Identification of glycosphingolipid receptors for pierisin-1, a guanine-specific ADP-ribosylating toxin from the cabbage butterfly.

Pierisin-1, a cytotoxic protein found naturally in the cabbage butterfly, induces apoptosis of mammalian cells. Our recent studies suggest that pierisin-1 consists of an N-terminal ADP-ribosyltransferase domain, and a C-terminal region that binds to receptors on the surfaces of target cells and incorporates the protein into cells. The present study was undertaken to identify receptors for pierisin-1. The cross-linking and cloning experiments suggested that the proteins on cell membrane had no binding ability to pierisin-1. Inhibitory assays of fractionated lipids from human cervical carcinoma HeLa cells, which are highly sensitive to pierisin-1, indicated neutral glycosphingolipids on the cell surface to show receptor activity. Inhibitory assays and TLC immunostaining using anti-pierisin-1 antibodies demonstrated two neutral glycosphingolipids as active components. Analysis of their structures with glycosphingolipid-specific antibodies and negative secondary ion mass spectrometry identified them as globotriaosylceramide (Gb3) and globotetraosylceramide (Gb4). The receptor activities of Gb3 and Gb4 for pierisin-1 were also confirmed with these authentic compounds. Pierisin-1-insensitive mouse melanoma MEB4 cells were found to lack pierisin-1 receptors, including Gb3 and Gb4, but pretreatment of the cells with glycosphingolipid Gb3 or Gb4 enhanced their sensitivity to pierisin-1. Thus, Gb3 and Gb4 were proven to serve as pierisin-1 receptors. The C-terminal region of pierisin-1 consists of possible lectin domains of a ricin B-chain, containing QXW sequences, which are essential for its structural organization. Alteration of QXW by site-directed mutagenesis caused marked reduction of pierisin-1 cytotoxicity. Thus, our results suggest that pierisin-1 binds to Gb3 and Gb4 receptors at the C-terminal region, in a manner similar to ricin, and then exhibits cytotoxicity after incorporation into the cell.

Pierisin-1 is a 98-kDa protein present in the cabbage butterfly that has potent cytotoxic activity (1,2). Among 13 mamma-lian cell lines so far tested, human cervical carcinoma HeLa cells are the most sensitive to the cytotoxic effects of pierisin-1, whereas mouse melanoma MEB4 cells are the least sensitive, with an IC 50 value ϳ5000 times higher (3)(4)(5). Pierisin-1 is a potent inducer of apoptosis of mammalian cells, which is accompanied by cleavage of DNA to nucleosome units and of poly(ADP-ribose) polymerase (2,3). Cloning of a complementary DNA (cDNA) of piersin-1 from Pieris rapae revealed that pierisin-1 shares sequence homology with the enzyme units of ADP-ribosylating toxins, including the A-subunit of cholera toxin in its 27-kDa N-terminal region (6). Furthermore, substitution of a glutamic acid residue at a presumed NAD-binding site caused loss of cytotoxic activity, suggesting an essential role for ADP-ribosylating activity in exerting the cytotoxicity of piersin-1 (6). Similarly, pierisin-2 from Pieris brassicae has been suggested to exert its action through ADP-ribosyltransferase (5). Recently, we reported that the target molecule for mono(ADP-ribosyl)ation catalyzed by pierisin-1 is a DNA, but not a protein, providing a contrast to bacteria-derived ADPribosylating toxins such as cholera toxin and pertussis toxin (7). Pierisin-1 efficiently catalyzes the ADP-ribosylation of double-stranded DNA. The ADP-ribose moiety of NAD is transferred by pierisin-1 to the amino group at N 2 of the deoxyguanosine base (7).
An in vitro expressed peptide consisting of only the N-terminal region exhibited cytotoxicity and apoptosis-inducing activity when it was incorporated by electroporation (4). However, the N-terminal peptide alone could not be incorporated into the cells. The remaining 71-kDa C-terminal region plays a role in binding and internalization of the whole protein into the target mammalian cells (4). The C-terminal region of pierisin-1 shares sequence similarity with HA-33 (or HA1), a subcomponent of hemagglutinin of botulinum toxin (8,9). Recent reports of a requirement for sialic acid or galactose moieties for binding of HA-33 suggest that glycolipids or glycoproteins might similarly play a role in binding of pierisin-1 to cells (10,11). Indeed, we found that addition of the lipid fraction prepared from HeLa cells, being highly sensitive to pierisin-1, to the medium inhibited the cytotoxic activity of pierisin-1 as did the membrane fraction (4). Therefore, components of the glycolipids on the membrane in HeLa cells might act as receptors for pierisin-1.
