Shiga Toxin Binds Human Platelets via Globotriaosylceramide (P^k Antigen) and a Novel Platelet Glycosphingolipid

Abstract

Hemolytic-uremic syndrome is a clinical syndrome characterized by acute renal failure, microangiopathic hemolytic anemia, and thrombocytopenia that often follows infection by Shiga toxin- or verotoxin-producing strains of Escherichia coli. Because thrombocytopenia and platelet activation are hallmark features of hemolytic-uremic syndrome, we examined the ability of Shiga toxin to bind platelets by flow cytometry and high-performance thin-layer chromatography (HPTLC) of isolated platelet glycosphingolipids. By HPTLC, Shiga toxin was shown to bind globotriaosylceramide (Gb₃) and a minor platelet glycolipid with an R_f of 0.03, band 0.03. In a survey of 20 human tissues, band 0.03 was identified only in platelets. In individuals, band 0.03 was expressed by 20% of donors and was specifically associated with increased platelet Gb₃ expression. Based on glycosidase digestion and epitope mapping, band 0.03 was hypothesized to represent a novel glycosphingolipid, IV³-β-Galα1-4galactosylglobotetraosylceramide. Based on incidence, structure, and association with increased Gb₃ expression, band 0.03 may represent the antithetical Luke blood group antigen. By flow cytometry, Shiga toxin bound human platelets, although the amount of Shiga toxin bound varied in donors. Differences in Shiga toxin binding to platelet membranes did not reflect differences in platelet Gb₃ expression. In contrast, there was a loose association between Shiga toxin binding and decreasing forward scatter, suggesting that Shiga toxin and verotoxins bind more efficiently to smaller, older platelets. In summary, Shiga and Shiga-like toxins may bind platelets via specific glycosphingolipid receptors. Such binding may contribute to the thrombocytopenia, platelet activation, and microthrombus formation observed in hemolytic-uremic syndrome.

HUS is a clinical syndrome characterized by acute renal failure, thrombocytopenia, and a microangiopathic hemolytic anemia (70). Although HUS can occur at any age, HUS is particularly common among young children, usually in association with a prodromal diarrheal illness (10). In addition to being the most common cause of acute renal failure in children, HUS is also a cause of chronic renal insufficiency and chronic renal failure in 15 and 5% of affected children, respectively (reference 10 and references therein).

Although multiple etiologies have been implicated in the pathogenesis of HUS (70), there is increasing evidence that most cases of community-acquired or idiopathic HUS may follow infection by certain strains of enterohemorrhagic Escherichia coli or VTEC (3, 10, 21). A commensal organism in the digestive tract of some cattle (42), VTEC can contaminate beef products during commercial meat processing. As a result, there have been several large outbreaks of VTEC-associated illnesses after ingestion of undercooked contaminated beef (22). In one recent outbreak in the Pacific Northwest, more than 700 cases of E. coli-related illnesses, including 55 cases of HUS and 4 deaths, were reported (22). Although the mortality rate from VTEC infection in the latter outbreak was less than 1%, mortality rates as high as 35% have been reported (21). Overall, E. coli is the fourth most common food-related illness in the United States, with an estimated incidence of 10,000 to 20,000 cases per year (3). Costs of enterohemorrhagic E. coli-associated illness, due to medical treatment and lost productivity, are reported to range from 216 and 580 million dollars annually (3). As a result of the threat and cost to public health, there is increasing interest by government agencies and food industry officials in the pathogenesis, treatment, and prevention of VTEC-associated illness.

Previous studies over the last two decades have shown that the causative agent of both E. coli-associated gastroenteritis and HUS is a bacterial toxin (73) known as verotoxin or Shiga-like toxin. Like Stx from Shigella dysenteriae, verotoxin is a multimeric toxin consisting of five identical B subunits (MW, 6,500) and a single, enzymatically active A subunit (MW, 32,000) (73). During infection, Stx and verotoxin bind to cell surface receptors via the toxin B subunit (49), followed by endocytosis of the toxin (81) and proteolytic cleavage of the A subunit. The latter then binds and inactivates rRNA, leading to inhibition of protein synthesis and apoptosis (48, 73).

The receptors for Stx and most verotoxins (Stx1 and Stx2) were previously identified as cell surface GSLs which display a terminal Galα1-4Gal or galabiosyl epitope (15). To date, three GSLs which can potentially serve as a receptor for Stx (Stx1 and Stx2) have been identified (15). These GSLs include galabiosylceramide, a gala family ceramide disaccharide, Gb₃ or the P^k antigen, and the P₁ antigen, a neolacto series pentaosylceramide (Table 1). Stx2e, the causative agent of pig edema disease, recognizes the P blood group antigen (20). Tissue-specific differences in the distribution and relative expression of galabiosylceramide, Gb₃, and P₁ antigens are hypothesized to play a role in the clinical spectrum of illness seen with S. dysenteriae and VTEC-related infections (49, 55, 100). The cells and tissues previously shown to bind verotoxins and Stx via cell surface GSLs include human erythrocytes (8, 49), endothelium (54, 62, 74, 91), and kidney (11).

TABLE 1.

Known GSL receptors for Stx and Shiga-like toxins^a

GSL family	Name (symbol)	GSL structureb
gala	Galabiosylceramide	Galα1→4Galβ1→1′Cer
globo	Globotriaosylceramide (Gb3, CTH, CD77, Pk)	Galα1→4Galβ1→4Glcβ1→1′Cer
neolacto	IV4-α-Galactosylneolactotetraglycosylceramide (P1)	Galα1→4Galβ1→4GlcNAcβ1→3Galβ1→4Glcβ1→1′Cer
globoc	Globotetraosylceramide (Gb4,P)	GalNAcβ1→3Galα1→4Galβ1→4Glcβ1→1′Cer

^aGSL classification, trivial names, and structures as recommended by the International Union of Pure and Applied Chemistry-International Union of Biology Commission on Biochemical Nomenclature (47). 

^bUnderlined oligosaccharides represent the critical terminal Galα1-4Gal disaccharide required for Stx, Stx1, and Stx2 recognition and binding. 

^cReceptor for Stx2e. 

Stx may also potentially bind human platelets (58). Multiple studies have provided laboratory evidence of widespread platelet activation in patients with HUS (4, 25, 36) including histologic evidence of platelet microthrombi occluding small renal vessels (70, 78). In vitro, bacterial supernatants containing verotoxin were shown to potentiate platelet aggregation (80). Furthermore, Gb₃ was reported to be a major platelet GSL antigen (28, 57, 87). To investigate whether platelets can bind Stx, we examined Stx binding to platelets by two-color immunofluorescence flow cytometry. We also examined Stx binding to platelet neutral GSLs isolated from pooled platelet concentrates and from individual platelet donors.

(Part of this work has been reported as a preliminary communication [28a].)

MATERIALS AND METHODS

Abbreviations.

BSA, bovine serum albumin; C, chloroform; CDH, lactosylceramide (Galβ1-4Glcβ1-1Cer); Cer, ceramide; FITC, fluorescein isothiocyanate; CV, coefficient of variation; DPA, diphenylamine; Fuc, fucose; Gal, galactose; GalNAc, N-acetylgalactosamine; Gb₃, globotriaosylceramide or P^k blood group antigen (Galα1-4Galβ1-4Glcβ1-1Cer); Gb₄, globotetraosyceramide or P blood group antigen (GalNAcβ1-3Galα1-4Galβ1-4Glcβ1-1Cer); Glc, glucose; GlcNAc, N-acetylglucosamine; GSL(s), glycosphingolipid(s); HPA, Helix pomatia lectin; HPLC, high-performance liquid chromatography; HPTLC, high-performance thin-layer chromatography; HUS, hemolytic-uremic syndrome; IgM, immunoglobulin M; Le, Lewis; Le^a, Lewis^a (Galβ1-3[Fucα1-4]GlcNAc-R); Le^b, Lewis^b (Fucα1-2Galβ1-3[Fucα1-4]GlcNAc-R); lyso-Gb₃, lysoglobotriaosylceramide or globotriaosylsphingosine; M, methanol; MAb, monoclonal antibody; MW, molecular weight; PBS, phosphate-buffered saline; PE, phycoerythrin; R, carbohydrate residue; R_f, relative mobility; SSEA-3, stage-specific embryonic antigen (Galβ1-3GalNAcβ1-3Galα1-4Galβ1-4Glcβ1-1Cer); Stx, Shiga holotoxin; Stx-B, Shiga toxin B subunits; VTEC, verotoxin-producing Escherichia coli.

