Structure-dependent Pseudoreceptor Intracellular Traffic of Adamantyl Globotriaosyl Ceramide Mimics

Background: Verotoxin internalization and retrograde transport to the Golgi/ER is mediated by Gb₃ glycolipid.

Results: Amphipathic Gb₃ mimics can alter binding, trafficking, and cytotoxicity of verotoxins.

Conclusion: The lipid moiety of Gb₃ analogues determines the trafficking of verotoxins.

Significance: Synthetic glycolipid analogues can function as membrane receptors to internalize bound ligand and subvert endogenous GSL traffic.

Abstract

The verotoxin (VT) (Shiga toxin) receptor globotriaosyl ceramide (Gb₃), mediates VT1/VT2 retrograde transport to the endoplasmic reticulum (ER) for cytosolic A subunit access to inhibit protein synthesis. Adamantyl Gb₃ is an amphipathic competitive inhibitor of VT1/VT2 Gb₃ binding. However, Gb₃-negative VT-resistant CHO/Jurkat cells incorporate adaGb₃ to become VT1/VT2-sensitive. CarboxyadaGb₃, urea-adaGb₃, and hydroxyethyl adaGb₃, preferentially bound by VT2, also mediate VT1/VT2 cytotoxicity. VT1/VT2 internalize to early endosomes but not to Golgi/ER. AdabisGb₃ (two deacyl Gb₃s linked to adamantane) protects against VT1/VT2 more effectively than adaGb₃ without incorporating into Gb₃-negative cells. AdaGb₃ (but not hydroxyethyl adaGb₃) incorporation into Gb₃-positive Vero cells rendered punctate cell surface VT1/VT2 binding uniform and subverted subsequent Gb₃-dependent retrograde transport to Golgi/ER to render cytotoxicity (reduced for VT1 but not VT2) brefeldin A-resistant. VT2-induced vacuolation was maintained in adaGb₃-treated Vero cells, but vacuolar membrane VT2 was lost. AdaGb₃ destabilized membrane cholesterol and reduced Gb₃ cholesterol stabilization in phospholipid liposomes. Cholera toxin GM1-mediated Golgi/ER targeting was unaffected by adaGb₃. We demonstrate the novel, lipid-dependent, pseudoreceptor function of Gb₃ mimics and their structure-dependent modulation of endogenous intracellular Gb₃ vesicular traffic.

Introduction

Verotoxin (VT)³ comprises a family of Escherichia coli-derived AB₅ subunit toxins (also termed Shiga toxins). Verotoxin cytopathology is targeted via the B subunit pentamer of the holotoxin binding to its receptor glycosphingolipid (GSL), globotriaosyl ceramide (Gb₃; also known as the p^k blood group antigen (1) and CD77, a human B cell marker (2)). VT1 and VT2 (60% identical at the nucleotide level (3)) are the primary verotoxins associated with clinical disease (4). Gastrointestinal infection with verotoxin-producing E. coli can result in the pathology of hemorrhagic colitis, which may precede the more severe hemolytic uremic syndrome (HUS), a renal pathology characterized by a triad of symptoms, thrombocytopenia, anemia, and renal glomerular microangiopathy (5). Hemorrhagic colitis is mediated via VT targeting Gb₃ within the submucosal microvasculature of the GI tract. Subsequent systemic verotoxemia results in toxin access to renal glomerular endothelial cells, which also express Gb₃ (6) to mediate endothelial cell damage, blood vessel occlusion, glomerular infarct, and subsequent hemolysis. HUS, primarily a disease of the very young and elderly (7), currently retains an approximately 5% mortality, and estimates of morbidity range as high as 30%. The recent German outbreak of enteroaggregative, VT2-expressing E. coli infections (8, 9) with an HUS incidence reaching 25% and a preponderance of female adult cases indicates major, unsuspected knowledge gaps in VT-induced pathology.