The present study was designed to explore this possibility. Two glycosphingolipids were thereby identified as receptor molecules. Immunostaining with antibodies against glycosphingolipids and their structural analysis by mass spectrom-etry indicated them to be the neutral glycosphingolipids Gb3 1 and Gb4. 2 The possible structure and function of the C-terminal region of pierisin-1 were also investigated by site-directed mutagenesis.

EXPERIMENTAL PROCEDURES
Cell Lines-Human cervical carcinoma HeLa cells and mouse melanoma MEB4 cells were obtained from the Institute of Physical and Chemical Research (RIKEN) cell bank (Tsukuba, Japan). The cells were cultured in RPMI 1640 medium supplemented with 5% heat-inactivated fetal bovine serum (Invitrogen, Carlsbad, CA) and 50 g/ml kanamycin sulfate (Invitrogen) unless otherwise described. Cell cultures were maintained at 37°C in an atmosphere of 95% air and 5% CO 2 .
Expression Cloning of a cDNA Encoding the Pierisin-1 Receptor-For the panning procedure, a modification of the procedure of Almenoff et al. (13,14) was employed. In the first round of screening, COS cells were transfected with a human liver cDNA library (Takara 9505). After 72 h, the cells were pooled and resuspended in panning buffer (PBS containing 5 mM EDTA and 0.02% NaN 3 5% fetal calf serum), and the panning plates were coated with pierisin-1 (44 g per plate in 50 mM Tris-HCl buffer, pH 9.5) for 3 h and blocked with PBS containing 0.1% BSA at 4°C for overnight. The transfected COS cells were then distributed into the coated plates, allowed to attach for 10 min at room temperature, then washed three times gently with panning buffer. Then cells remaining on the dishes were lysed, and plasmid DNA was recovered by the alkaline Miniprep method and amplified in Escherichia coli to obtain material for the next cycle of panning.
Purification of Glycosphingolipids-Receptor activity was monitored by inhibitory effects on cytotoxicity of pierisin-1 to HeLa cells using a method described previously (4). Total lipids were extracted from 5 ϫ 10 8 harvested cells according to the method of Murayama et al. (15). Briefly, packed cells were extracted twice with 2-propanol/hexane/water (55:25:20, v/v) by sonication for 5 min followed by centrifuging. The pellet was re-extracted with chloroform/methanol (2:1, v/v) and subsequently with chloroform/methanol (1:1, v/v). The combined extract was then evaporated, and the residue was dissolved in 1 ml of solvent A (chloroform/methanol/water, 30:60:8, v/v).
To remove phospholipids, mild alkaline degradation was performed in 0.5 N sodium hydroxide in methanol at 37°C for 1 h, followed by dialysis against water overnight and lyophilization. The lyophilized sample was dissolved in solvent A and applied to DEAE-Sephadex A-25 column (bed volume 16 ml, acetate form) (Amersham Biosciences, Buckinghamshire, UK). Neutral glycosphingolipids were recovered in the flowthrough fractions, while acidic glycosphingolipids were eluted with solvent A containing 0.8 M sodium acetate. For further purification of individual glycosphingolipids, the flowthrough fractions were developed on TLC plates of Silica Gel 60 (E. Merck, Darmstadt, Germany) with solvent B (chloroform/methanol/water, 65:25:4, v/v). Spots on TLC plates corresponding to each glycosphingolipid were scraped off, and the lipids were extracted with solvent B. The extracts were then evaporated, and the residue was suspended with water, dialyzed overnight against 10 mM EDTA, and then lyophilized. EDTA was removed by passage through a Sephadex LH-20 column (Amersham Biosciences) equilibrated with solvent A. The eluates were used for further experiments.
