Identification, isolation, and cloning of a Bacillus thuringiensis CryIAc toxin-binding protein from the midgut of the lepidopteran insect Heliothis virescens.

Bacillus thuringiensis toxins are insecticidal to a variety of insect species. The selectivity of the toxins produced by these bacteria is dependent on both the toxin structure and the receptor sites that are present in different insect species. One of these toxins, CryIAc, is highly insecticidal to the noctuid pest Heliothis virescens. Using toxin overlay assay, a 120-kDa glycoprotein was identified as a toxin-binding protein. This protein was partially purified, its N-terminal sequence was determined, and the full-length cDNA encoding this protein was isolated from a H. virescens midgut library. The B. thuringiensis toxin-binding protein, BTBP, has high homology to aminopeptidase N from eukaryotes and prokaryotes.

Bacillus thuringiensis toxins are insecticidal to a variety of insect species. The selectivity of the toxins produced by these bacteria is dependent on both the toxin structure and the receptor sites that are present in different insect species. One of these toxins, CryIAc, is highly insecticidal to the noctuid pest Heliothis virescens. Using toxin overlay assay, a 120-kDa glycoprotein was identified as a toxin-binding protein. This protein was partially purified, its N-terminal sequence was determined, and the full-length cDNA encoding this protein was isolated from a H. virescens midgut library. The B. thuringiensis toxin-binding protein, BTBP 1 , has high homology to aminopeptidase N from eukaryotes and prokaryotes.
Bacillus thuringiensis, a Gram-positive bacterium, produces insecticidal parasporal inclusions during sporulation (1). This insecticidal activity is dependent on a unique set of inclusion proteins or ␦-endotoxins produced by these B. thuringiensis strains (2). Most of these insecticidal proteins are usually produced as large protoxins, about 130 kDa, although smaller naturally truncated proteins are also observed. When ingested by susceptible insect larvae, the parasporal inclusions dissolve in the alkaline environment of insect midgut giving rise to soluble protoxins that are activated by midgut proteases (2,3). For example, the 130-kDa CryIAc protoxin is activated to a 65-kDa toxin in the midgut of lepidopteran Heliothis virescens and Manduca sexta larvae and is toxic to these insects.
These activated toxins then interact with the apical membranes of insect midgut columnar cells (4,5). Using preparations of columnar cell brush-border membrane vesicles (BBMV) 1 previous studies have demonstrated that in susceptible insects, high affinity toxin binding sites are correlated with the insecticidal activity of B. thuringiensis toxins (6,7). Changes in toxin structure, even minor, that are associated with decreased toxin binding result in decreased insecticidal activity (8 -10).
To identify the precise toxin binding targets in insects, midgut cell membranes have been electrophoretically separated and probed with radiolabeled toxins in toxin overlay assays. Several CryIAc binding proteins ranging from 150 to 50 kDa were observed in H. virescens (11)(12)(13)(14). We have identified the proteins involved in CryIAc toxicity to H. virescens using toxin overlay assays (15). A 120-kDa B. thuringiensis toxin-binding protein, BTBP 1 , from the midgut brush-border membrane that binds the CryIAc toxin was partially purified, and its N-terminal sequence was determined. A cDNA clone encoding this protein was isolated and characterized. The deduced amino acid sequence shows that the BTBP 1 belongs to the aminopeptidase N family of proteins.

MATERIALS AND METHODS
Toxin Isolation and Radiolabeling-Parasporal inclusions from B. thuringiensis sp. kurstaki strain HD-73 were isolated, the protoxin proteolytically was activated, and the 65-kDa activated toxin was purified on a fast protein liquid chromatography Superose 12 column (Pharmacia Biotech Inc.) as described previously (15). Radioiodination was performed as described previously (15) giving specific activities of 100 -200 Ci/mmol. The biological activity of the radiolabeled CryIAc toxin was comparable with that of the unlabeled toxin (15).
Partial Purification of the CryIAc Binding Proteins-Freshly prepared H. virescens BBMV (15) were solubilized in 2% CHAPS (Boehringer Mannheim Biochemicals) in solubilization buffer (20 mM Tris, pH 7.5, 1 mM EDTA, 1 mM Mg 2 SO 4 , 0.01% NaN 3 , and 10% glycerol) containing protease inhibitors (1 mM phenylmethylsulfonyl fluoride, 1 M aprotinin, 0.7 M leupeptin, and 1 M pepstatin). Following 1-h incubation at 4°C, the mixture was centrifuged for 1 h at 100,000 ϫ g. The supernatant was applied to a 1-ml fast protein liquid chromatography MonoQ column (Pharmacia). After washing with solubilization buffer the proteins were eluted with a 0 to 500 mM KCl linear gradient in solubilization buffer with 1% CHAPS. Fractions were assayed for toxin binding ability using a solid-phase binding assay (15). Fractions eluting at 250 mM KCl that bound the CryIAc toxin were pooled and are called the Pool II fraction.
