid sequences at the N-and C-terminal regions. Multiple SBE isozymes have been found in individual plant species and are encoded by two gene families (families A and B) based on the primary sequences. Members of the two families display distinct enzymatic properties, presumably because of the differences in N-and C-terminal regions. Several studies have shown that the N-terminal region is important for specificity of transferred chain length and is required for maximum enzyme activity (5,6), whereas the C-terminal region is involved in substrate specificity (5,7).
Starch synthases (SS), except granule-bound starch synthase I (GBSSI), exist as both soluble and starch granulebound forms (1,8,9). Likewise, SBEs from maize (10,11), pea (8), and wheat (12)(13)(14) also occur in both soluble and granule fractions. However, the starch-bound SBEs and SSs are not considered active in starch biosynthesis. Their association with the starch granule has been suggested to be due simply to their being trapped within the starch granule during its growth and maturation (8,11). Recently a novel granule-bound SBE has been isolated from wheat endosperm (15). This SBE is much larger (152 kDa) than previously characterized family A-type SBEs and contains not only the intact SBEI primary sequence but also an extra N-terminal domain formed by partial duplication of the SBEI sequence located near the N terminus. This N-terminal domain is likely responsible for binding of this SBE isoform to the starch granule.
To investigate the relationships between structure and function of SBEs, as well as its mode of action in starch biosynthe-sis, we isolated and characterized a SBE isozyme (PvSBE2) from immature seeds of the kidney bean (Phaseolus vulgaris L.) (16). A comparison of the determined N-terminal sequence of purified PvSBE2 to the primary sequence derived from pvsbe2 cDNA suggested that PvSBE2 was produced from a preprotein with a plastid signal leader sequence of 156 residues (16,17). Additionally, comparison of the primary sequence of PvSBE2 to other plant SBEs such as pea SBEI (18,19), maize BEIIb (20,21), and rice RBE3 (22) showed that despite sharing considerable sequence homology with these other SBEs, PvSBE2 is shorter by about 50 -100 amino acid residues (Fig. 1,  A and B). Moreover, unlike the pea SBEI, which is observed in both the soluble and starch granule fractions (8), PvSBE2 protein is present only in the soluble fractions during seed development (17). Based on these observations, we suspected the existence of a larger form of PvSBE2 for which the molecular mass would be closer to pea SBEI. Here, we describe the occurrence of a larger form (LF-PvSBE2) of PvSBE2 containing an extended N terminus of 111 residues. This larger enzyme form is observed in both the soluble and starch granule fractions of kidney bean developing seeds. Immunoblot and molecular analyses suggest that there are two possible mechanisms for generating LF-PvSBE2 and PvSBE2: a post-translational modification and alternative splicing. In addition to their different subcellular locations, LF-PvSBE2 and PvSBE2 have distinct enzymatic properties. Based on the results obtained in this study, the relationship between the enzymatic properties and subcellular locations of both isoforms is discussed.

EXPERIMENTAL PROCEDURES
Plant Materials-Kidney bean (P. vulgaris L. cv. toramame) seeds were harvested from plants grown at an experimental field of Hokkaido University in Sapporo, Japan. Seeds were isolated at different stages of development and separated as small-(4 -8 mm), mid-(10 -12 mm), and FIG. 1. N-terminal primary sequences of several SBEs and identification of the large form of PvSBE2. A, alignment of mature N-terminal primary sequences of kidney bean PvSBE2 (16,17), pea SBEI (19), rice RBE3 (22), and maize BEIIb (21). Regions I and II represent two of the four regions conserved among ␣-amylase family enzymes. B, sequence alignment of N-terminal regions of preproteins deduced from cDNA sequences of kidney bean pvsbe2 (Pv (17)) and pea sbeI (Pea (19)). Identical amino acids are indicated as white letters on a black background. Arrows indicate the Nterminal sites determined by protein sequencing. C, immunoblot analysis of different fraction of developing kidney bean seeds. Proteins from soluble and granule fractions of kidney bean seeds (mid-size) were subjected to 8% SDS-PAGE and blotted. Blots were developed with anti-rPvSBE2. Lane Pv2, purified rPvSBE2 protein (0.35 g) as a control; lanes SF and GF, soluble proteins (250 g) and proteins from 5 mg of starch, respectively. D, analysis of N-terminal sequence for an ϳ100-kDa granule-bound polypeptide. Starch granule-bound proteins were prepared from about 50 mg of starch of midsize seeds, subjected to 8% SDS-PAGE, and stained with Coomassie Brilliant Blue. The 100-kDa polypeptide band (arrow) was blotted onto a polyvinylidene difluoride membrane and then sequenced. The arrowhead indicates the 59-kDa polypeptide corresponding to granulebound starch synthase I (GBSSI).
large-size (14 -16 mm) developing seeds and mature seeds; they were stored at Ϫ20°C until use.