Tissue preparation.

Outdated (72- to 96-h-old), whole-blood-derived, nonapheresis platelet concentrates (1) from volunteer blood donors were purchased from Siouxland Red Cross (Sioux City, Iowa). Outdated, single-donor, apheresis platelet concentrates, erythrocytes (blood group AB+, P₁+, Le^b+ and B+, P₁−, Le^a+, Le^b−), lymphocyte, monoblast, and granulocyte concentrates were obtained from the DeGowin Blood Center (Department of Pathology, University of Iowa, Iowa City) as previously described (28, 30). From 13 apheresis platelet donors, an erythrocyte sample was also collected for P₁ typing and erythrocyte GSL analysis. Human kidney, aorta, intestine, heart, lung, liver, brain, and cauda equina were obtained from the hospital autopsy service (Robert Schelper and Van Savell, Department of Pathology). Human synovium was obtained from patients undergoing prosthetic joint replacement and were a kind gift from Stanley Naides, John Callahan, and C. Peradones (Departments of Internal Medicine and Orthopedic Surgery, University of Iowa). Human bronchial epithelial cells were obtained at bronchoscopy from paid volunteers and were supplied as isolated neutral GSLs from Doug Hornick, Department of Internal Medicine. Rabbit erythrocyte GSLs were a gift from Jim Greer (Department of Pathology). Signed consent was obtained from apheresis platelet donors at the time of donation. All tissue procurement was in accordance with the institutional human investigation review committee.

All cells and tissues except platelets, granulocytes, lymphocytes, monoblasts, and plasma were washed extensively with isotonic PBS (pH 7.4) prior to lipid extraction. Aortic endothelium was isolated by scraping the luminal surface of 15 thoracic aortas with a glass microscope slide. The aortas were then dissected free of serosa to obtain aortic smooth muscle. Intestinal smooth muscle and large and small bowel mucosa were isolated similarly. Platelets were isolated as previously described (28, 30, 57) by initially centrifuging each whole-blood-derived platelet concentrate at 350 × g for 15 min to remove contaminating leukocytes and erythrocytes. The resulting platelet-rich supernatants were then carefully decanted and centrifuged at 4,500 × g for 30 min to obtain a platelet pellet. Erythrocyte and leukocyte contamination of the platelet pellet was less than 0.01% each (29, 30). The platelet pellet was then washed twice with isotonic ammonium bicarbonate (154 mM) to osmotically lyse residual erythrocytes, frozen, and lyophilized dry. Granulocytes, lymphocytes, and monoblasts were initially diluted with 2 volumes of isotonic ammonium bicarbonate, centrifuged (4,500 × g, 30 min) to obtain a cell pellet, washed twice more with ammonium bicarbonate, frozen, and lyophilized. Platelet-poor plasma obtained from the isolation of platelets was also frozen and lyophilized dry prior to extraction.

GSL isolation.

The total neutral GSL fraction of each tissue or cell type was isolated by the procedure of Ledeen and Yu (59) for all tissues and cells, except platelets, erythrocytes, granulocytes, monoblasts, lymphocytes, and plasma, which were isolated by a modified procedure (30, 57). For single-donor apheresis platelets, single-donor whole-blood-derived platelets and individual erythrocyte samples (2 ml of packed erythrocytes), washed and lyophilized cell pellets were extracted in 100 ml of C-M (1:1 [vol/vol]) for 48 h. The total lipid extract was dried, resuspended in 200 ml of C-M-water (30:60:8 [vol/vol]), and applied to a DEAE-Sephadex column (A25, 10-ml bed volume; Sigma, St. Louis, Mo.). Neutral lipids were eluted with 200 ml of C-M-water (30:60:8) [vol/vol]), evaporated in vacuo, saponified with 35 ml of 0.3 N methanolic sodium hydroxide, and dialyzed against distilled water (MW cutoff, 14,000; Spectra-Por, Houston, Tex.). The dialysis retentate was lyophilized dry, resuspended in 45 ml of C and applied to a silicic acid column (40-μm-diameter, 10-ml bed volume; J. T. Baker, Phillipsburg, N.J.). The column was sequentially washed with 100 ml of C and 50 ml of ethyl acetate to remove cholesterols and steroids, respectively. Neutral GSLs were then isolated by batch elution with 50 ml each of acetone-methanol (9:1 and 7:3 [vol/vol]). The last two fractions were pooled, dried, and resuspended in C-M (1:1 [vol/vol]) prior to use.

Preparation of lyso-Gb₃.

Lyso-Gb₃ was prepared from Gb₃ (Sigma) as described by Schwarzmann and Sandhoff (84). Briefly, 500 μg of Gb₃ was dissolved in 0.8 M potassium hydroxide in dry methanol and incubated for 20 h at 100°C. The solution was cooled to 20°C, neutralized with acetic acid, dried under nitrogen, and dialyzed against distilled water for 6 h at room temperature. The dialysate was lyophilized dry, resuspended in C-M (1:1 [vol/vol]) and examined by HPTLC as described below.

HPTLC.

HPTLC was performed by published procedures (57) using aluminum-backed HPTLC plates (E. Merck, Darmstadt, Germany). Neutral GSLs were spotted onto HPTLC plates and then developed in solvent A (C-M-water, 65:25:4 [vol/vol]) or solvent B (C–M–0.2% aqueous CaCl₂, 55:45:10 [vol/vol]). GSLs were detected by spraying with DPA reagent (Sigma) (59) or by immunostaining as described below. GSL bands were characterized by intensity (percent total staining density) and R_f by scanning densitometry at 370 nm (Schimadzu Instruments, Columbia, Md.). Error in mobility measurements was less than 0.01 unless otherwise stated. The concentrations of neutral GSL spotted per HPTLC lane was 80 to 100 μg for pooled platelets, erythrocytes, lymphocytes, granulocytes, intestine, heart, lung, liver, kidney, synovium, brain, cauda equina, and aorta. For single-donor erythrocyte- or whole-blood-derived, nonapheresis platelet GSL samples (≅5.5 × 10¹⁰ platelets total) (1), 1/5 of each sample was spotted per HPTLC lane. For single-donor, apheresis platelet concentrates (≥3.0 × 10¹¹ platelets total) (1), the total neutral GSL was resuspended in 500 μl of C-M (1:1 [vol/vol]) and between 1/100 and 1/50 of each sample was spotted per HPTLC lane. The amount of GSL spotted was normalized relative to the amount of lactosylceramide in each sample, as determined by scanning densitometry (mean area in square millimeters) of DPA-stained HPTLC plates.

Toxins and immunologic reagents.

Stx (from S. dysenteriae), isolated Stx-B, and rabbit anti-Stx polyclonal antibody were a kind gift from Arthur Donohue-Rolfe, Tufts University, Boston, Mass. (32). MAb 38.13 (anti-Gb₃ or P^k, rat IgM) (72), PE-labeled anti-platelet glycoprotein IIb/IIIa (MAb P2, CD41; mouse IgG) and a biotinylated anti-rat IgM were purchased from Immunotech, Inc., Westbrook, Maine. Monoclonal antibodies Pk002 (anti-P^k and P₁, mouse IgM) and P001 (anti-globo: anti-P^k, P, P₁>Forssman, SSEA-3; mouse IgM) were purchased from Accurate Chemical and Scientific Corp., San Diego, Calif. (14). Monoclonal anti-blood group P or Gb₄ (MAb MC631, mouse IgM) was purchased as a hybridoma supernatant from the Developmental Studies Hybridoma Bank maintained by the Department of Biology at the University of Iowa, Iowa City, under contract no. N01-HD-2-3144 (51, 52). Monoclonal anti-Forssman (MAb M1/22.25.8; TIB-121, mouse IgM) was purchased from the American Type Culture Collection (Rockville, Md.) and used as a hybridoma supernatant (96). Gal-13 (mouse IgM), an anti-Galα1-3Galβ1-4GlcNAc-R MAb (38), was a kind gift from Uri Galili, University of California at San Francisco, San Francisco. SC802, a recombinant P-fimbriated E. coli strain was metabolically labeled with [³⁵S]methionine by the method of Bock et al. (9) and was provided by Steven Clegg and Doug Hornik (Departments of Microbiology and Internal Medicine, University of Iowa). Monoclonal anti-blood group A (Birma-1, mouse IgM) (64), anti-blood group Le^a (LM112/161, mouse IgM) (37), and anti-Le^b (LM119/181) (40) were purchased from Gamma Biologicals, Houston, Tex. Biotinylated anti-mouse IgM, anti-rabbit IgG, avidin-linked alkaline phosphatase (ABC kit) and alkaline phosphatase substrate (SK-5000) were purchased from Vector Laboratories (Burlingame, Calif.).