For reasons as yet unclear, VT2 is more frequently associated with clinical disease than VT1 (10, 11), despite the fact that VT1 is a more potent cytotoxin in vitro (12) and both toxins bind to the same receptor (13). Despite a common receptor Gb₃, VT1 and VT2 preferentially bind different and shared epitopes within the Gb₃ carbohydrate (14, 15), which may be differentially available within different lipid contexts (12). Such differential receptor binding results in coincident but also discreet VT1 and VT2 binding sites on the surface of sensitive cells (12, 16) and within human renal tissue (15, 17). Cholesterol within human renal glomeruli can mask Gb₃ to prevent VT1 and VT2 binding (15, 17). Unlike VT1, VT2 can induce the formation of intracellular vacuoles in a subfraction of susceptible renal epithelial cells (12).

Cell membrane GSL carbohydrate presentation for ligand binding is complex, being a function of both the highly heterogeneous composition of the membrane-embedded ceramide and a lateral association with other membrane lipids, most notably cholesterol (18), to form domains of differential membrane order (19). Molecular simulation shows that the cholesterol-GSL interaction can alter the GSL carbohydrate conformation (18, 20) from a membrane-perpendicular to -parallel format. Cholesterol can mask GSLs to prevent appropriate ligand binding in tissues (15, 17, 20) and in model and cell membranes (20, 21). Nevertheless, to mediate cell cytotoxicity in vitro, Gb₃ must be expressed in what are termed cell surface lipid microdomains or rafts (22, 23), in which the concentration of GSLs and cholesterol and membrane order are significantly increased (24). The renal glomerulus is the target of the VT-induced pathology of HUS, and glomerular Gb₃ is within such domains (17). We propose that adaGSLs, unlike the parent GSL, will not interact with cholesterol; indeed, the adamantane frame may partially substitute for cholesterol to provide a mimic of the GSL-cholesterol complex (25).

Membrane Gb₃ within such domains mediates both clathrin-dependent (26) and -independent (27) VT internalization and subsequent “retrograde transport,” from the cell surface through endosomes, trans-Golgi network (TGN), and Golgi to the ER (28), where the A subunit is translocated to the cytosol for inhibition of protein synthesis (29). When VT binds to non-raft Gb₃, internalization of the toxin receptor complex mediates the transport of the toxin to the lysosome for degradation, without the induction of cytotoxicity (22, 30). This may be similar to the abnormal transport of accumulated GSLs and cholesterol to lysosomes in GSL storage diseases (31). Cholesterol depletion (27) or modulation (32) can also prevent the GSL endosome-TGN transition. A balance of GSL-cholesterol interaction may be required for retrograde transport.

Verotoxin binding to the Gb₃ oligosaccharide is modulated by the lipid moiety of Gb₃ (33, 34) and the membrane environment in which Gb₃ is presented (35). This has been termed “aglycone modulation of GSL receptor function” (36). The requirement for retrograde transport of the toxin receptor complex to the endoplasmic reticulum is shared by the cholera toxin/GM1 receptor interaction (37, 38). Exogenous GSL analogues in which the native fatty acid is replaced by a fluorescent group (e.g. BODIPY or NBD) also demonstrate retrograde transport from the cell surface to the Golgi (39).

Several groups have developed receptor analogues based on the Gb₃ carbohydrate sequence coupled to polymeric or pentameric scaffolds (40–42) to develop specific receptor-based means to prevent the in vivo cytotoxicity, which may follow verotoxin-producing E. coli infection. The binding affinity of the VT B subunit pentamer for the lipid-free oligosaccharide is much reduced compared with native Gb₃ glycolipid (43), but this can be largely countered by multivalency (41), particularly when tailored to accommodate the pentameric geometry of the receptor B subunits displayed within the VT holotoxin (40, 42).