Extraction of Soluble and Granule-bound Proteins from Kidney Bean-Starch granules were extracted from developing seeds of kidney bean essentially according to the procedure described by Smith (23) with the exception that starch granules were allowed to sediment under gravity at room temperature for 1 h instead of centrifugation. The supernatant obtained by centrifugation (13,000 ϫ g for 30 min at 4°C) of the kidney bean homogenates was used as a soluble fraction. To identify the occurrence of LF-PvSBE2 in developing seeds and to determine the N-terminal amino acid sequence of LF-PvSBE2, starch granule-bound proteins were prepared from isolated starch granules of mid-size developing seeds according to the procedure described by Båga et al. (15). For the semi-quantification and distribution of each polypeptide by immunoblot analysis, starch granule-bound proteins were obtained by dissolution of the isolated starch granules in dimethyl sulfoxide and subsequent dilution with deionized water.
Protein Analyses-The N-terminal amino acid sequence of LF-PvSBE2 was determined by a protein sequencer after the peptide was transferred onto a polyvinylidene difluoride membrane from SDS-PAGE (24). Immunoblot analysis was performed with rPvSBE2 antiserum as described previously (17).
Isolation of the Partial pvsbe2 Gene Fragment and Nucleotide Sequencing-Genomic DNA was isolated from the expanded leaves of kidney bean plants by the method of Murray and Thompson (28). The DNA was partially digested with Sau3AI and fragments (10 -20 kb in size) were inserted into the BamHI site of EMBL3 (Stratagene). The recombinant DNAs were packaged into bacteriophage particles (GIGA pack Gold-III; Stratagene) and grown on Escherichia coli LE392 (P2). The genomic library consisting of 3.6 ϫ 10 5 was screened with a digoxigenin-labeled probe using a digoxigenin DNA labeling kit (Roche Molecular Biochemicals). The probe (about 1.4 kb) was prepared by PCR using genomic DNA as a template and specific primers within the second and third exons of the pvsbe2 gene (5Ј-CCT TGC TGT AAA ATC TTC TCA TGA TTC TG-3Ј and 5Ј-CAT AAC CTA GAG GAT CTA ACC ATG GAA G-3Ј), based on the cDNA sequence. The methods used for hybridization and detection of the signals were as described previously (17). The SalI/BglII genomic fragment containing the partial promoter region and the third intron was isolated and sequenced. DNA was sequenced by the dideoxy chain termination method (29) using a Thermo Sequenase Cycle Sequencing kit (United States Biochemical Corp.).
5Ј Rapid Amplification of cDNA Ends (5Ј-RACE) and Reverse Transcription Polymerase Chain Reaction (RT-PCR)-Total RNA was isolated from small-, mid-, and large-size developing seeds as well as mature seeds of the kidney bean as described previously (17). 5Ј-RACE was done using total RNA from mid-size developing seeds with a 5Ј-RACE System for Rapid Amplification of cDNA Ends kit (Version 2.0; Invitrogen) according to the manufacturer's protocol. The first strand synthesis was primed with the specific primer to pvsbe2 cDNA (945ANTI, 5Ј-AAT TGG TGG TGA ACC ATC CAC ATT GTT TG-3Ј). Terminal deoxynucleotidyl transferase was used to add homopolymeric (C) tails to the 3Ј-end of the first strand cDNAs. For amplification of target cDNA, the first round PCR was performed with the dC-tailed cDNAs as templates using the Abridged Anchor Primer (5Ј-GGC CAC GCG TCG ACT AGT ACG GGI IGG GII GGG IIG-3Ј, where I is inosine) and the 945ANTI primer. These amplified products were then subjected to a second round of PCR with the Abridged Universal Amplification Primer (5Ј-GGC CAC GCG TCG ACT AGT AC-3Ј) and the nested specific primer (745ANTI, 5Ј-ATG CAT CCA GGC CAC CTT CAT GCT TG-3Ј). The amplified DNA was separated on a 2% agarose gel, purified, subcloned into pT7Blue TA-vector (Novagen), and then sequenced.
RT-PCR was performed using an RNA LA-PCR kit (Takara Shuzo, Kyoto, Japan) according to the manufacturer's protocol. The RT products were prepared from total RNA of small-, mid-, and large-size developing seeds and mature seeds using an oligo-dT adaptor primer and avian myeloblast virus reverse transcriptase. The RT products for pvsbe2 transcripts were amplified using the specific primers (231 SEN, 5Ј-TCT CAG AAA GAA CAA CTT CTC TC-3Ј and 745ANTI). RT-PCR for kidney bean actin transcripts was also done using the specific primers (KBACT-N, 5Ј-GGA CGA GGC TCA ATC GAA GA-3Ј; and KBACT-C, 5Ј-ACT GAC ACC GTC TCC GGA GT-3Ј).