HPTLC-immunostaining.

HPTLC-immunostaining was performed as previously described (18). Briefly, air-dried, solvent-developed HPTLC plates were dipped in a hexane solution of 0.04% poly(isobutyl)methacrylate (Polysciences, Warrington, Pa.) for 60 s and then air dried. The plates were blocked with buffer A (40 mM Tris-HCl, 150 mM NaCl, 1% BSA, 0.1% sodium azide [pH 7.8]) for 45 min. For Stx overlays, blocked plates were incubated with Stx (final concentration, 50 ng/ml) or isolated Stx-B (final concentration, 0.5 μg/ml) in buffer A for 2 h at room temperature. A toxin-free plate was also included as a negative control. Plates were washed with PBS (pH 7.4) and then incubated with rabbit anti-Stx polyclonal antibody diluted in PBS–1% BSA (pH 7.4) (final concentration, 1:5,000) for 1 h. The plates were washed again with PBS, incubated with biotinylated anti-rabbit IgG in PBS–1% BSA (1 h), washed, and then allowed to react with an avidin-linked alkaline phosphatase (Vector) for 30 min. Bound antibody was detected with an alkaline phosphatase substrate (Vector) in cold (4°C) 100 mM Tris-HCl, pH 9.5.

For immunostaining with anticarbohydrate MAbs, HPTLC plates were incubated with primary antibody diluted in buffer A for 1 h, followed by a biotinylated anti-mouse or anti-rat secondary antibody, alkaline phosphatase, and substrate as described above.

Glycosidase digestion.

Platelet neutral GSLs were digested with α-galactosidase from green coffee beans (EC 3.2.1.22) by the method of Ariga et al. (5). Briefly, approximately 50 μg of total platelet neutral GSL or 15 μg of Gb₃ (Sigma) was suspended in 100 μl of sodium citrate buffer (100 mM, pH 5.0) containing 10% sodium taurodeoxycholate and 20 μl of α-galactosidase (50 U/1.0 ml of 3.2 M ammonium acetate; Boehringer Mannheim, Indianapolis, Ind.). The mixture was incubated for 24 h at 37°C. The reaction was terminated by the addition of 100 μl of C-M (1:1 [vol/vol]) and dried under nitrogen. For analysis, samples were resuspended in C-M (1:1 [vol/vol]), spotted onto HPTLC plates, and examined by DPA and HPTLC-immunostaining with Stx. An enzyme-free, buffer control was also included.

For digestion with glycan ceramidase (60), 50 μg of total platelet neutral GSL or 15 μg of Gb₃ was suspended in 100 μl of sodium acetate buffer (50 mM, pH 5.0) containing sodium taurodeoxycholate (1 mg/ml) and 20 μl of glycan ceramidase (5 U/100 μl; V-Labs, Covington, La.) from the leech Marobdella decora (60) and incubated for 48 h at 37°C. As before, reactions were terminated by the addition of 100 μl of C-M (1:1 [vol/vol]), and the samples were dried under nitrogen and subjected to HPTLC.

Fluorescein labeling of Stx-B.

For flow cytometry, Stx-B were labeled as previously described (8). Briefly, 100 μg of lyophilized Stx-B was resuspended in 100 μl of sodium bicarbonate buffer (0.5 M, pH 9.5) containing 2 mg of FITC (isomer I; Sigma) per ml and incubated for 2 h at 25°C in the dark. Unbound FITC was removed by filter centrifugation (Centricon 30; Amicon, Beverly, Mass.). FITC-labeled Stx-B were resuspended in 1 ml of PBS–1% BSA (pH 7.4), aliquoted, and stored at −20°C until use. The toxin-free filtrate was diluted 1/5 with PBS–1% BSA and saved as a negative control.

Flow cytometry.

Whole-blood-derived, single-donor platelet concentrates from 16 blood group A and two blood group O donors were leukodepleted by differential centrifugation (350 × g, 15 min), and the resulting platelet-rich supernatant was counted on a Coulter counter (model Z; Coulter, Hialeah, Fla.). For flow cytometry, one million platelets were incubated with 5 μl of PE-labeled anti-platelet glycoprotein IIb/IIIa (MAb P2; CD41), 20 μl of FITC-labeled Stx-B (100 μg/ml stock), and PBS buffer (PBS with 1% BSA, 0.2% EDTA, and 0.1% sodium azide [pH 7.4]) in a final volume of 100 μl for 2 h at 4°C. A toxin-free FITC filtrate was also included as a negative control. For some samples, platelets were stained in parallel with MAb P2 and 5 μl of FITC-labeled HPA (10 μg/ml working dilution; Sigma). Platelets were washed twice with PBS buffer (1,000 × g, 10 min) and resuspended in 400 μl of PBS–1% paraformaldehyde. All samples were performed in duplicate and analyzed on a Becton Dickinson 440 flow cytometer equipped with an argon laser. A minimum of 10,000 events were measured per sample. FITC and PE spectral overlaps were corrected with electronic compensation. Data was collected on a VAX station 3200 computer equipped with DESK software (kindly supplied by Wayne Moore, Stanford University). For histograms, platelets were gated on forward scatter, orthogonal scatter, and MAb P2 positivity and the percent platelets positive for Stx-B or HPA was determined. Results were recorded as the mean (n = 2) percent platelets positive for CD41 and Stx-B. Final graphic output was performed with Canvas software for the MacIntosh computer.

Serology.

In apheresis platelet donors, an erythrocyte sample (5 ml) was collected for erythrocyte phenotyping. To type donors for the P₁ and P₂ phenotypes, a drop of 2% erythrocyte suspension was incubated with 2 drops of human polyclonal anti-P₁ antisera (Gamma Biologics) for 15 min at 4°C according to the manufacturer’s instructions. Donors with erythrocytes which agglutinated with anti-P₁ were considered P₁ positive. Donors with erythrocytes which failed to agglutinate were considered to be of the P₂ phenotype. The latter was confirmed by isolating the erythrocyte neutral GSLs from P₂ donors and demonstrating the presence of both P^k and P antigens (67).

Statistics.

Platelet GSL expression was compared by using a two-tailed Student t test, Mann-Whitney U-test, Spearman rank sum, and Pearson product-moment correlation (17). A P of <0.05 was considered statistically significant.

RESULTS

Stx binds Gb₃ and a novel GSL in human platelets.

Gb₃ or the P^k blood group antigen, the physiologic receptor for Stx on human umbilical vein endothelium (62, 74, 75, 92), rabbit intestinal mucosa (49), Vero (61), HeLa (49), MDCK (82), and Daudi cells (26), was previously reported to be present on human platelets (28, 57, 87). To confirm the latter, we isolated the total neutral GSL fraction from pooled, type-specific platelet concentrates from more than 100 whole-blood donors. Following isolation, the total platelet neutral GSL fraction was separated by HPTLC and chemically stained with DPA reagent, which nonspecifically reacts with the oligosaccharide moiety on GSLs (59).