TLC Immunostaining-TLC immunostaining for detection of molecules bound with pierisin-1 was performed as follows. Glycosphingolipids were developed on a plastic TLC plate (Polygram SIL G, Macherey-Nagel, Germany) using solvent B. Two chromatograms were developed in parallel on the same sheet. One was visualized with orcinol reagent for chemical detection of glycosphingolipids. The other was soaked overnight in phosphate-buffered saline containing 1% bovine serum albumin (PBS/BSA) at 4°C to block nonspecific antibody binding. The plate was incubated for 1 h with pierisin-1 (2 g/ml) in PBS/BSA. After rinsing with PBS, the plate was soaked for 10 min at room temperature in PBS containing 4% formaldehyde. After washing four times with PBS, the plate was incubated for 2 h at room temperature with 1:1000 diluted rabbit anti-pierisin-1 antisera (4) in PBS/BSA containing 5% goat serum. The plate was then washed five times with PBS containing 0.05% Tween 20 and washed once with PBS/BSA. Pierisin-1, bound with anti-pierisin-1 antibody, was incubated for 1 h at room temperature in 1:2000 diluted horseradish peroxidase-conjugated goat antirabbit IgG antibody (#NA9340, Amersham Biosciences) in PBS/BSA. After washing five times with PBS, the bound antibody was visualized with an ECL chemiluminescence kit (Amersham Biosciences).
For conversion of Gb4 to Gb3, the isolated glycosphingolipids were incubated with ␤-N-acetylhexosaminidase from jack beans (Seikagaku Kogyo, Tokyo, Japan). The resulting products were also assayed by TLC immunostaining.
TLC Blotting⁄Secondary Ion Mass Spectrometry-Glycosphingolipids that were separated on the TLC plate were transferred to a polyvinylidene difluoride membrane by the TLC blotting method (18). The appropriate position, sized at about 2 mm in diameter, was cut out and placed on the SIMS target tip with 0.5 l of triethanolamine as the SIMS matrix. The SIMS spectra of the glycosphingolipids in negative ion mode were obtained using a TSQ70 triple-quadrupole mass spectrometer (ThermoFinnigan, San Jose, CA) equipped with a 20-KeV cesium ion gun.
Effects of Glycosphingolipid on Cytotoxicity of Pierisin-1 in MEB4 Cells-MEB4 cells, maintained in Opti-MEM (Invitrogen)-reduced serum (0.5% FBS (HyClone)) medium for at least 1 week, were trypsinized and suspended in a reduced serum medium at a density of 5 ϫ 10 4 cells/ml. In the next step, a 100-l aliquot of the cell suspension was dispensed in each well of the 96-well plate and cultured for 18 h. The cells in each well were then incubated for 6 h with 100 l of the reduced serum medium containing 100 g/ml (Gb4, Gb5, or GM3) or 10 g/ml (Gb3) of glycosphingolipid (Sigma Chemical Co., St. Louis, MO). After removal of the glycosphingolipid-containing medium, the cells in each well were washed with the medium and incubated for 1 h with 100 l of the medium containing pierisin-1. After washing they were further cultured in the same medium for 48 h, and the effects of the incorporated glycosphingolipid on the cytotoxicity of pierisin-1 were examined by phase-contrast microscopy and WST-1 cell proliferation assay described previously (2).
Following the amplifications, these PCR products were subjected to a second round of PCR to obtain full-length mutated DNA fragment. The 25-l reaction mixture contained 5 l each of PCR solution of 5Ј-and 3Ј-fragments. The thermocycle conditions were 5 cycles of denaturation at 98°C for 15 s, re-annealing at 60°C for 30 s, and extension at 72°C for 2 min. Using this protocol, more than half the 5Ј-and 3Ј-fragments could be converted to full-length DNA fragments. The resultant DNA was used as the template for the in vitro expression system described previously (6) using MEGAscript and rabbit reticulocyte lysate (Ambion, Austin, TX). The PCR primer for attachment of T7 promoter sequence was 5Ј-TAATACGACTCACTATAGGGCGAATTGCCACCAT-GGCTGACCGTCAACCTTA-3Ј. The cytotoxicity of each translated protein in HeLa cells was assessed by the WST-1 cell proliferation assay (2).