Production of Recombinant Large Form of PvSBE2 (rLF-PvSBE2) in E. coli-Plasmids for the expression of LF-PvSBE2 protein in E. coli cells were constructed as follows. A 1,092-bp cDNA fragment was amplified by PCR with LF-SEN (5Ј-CCC CAT GGT AAA ATC TTC TCA TGA TTC TG-3Ј) and SBE2-A (5Ј-GAC TTA GGA TCC TCT GGG GTT CC-3Ј) primers, using the pvsbe2 cDNA cloned into pBluescript II SK (pBS-PvSbe2; 17) as a template. The amplified fragment was digested with NcoI/KpnI and then cloned together with the KpnI/SacI fragment from pBS-PvSbe2 into the NcoI/SacI sites of pET23d(ϩ) (Novagen, Inc.). The resulting plasmid, designated pET-LFPvSbe2, was introduced into E. coli BL21(DE3) strain (Novagen). Crude extract was obtained according to the previously described method (17) from cell cultures grown at 25°C.
Purification of rLF-PvSBE2-After dialysis against buffer A (20 mM Tris-HCl buffer (pH 7.5) containing 1 mM dithiothreitol and 1 mM EDTA), the crude extract was applied onto a DEAE-Sepharose CL-6B column (2.0 ϫ 11.5 cm). The column was washed with buffer A, and then the enzyme was eluted using a 0-0.4 M NaCl gradient. Enzyme active fractions were dialyzed against buffer B (20 mM Tris-HCl buffer (pH 6.8) containing 1 mM dithiothreitol and 1 mM EDTA) and then chromatographed on a DEAE-Sepharose CL-6B column (1.3 ϫ 10 cm) using a 0 -0.3 M NaCl gradient. Active fractions were pooled, concentrated, and fractionated on a Bio-Gel P-200 column (2.9 ϫ 71 cm) equilibrated with buffer B containing 0.1 M NaCl. The active fractions were dialyzed against 5 mM sodium phosphate buffer (pH 8.0) containing 1 mM dithiothreitol and 1 mM EDTA and applied to a hydroxyapatite (Seikagaku Corp, Tokyo) column (1.3 ϫ 8.0 cm). The enzyme solution was eluted with a 5-200 mM linear gradient in sodium phosphate buffer.
Assay of SBE Activity-SBE activity was determined by the following two methods. The iodine staining assay (Assay I) was performed ac- cording to the method of Boyer and Preiss (20), which monitors the decrease in absorbance at 660 nm for amylose (potato amylose type III, Sigma) or at 540 nm for amylopectin (potato amylopectin, Sigma) as substrates. One unit of enzyme activity was defined as the amount that decreased the absorbance of 0.1 at 660 or 540 nm/min at 30°C.
The branching linkage assay (Assay BL) was done by the method of Takeda et al. (25) using reduced amylose as a substrate. One unit of the enzyme activity was defined as 1 mol of branch linkages formed/min at 30°C.
Analysis of Enzymatic Properties-pH stability, optimum tempera-

FIG. 3. Analysis of the 5-RACE and RT-RCR products and possible alternative splicing for the expression of the pvsbe2 gene.
A, the 5Ј-RACE products (lane P) were subjected to 2% agarose gel electrophoresis. Lane Std, DNA marker containing the 1-kb and 100-bp ladders (Invitrogen). The arrows with "(I)" and "(II)" indicate the amplified bands. B, comparison of the nucleotide and deduced amino acid sequences between the 5Ј-RACE products I and II. The deduced amino acids from the I and II fragments are shown above and below, respectively, each nucleotide sequences. The arrowheads are the inserted positions of introns. The white letters in black circles correspond to the N termini of mature LF-PvSBE2 (Val) and PvSBE2 (Ile). The lines between sequences I and II represent the specific primers used for RT-PCR (see also legend for panel D). C, schematic illustration of the generation of two mRNAs from the pvsbe2 gene. Each exon is shown by boxes. The open boxes indicate 5Ј-untranslated regions, and black and gray boxes represent regions translated into putative transit peptides and mature proteins, respectively. Two transcripts with different 5Ј-ends could be produced by alternative splicing, thereby generating two preproteins with different N-terminal structure. D, analysis of RT-PCR products for pvsbe2 and actin fragments during seed development. The RT products from each developing stage (S, M, L, or Ma) were used as templates. The products were separated on a 1.5% agarose gel. An actin fragment was amplified and used as a control (lower panel). The arrows preceded by "(I)" and "(II)" in the upper panel correspond to the amplified DNAs with the reasonable sizes predicted from sequences I and II in panel B, respectively. The middle panel represents the comparative levels estimated from each band strength by NIH Image software as the ratios to the estimated level of amplified fragment (I) in small seeds. E, the relative levels of total transcripts and total protein accumulation estimated by NIH Image software (Figs. 2B and 3D). The closed and open circles indicate total RNA expression by RT-PCR and total protein accumulation profiles by immunoblot, respectively. The gray triangles show the RNA expression profile by Northern analysis (17) and subsequent estimation from the band-strength by NIH Image software. In the each profiles, the maximum level is defined as 1.