Consistent with previous reports (28, 57, 87), platelets expressed three major DPA-positive neutral GSLs with R_fs of 0.40, 0.25, and 0.16, respectively (Fig. 1, lane 1 [solvent A]). Based on their mobilities relative to those of authentic GSL standards by HPLC (data not shown) and HPTLC, platelet DPA bands 0.40, 0.25, and 0.16 were identified as lactosylceramide (CDH), Gb₃, and Gb₄ or P antigen, respectively (Table 2). Their identities were subsequently confirmed by one- and two-dimensional nuclear magnetic resonance of the isolated GSLs (29). Together, these three GSLs comprised more than 90% of the total neutral GSL expressed by platelets. Based on scanning densitometry of DPA-stained HPTLC plates, the relative proportions of CDH, Gb₃, and Gb₄ were 45.0% ± 3.0%, 20.4% ± 1.4%, and 24.9% ± 1.7% (mean percent total neutral GSL ± standard deviation, n = 4), respectively. A comparison of the neutral fractions from four different platelet GSL preparations (blood group A, n = 2; blood group O, n = 2) showed no significant differences in the distribution of CDH, Gb₃, and Gb₄ for different platelet preparations or donor ABO type (data not shown).

FIG. 1.

Stx binding to platelet neutral GSLs. The total neutral GSL fraction of human platelets was spotted onto HPTLC plates (80 μg per HPTLC lane). GSLs were visualized chemically with DPA (lane 1) or by HPTLC-immunostaining with Stx (lane 2), Stx-B (lane 3), or toxin-free, buffer control (lane 4). Major DPA and Stx-positive GSL bands were characterized by their mobility relative to that of the solvent front. R_fs are shown to the right of the gel. GSLs were separated in solvent A. OR, origin.

TABLE 2.

Mobility of band 0.03 relative to those of other platelet GSL antigens

GSL namea	No. of glycan residues	Mobility (Rf)b		GSL structure
		Solvent A	Solvent B
GlcCer	1	0.59	0.85	Glcβ1→1′Cer
LacCer	2	0.40	0.75	Galβ1→4Glcβ1→1′Cer
Gb3	3	0.25	0.65	Galα1→4Galβ1→4Glcβ1→1′Cer
Gb4	4	0.16	0.55	GalNAcβ1→3Galα1→4Galβ1→4Glcβ1→1′Cer
Lea-5	5	0.06	0.46	Galβ1→3(Fucα1→4)GlcNAcβ1→3Galβ1→4Glcβ1→1′Cer
A-6	6	0.05	0.48	GalNAcα1→3(Fucα1→2)Galβ1→3/4GlcNAcβ1→3Galβ1→4Glcβ1→1′Cer
Leb-6	7	0.04	0.40	Fucα1→2Galβ1→3(Fucα1→4)GlcNAcβ1→3Galβ1→4Glcβ1→1′Cer
A-8	8	0.03	0.35	GalNAcα1→3(Fucα1→2)Galβ1→(4GlcNAcβ1→3Galβ1→)24Glcβ1→1′Cer
Lea-7	7	0.01	0.24	Galβ1→3(Fucα1→4)GlcNAcβ1→3Galβ1→4GlcNAcβ1→3Galβ1→4Glcβ1→1′Cer
Leb-8?c	8	NA	0.14	Fucα1→2Galβ1→3(Fucα1→4)GlcNAcβ1→3Galβ1→4GlcNAcβ1→3Galβ1→4Glcβ1→1′Cer
Lea-9?c	9	NA	0.08	Galβ1→3(Fucα1→4)GlcNAcβ1→3(Galβ1→4GlcNAcβ1→3)2Galβ1→4Glcβ1→1′Cer
Band 0.03	6–8	0.03	0.19	Galα1→4Gal-(unknown oligosaccharide)-Galβ1→4Glcβ1→1′Cer

^aBy using the conventions of Henry et al. (44), the names of ABO and Le GSLs were abbreviated by listing the activity of the GSL (ABO or Le) followed by the size or total number of carbohydrate residues in the oligosaccharide moiety. 

^bR_fs (mean ± 0.01) of GSL bands after separation in solvent A (C-M-water, 65:25:4 [vol/vol]) or solvent B (C–M–aqueous 0.2% CaCl₂, 55:45:10 [vol/vol]). NA, not applicable. 

^cThe GSL bands hypothesized to reflect a possible Le^a-8 and Le^b-9 GSL structure were observed only after separation in solvent B. 

To determine whether Stx would bind Gb₃ in human platelets, platelet neutral GSLs were immunostained with Stx on HPTLC plates (Fig. 1, lane 2). As expected, Stx bound a neutral GSL (R_f, 0.25), which is consistent with Gb₃ or the P^k blood group antigen in platelets. Stx did not bind either Gb₄ or CDH, which is consistent with previous results (61). Interestingly, Stx also bound a slow-moving, unknown, neutral GSL with an R_f of 0.03 (band 0.03). A comparison of band 0.03 with the DPA control showed that band 0.03 was a minor platelet GSL, representing less than 1% of the total neutral GSL present. Gb₃ and band 0.03 also bound isolated Stx-B (lane 3). No binding was observed with a toxin-free, buffer control (lane 4).

Band 0.03 is a platelet-specific glycolipid.

To determine whether Stx band 0.03 was specific for human platelets or was expressed by other human tissues, we screened the neutral GSL fraction from 20 human tissues for GSLs capable of binding Stx. Tissues of interest included hematopoietic cells as well as several tissues potentially affected in HUS and thrombotic thrombocytopenia purpura. On HPTLC-immunostaining, Stx recognized four GSL species with R_fs of 0.40, 0.25, 0.06, and 0.03 (Table 3). Stx band 0.25 was identified in the vast majority (17 of 20) of tissues screened, whereas Stx bands 0.40, 0.06, and 0.03 showed marked tissue restriction: Stx band 0.40 was identified only in kidney and small bowel mucosa, Stx band 0.06 was specifically expressed in P₁ (but not P₂) erythrocytes, and (Stx) band 0.03 was identified only in the GSL fraction of human platelets. Band 0.03, therefore, appeared to represent a potentially platelet-specific GSL.

TABLE 3.

Distribution of Stx-positive GSLs in human cells and tissues

Cell or tissue	Stx-positive GSL band with Rf ofa:
	0.40	0.25	0.06	0.03
Platelets	−	+	−	+
Erythrocytes (P1+)	−	+	+	−
Erythrocytes (P1−)	−	+	−	−
Lymphocytes	−	+	−	−
Granulocytes	−	−	−	−
Monoblasts	−	+	−	−
Plasma	−	+	−	−
Kidney	+	+	−	−
Heart	−	+	−	−
Lung	−	+	−	−
Liver	−	+	−	−
Synovium	−	+	−	−
Aortic endothelium	−	+	−	−
Tracheal epithelium	−	+	−	−
Small intestinal mucosa	+	+	−	−
Large intestinal mucosa	−	+	−	−
Gastrointestinal smooth muscle	−	+	−	−
Aortic smooth muscle	−	+	−	−
Brain	−	−	−	−
Cauda equina	−	−	−	−

^aRelative mobility (R_f) of Stx-positive GSLs in solvent A (C-M-water, 65:25:4 [vol/vol]). Symbols: +, band identified; −, band not identified. 

Band 0.03 is a novel Stx-binding GSL, distinct from galabiosylceramide, Gb₃, and P₁ antigens.

The ability of Stx bands 0.40, 0.25, 0.06, and 0.03 to bind Stx suggested that these GSLs, by definition, possessed a terminal Galα1-4Gal-R or galabiosyl epitope. In order to identify and compare these GSLs, the neutral GSL fractions from platelets (Stx bands 0.25 and 0.03), kidney (Stx bands 0.40 and 0.25), and P₁ erythrocytes (Stx bands 0.25 and 0.06) were immunostained with several anti-globo MAbs and lectins with overlapping epitopes (Table 4). Additional controls included rabbit erythrocytes and commercially available Gb₃, Gb₄, and Forssman antigen GSL standards.

TABLE 4.

Immunologic characterization of Stx-positive GSLs

Reagent	Epitope	Stx-positive GSL banda with Rf of:
		0.40	0.25	0.06	0.03
Stx	Galα1→4Gal-R	+	+	+	+
E. coli SC802	R-Galα1→4Gal-R	+	+	+	+
MAb P001	(R)-Galα1→4Gal-R	−	+	+	+
MAb 38.13	Galα1→4Galβ1→4Glc-R	−	+	−	−
MAb Pk002	Galα1→4Galβ1→4Glc/GlcNAc-R	−	+	+	−
MAbMC631	(R)-GalNAcβ1→3Galα1→4Gal-R	−	−	−	+
MAb Gal-13	Galα1→3Galβ1-4GlcNAc-R	−	−	−	−
MAb M1/22.28.8	GalNAcα1→3GalNAcβ1→3Galα1→4Galβ1→4Glc-R	−	−	−	−

^aSymbols: +, band identified; −, band not identified. 