An additional approach is to try to utilize the inherent high affinity binding of the native Gb₃ glycolipid. Substitution of the Gb₃ fatty acid with an adamantane frame provided a water-soluble analog of Gb₃ that retained high affinity VT1 binding in an aqueous environment (44, 45). Although this analog proved an effective competitor to prevent VT1 and VT2 cytotoxicity in vitro (12), in vivo, the analog was found to augment rather than reduce VT2 cytopathology (46).

Our present studies indicate that this is due to a previously unrecognized property of such GSL analogues. We now show that adaGb₃ can partition into receptor-negative cells to render them VT1- and VT2-sensitive. This pseudoreceptor function is mediated via a novel intracellular routing pathway involving early endosomes. Moreover, we find that this adaGb₃ pathway can also hijack the endogenous cellular Gb₃-mediated retrograde transport of VT in sensitive cells to reroute traffic and induce resistance. This may relate to loss of interaction with cholesterol that we show for the adaGSL mimic. We have designed several new soluble adaGb₃ analogues and show this dominant rerouting of native intracellular GSL trafficking is prevented by chemical substitution within the adamantane frame, suggesting the existence of intramembrane vesicular trafficking cues.

A dimeric Gb₃ analog retains the VT1/VT2 inhibitory activity of adaGb₃ in solution but is unable to insert into cell membranes to show VT1/VT2 pseudoreceptor function. These studies demonstrate the importance of the lipid chemistry of Gb₃ in membrane incorporation and intracellular trafficking and illustrate a new approach against VT-induced cytopathology. Exogenous GSL mimics can be functionally trafficked in cells and, according to their lipid structure, subvert endogenous intracellular membrane GSL trafficking pathways.

EXPERIMENTAL PROCEDURES

Synthetic Compounds

Synthesis of the adamantyl analogues of Gb₃ (Fig. 1) is described in the supplemental material. Analogues were prepared for cell insertion as follows. Compounds were dried from solution in CHCl₃-CH₃OH (2:1, v/v) under N₂, resuspended in ethanol, sonicated briefly, dried under N₂, and resuspended in water at a concentration of 100 μm. To allow the analogues to reach an equilibrium state in solution, solutions were vortexed and sonicated for 30 s and then incubated at 37 °C for 2 h. Aliquots were dispensed into glass tubes, rapidly frozen on dry ice, and lyophilized overnight. The compounds were redissolved in chilled serum-free culture medium immediately before the addition to cells.

FIGURE 1.

Scheme for adaGb₃, carboxyadaGb₃, urea-adaGb₃, OHEtadaGb₃, and adabisGb₃.

Reagents and Antibodies

Verotoxin 1 (VT1) and VT2 and VT1 B-subunit were purified as described previously (34, 46). Antibodies used were as follows. Mouse anti-VT1 mAb PH-1, reactive against the VT1 B-subunit and polyclonal rabbit anti-VT1 B-subunit, were prepared in our laboratory. Rabbit anti-VT2 was a generous gift of Dr. Glen Armstrong (University of Calgary). Goat anti-EEA1 (Santa Cruz Biotechnology, Inc.), mouse anti-Lamp-2 (Developmental Studies Hybridoma Bank, clone H4B4), rabbit anti-Rab6 (Santa Cruz Biotechnology, Inc.), and rabbit anti-calnexin (Enzo Life Sciences) were obtained as indicated. Brefeldin A, cholera toxin B-subunit, sphingomyelin (SPM), methyl-β-cyclodextrin (MβCD), cholesterol, egg phosphatidylcholine (PC), and crystal violet were from Sigma. Fluorescence mounting medium was from Dako, and paraformaldehyde was from EM Sciences. [³H]cholesterol and Cy3 were from GE Healthcare. Alexa Fluor 488 pentafluorophenyl ester, DiI LDL, DAPI, and Texas Red sulfonyl chloride were from Molecular Probes. Proteins were labeled with fluorophores using standard conditions as recommended by the manufacturer and isolated by gel filtration using G-25.