ture, thermal stability, citrate effects, and kinetic parameters were examined as described previously (16).
Absorption on a Raw Starch-Mixtures (500 l) of rPvSBE2 or rLF-PvSBE2 containing 50 mg of raw potato starch (Sigma) were incubated at 25°C for 15 min and then briefly centrifuged (10,000 ϫ g for 30 s). The amount of SBE activity remaining in the supernatant was then measured by Assay I as described above. Rhizopus sp. glucoamylase (Nagase Biochemicals, Osaka, Japan) was used as a positive control of the enzyme containing a raw starch-binding domain. Glucoamylase activity was assayed by determining the amount of glucose liberated from a substrate (potato soluble starch) by the modified Tris-glucose oxidase-peroxidase method (26) using a Glu ARII kit (Wako Pure Chemical, Ltd., Osaka, Japan) as described previously (26).
Analysis of Chain Transfer Patterns-Chain transfer patterns of rPvSBE2 and rLF-PvSBE2 were determined by a modified method described by Hanasiro et al. (27). Reduced amylose was incubated in 500 l of 25 mM MOPS (pH 7.5) at 30°C with 1 milliunit (determined by Assay BL) of rPvSBE2 or 2-10 milliunits of rLF-PvSBE2. After being incubated for 4 h, the samples were boiled for 1.5 min to inactivate the enzyme activity and then debranched by incubation with isoamylase (0.3 units) in 0.1 M sodium acetate buffer (pH 3.5) for 2.5 h at 45°C. The debranching reaction was terminated when the samples were boiled for 3 min. The reaction mixtures were passed through a 0.2-m filter and then analyzed by HPAEC-PAD.

RESULTS
Occurrence of an Alternative Form for PvSBE2-Immunoblot analysis of the soluble and starch granule fractions prepared from developing kidney bean seeds (mid-size) showed that anti-rPvSBE2 reacted with the 82-kDa PvSBE2 located in the soluble fraction (Fig. 1C, lane SF). PvSBE2 was not detected in the starch granule fraction, although a 100-kDa protein associated with the starch granule fraction was readily evident (Fig. 1C, lane GF). To determine the nature of the 100-kDa starch granule-associated protein, its N-terminal sequence was determined by Edman degradation (Fig. 1D). The generated amino acid sequence, VKSSHDSD, of the 100-kDa polypeptide aligns with the primary sequence deduced from the pvsbe2  at 280 nm is 10. b These activities were measured by Assay I method using amylose as a substrate.

TABLE II
The abilities of rLF-PvSBE2, rPvSBE2, and glucoamylase for adsorption to a raw starch After mixture of each enzyme preparation and raw potato starch was incubated at 25°C for 15 min, enzyme activity remaining in the supernatant was measured. SBE activity was measured by Assay I using amylose as a substrate. Glucoamylase activity was measured by the modified Tris-glucose oxidase-peroxidase method using soluble starch as a substrate. Results are the average of at least three experiments and are given Ϯ the standard deviations. cDNA (17) beginning at residue position 46. This peptide sequence was 111 residues upstream of the mature N terminus of purified PvSBE2 (Fig. 1B). These results suggest that the 100-kDa starch granule-associated polypeptide in mid-size seeds is a larger isoform of PvSBE2 (designated LF-PvSBE2).

Distribution and Amounts of LF-PvSBE2 and PvSBE2 during Seed Development-The spatial localization of LF-PvSBE2
and PvSBE2 during the development of kidney bean seeds was examined by immunoblot analysis (Fig. 2A). PvSBE2 was detected only in soluble fractions from small-, mid-, and large-size developing seeds but not in mature seeds. In contrast, LF-PvSBE2 was found mostly in the starch granule and soluble fractions at the latter stages of seed development when the rates of starch biosynthesis are at their maximal levels and then begin to slow down. The relative amounts of each SBE2 isoform were estimated from the intensities of known amounts of loaded rPvSBE2 protein (Fig. 2B). The levels of PvSBE2 protein peak at the mid-stage of seed development and then drop off at later stages when starch levels begin to increase rapidly. In contrast, the levels of LF-PvSBE2 protein parallel the increase in starch levels (17). They are relatively low in mid-developing seeds and then increase dramatically at the later stage when starch levels also increase. More than 75% of the total SBE2 protein is contributed by LF-PvSBE2 protein, whereas PvSBE2 comprises less than 25%. About half of the amount of LF-PvSBE2 was located in the starch granule fractions at the later stages of seed development. These results indicate that kidney bean seeds contain two SBE2 forms: a smaller PvSBE2 isoform, which attains maximum levels when starch levels begin to increase, and a larger LF-PvSBE2 isoform in which the temporal accumulation pattern coincides with starch synthesis.