As shown in Table 4, there were marked differences in the activity of Stx bands 0.40, 0.25, 0.16, and 0.03 with different anti-globo reagents. By definition, all four GSL bands bound Stx. All four Stx bands were also positive with SC802, a recombinant P-fimbriated E. coli strain which recognizes the galabiose ([R]-Galα1-4Gal-R) disaccharide (9). With the single exception of band 0.40, most Stx-positive GSL bands were also positive with MAb P001. Like E. coli SC802, MAb P001 recognizes most globo series GSLs through recognition of terminal and subterminal galabiosyl motifs.

We also tested the activities of two anti-P^k MAbs, 38.13 and Pk002, against our four Stx GSLs. Like Stx, MAbs 38.13 (72) and Pk002 (14) recognize oligosaccharides expressing a terminal Galα1-4Gal-R epitope. In contrast to Stx, however, MAbs 38.13 and Pk002 recognize a trisaccharide epitope so that MAb recognition is also dependent on the identity and anomeric linkage of the adjacent upstream carbohydrate residue (R). In the case of MAb 38.13, R must be a β1-4Glc or the Gb₃ epitope. For MAb Pk002, R may be either a β1-4Glc (Gb₃) or β1-4GlcNAc (P₁). As shown in Table 4 and Fig. 2, MAb 38.13 recognized only band 0.25, whereas MAb Pk002 recognized Stx bands 0.25 and 0.06 in the neutral GSL fraction of P₁ erythrocytes (Fig. 2, lane 2). Neither MAb recognized band 0.03 in human platelets (Fig. 2, lane 1). Band 0.03 also failed to bind MAb Gal-13 which recognizes a terminal isoglobo-like epitope (38). Therefore, band 0.03 appeared to be a Galα1-4Gal-R-Cer, where R is an oligosaccharide not beginning in Glc or GlcNAc.

FIG. 2.

GSL specificity of Stx, MAb Pk002, and MAb 38.13. The total neutral GSL fraction from pooled human platelets (lane 1) and a P₁+ erythrocyte control (lane 2) were spotted onto HPTLC plates and immunostained with Stx (ST), MAb Pk002 (P_K002) (anti-P^k and anti-P₁), or MAb 38.13 (anti-P^k only). Numbers to the left are the R_fs of the MAb or Stx-positive GSL bands. The apparent binding of Stx to Gb₄ and MAb 38.13 to CDH and GlcCer (near top of plate) represents nonspecific binding. GSLs were separated in solvent A.

On the basis of immunologic activity, mobility and tissue distribution, we were able to characterize and identify three of four Stx-positive GSLs. In summary, Stx band 0.40 was a dihexosylceramide possessing a terminal galabiosyl epitope based on the following: (i) mobility similar to that of lactosylceramide, another ceramide disaccharide; (ii) positivity with SC802; and (iii) positivity with Stx. Stx band 0.40, therefore, must be galabiosylceramide. Galabiosylceramide has been isolated from human kidney (63) and was recently identified as the receptor for Staphylococcus aureus enterotoxin B in human kidney and small bowel (23, 24). As shown in Table 3, this was consistent with the distribution of Stx band 0.40 in our study.

Similarly, Stx bands 0.25 and 0.06 were identified as Gb₃ and P₁ antigen, respectively. Stx band 0.25 was (i) a trihexosylceramide based on its mobility relative to those of CDH, Gb₃, and Gb₄ standards; (ii) positive with SC802 and MAb P001, confirming the presence of at least one galabiosyl motif; (iii) positive with Stx, consistent with a terminal galabiose epitope; and (iv) positive with two anti-P^k MAbs, 38.13 and Pk002. Stx band 0.25, therefore, was a trihexosylceramide in which the oligosaccharide was a Galα1-4Galβ1-4Glc or Gb₃ antigen. Likewise, band 0.06 was (i) positive with SC802 and MAb P001; (ii) positive with Stx; (iii) positive with MAb Pk002; (iv) negative with MAb 38.13; (v) had a mobility greater than Gb₄, a tetraosylceramide (Table 2); and (vi) was expressed by P₁+ but not P₁− erythrocytes (Table 3). Stx band 0.06 was, at minimum, a ceramide pentasaccharide in which the oligosaccharide terminated in a Galα1-4Galβ1-4GlcNAc-R trisaccharide or the P₁ antigen (Table 1).

Band 0.03, on the other hand, appeared to be a novel Stx-binding GSL, immunologically distinct from galabiosylceramide, Gb₃, and P₁ antigen. Band 0.03 had the following characteristics: (i) positive with MAb P001, SC802, and Stx; (ii) negative with MAbs Pk002 and 38.13; (iii) a slower mobility than that of band 0.06 or the P₁ antigen; and (iv) weakly reactive with MAb MC631, which recognizes GSLs with a globoside or type 4 oligosaccharide core (Table 4) (51, 52).

Band 0.03 is not lyso-Gb₃.

Although it appeared that band 0.03 was unrelated to Gb₃, we could not exclude the possibility that band 0.03 was a lyso-Gb₃: Lyso-Gb₃ has a much slower mobility by HPTLC (R_f 0.04) than Gb₃ has (R_f, 0.25) due to the loss of an N-acyl fatty acid moiety. To determine if band 0.03 was lyso-Gb₃, lyso-Gb₃ was prepared from Gb₃ as previously described (84) and then immunostained with MAb 38.13. Unlike band 0.03, Gb₃ and lyso-Gb₃ were recognized by MAb 38.13 (data not shown).

Glycosidase digestion of band 0.03.

To help confirm that band 0.03 was a Galα1-4Gal-R-Ceramide, the total neutral platelet GSL fraction was incubated with α-galactosidase, an exoglycosidase specific for terminal α-linked galactose residues (5, 31). As shown in Fig. 3, digestion of Gb₃ with α-galactosidase (lane 2) resulted in a 77% conversion of Gb₃ to CDH on DPA (Fig. 3A), with a concomitant loss of Stx binding by HPTLC-immunostaining (Fig. 3B) compared to the buffer control (lane 1). Similarly, digestion of platelet neutral GSL with α-galactosidase resulted in a complete loss of Stx binding to band 0.03 (Fig. 3B, lane 4).

FIG. 3.

Sensitivity of band 0.03 to α-galactosidase. The total platelet neutral GSL fraction (lanes 3 and 4) and a Gb₃ standard (lanes 1 and 2) were incubated with buffer only (lanes 1 and 3) or α-galactosidase (lanes 2 and 4). For analysis, GSLs were separated on HPTLC plates (solvent A) and visualized with DPA (A) or by HPTLC-immunostaining with Stx (B).

The resistance of band 0.03 to base hydrolysis strongly suggested that band 0.03 was a GSL (59). To confirm this, the total platelet neutral GSL was digested with glycan ceramidase, a glycosphingolipid hydrolase which recognizes the Glcβ1-1′Cer linkage characteristic of most GSLs (60). As before, band 0.03 failed to bind Stx after glycan ceramidase digestion of platelet neutral GSLs (data not shown).

Band 0.03 is a hexa- to octaosylceramide.

Because band 0.03 appeared to be at least a hexosylceramide based on its mobility relative to that of the P₁ antigen, we compared the mobility of band 0.03 in two different solvents relative to other blood group-active GSL antigens present in the neutral GSL extract of platelets (Table 2) (29, 46). Based on R_f alone, band 0.03 appeared to be a hexa- to octaosylceramide. Because of the generally greater mobility of fucosylated GSLs (12), the lower mobility of band 0.03 relative to blood group A (A-6) and Lewis GSLs (Le^b-6) in solvent B may overestimate the size of the oligosaccharide moiety on band 0.03.

Gb₃ and band 0.03 expression by individual platelet donors.