This Pool II toxin binding fraction was then applied to a Ricinis communis agglutinin (RCA) agarose column in 20 mM Tris, pH 7.5, 150 mM NaCl, 1 mM EDTA, 1 mM MgSO 4 , 1% CHAPS (w/v) containing protease inhibitors at 4°C overnight, with gentle mixing. The preparation was then poured into a column, washed extensively, and eluted with 200 mM lactose in the appropriate buffer, and the eluted material was concentrated in a stirred cell using PM10 filters (Amicon).
Protein Determination and Enzyme Assays-Protein concentrations were determined using the detergent compatible protein assay kit (Bio-Rad) with bovine serum albumin as a standard. Alkaline phosphatase (16) and aminopeptidase activities were assayed as described previously (17).
N-terminal Sequence Analysis-To identify the N-terminal sequences, the 170-, 120-, and 105-kDa Mono Q-purified proteins in Pool II were separated by SDS-PAGE, transferred to an Immobilon membrane in 50 mM CAPS, 20% methanol, and 0.1% EDTA. The membrane was then washed, stained, and destained as described previously (19). Individual bands were excised and subjected to N-terminal amino acid gas-phase sequencing at the Biotechnology Instrumentation Facility at the University of California, Riverside.
cDNA Library Construction-Size fractionated poly(A ϩ ) RNA, Ͼ3 kilobases, obtained from the midguts of early fifth instar H. virescens larvae was used for the construction of a unidirectional cDNA using the Life Technologies, Inc. SuperScript Plasmid System kit and an oligo(dT) primer in a modified pCDM8 vector (20). The library consisting of about 10 5 transformants was plated on 15 LB/Amp (100 g/ml) plates. Plasmid preparations were made from these 15 library pools.
cDNA Library Screening-Briefly, the cDNA library was screened by initially performing PCR using plasmid DNA as template from each of the 15 library pools (21). The primers used were degenerate primers H. DNA from the plasmid pool showing the presence of the expected size product by PCR was then used for electroporation of E. coli DH10B cells. Transformed cells were used for additional screening by PCR as described previously (21). Successive rounds of this protocol yielded two clones.
Dideoxy double-stranded sequencing of the cDNA insert was performed using Sequenase 2.0 as described by the manufacture (U. S. Biochemical).
In Vitro Transcription and Translation-In vitro transcription was performed with the cDNA clone linearized with NotI using the mMessage mMachine kit (Ambion) catalyzed by T7 polymerase. The cRNA was translated in a rabbit reticulocyte cell-free translation system (Promega, WI) using [ 35 S]methionine (Amersham, Corp.). The translated products were immunoprecipitated by the addition of anti-H. virescens BBMV antibody followed by Protein A-Sepharose (Sigma). The precipitates were washed extensively in buffer containing 150 mM NaCl, 10 mM Tris-HCl, pH 7.5, 2 mM EDTA, and 5% Nonidet P-40 followed by 3 washes in a similar buffer but containing 500 mM NaCl. The final pellets were resuspended in SDS-PAGE loading buffer and analyzed on 8% gels SDS-PAGE under reducing conditions and visualized by fluorography (DuPont NEN).

RESULTS
Purification of the CryIAc Binding Protein-We previously (15) identified a number of proteins in H. virescens midgut BBMV and CHAPS-solubilized BBMV that bind iodinated CryIAc toxin. An attempt was made to identify one of these toxin-binding proteins. CHAPS-solubilized BBMV, when separated on Mono Q anion-exchange chromatography, gave rise to three toxin-binding fractions (15). One of these fractions, Pool II (Fig. 1, lane 4), from this Mono Q separation, eluting at 250 mM KCl, was then further separated using RCA chromatography. Combination of anion-exchange and lectin chromatography resulted in the purification of 120-and 105-kDa proteins (Fig. 1, lane 6). Minor amounts of a 170-kDa protein were also detected (lane 6). Purification using RCA chromatography demonstrates that these three proteins are all glycosylated.