Isolation of Pierisin-1 Receptors from HeLa Cells-Pierisin-1
exhibits cytotoxicity against mammalian cells after being incorporated into the cells by interaction of its C-terminal region with the receptor on the cell membrane. To identify the possible receptor protein, we performed cross-linking experiments. HeLa cells were incubated with pierisin-1, cross-linked at each amino group by disuccinimidyl suberate, and then subjected to Western blotting. However, no cross-reacting bands were observed. Specific binding of pierisin-1 to a membrane protein was further examined by expression cloning of a cDNA encoding a possible receptor from a human liver cDNA library through the affinity panning system. The selectivity for pierisin-1 receptor expression was analyzed using COS cells transiently transfected with the human liver cDNA library. To select COS clones expressing receptor cDNA, the transfected cells were placed on panning plates coated with pierisin-1. However, positive cells were not detected. These results suggested that the protein fraction has no ability bind to pierisin-1.
It is plausible that receptor molecules competitively inhibit the binding of pierisin-1 to the cells, thereby inhibiting cytotoxicity. We have reported that membrane fractions from HeLa cells and total lipid fraction of HeLa cells inhibited cytotoxic activity of pierisin-1 (4). Treatment of HeLa cells with 2 ng/ml pierisin-1 at a pulse duration of 15 min induced cell death in about 50% of the cells. Contrary to this, preincubation of 2 ng/ml pierisin-1 with about 50 g of total lipid fraction from HeLa cells before the treatment caused cell death in only 20%. Furthermore, an inhibitory assay of fractionated lipid from HeLa cells suggested that a polar lipid fraction, which contains glycolipids and phospholipids, has a binding ability to pierisin-1 (data not shown). To determine which of the polar lipids on HeLa cells is the receptor candidate, we first obtained glycosphingolipids from the total lipids of HeLa cells by mild alkaline degradation and dialysis and then analyzed the inhibitory effects on cytotoxicity. The inhibitory effects of the glycosphingolipids on cytotoxicity were stronger than for the parent material, the total lipid fraction, suggesting that the glycosphingolipids are the major source of pierisin-1 receptor. When the glycosphingolipids were further separated into neutral and acidic glycosphingolipids by DEAE-Sephadex A-25 column chromatography, about 75% of the inhibitory activity was recovered in flowthrough fractions, whereas no such activity was

CATGCTATATTCTATCGTCCCGAAATATCTCAGATCTCC GGAGATCTGAGATATTTCGGGACGATAGAATATAGCATG
observed in any of the eluted fractions, which contain acidic glycosphingolipids. These findings indicate that neutral glycosphingolipids are pierisin-1 receptor candidates. None of the samples isolated from pierisin-1-insensitive MEB4 cells with the same methods exhibited any inhibitory effects. The neutral glycosphingolipids from HeLa cells were developed on a TLC plate using solvent B (Fig. 1). Of the various glycosphingolipids in the fraction stained with orcinol reagent, two major doublet bands, each doublet probably representing different sugar residues with two different classes of ceramide, were clearly detected by TLC immunostaining with anti-pieri-sin-1 antibodies (Fig. 1). The mobility of these two sets of bands was similar to those of authentic Gb3 and Gb4, which also exhibited binding ability with pierisin-1.
In the next step, we purified both positive doublet bands, designated as fractions I and II (Fig. 1), by preparative TLC and analyzed their effects on pierisin-1 cytotoxicity. As expected, a decrease was noted following addition of either of the two fractions (data not shown), suggesting Gb3 and Gb4 to be receptor glycolipids, present in HeLa cells.
Structural Analyses of Receptor Glycosphingolipids-Fractions I and II showed positive bands on TLC immunostaining FIG. 3. Negative SIMS spectra of the two neutral glycosphingolipids with binding activity to pierisin-1. Fractions I and II were prepared from 5 ϫ 10 8 harvested HeLa cells and dissolved in 500 l of solvent A. Samples (5 l) that were separated on the TLC plate were transferred to a polyvinylidene difluoride membrane by the TLC blotting method (18). SIMS spectra for glycosphingolipids from fractions I and II are shown in A and B, respectively, with interpretations of the fragment ions. Those at m/z 595, 744, 893, and 1042 are cluster ions of the triethanolamine used as a matrix. Although an additional ion at m/z 1024.8 is apparent in spectrum A, its presence may be due to impurity. with anti-Gb3 and anti-Gb4 antibodies, respectively (Fig. 2). Moreover, degradation products of both fraction II and authentic Gb4 with ␤-N-acetylhexosaminidase were confirmed to be Gb3 by TLC immunostaining using anti-Gb3 antibody (data not shown). These data strongly suggested that sugar residues of the glycosphingolipids isolated from HeLa cells, exhibiting receptor activity, are identical to those of Gb3 and Gb4, respectively.