In an effort to determine whether Gb₃ and/or band 0.03 expression differed in individuals, we examined the neutral GSL extracts isolated from 25 individual, single-donor apheresis platelet donors. As shown in Table 5, the donors ranged from 19 to 56 years of age and included nearly equal numbers of male (n = 14) and female (n = 11) donors. With the exception of blood groups O and B, the distribution of ABO types among our sample of platelet donors was similar to the expected distribution of ABO phenotypes among Caucasian donors (2). As before, the total platelet neutral GSL fraction from each individual donor apheresis platelet concentrate was isolated and examined by scanning densitometry of DPA-stained HPTLC plates (Table 5).

TABLE 5.

Expression of Gb₃ and band 0.03 by individual platelet donors

Donor	Donor characteristics				% Platelet neutral GSLa			Ratioa,b		Stx-positive bands with Rf ofc:
	ABO/Rhd	P1d	Sexe	Age (yr)	CDH	Gb3	Gb4	Gb3/CDH	Gb3/Gb4	0.03	0.25
Individual donors
1	A+		F	27	19.2	41.5	31.3	2.15	1.32	+	+
2	A+	+	M	29	23.8	38.6	33.0	1.63	1.17		+
3	O−	+	M	36	67.6	8.7	9.2	0.10	0.94		−
4	A−	+	M	44	29.5	29.8	29.0	1.01	1.03	+	+
5	A+		F	21	55.3	14.6	20.1	0.26	0.73		+
6	A−		M	22	54.3	10.6	23.2	0.19	0.46		+
7	A+		F	39	25.9	32.6	26.4	1.26	1.23	+	+
8	A+	+	M	46	78.2	1.5	6.8	0.02	0.22		+
9	O+		M	28	44.6	17.2	27.1	0.39	0.63		+
10	O+		F	56	43.4	19.8	27.7	0.47	0.71		+
11	O+		F	47	48.7	14.5	26.0	0.30	0.55		+
12	A+		F	42	59.2	17.6	20.5	0.30	0.86		+
13	O−		F	23	25.5	28.8	34.0	1.14	0.85	+	+
14	O+	−	M	35	20.1	31.6	35.8	1.59	0.88		+
15	B−		F	22	37.4	24.4	29.3	0.62	0.83		+
16	B−	+	M	32	47.9	18.0	23.7	0.37	0.76		+
17	B+	+	M	40	36.4	21.4	22.6	0.59	0.95		+
18	AB−	−	F	30	21.8	35.6	33.4	1.62	1.06	+	+
19	A+	+	F	42	59.7	22.3	14.9	0.37	1.50		+
20	A+	+	M	39	60.0	15.5	8.1	0.26	1.91		+
21	O−	+	M	36	52.9	19.8	12.8	0.37	1.55		+
22	O+	+	M	25	25.8	5.8	10.3	0.23	0.56		+
23	O+	+	M	36	19.5	37.2	38.3	1.90	0.09		+
24	B−		F	19	33.9	27.4	28.7	0.81	0.95		+
25	B+	+	M	41	22.6	39.0	30.8	1.72	1.27		+
Groups of donorsf
All donors (n = 25)					40.5 ± 17.2	22.9 ± 11.0	24.1 ± 9.2
Band 0.03-positive donors (n = 5)					24.4 ± 4.0	33.7 ± 5.2	30.8 ± 3.1	1.44 ± 0.5	1.09 ± 0.2
Band 0.03-negative donors (n = 20)					44.6 ± 16.8	20.3 ± 11.0	22.4 ± 9.4	0.63 ± 0.6	0.91 ± 0.4
Pg					<0.02	<0.02	>0.05	<0.01	>0.10

^aScanning densitometry of DPA-stained HPTLC plates (n = 2). 

^bRatio of area (in square millimeters). 

^cSymbols: +, band identified; −, band not identified. 

^dDonor erythrocyte phenotype. 

^eF, female; M, male. 

^fValues shown for groups of donors are the means ± standard deviations. 

^gStudent t test. P values in blood that were also found to be statistically significant (P < 0.05) by the Mann-Whitney U-test (17) are shown in bold type. 

Unlike platelet GSLs isolated from large pools of donors, individual donors exhibited marked heterogeneity in the relative distribution of the three major platelet neutral GSLs. Gb₃ varied the most, ranging from 1.5 to 41.5% of the total platelet neutral GSL, a nearly 28-fold difference. The difference was also reflected in a large CV (48%). CDH and Gb₄ expression was likewise diverse (CDH range, 19.2 to 78.2%, CV = 42%; Gb₄ range 6.8 to 38.3%, CV = 38%). There was no correlation between platelet GSL expression and donor sex, ABO, or Rh phenotype (data not shown).

To confirm the presence of Gb₃ and band 0.03 in our individual donor platelet GSL samples, neutral GSLs were immunostained with MAb Pk002 and Stx as described above. MAb Pk002 recognized Gb₃ in all donor GSL samples. MAb Pk002 also recognized the P₁ antigen in the P₁+ erythrocyte control (lane R). No P₁ antigen was detected in any single donor platelet GSL samples with MAb Pk002, even in P₁+ platelet donors.

By HPTLC-immunostaining, Stx recognized Gb₃ in most donor platelet GSL samples. The single exception was donor 3 in which Gb₃ comprised less than 9% of the total platelet neutral GSL expressed (Table 5). In addition to Gb₃, Stx also bound a second minor platelet GSL (R_f, 0.03) in five (20%) of our platelet GSL samples (donors 1, 4, 7, 13, and 18). Based on mobility and lack of activity with MAb Pk002, the second Stx-positive GSL in our single donor platelet GSL samples was identified as band 0.03.

Platelet band 0.03 is related to platelet CDH and Gb₃ expression.

To determine whether there was any relationship between band 0.03 expression and other globo series GSLs, we compared the mean expression of CDH, Gb₃, and Gb₄ in band 0.03-positive and -negative platelet donors. Differences between band 0.03-positive and -negative donors were considered statistically significant if the P was <0.05 by the Student t test or the Mann-Whitney U-test. As shown in Table 5, the percent Gb₃ and the Gb₃/CDH ratio were significantly higher in band 0.03-positive donors. In contrast, there was no association between band 0.03 and the percent Gb₄ or the Gb₃/Gb₄ ratio. The increased percent Gb₃ and Gb₃/CDH ratio suggest a possible increased conversion of CDH to Gb₃ in band 0.03 donors.

Platelet band 0.03 expression is not related to donor P₁ or P₂ phenotype.

The frequency of band 0.03 (20%) in our sample of 25 platelet donors suggests that band 0.03 may be related to a P₂ (P₁-negative) erythrocyte donor phenotype. Among Caucasian blood donors, the P₂ phenotype is observed in approximately 21% of donors (2). Furthermore, studies of GSLs isolated from P₁ and P₂ donor erythrocytes have shown an increase in the mean percent Gb₃ and the Gb₃/CDH ratio in P₂ erythrocytes (35). Of 13 donors for which an erythrocyte sample was available, 11 (85%) donors were P₁ positive and 2 (15%) were P₁ negative or P₂ (Table 5).

Although the mean percent Gb₃ (percent total platelet neutral GSL) and the Gb₃/CDH ratio were higher in P₂ platelet donors than in P₁ platelet donors, the differences were statistically insignificant due to an insufficient number of P₂ donors (P > 0.05, Spearman rank sum). Among band 0.03-positive donors with a P₁ or P₂ phenotype, one donor was P₁ (donor 4) and another P₂ (donor 18). In addition, band 0.03 was not observed in donor 14, another P₂ donor. Based on this small sample, band 0.03 expression did not appear to be directly related to the donor P₁ or P₂ erythrocyte phenotype.

Stx binding to platelets is unrelated to the relative fraction of Gb₃.

To determine whether individual differences in Gb₃ and band 0.03 expression affected the ability of Stx to bind circulating platelets, we compared platelet GSL expression to Stx binding in 18 blood donors by HPTLC-immunostaining and two-color immunofluorescence flow cytometry. For flow cytometry, whole-blood-derived platelet concentrates from 16 type A and 2 type O blood donors were incubated with FITC-labeled Stx-B and a PE-labeled anti-platelet glycoprotein IIb/IIIa MAb (CD41; MAb P2). For a negative control, platelets were incubated with a toxin-free FITC filtrate. Platelets were gated on CD41 positivity, forward scatter, and orthogonal scatter. Results were reported as the percent platelets positive for anti-CD41 and Stx-B (Table 6). For GSL analysis, the remaining platelet concentrate was washed, lyophilized, and isolated as described before. Isolated GSLs were analyzed by scanning densitometry of DPA-stained HPTLC plates.