Toxin overlay assay (15) showed that the 120-and 105-kDa RCA-purified proteins bound iodinated CryIAc toxin (Fig. 2,  lane 2) and that addition of excess unlabeled CryIAc toxin significantly decreased binding (Fig. 2, lane 3). Low level bind-ing was also observed with a 140-kDa protein, but this binding was not as readily displaced with excess unlabeled CryIAc toxin. The 140-kDa protein is not visible in the SDS-polyacrylamide gel (Fig. 1, lane 6). The 170-kDa protein did not bind the CryIAc toxin. Similar results were obtained with Mono Q Pool II fractionated proteins, except the 170-kDa protein also bound the CryIAc toxin (15).
Analysis of enzyme activity in the various fractions during purification through the Mono Q anion-exchange chromatography shows that alkaline phosphatase and aminopeptidases are enriched in the BBMV, solubilized BBMV, and in Pool II (Table  I). The alkaline phosphatase activity was enriched about 17ϫ from that found in crude homogenates, while the activities of leucine, lysine, and phenylalanine aminopeptidases were enriched 10, 14 and 9ϫ, respectively. Leucine aminopeptidase, an enzyme used as a marker for insect BBMV (22), had the highest specific activity in the Mono Q pooled fraction, Pool II.
N-terminal Sequencing and cDNA Isolation-N-terminal sequencing of the 170-, 120-, and 105-kDa proteins from Pool II (Fig. 1, lane 4) was performed. The N termini of the 170-and 120-kDa proteins are DPAYRLPTL and NV(V/A)ASPYRLPT, respectively (Table II). The N-terminal sequence of the 120-kDa protein purified from the RCA column was identical to that of the 120-kDa protein from the Mono Q column. The N terminus of the 105-kDa protein from the pooled fraction was different. The 170-and 120-kDa protein N-terminal sequences were compared with that of the N-terminal sequence of M. sexta CryIAc toxin 120-kDa binding protein (Table II) (23) and to known aminopeptidase N (24), since it has previously been established that the CryIAc binding protein in M. sexta appears to be aminopeptidase N-like (23,25).
The sequence YRLPT was conserved in all three insect CryIAc toxin-binding proteins, and the sequence YRLP was conserved in all of the proteins compared (Table II). Oligonucleotide primers, forward and reverse, were made ensuring the sequence YRLPT was present at the 3Ј end of the primers. When used in conjunction with a primer to conserved regions of aminopeptidase N (AMN2R) in PCR screening, both the M. sexta (MS1F) and H. virescens (HV1F) primers gave the same size product, 0.85 kilobase. Reamplification of these PCR products with AMN1F and AMN1R gave products of the expected size, 170 bp. Amplification of one of the fractionated libraries, number 3, using the primers V1F and AMN1R, followed by reamplification using V2F and HV1R, showed the presence of prominent bands between 220 and 300 bp, suggesting the presence of cDNA sequence upstream of the N-terminal sequences obtained. Electroporation of fraction 3 plasmid DNA into DH10B cells and successive rounds of PCR screening (21) resulted in the isolation of two clones. Nucleotide sequencing of these two clones showed that one clone contained a sequence that matched that of the 120-kDa protein. This clone had an insert of 3419 bp, with a 3027-bp open reading frame encoding a 1009-amino acid protein (Fig. 3). The putative translation start site at nucleotide 35 contains a consensus Kozak sequence, AAGATGG (26). A polyadenylation sequence, AATAAA   (27), at nucleotide 3385 precedes the poly(A) tail, which is 337 bp downstream of the termination codon. The protein, BTBP 1 , has a calculated molecular weight of 113,461 Da and a pI of 5.29. The N-terminal sequence obtained from protein sequencing is between amino acids 53 and 63 and indicates that the mature BTBP 1 isolated from H. virescens BBMV has 52 amino acids cleaved from the N terminus. Both the N and C termini are hydrophobic. Two potential N-glycosylation sites are observed at amino acid residues 581 and 906. A metal binding motif, HEXXH (28,29) is observed at 374 -378.
Role of Glycosylation in Toxin Binding and Immunoprecipitation with Antibodies-The CryIAc toxin has been shown to bind to carbohydrate moieties (11)(12)(13)(14). Fig. 4 shows CryIAc binding to RCA purified proteins (lane 1) and its displacement by the presence of 200 mM N-acetylgalactosamine (lane 2) confirming the role of carbohydrate moieties in toxin binding.