Cell Culture

Vero cells were grown in Eagle's minimum essential medium, 5% FCS, CHO; HEK-293 cells were grown in DMEM, 10% FCS; and Jurkat cells were grown in RPMI 1640, 10% FCS at 37 °C, 5% CO₂. All tissue culture media and buffers were obtained from Wisent Inc.

Insertion of GSL Analogues into Cells

Cells were washed twice with serum-free RPMI 1640 with 20 mm Hepes (H-RPMI) and chilled on ice. Cells were then incubated with freshly dissolved adaGb₃-solution for 1 h at 4 °C. Serum-containing medium was added for 37 °C incubations exceeding 1 h.

Cell Cytotoxicity Assays

Vero and CHO cells seeded in 96-well cell culture plates (3 × 10⁴ cells/well) were grown at 37 °C overnight. Jurkat cells were centrifuged, washed twice with chilled H-RPMI, distributed in 96-well plates (5 × 10⁴ cells/well), and chilled on ice.

Dilutions of adamantyl analogues were added to the cells and incubated on ice for 1 h. 10-fold serial VT1/VT2 or ricin dilutions were then added to the cells and incubated at 37 °C. After 4 h, serum was added to a final concentration of 5%, and cells were incubated at 37 °C for another 68 h. In some experiments, 0.5 μg/ml brefeldin A (BFA) was added for 30 min at 37 °C before the addition of VT and maintained throughout. Due to the long term toxicity of BFA, cytotoxicity was measured after 18 h of toxin treatment.

At the end of the incubation period, live cells were fixed onto the wells using 2% formalin in PBS and stained with crystal violet as described (47). Dye was solubilized with 100 μl of 10% acetic acid, and optical density was read at 560 nm using an ELISA plate reader. Cell viability was expressed as a percentage of control cells, which were treated with neither VT nor adaGb3 analogues.

The ability of the adamantyl Gb₃ analogues in solution to block VT cytotoxicity to Gb₃+ve cells was assessed by prebinding the toxin and analogues prior to the addition to Vero cells. Dilutions (50 ng/ml to 0.03 pg/ml) of VT1 or VT2 in Eagle's minimum essential medium were prepared and mixed with an equal volume of adaGb₃ or adabisGb₃ (100, 50, 25, 12.5, 6.3, and 3.1 μm). After incubation at 37 °C for 60 min, 50 μl of the mixture was added to Vero cells in 96-well plates and incubated for 1 h at 37 °C. Cells were washed with Eagle's minimum essential medium and then incubated in complete medium for 72 h at 37 °C. Cell viability was measured as described above by crystal violet staining.

Confocal Fluorescence Microscopy

CHO and Jurkat cells were grown to 80% confluence on 12-mm gelatin-treated glass coverslips. Cells were washed twice with H-RPMI, chilled on ice, and then incubated with 50 μm adabisGb₃ or 20 μm adaGb₃ (CHO) or 10 μm adaGb₃ (Jurkat) in H-RPMI for 1 h on ice. For labeling of the cells with VT, 4 μg/ml Alexa Fluor 488-VT1 or Texas Red-VT2 was bound on ice for 1 h. Cells were washed with cold PBS and fixed with 4% paraformaldehyde in PBS. To observe the intracellular trafficking of VT, bound VT was internalized at 37 °C for 10 min, 1 h, or 6 h. At the end of the incubation period, the cells were fixed with 4% paraformaldehyde in PBS, permeabilized for 15 min at ambient temperature with 0.2% Triton X-100, and then blocked with 2% BSA. Verotoxins were detected with VT1- or VT2-specific antibodies and the organelle-specific antibodies to EEA1 (early endosomes), Rab6 (Golgi), Lamp-2 (late endosome/lysosome), or calnexin (ER), followed after washing with Alexa Fluor 488 or 546 anti-goat or 594 anti-rabbit secondary antibody.