For further confirmation of the structures of the isolated glycosphingolipids, each glycosphingolipid from fractions I and II was analyzed by negative ion SIMS (Fig. 3). Deprotonated molecules ([M-H] Ϫ ) were observed at m/z 1106.9 and 1132.9 in the spectrum for that from fraction I, as shown in Fig. 3A. The ion of m/z 1106.9 corresponded to Gb3, consisting of sphinge-nine (d18:1) and docosanoic acid (C22:0) as ceramide. The ion of m/z 1132.9 corresponded to Gb3 consisting of d18:1 and tetracosenoic acid (C24:1) as ceramide. The fragment ions were weakly observed at m/z 808.7 and 970.7, corresponding to Glc-Cer and LacCer (d18:1/C24:1), respectively. Thus, the structure for both compounds was determined to be Gb3 (Gal-Gal-Glc-Cer). A deprotonated molecule was observed at m/z 1336.1 in the spectrum of the glycosphingolipid from fraction II, as shown in Fig. 3B, thus corresponding to Gb4, consisting of d18:1 and C24:1 as ceramide. Fragment ions observed at m/z 646.6, 808.8, 970.8, and 1133.0 corresponded to ceramide, Glc-Cer, LacCer, and Gb3 (d18:1/C24:1), respectively. Accordingly, the structure of the compound was concluded to be Gb4 (GalNac-Gal-Gal-Glc-Cer).

FIG. 4. Effects of glycosphingolipids on the cytotoxicity of pierisin-1 in MEB4 cells.
A, MEB4 cells treated with Gb3 (10 g/ml) or Gb4 (100 g/ml) for 6 h were incubated with various doses of pierisin-1 for 1 h at 37°C. After further incubation for 48 h, pierisin-1 cytotoxicity was examined with a WST-1 cell proliferation assay (2). Gb3 (closed triangle), Gb4 (closed circle), control (open circle). Each experiment was carried out in triplicate. B, MEB4 cells with Gb4 (lower part) or without Gb4 (upper part) were incubated for 1 h with 250 ng/ml pierisin-1, and 48 h later, changes in cell morphology were assessed by phase-contrast microscopy.

TABLE II
Inhibitory effects of various glycolipids on cytotoxicity of pierisin-1 All glycosphingolipids and their sugar moieties were obtained from Sigma. A sample of 0.5 ng of native pierisin-1 was preincubated in a 50-l aliquot solution containing 5 g each authentic glycolipid sample for 1 h at 4°C. Then the appropriate volume of preincubated mixture was subjected to WST-1 cell proliferation assay to determine the inhibitory effect of the sample on the cytotoxicity of pierisin-1 to HeLa cells (4). The data were obtained from two independent assays. ϩ, more than 50% of cytotoxic activity was suppressed by the sample; Ϫ, no suppression was observed.

Neutral glycosphingolipids
GalCer Inhibitory Potential of Various Glycolipids on Cytotoxic Activity of Pierisin-1-We then examined the inhibitory effects of a series of glycolipids on the cytotoxicity of pierisin-1 to HeLa cells. As shown in Table II, authentic Gb3 and Gb4 exhibited clear inhibition. Furthermore, authentic Gb5 exhibited similar activity as well as binding to pierisin-1 on TLC immunostaining with anti-pierisin-1 antibody (data not shown). However, GalCer, GlcCer, and LacCer had no inhibitory activity. An oligonucleotide sugar, Gal␣1-4Gal␤1-4Glc, corresponding to the terminal sugar sequence of Gb3, also showed no such inhibitory effect. Other glycolipids such as gangliosides GM1, GM2, and GM3, as well as asialo-GM1 and asialo-GM2 did not affect the cytotoxicity of pierisin-1. Thus, Gb3 might be the minimal requirement structure for the pierisin-1 receptor.