In vitro transcription and translation of BTBP 1 cDNA gives a protein of about 113 kDa (Fig. 5, lane 3). This size, as predicted from the deduced amino acid sequence, lends support to the putative start site. The in vitro translated CryIAc binding protein was immunoprecipitable with anti-H. virescens BBMV antibodies (Fig. 5, lane 4), but not with preimmune antibodies (lane 5). The smaller and weakly labeled products (lane 3) are derived either from incomplete transcription and translation or are proteolytically cleaved BTBP 1 , since these products all are immunoprecipitable only by anti-BBMV antibodies (lane 4) and not by preimmune antibodies (lane 5). No products were obtained when only the vector was used for in vitro transcription and translation (Fig. 5, lanes 1 and 2). Immunolocalization studies show these antibodies react with the midgut brushborder membrane of H. virescens (data not shown), demonstrating that the cloned BTBP 1 is apparently localized to the brushborder membrane of H. virescens midgut. The in vitro translated BTBP 1 protein does not bind CryIAc toxin (data not shown) consistent with previous data showing the CryIAc toxin binding to carbohydrate moieties of this 120-kDa protein.

DISCUSSION
Protein purification, N-terminal sequencing, and characterization of the cDNA isolated identify the CryIAc toxin-binding protein from the midgut of H. virescens as an aminopeptidase N-like protein (Fig. 6). An aminopeptidase N from M. sexta, that also binds the CryIAc toxin, has recently been reported (30). The H. virescens BTBP 1 has 42 and 62% identity and similarity, respectively, to the M. sexta protein. Comparison of these two sequences suggests the M. sexta CryIAc binding protein probably has an additional 10 -25 amino acids upstream of the reported N-terminal sequence (30). Similar levels of identity, although slightly lower, are observed with other eukaryote aminopeptidases N, zinc-dependent metalloproteases (28,29). For example, H. virescens BTBP 1 has 32 and 52% identity and similarity, respectively, to human aminopeptidase N.
The BTBP 1 N terminus is quite divergent from that of other aminopeptidases. The first 15 amino acids are highly hydrophobic (2.6) and probably contain an appropriate signal sequence that facilitates membrane targeting (31,32). However, unlike mammalian aminopeptidases, where the mature protein undergoes limited truncation at the N terminus (33), the first 52 amino acid residues of H. virescens BTBP 1 are cleaved. This can be explained, in part, by the different membrane anchors used by H. virescens BTBP 1 , and those used by ami-nopeptidase N in mammals. In the latter, aminopeptidase N is membrane-bound apparently by a hydrophobic N terminus (33), and trypsin treatment results in cleavage of the membrane anchor, with the resulting soluble protein beginning at amino acid residue 40 (33). In insects, aminopeptidase N is apparently membrane bound at the C terminus. The soluble form of the insect leucine aminopeptidase is obtained by treatment with phosphatidylinositol-specific phospholipase C (34,35). Immunoprecipitation of BTBP 1 by BBMV-specific antibodies suggests that this protein is localized to the midgut cell brush border of H. virescens. Leucine aminopeptidase activity is often used as a marker for insect midgut columnar cell brush border (22,36).
Unlike the other aminopeptidases, the H. virescens BTBP 1 has a long hydrophobic C terminus. In two insect species, Bombyx mori and M. sexta, the midgut aminopeptidase N activity is membrane-linked via a glycosylphosphatidylinositol (GPI) anchor (34,35). The presence of a long hydrophobic C terminus preceded by hydrophilic residues suggests that the BTBP 1 is similarly linked by a GPI anchor. A putative sequence signaling the addition of GPI anchors (37), DSA, is observed. The aspartic acid at 987 is a likely site for attachment of the GPI anchor, with cleavage of the hydrophobic tail between Asp 987 and Ser 988 . The presence of a GPI anchor in H. virescens, however, needs to be established, and alternative mechanisms of BTBP 1 membrane anchoring, however, cannot be excluded. N-and C-terminal proteolytic cleavage will result in a 105-kDa protein containing 935 amino acid residues. The BTBP 1 isolated from H. virescens midgut, however, is of 120 kDa, consequently the difference in the molecular masses of these two proteins results from carbohydrate or other posttranslational modifications. The M. sexta CryIAc binding protein similarly has a putative GPI signal sequence, a hydrophobic N terminus, and identical molecular masses for the processed and isolated proteins (30).