Generation of Two Transcripts by Alternative Splicing-To determine whether LF-PvSBE2 and PvSBE2 were differently spliced products of the same gene, 5Ј-RACE was performed using total RNA from mid-size developing seeds of kidney bean (Fig. 3A). Two distinct bands of about 0.75 and 0.55 kb (denoted as I and II, respectively, in Fig. 3A) were detected in the 5Ј-RACE products, which were sequenced together with a partial pvsbe2 gene fragment corresponding to the 5Ј coding sequences. The nucleotide sequence alignment between the two amplified cDNAs and the partial gene fragment demonstrated that both amplified cDNAs were derived from the same pvsbe2 gene and that the two transcripts were generated by alternative splicing (Fig. 3, B and C). The sequence of the larger RNA transcript I was identical to the previously determined pvsbe2 cDNA. In contrast, the sequence of the smaller transcript II was composed of truncated exons 1 (exon 1Ј) and 2 (exon 2Ј) as well as intact exon 3 at the 5Ј-end, suggesting that the translation commences with the Met residue at position 99 of the derived pvsbe2 primary sequence. Overall, pre-LF-PvSBE2 and PvSBE2 are coded by two distinct RNA transcripts formed by alternative splicing of introns during transcription.
To access the relative levels of the two prsbe2 transcripts during kidney bean seed development, RT-PCR was performed using RNA isolated from seeds at different stages of development (Fig. 3D). The amplified DNA levels derived from actin RNA were constant during seed development and used as an internal control (Fig. 3D, lower panel). Two amplified bands corresponding to transcripts I and II were evident at all stages of seed development although their temporal accumulation patterns were distinct. Transcript I peaked during the midstage of seed development, whereas transcript II showed high levels at the mid-and late-stages of seed development (Fig. 3D,  upper and middle panels). The temporal RNA accumulation patterns displays by transcript I (pre-LF-PvSBE2) and II (pre-PvSBE2) were different from those observed for the SBE2 isoforms (Fig. 3E), suggesting that the regulation of the two SBE2 isoforms occurs post-translationally.
Preparation of Recombinant LF-PvSBE2-rLF-PvSBE2 enzyme was produced in the E. coli cells and purified by four successive column chromatography steps (Table I) in order to investigate its enzymatic properties. The final enzyme preparation from 1.2 liters of cultured cells contained 9.0 mg of rLF-PvSBE2 with a specific activity of 170 units/mg using amylose as a substrate. The specific activity was about 80% of that (214 units/mg) measured for purified rPvSBE2. The specific activity of purified rPvSBE2 under the same assay conditions was nearly identical (223 units/mg) to that of the native PvSBE2 purified from immature seeds (16). Purified rLF-PvSBE2 migrated as a single polypeptide band with the same mobility as the immunoreacted protein from starch granule fractions on SDS-PAGE (Fig. 4, A and B).
The estimated molecular mass (100 kDa) of purified rLF-PvSBE2 obtained by SDS-PAGE was significantly larger than the 93.8 kDa predicted from the deduced primary sequence. Such electrophoretic behavior on SDS-PAGE has been also observed for the pea SBEs and for another kidney bean SBE protein (17,19). Unlike rLF-PvSBE2, the estimated molecular mass (82 kDa) of rPvSBE2 agrees well with that predicted from the primary sequence. The anomalous electrophoretic behavior of the pea SBEs has been suggested to result from the high content of negatively charge residues. Indeed, about 21% of the extra 111 amino acids of rLF-PvSBE2 consist of negative charged residues (14 Asp and 9 Glu). These observations suggest that the additional N-terminal domain of rLF-PvSBE2 is responsible for its higher than expected mobility on SDS-polyacrylamide gels than if based on molecular size alone.