Microscope Image Acquisition

Fluorescently stained cells were viewed with a Leica DMRE2 confocal microscope under oil at ×63 with a numerical aperture of 1.4 at ambient temperature. Fluorophores used were DAPI for nuclear staining, Alexa488 (for VT), Cy3 (for cholera toxin (CT)), Texas Red (for VT2), and Alexa 594 (secondary antibody). Images were captured with a Hamamatsu EM-CCD C9100 digital camera using Volocity 5.5.0. Confocal stacks were deconvolved with Volocity software by iterative restoration using calculated point spread functions. Composite images were assembled using Photoshop CS4 and Zeiss Image Examiner.

Isolation of Total Lipids from Cells and Detection by TLC

CHO cells (1 × 10⁷ cells) and Jurkat cells (2.4 × 10⁶ cells) were treated with 20 μm adaGb₃ or 50 μm adabisGb₃ for 1 h on ice and then washed with PBS. Cell suspensions in water were transferred into chloroform/methanol (2:1, v/v) and shaken vigorously overnight. After filtering the cell debris, the solvents were dried down, resuspended in methanolic NaOH for saponification, and then neutralized with NH₄HOAc and HCl. The samples were desalted using Sep-Pak C₁₈ cartridges (Waters, Milford, MA). The lipids were eluted with methanol and chloroform/methanol (2:1), dried down, and redissolved in 100 μl of chloroform/methanol (2:1). The samples (10 μl each) were separated on two identical TLC plates (chloroform/methanol/water; 65:25:4), one for detection with orcinol and one for VT binding.

VT1 and VT2 Binding to Gb₃ and Analogues by TLC Overlay

After GSL separation, the TLC plates were dried and incubated with 1% fish gelatin in TBS for 3 h at room temperature and washed twice with TBS. The plates were incubated with VT1 B-subunit (0.35 μg/ml) or VT2 (2.5 μg/ml) overnight at 4 °C, with polyclonal rabbit anti-VT1 B-subunit (VT1B) or anti-VT2 for 3 h at room temperature, and with HRP-conjugated goat anti-rabbit IgG for 1 h at room temperature. Bound toxin was detected by development using 0.6 mg/ml 4-chloro-1-naphthol, 0.015% H₂O₂ in TBS.

Liposomal Assay of Glycosphingolipid-Cholesterol Interaction

The ability of Gb₃ as compared with adaGb₃ to interact with cholesterol in membranes was quantitated in liposomes by measuring the induced resistance to cholesterol extraction by MβCD, based on previous studies (48, 49). Multilamellar PC, [³H]cholesterol liposomes with or without Gb₃, SPM, or adaGb₃ were prepared. (a) 0.2 μmol of PC + 0.07 μmol of cholesterol (500,000 dpm), (b) 0.14 μmol of PC + 0.07 μmol of cholesterol + 0.07 μmol of Gb₃, and (c) 0.07 μmol of PC + 0.07 μmol of cholesterol + 0.07 μmol of Gb₃ or SPM + 0.07 μmol of adaGb₃ were dried together from organic solvent under N₂, freeze-dried, and vortexed in PBS for 30 min at room temperature to give a total lipid concentration of 500 μm in a total volume of 500 μl. Liposomes were briefly sonicated and incubated at 85 °C for 30 min with vortexing every 5 min, cooled to room temperature, centrifuged at 11,000 × g for 10 min, and washed in PBS. 50-μl aliquots were mixed with 50 μl of PBS with or without 0.5 mm MβCD at room temperature for 60 min with frequent vortexing. The suspension was centrifuged at 11,000 × g for 10 min, and tritium in the supernatant was counted in a scintillation counter.