Effects of Glycosphingolipids on Cytotoxicity of Pierisin-1 in MEB4 Cells-In pierisin-1-insensitive MEB4 cells, no glycolipids binding to pierisin-1, including Gb3 and Gb4, were detected by TLC immunostaining (data not shown). Authentic Gb4 was added to MEB4 cells in reduced serum medium and incubated. After removal of Gb4, the cells were treated with pierisin-1. Cytotoxic assay demonstrated that sensitivity to pierisin-1 was clearly enhanced (Fig. 4A). Similar effects were observed with one-tenth the concentration of authentic Gb3 (Fig. 4A). Similarly, Gb5 enhanced cell sensitivity, although GM3 did not show any effect. No morphological changes were observed on treatment of cells with any glycosphingolipids in the absence of pierisin-1. Fig. 4B illustrates pierisin-1-induced morphological changes of MEB4 cells, with or without Gb4 pretreatment. Concentrations of pierisin-1 that exhibited little effect on control MEB4 cells caused shrinkage detachment after pretreatment with Gb4. The results thus suggest that authentic glycosphingolipids, Gb3 and Gb4, also serve as pierisin-1 receptors.
Role of the QXW Sequence in the C-terminal Region of Pierisin-1-Although the C-terminal region of pierisin-1 is known to share sequence homology with HA-33, the structure of HA-33 remains unclear. Recent data base searches indicated pierisin-1 and HA-33 to exhibit sequence similarity with the lectin domain of ricin B-chain. The QXW sequence pattern is present in each subdomain of the lectin domain of ricin B-chain and is important for its structural organization and function (21,22). This QXW sequence was found in the C-terminal region of pierisin-1 (Fig. 5). Tryptophan residue at the QXW sequence pattern was conserved at 11 of 12 subdomains. To clarify whether the QXW sequence of pierisin-1 is essential as in ricin, site-directed mutagenesis was conducted to alter the conserved tryptophan residue (Table III). Replacement of the residue by glycine in each ␤ subdomain, the most conserved among the ␣, ␤, and ␥ subdomains (see Fig. 5), in domains 1, 2, 3, and 4 (W354G, W505G, W656G, and W801G, respectively) showed a loss of at least 95% of the cytotoxic activity against HeLa cells. This implies necessity of all four domains for the exertion of cytotoxicity of pierisin-1. Next, the conserved tryptophan in subdomains 3␣ and 3␥ was substituted with glycine. The cytotoxic activities of the resultant mutant proteins (W607G and W704G) decreased to less than 5% of the parent level, similar to the 3␤ mutant, W656G. Furthermore, replacement of tryptophan 656 in the 3␤ subdomain by other than glycine, phenylalanine (W656F) or histidine (W656H), also resulted in a marked reduction, by 82 and 89%, respectively, of cytotoxic activity against HeLa cells. These results suggest that the QXW sequence in each subdomain of every domain in the C-terminal region of pierisin-1 plays a structural role in the cytotoxicity. DISCUSSION In the present study, two active neutral glycosphingolipids exerting inhibitory effects on the cytotoxicity of pierisin-1 were

TABLE III Cytotoxic activity of pierisin-1 mutated in the conserved QxW sequence
Mutated DNA templates were obtained by PCR using mutated primers. Each mutated protein was prepared in vitro using T7 RNA polymerase and rabbit reticulocyte lysate. HeLa cells were incubated with various doses of mutated proteins for 72 h at 37°C, and subjected to WST-1 cell proliferation assay (2). Data represent the cytotoxic activity of each mutated pierisin-1 relative to the non-mutated control protein.