Starch Binding Ability of rLF-PvSBE2-Because the subcellular distribution profiles of LF-PvSBE2 and PvSBE2 suggested that the N-terminal region of LF-PvSBE2 participates in the interaction with starch granules, we examined whether this protein had the capacity to bind to raw starch (Table II). This was accomplished by incubating rLF-PvSBE2 (as well as rPvSBE2 to serve as a control) with raw starch and then assaying for the amount of SBE activity remaining in solution after removal of the starch granules by centrifugation. Such assays showed no significant amount of binding by either SBE form. In contrast, glucoamylase, which is capable of binding directly to raw starch (30,31), showed a 59% decrease in enzyme activity in the soluble fraction under the same conditions. These results suggest that rLF-PvSBE2 is unable to directly bind to raw starch in vitro.
Enzymatic Properties of rLF-PvSBE2-To determine the effect of the N-terminal extension in rLF-PvSBE2 on enzyme function, the physical properties of purified rLF-PvSBE2 were measured and compared with those of rPvSBE2 (Figs. 5 and 6). rLF-PvSBE2 was fairly stable between pH 6.5 and 9.5, as more than 90% of the original activity remained after incubation at 4°C for 16 h. The pH stability range was a little broader than that for rPvSBE2 (pH 7.0 -9.0). rLF-PvSBE2 and rPvSBE2 were most active at around 30°C. Nevertheless, almost no inactivation occurred when the enzymes were kept at pH 8.0 for 15 min at up to 45°C for rLF-PvSBE2 as compared with a 50°C level for rPvSBE2. The inconsistent data between the optimum temperature and thermal stability have been interpreted as the temperature-dependent stability of a secondary and tertiary structure of amylose as a substrate (3,32); namely, the structure of amylose, probably a helical structure, that can be utilized by SBEs is stable at a lower temperature. Both enzymes were noncompetitively activated by citrate (data not shown). In the presence of 0.3 M citrate, rPvSBE2 activity increased 3.3-fold, whereas rLF-PvSBE2 was stimulated 3.7fold. These results suggest that the extended N-terminal region has only a moderate effect on the thermal stability and citratestimulating properties of rLF-PvSBE2 enzyme (Fig. 5).
The chain-length distribution pattern of debranched products was analyzed by HPAEC-PAD after incubation of rLF-PvSBE2 with reduced amylose followed by isoamylase treatment to generate branched chains (17). The final chain-length distribution pattern generated by rLF-PvSBE2 was very similar to that generated by rPvSBE2 (Fig. 6A). Both enzymes preferentially transferred short chains similar to those observed for maize BEII and rice RBE3 of family A SBEs. However, unlike rPvSBE2, where only 1 milliunit of enzyme activity was needed to produce sufficient product for chain-length distribution analysis, substantially higher amounts of rLF-PvSBE2 were required to obtain sufficient product for HPAEC-PAD analysis. Close examination of the reaction kinetics of the rLF-PvSBE2 over extended incubation periods showed considerable differences from that displayed by rPvSBE2. Unlike the near linear kinetics displayed by rPvSBE2, the rLF-PvSBE2mediated reaction was linear for only 20 -30 min before the activity began to slow down (Fig. 6B). As mock reactions in the absence of substrate indicated that rLF-PvSBE2 was stable when incubated at 30°C for 4 h (data not shown), it is likely that rLF-PvSBE2 is unable to act on a substrate that becomes increasing amylopectin-like over time. This is supported by the kinetic properties of rLF-PvSBE2 for amylopectin, as described the next section.
Affinity of rLF-PvSBE2 to Amylose and Amylopectin-The kinetic parameters of rLF-PvSBE2 were analyzed with amylose or amylopectin as a substrate and compared with those obtained for rPvSBE2 (Table III). Lineweaver-Burk plots of the reaction catalyzed by rLF-PvSBE2 and rPvSBE2 showed Michaelis constants (K m ) for amylose to be 4.80 and 1.27 mg/ ml, respectively, and the maximum velocities (V max ) to be 396 and 242 units/mg protein, respectively. In contrast, the K m and V max values for amylopectin were calculated to be 4.4 mg/ml and 135 units/mg protein for rLF-PvSBE2 and 18.4 mg/ml and 561 units/mg protein for rPvSBE2. The K m and V max values for amylopectin by rLF-PvSBE2 are about one-quarter of those of rPvSBE2. This indicates that the Lineweaver-Burk plots of the reactions catalyzed by rLF-PvSBE2 and rPvSBE2 are parallel. An alternative way of looking at these enzymes is that the rLF-PvSBE2 reaction is an uncompetitive type of that catalyzed by rPvSBE2. Hence, the nonlinear kinetics exhibited by rLF-PvSBE2 reaction is due to uncompetitive product inhibition. The V max /K m value for amylopectin is almost identical for rLF-PvSBE2 and rPvSBE2, indicating that both enzymes can act comparably on amylopectin or amylopectin-like glucans at a low concentration of the substrate. However, since it is reasonable to consider that the concentration of amylopectin in developing seeds is far higher than that in our assay, both isoforms with different kinetic parameters contribute to in vivo amylopectin biosynthesis. DISCUSSION SBE isoforms have a dominant role in determining the structure of amylopectin. Our previous data have shown that PvSBE2, one of the SBE isozymes detected in kidney bean seeds, is classified as a family A type on the basis of enzymatic properties and the predicted primary sequence (16,17). This PvSBE2 protein was observed only in the soluble fraction during seed development (17). The data presented in this study show that a larger form of PvSBE2, LF-PvSBE2, exists that contains an extended N-terminal region. Unlike the soluble location of PvSBE2, LF-PvSBE2 is observed in both the soluble and starch granule fractions. Our studies also show that the extended N-terminal region in LF-PvSBE2 alters not only its subcellular location but also its kinetic properties as well. The two isoforms are encoded by the same gene, which produces two distinct transcripts generated by alternative splicing of the first two exons.