The hydrophobic tail and the GPI-linked anchor could play a crucial role in insects resistant to B. thuringiensis toxins. Elevation of endogenous phosphatidylinositol-specific phospholipase C, or B. thuringiensis phosphatidylinositol-specific phospholipase C could result in decreased levels of membranebound BTBP 1 or aminopeptidase N causing increased tolerance and/or resistance to the CryIAc toxin. Alternatively down-regulation of BTBP 1 expression in insect midgut will result in the loss of toxin binding sites, and decreased CryIAc toxicity. However, H. virescens resistance to the CryIAc toxin does not appear to be correlated with decreased toxin binding (38), and consequently alternative mechanisms of toxin resistance are likely. On the other hand, in Plutella xylostella and Plodia interpunctella, resistance to the CryIAb toxins is correlated with the availability of toxin binding sites (39 -41).
A CryIAb toxin-binding protein has recently been cloned and shown to be a cadherin-like 210-kDa membrane glycoprotein (42). Unlike the CryIAb-binding protein, which apparently has an intracellular component, BTBP 1 is entirely extracellular and is probably GPI-anchored. Hence it is unlikely to directly participate in any intracellular changes that have been observed with B. thuringiensis intoxication (43), although it could affect external mediators of signal transduction.
The precise mechanism of B. thuringiensis toxicity is not known. The generally accepted model is that following toxin binding to a receptor protein or macromolecule, the toxin un- FIG. 6. Comparison of H. virescens BTBP 1 with known aminopeptidases. BTBP 1 sequence was compared with nucleotide sequences (Genbank, release 88.0) and with protein sequences (Swiss Protein, release 31.0). The greatest similarity is observed with zinc-dependent metalloproteases, which includes aminopeptidases N. Only some sequences are used for comparison and include human (24), Caenorhabditis elegans (54), mouse (51), Saccharomyces (55), and Lactobacilli (56). The M. sexta sequence used was as published recently (30). dergoes a conformational change that facilitates toxin insertion into the apical cell membrane of insect midgut columnar cells. This initial binding is then followed by oligomerization of the bound toxin (3,44,45). The cation ion selective pore formed by this oligomer causes a disruption of osmotic balance in the midgut epithelial layer (46 -48). This disruption of midgut function ultimately leads to the insect's death (3). Potentially the 120-kDa toxin-binding proteins isolated here and that in M. sexta (23,25), since they have an ability to bind the CryIAc toxin, function to localize the toxin to the columnar cell membrane in insect midguts. Indeed phospholipid vesicles containing a 120-kDa aminopeptidase N from M. sexta enhanced Rb 2ϩ permeability when challenged with the CryIAc toxin (25). The CryIAc binding in M. sexta, like that in H. virescens, is partially blocked by N-acetylgalactosamine, suggesting that the binding occurs on the carbohydrate moieties of this protein. Two putative N-glycosylation sites are observed in the H. virescens BTBP 1 protein. Both sites occur in predicted turns in protein structure (49) and are likely glycosylated. Moreover, just as in M. sexta (30), the H. virescens BTBP 1 protein sequence preceding the GPI signal sequence is rich in Ser/Thr residues, which could also provide O-linked glycosylated residues required for CryIAc binding. Furthermore, the addition of a GPI anchor provides additional carbohydrate moieties for toxin interaction. The nature of these carbohydrate moieties, at the N-and Oglycosylation sites, and at the GPI anchor, is not known. Their characterization should provide a better understanding of the selectivity of B. thuringiensis toxins.
If B. thuringiensis CryIAc toxin activity is mediated predominantly by the carbohydrate moieties, it is likely that the differing responses observed between insects is due to heterogeneity in glycosylation. This heterogeneity in glycosylation could in part explain the rapid recovery of susceptibility from previously resistant insects (41). A variety of glycoproteins could function as toxin-binding proteins, hence facilitating toxin interaction with the midgut epithelium.
Aminopeptidases N are not only widely distributed in a number of mammalian cell types, but they also appear to play various roles in these cell types (50). In mice they are involved in hematopoeisis (51) and in the degradation of extracellular matrix and tumor cell invasion (52). In intestinal epithelia, aminopeptidase N is involved in peptidyl bond cleavage releasing amino acids that are transported across the brush-border membrane (50). Their high level expression in a variety of cell types facilitates the entry of certain coronaviruses in humans (53). Hence the interaction observed here with the CryIAc toxin similarly enables B. thuringiensis to be insecticidal.