The data were obtained from two independent assays. The two mutated pierisin-1 (W656G and W656F) were shown to possess an active Nterminal enzyme domain, as in the case of the non-mutated control protein, by ADP-ribosyltransferase assay with a slight modification, as previously reported (7) isolated. The binding abilities of the glycosphingolipids to pierisin-1 were demonstrated by TLC immunostaining. Their identities as Gb3 and Gb4 were revealed using anti-Gb3 and -Gb4 antibodies and confirmed by negative ion SIMS. As noted above, Gb3 and Gb4 may serve as pierisin-1 receptors, and this appears to be the case for Gb5, whereas several structurally related molecules, including LacCer, GM3, asialo-GM2, and globotriaose, were found to be without activity. Accordingly, neutral glycosphingolipids with terminally or internally located saccharide chains of Gal␣1-4Gal or Gal␣1-4Gal␤1-4Glc might be required for binding of pierisin-1. Because oligosaccharides sugars themselves showed no receptor activity, a ceramide moiety appears essential for receptor function.
Our results provided strong evidence that the lack of pierisin-1 receptors such as Gb3 and Gb4 in MEB4 cells is the reason for poor incorporation of pierisin-1 and hence their insensitive phenotype. In addition, among seven mammalian cell lines with different sensitivities to pierisin-1, binding and incorporation correlated with the sensitivity of the cells to the toxic effects of pierisin-1. In fact, we have also confirmed the presence of abundant amounts of Gb3 and Gb4 receptors in pierisin-1-sensitive human gastric carcinoma TMK-1 cells and human breast carcinoma MCF-7 cells, in addition to HeLa cells (data not shown). Thus, the presence of pierisin-1 receptor on the cells is an important factor affecting pierisin-1 sensitivity.
Ricin, a toxic protein found in the castor bean Ricinus communis, is composed of a sugar-binding subunit B, which attaches to receptors on the surfaces of target cells, and a subunit A, which acts as an N-glycosidase inactivating cellular ribosomes (23,24). Ricin binds to both glycoproteins and glycolipids with terminal galactose units and can therefore interact with a large number of different molecules on cell surfaces. The ricin B-chain has two sugar-binding domains, each of which is composed of three copies (␣, ␤, and ␥) of a galactose-binding subdomain of about 40 amino acid residues (21). The most characteristic sequence feature is the presence of a Gln-X-Trp pattern, where X is any amino acid residue. These Trp residues constitute the hydrophobic core of the sugar-binding domains and stabilize the C-terminal hook of each subdomain (21,(25)(26)(27). Abrogation of lectin activity is observed with substitution of tryptophan in the QXW sequence of the ricin B chain (22). The C-terminal region of pierisin-1 is composed of four presumed lectin-like domains, and each includes three subdomains ␣, ␤, and ␥. The QXW pattern is conserved partially or completely conserved in all of the 12 subdomains of pierisin-1. Site-directed mutation of C-terminal pierisin-1 by replacement of tryptophan at any conserved QXW sequence in the present study resulted in markedly reduced cytotoxic activity to HeLa cells. These results suggest that the conserved QXW sequence in all of ␣, ␤, and ␥ subdomains in each domain of C-terminal region of pierisin-1 might have an important structural role, and all four bind to receptors. This would ensure efficient incorporation of pierisin-1 into cells. Thus, structure of the C-terminal region of pierisin-1 and its binding to receptors may resemble those of the ricin B-chain.
Pierisin-1 is present naturally in the cabbage butterfly. Because levels of the protein increases in the fifth instar larvae (6), elimination of cells in larval tissues by expression of a receptor is plausible. However, the glycolipid composition in insects, including the cabbage butterfly, has not yet been fully elucidated. Analyses of several Dipteran insect species, such as the fruit fly, and other invertebrates such as the Biwa pearly mussel, suggest that invertebrates possess characteristic man-nose-containing glycolipids as the core sequence of Man␤1-4Glc␤1-1ЈCer (28 -31). Any receptor activity of these particular glycolipids for pierisin-1 and whether they are present in the cabbage butterfly are unknown. It should be noted, however, that globo-series glycosphingolipids, such as Gb3, are relatively primitive glycolipids. Hence, it is possible that they might be found in insects. Alternatively, invading organisms with pierisin-1 receptors might be the natural targets of pierisin-1. It is also interesting to know the origin of pierisin-1, a unique guanine ADP-ribosyltransferase combined with lectin domain of the ricin B-chain. A full understanding of the evolutional aspect of pierisin-1 as well as determination of the presence and distribution of its receptors in insects and possible target bio-organisms are important issues for the characterization of the biological roles of pierisin-1.