Although the pvsbe2 cDNA encodes an open reading frame similar in length to the pea sbeI cDNA, and the deduced amino acid sequences share a high homology each other, purified PvSBE2 from immature seeds of the kidney bean was significantly shorter than pea SBEI (8,16). Additionally, PvSBE2 polypeptide is found only in the soluble fraction in developing seed homogenates (17), whereas pea SBEI is found in both soluble and starch granule fractions (8). This close sequence homology between the pea and kidney bean SBE isoforms, but with differences in the molecular size and the subcellular distribution of PvSBE2 and pea SBEI, triggered a search for a larger form of PvSBE2 in developing kidney bean seeds. Indeed, a larger polypeptide, which cross-reacted with anti-rPvSBE2, was found in the starch granule fractions of kidney bean developing seeds (Fig. 1C). The determined N-terminal sequence of the polypeptide suggested that the larger PvSBE2 form is also encoded by the pvsbe2 gene, which is present as a single copy in the kidney bean genome (data not shown). Accordingly, two mechanisms can be postulated for the production of the two different proteins from a single gene: alternative splicing of RNA transcript or a post-translational modification. Analysis of the 5Ј-RACE products (Fig. 3) shows the presence of two RNA transcripts, I and II, which code for pre-LF-PvSBE2 and pre-PvSBE2, respectively. When analyzed by the PSORT network program (33), both the N-terminal regions of pre-LF-PvSBE2 (residues 1-45) and pre-PvSBE2 (residues 99 -156) were predicted to be plastid-targeting leader sequences. There are many reports on alternative splicing in higher plants (34), although in most cases the differences in the biological functions of variant precursors have not been elucidated. For example, the sbeI gene for a wheat SBE isoform transcribes RNAs that code for three variant preproteins with different transit peptides. These putative proteins, however, have yet to be identified in the wheat endosperm (35). The present study indicates the possibility that alternative splicing of a single gene can give rise to two different SBEs displaying distinct catalytic properties and located in different distribution patterns within the plastid. Although RNA transcript levels are not precisely estimated by RT-PCR, the expression profiles obtained by this technique are distinct for the two pvsbe2 transcripts (Fig. 3D). Northern analysis for pvsbe2 mRNA (17) does not distinguished these transcripts because of the small difference in size (about 0.2 kb). Nevertheless, the total transcript levels during seed development, as assessed by RT-PCR (Fig. 3D), agree with the RNA expression profiles obtained by Northern blotting (Fig. 3E). Additionally, the RNA expression profiles parallel the total SBE protein accumulation profiles, including soluble and granule-bound forms, estimated by immunoblot analysis (Figs. 2 and 3E). However, no correla-tion is observed between the temporal accumulation patterns of RNA transcripts I and II and the corresponding protein accumulation profiles (refer to Figs. 2B and 3, D and E). From these results, it is reasonable to assume that the pvsbe2 gene is subject not only to transcriptional regulation but to post-translational as well. In fact, our preliminary experiments suggest that a proteolytic processing activity generating PvSBE2 protein from LF-PvSBE2 protein as a substrate occurs in the amyloplast fraction isolated from mid-size seeds of kidney bean. The isolation and identification of this processing activity are under way.