TABLE 6.

Stx binding to platelets versus platelet Gb₃ expression in individual platelet donors by flow cytometry and HPTLC^a

Sample	% Gb3b,c	Gb3/CDHb,d	% Stx-B positivee
1	14.8	0.42	71.6
2	24.5	0.46	38.9
3	40.5	1.53	49.3
4	21.5	0.35	28.8
5	31.6	0.74	46.5
6f	ND	ND	30.4
7	38.4	0.93	25.7
8f	ND	ND	33.6
9f	ND	ND	49.4
10	12.3	0.07	0.03
11f	ND	ND	22.0
12	21.1	0.35	7.0
13f	ND	ND	19.9
14	45.6	0.99	12.4
15	34.0	0.72	10.6
16	25.7	0.28	19.6
17	27.8	0.41	18.6
18	11.0	0.16	16.9

^aPercent Gb₃ versus Stx-B positivity: r = +0.03 by Pearson product-moment correlation (17). 

^bScanning densitometry of DPA-stained HPTLC plates. ND, not done. 

^cPercent Gb₃ as percent total platelet neutral GSL. 

^dRatio of the area (in square millimeters) of Gb₃ and CDH, as determined by scanning densitometry. 

^ePercent platelets positive with FITC-labeled Stx-B subunits and PE-labeled MAb P2 (anti-CD41). 

^fInadequate sample for analysis. 

As shown in Table 6, the percent platelets labeled by Stx varied in the donor samples, ranging from 0.3 to 71.6%. Likewise, the percent Gb₃ varied in these same donor platelet GSL samples, ranging from 11.0 to 45.6% of the total platelet neutral GSL. Although the mean percent Gb₃ (26.8%, n = 13) and the mean percent Stx-B positive platelets by flow cytometry (27.8%, n = 18) were similar, there was no correlation between percent Gb₃ and percent Stx-B positivity among individual donors (r = 0.03, Pearson product-moment correlation). Similarly, no relationship was observed between the Gb₃/CDH ratio and Stx positivity.

Stx binding to platelet membranes may, however, be related to platelet size. A comparison of percent Stx-B positive platelets by forward scatter showed a trend toward increased Stx binding with decreasing forward scatter among individual platelet samples (Fig. 4). Because forward scatter is a function of cell size (85), these results suggested that Stx may bind with greater efficiency to smaller and older platelets (79, 85), possibly reflecting a decrease in the platelet glycocalyx with unmasking of otherwise cryptic GSL receptors. To test the latter hypothesis, type A platelets were stained in parallel with HPA lectin, which recognizes the blood group A antigen (86). Since a large portion of A antigen is expressed on platelet glycoproteins (83), there should be a decrease in A-antigen expression on platelets with decreasing platelet size. As expected, HPA fluorescence decreased with decreasing forward scatter.

FIG. 4.

Stx and HPA binding to platelets is related to platelet size. The mean (FITC) fluorescence intensity of HPA and Stx binding relative to platelet forward scatter (x axis) was plotted for platelets from five donors. As shown, there was an increase in Stx and decrease in HPA binding with decreasing forward scatter. Mean fluorescence intensity represents the mean of two independent samples per lectin (HPA and Stx) per donor (standard error of mean = 2.4 [17]).

We also wished to examine the effect of band 0.03 expression on Stx-B binding to intact platelets. Given its apparent size, band 0.03 should be less susceptible than Gb₃ to the effect of glycoprotein masking. Unfortunately, due to the small quantity of GSL isolated from individual whole-blood-derived platelet concentrates, there was insufficient material for HPTLC-immunostaining with Stx. To assess the effect of band 0.03 on Stx binding to intact platelets will require repeat studies using single-donor, apheresis platelets which contain at least six times more platelets than whole-blood-derived platelet concentrates (1).

DISCUSSION

As previously reported (57, 87), Gb₃ is a major neutral GSL of human blood platelets and can be shown to bind Stx by HPTLC-immunostaining. Platelets also express a second Stx receptor, band 0.03. The latter appears to represent a potentially novel Stx-binding GSL, immunologically distinct from either Gb₃, P₁, or galabiosylceramide. Furthermore, band 0.03 may be platelet specific: in a survey of 20 human tissues, band 0.03 is identified only in platelets. In contrast, Gb₃, the putative physiologic receptor for Stx, is ubiquitously expressed in most tissues arising from the embryonic mesoderm (28).

In HUS (10, 19), Stx binding to either Gb₃ or band 0.03 could contribute to microthrombus formation (70, 78), thrombocytopenia, and platelet activation (4, 25, 36), possibly through induction of the sphingomyelin/ceramide signal transduction pathway (43, 99). Evidence supporting Stx-induced platelet activation includes a study by Rose et al. (80) which showed an increase in agonist-induced platelet aggregation after incubation of plasma with verotoxin. Stx binding to platelet membranes may also contribute to thrombocytopenia via splenic removal of toxin-coated platelets.

We also observed a reasonably good correlation between clinical disease and the expression of Stx receptors on other tissues. Both renal failure (11) and toxin-mediated diarrhea (49, 50, 55, 65) could reflect Stx binding to Gb₃ and/or galabiosylceramide (Stx band 0.40) on kidney and small intestinal mucosa. On vascular endothelium, Stx binding to Gb₃ is believed to play a critical role in the pathogenesis of VTEC-associated HUS and thrombotic thrombocytopenic purpura through upregulation of vascular addressins such as E- and P-selectin (68), decreased synthesis of tissue plasminogen activator and inhibitor (90), and endothelial cell death (54, 62, 74, 91, 92) with exposure of the thrombogenic subendothelium. On monocytes, Stx binding can lead to a release of inflammatory cytokines such as tumor necrosis factor alpha (56, 77, 93, 94). Tumor necrosis factor alpha and other inflammatory cytokines may further enhance Stx-mediated cell injury via cytokine-induced increases in Gb₃ synthesis (54, 75, 77, 91–94). On erythrocytes, Stx binding to Gb₃ and P₁ antigen may act synergistically with RTX toxin to promote erythrocyte hemolysis (6–8). Likewise, Stx binding to Gb₃ on other tissues (liver, heart, and lung) could contribute to the etiology of thrombotic thrombocytopenia purpura (76, 97) and multisystem organ failure sometimes seen with VTEC infections (10).

Like previous investigators (87), we initially examined GSLs isolated from a large pool of platelet donors. However, a number of small studies have suggested normal biologic variability in GSL expression in individuals. In a small study of two apheresis platelet donors, Holgersson et al. (46) demonstrated clear, donor-specific differences in platelet globo-GSL expression. Likewise, Fletcher et al. (35) noted differences in erythrocyte Gb₃ and CDH expression which were related to the donor P₁ or P₂ phenotype. In human kidney (11), Gb₃ levels can reportedly vary from 0.254 to 0.896 nmol of Gb₃/mg of kidney cortex. Likewise, Obrig and colleagues (62, 75) observed marked differences in Stx-induced endothelial cytotoxicity among different endothelial cell lines, possibly reflecting a difference in Gb₃ expression by different donors.

To determine whether individuals may differ in the number and/or type of Stx receptors expressed on platelet membranes, we isolated and examined the GSLs from 25 platelet donors. In agreement with Holgersson’s (46) findings, there is significant heterogeneity in the relative expression of the three major platelet neutral GSLs in individual donors, particularly for Gb₃. Band 0.03 expression also varied in donors and may represent a low-frequency platelet alloantigen. Although there may be a relationship between donor P₁ or P₂ phenotype and Gb₃ expression on platelets, there is no correlation between a P₂ phenotype and band 0.03 expression. There is, however, a significant correlation between band 0.03 expression and increased percent Gb₃ and Gb₃/CDH ratios. The latter may suggest a possible association between Gb₃ and band 0.03 synthesis.

In erythrocytes (8) and vascular endothelium (91, 92), differences in Stx receptors have been shown to correlate with Stx binding and cytotoxicity. To determine whether the same were true for platelets, we compared platelet Gb₃ content with the binding of Stx to intact platelet membranes by flow cytometry and HPTLC. As expected, there are notable differences in both platelet Gb₃ content and Stx binding between donors. In individual donors, however, there is no correlation between Gb₃ expression and the ability of Stx to bind platelets (percent Stx-positive platelets).

Interestingly, there is a loose association between Stx binding and platelet size. Specifically, mean fluorescence intensity (FITC-Stx-B) tends to increase with decreasing forward scatter. Since forward scatter is related to cell size (85), these results suggest that Stx binds with greater efficiency to smaller and older platelets (79), possibly reflecting a decrease in the platelet glycocalyx as platelets age. Evidence for the latter includes parallel experiments in which type A platelets were labeled with HPA, a lectin specific for the blood group A antigen (86). In contrast to Stx, there is a parallel decrease in forward scatter and HPA fluorescence. Because a large portion of blood group A antigen is expressed on platelet glycoproteins (46, 83), this decrease in HPA positivity would support a decrease in platelet glycoproteins with decreasing platelet size (and increasing platelet age). Similar findings have been reported for aging erythrocytes (34).

This age-associated decrease in platelet glycoproteins may be important in the pathophysiology of Stx infections. Due to their size and proximity to the cell membrane, short-chain GSLs are frequently cryptic antigens on the surfaces of many cells (41, 95). As a consequence, many GSL antigens must be exposed through enzymatic modification of cell membranes by proteases or neuraminidase (41). This was specifically demonstrated for Gb₃ in ARH77 cells (95) and MDCK cells (82). In the latter study, treatment of MDCK cells with either pronase or neuraminidase resulted in a nearly twofold increase in Stx binding. Unmasking of GSL receptors on platelets and other tissues may also occur in vivo. At least two studies have reported the identification of cysteine proteases in the sera of patients with HUS and thrombotic thrombocytopenia purpura (33, 69).

Although we were able to demonstrate the presence of two Stx receptors on platelets, only Stx band 0.25 or Gb₃ was positively identified. Gb₃ was identified by both physical and immunologic methods including HPLC, HPTLC, HPTLC-immunostaining, and one- and two-dimensional nuclear magnetic resonance of the isolated GSL (29). In contrast, the low incidence of band 0.03 (20% of donors), coupled with the fact that band 0.03 was a minor platelet GSL, even in band 0.03-positive donors, made isolation of band 0.03 for biophysical studies difficult. As a consequence, the structure of band 0.03 was deduced by glycosidase digestion and epitope mapping with a series of anti-globo MAbs and lectins with differing specificities.

In summary, band 0.03 is positive with Stx, E. coli SC802, and MAb P001, consistent with the presence of a terminal galabiose epitope. In addition, Stx binding to band 0.03 is sensitive to both α-galactosidase and glycan ceramidase, also consistent with a R-Glcβ1-1′Cer GSL terminating in an α-linked Gal (31, 60). Band 0.03 is negative with MAbs 38.13, Pk002, and Gal-13, suggesting that band 0.03 is a Galα1-4Gal-R-Cer, where R is an oligosaccharide not beginning with Glc or GlcNAc. Therefore, R may be an oligosaccharide beginning with either a Gal or GalNAc. In support of the latter, band 0.03 is weakly positive with MAb MC631, a MAb which recognizes globoside (P antigen) as well as extended globo-series GSLs such as the SSEA-3, SSEA-4, globo-H, and Forssman antigens (13, 51, 52). Finally, a comparison of band 0.03 with other platelet (Table 2) and erythrocyte (P₁; R_f, 0.06) GSLs suggest that band 0.03 is a hexa- to octaosylceramide. The proposed structure for band 0.03, based on available data, is a ceramide hexosylceramide with a globoside or type 4 chain oligosaccharide core and a terminal galabiosyl epitope (IV³-β-Galα1-4Gal-Gb₄) (Fig. 5).

FIG. 5.

Proposed structure of platelet band 0.03 based on enzyme digestion and epitope mapping with a series of anti-globo series MAbs and lectins.

The incidence, as well as the structure of band 0.03, suggests that band 0.03 may be related to the Luke blood group antigen. A high-frequency antigen on human erythrocytes, the Luke blood group antigen is an extended globo-series ganglioside related to the P blood group family (see below) (67, 89). In studies of volunteer blood donors, Luke antigen can be detected on 98% of donors, with 80% of donors having a strong Luke phenotype and 18% of the donors with a weak Luke phenotype (66, 89). A Luke-negative phenotype is observed in 0.2 to 2% of P₁ and P₂ donors (16, 66, 89). A Luke-negative phenotype is also seen in the rare p (Gb₃−, Gb₄−, P₁−) and P^k (Gb₃+, Gb₄−) phenotypes, reflecting the absence of globo-GSL precursors necessary for the synthesis of the Luke blood group antigen (67).

Interestingly, studies of families of Luke-negative donors show a reciprocal relationship between expression of the P^k and Luke antigens (16). Specifically, donors lacking Luke antigen have significantly increased P^k expression, as determined by erythrocyte agglutination. This apparent increase in P^k was suggested to reflect the presence of a P^k-like GSL, related to Luke antigen, on Luke-negative erythrocytes (88). As stated by Mollison et al. (67) “…it has been postulated that the terminal neuraminic acid of the hypothetical Luke structure could be replaced by galactose in such individuals, giving rise to a Galα1-4Gal-globoside structure which could react with anti-P^k.” The latter GSL is essentially identical to that postulated for band 0.03. (Luke antigen, NeuAcα1-3Galβ1-3GalNAcβ1-3Galα1-4Galβ1-4Glcβ1-1′Cer; antithetical Luke antigen, Galα1 - 4Galβ1 - 3GalNAcβ1 - 3Galα1-4Galβ1-4Glcβ1-1′Cer.) The incidence of the band 0.03 phenotype (20%) suggests that band 0.03 may be expressed on platelets of weak Luke and Luke-negative donors.

The discovery of a potentially novel P^k-like GSL, related to the Luke antigen system, may have important implications in the diagnosis, prognosis, and pathogenesis of VTEC-associated HUS. Unlike Gb₃ or the P^k antigen, which is ubiquitously expressed by most tissues arising from the embryonic mesoderm (28), the Luke antigen has been identified in only a few tissues and cell types. In addition to erythrocytes (89), Luke antigen has been identified in teratocarcinoma (51, 52), endothelium, smooth muscle (39), dorsal root ganglion (45), platelets (29), and human kidney (53). In Luke-negative or weak Luke individuals, therefore, band 0.03 or an “antithetical” Luke antigen may serve as an additional and, due to its size, more accessible receptor for Stx in vascular endothelium, erythrocytes, platelets, and kidney.

Luke-negative and weak Luke antigen donors could also be at increased risk from Stx-mediated disease due to a constitutive increase in Gb₃ expression on target cell membranes (16, 27). Previously, Newburg et al. (71) demonstrated an association between VTEC-associated HUS and Gb₃ levels in human erythrocytes. In endothelial cells, cytokine-stimulated increases in Gb₃ synthesis and expression are associated with increased Stx cytotoxicity (54, 62, 75, 82, 91–93). Luke-negative and weak Luke individuals, therefore, may have an inherited increased sensitivity and susceptibility to Stx due to increased membrane Gb₃ expression, analogous to the effects of inflammatory cytokines. If true, a negative or weak Luke phenotype may represent a previously unrecognized risk factor in Shigella and VTEC infections.

In summary, we have shown that Stx can bind human platelets. Since the Galα1-4Gal disaccharide linkage essential for Stx binding is expressed only on GSLs (98), Stx presumably bind platelets via GSL receptors on the platelet cell membrane. Platelets possess two GSL receptors for Stx, Gb₃, and a novel GSL, band 0.03. Preliminary data suggest that the latter may represent the antithetical Luke antigen and, as such, may serve as a fourth GSL receptor for Stx on platelets, kidney, endothelium, and erythrocytes. Binding of Stx to platelets may play a role in the etiology of thrombocytopenia in S. dysenteriae (58) and VTEC infections (10) through either platelet activation or splenic removal of toxin-coated platelets.