A comparison of N-terminal primary sequences between family A and B SBEs has indicated that there is an extra flexible domain in some family A SBEs such as pea SBEI, maize BEIIb, and rice RBE3. Burton et al. (19) predicted that this extra domain might be involved in the interactions between SBE and starch or in determining the type of glucan chain the enzyme can utilize as a substrate. The N-terminal extension of LF-PvSBE2 corresponds precisely to this extra domain seen in these other SBEs. The difference in subcellular localization between LF-PvSBE2 and PvSBE2 ( Fig. 2A) indicates that the N-terminal domain is responsible for the association with the starch granule. However, since rLF-PvSBE2 does not directly bind to raw starch in vitro (Table II) and lacks a starch-binding domain (29,30,36), it is unlikely that LF-PvSBE2 binds directly to the starch granule. Additionally, rLF-PvSBE2 does not bind stably to amylopectin-like glucans formed during the in vitro branching assay described in Fig. 6B (data not shown). Despite the apparent lack of direct binding of rLF-PvSBE2 to starch, we propose an explanation that can account for the association of LF-PvSBE2 with starch granules during seed development. One possibility is that LF-PvSBE2 becomes buried within the growing amylopectin structure as it becomes converted into a crystallized starch granule because of its high affinity for amylopectin but low catalytic turnover ( Fig.  6B and Table III). This may provide a basis for the distribution of LF-PvSBE2 to soluble and starch granule fractions, particularly in the late-stage of seed development. This explanation is consistent with the hypothesis that entrapment of maize SSI within the starch granule is affected by its high affinity for a longer glucan (37).
It has been predicted from a comparison of primary sequences between SSs and granule-bound starch synthases or bacterial glycogen synthases that SSs also contain a flexible FIG. 7. A hypothetical model for generation and localization of LF-PvSBE2 and PvSBE2 during amylopectin synthesis. Two precursors translated from alternative spliced RNAs would be localized to the plastid stroma. PvSBE2 could be also converted from LF-PvSBE2 by an unidentified processing activity (broken arrow). Because of the catalytic properties due to the extensive N-terminal region, LF-PvSBE2 distributes to the starch granule and soluble fractions. In contrast, PvSBE2 without the extensive N-terminal region is localized only in the soluble fraction. N-terminal domain (38). Analyses of the recombinant truncated N-terminal forms of pea SSII (39) and maize SSI (40) revealed that the flexible domain does not participate in the catalysis but may be involved in regulating the association between enzymes and ␣-glucans, and/or in partitioning between soluble and granule-bound phases. In particular, the truncated form of maize SSI has reduced affinity for amylopectin, compared with the extended form (40). Additionally, in developing pea embryo, there are truncated and native forms of SSII in both soluble and granule fractions (39). These characteristics of native and truncated SSs are similar to those of rLF-PvSBE2 and rPvSBE2, suggesting the possibility that branching enzyme activity is coordinated with SS polymerizing activity. The possible interaction between SBEs and SSs has been suggested from the analysis of the amylose-extender mutants of maize (41) and rice (42), where genetic disruption results not only in a loss or decrease of branching enzyme but also of starch synthase activity as well.
Comparison of the enzymatic properties between rLF-PvSBE2 and rPvSBE2 suggests that the N-terminal extension of LF-PvSBE2 has a significant effect on the kinetic parameters but is not essential for enzyme catalysis (Figs. 5 and 6 and Table III). A similar situation was also reported for the maize BEIIb (6); the truncated form lacking the N-terminal 39 residues of mature maize BEIIb not only possessed about 70% net catalytic activity but also showed a chain transfer pattern identical to that obtained for the mature enzyme. These observations indicate that the 39 residues of the N terminus of maize BEIIb are not required for catalysis. Interestingly, the N-terminal of the truncated form of maize BEIIb is very close to that of PvSBE2.
Results from this study lead to a hypothetical model for production and distribution of LF-PvSBE2 and PvSBE2 (Fig.  7). The pvsbe2 gene transcribes two different mRNAs by alternative splicing, and the protein products of this gene are targeted to plastids. Because of their enzymatic properties, part of LF-PvSBE2 is entrapped within starch granule, whereas PvSBE2 distributes only in soluble phase. In the mid-stage, when starch biosynthesis begins to be promoted, PvSBE2 isoform, in which the enzymatic properties are affected only slightly by changes in the concentration and structure of the substrate, is mainly responsible for formation of ␣-1,6 branch points in amylopectin biosynthesis. The conversion of LF-PvSBE2 to PvSBE2 by an unidentified processing enzyme activity is probably important for the effective increase in the number of nonreducing ends for starch synthases. In the later stage, the level of PvSBE2 protein reduces in concert with the transcript level, and probably the processing enzyme activity is also decreased. Instead, LF-PvSBE2 protein not only is entrapped within the starch granule but also becomes accumulated in the soluble phase.
This study demonstrates that two SBE isoforms encoded by a single gene have different subcellular localization and protein accumulation profiles as well as distinct enzymatic properties. Our results shed new insights on our understanding of the regulatory mechanism of amylopectin biosynthesis. A proteolytic processing enzyme will be particular significant because the en-zyme could control the activities and localization of not only SBEs but also SSs by the removal of their N-terminal domains.