RESULTS

Adamantyl Gb₃ Analogues

Five soluble adamantyl Gb₃ analogues were synthesized from deacyl(lyso)-Gb₃ (see supplemental material). The structures of these species are shown in Fig. 1. Different substitutions at the 2-position of the adamantane frame generated an acidic carboxyadaGb₃, a basic urea-adaGb₃, and a neutral hydroxyethyl adaGb₃ (OHEtadaGb₃). Coupling a second lyso-Gb₃ to carboxyadaGb₃ generated the dimeric adamantylbisGb₃ (adabisGb₃).

VT1/VT2 Binding to Synthetic Gb₃ Analogues

The VT binding activities of Gb₃, adaGb₃, OHEtadaGb₃, carboxyadaGb₃, urea-adaGb₃, and adabisGb₃ were compared by TLC overlay (Fig. 2). Gb₃ and each adaGb₃ analog were bound by VT1 and VT2. VT2/Gb₃ binding was weaker than that of VT1, as shown previously (46). However, VT2 bound all adaGb₃ species in preference to native Gb₃. AdaGb₃ and adabisGb₃ bound strongly to both VT1 and VT2.

FIGURE 2.

VT1 and VT2 binding to Gb₃ and adaGb₃ analogues by TLC overlay. The glycolipids (1.3 μg each) were separated by TLC and detected by orcinol or by binding of either VT1 or VT2. The binding to adaGb₃, OHEtadaGb₃, carboxyadaGb₃, urea-adaGb₃, and adabisGb₃ relative to Gb₃ was calculated from the band intensities: VT1, 80, 20, 6, 35, and 68%; VT2, 400, 275, 140, 310, and 315%.

AdaGb₃ Induction of VT1/VT2 Cytotoxicity to Gb₃-negative Cells

To assess the potential pseudoreceptor function of these adaGb₃ analogues, the Gb₃-negative, VT-resistant Jurkat and CHO cell lines were selected as potential targets for receptor “reconstitution.”

Cells were incubated in the presence of adaGb₃ and then treated with VT1 or VT2. The resulting cytotoxicity curves (Fig. 3A) show that untreated CHO cells were resistant to VT1 and VT2 below 10 μg/ml and that Jurkat cells were resistant at <1 μg/ml, whereas 10–20% cells were killed at 10 μg/ml. Cells treated with adaGb₃, carboxyadaGb₃, OHEtadaGb₃, or urea-adaGb₃ all showed increased susceptibility to VT1 and VT2 cytotoxicity. Significantly greater VT sensitivity was induced in Jurkat, as compared with CHO cells. Indeed, reconstituted CHO cells were only sensitive to the highest dose of VT1 and VT2 tested and essentially only at the highest concentration of adaGb₃ analog added. The receptor-incorporated CHO cells were in general slightly more sensitive to VT1 than VT2. However, Jurkat cell VT1 and VT2 sensitivity was significantly increased following treatment with any of the adaGb₃ analogues. Again susceptibility to VT2 was slightly less than that of VT1, and the different adaGb₃ analogues showed varying degrees of induction of VT1 and VT2 sensitivity. Cells in which adaGb₃ was incorporated showed the greatest response to VT1 and VT2, with a CD₅₀ of between 0.1 and 10 ng/ml according to adaGb₃ dosage. CarboxyadaGb₃-treated cells and OHEtadaGb₃-treated cells showed less VT1/VT2 sensitivity, with CD₅₀ ranging from 1 to 100 ng/ml for VT1 and from 50 to 10,000 ng/ml for VT2 according to adaGb₃ analog dosage. Cells treated with urea-adaGb₃ were yet less sensitive to VT1 and VT2.

FIGURE 3.

A, induction of VT1/VT2 toxicity in adaGb₃ analog-treated CHO and Jurkat cells. CHO or Jurkat cells were incubated with 0, 10, 20, or 50 μm adaGb₃ at 4 °C for 1 h. Cells were treated with 10-fold serial diluted VT1 or VT2 and incubated at 37 °C. Cell viability was monitored after 72 h and expressed as a percentage of control cells, which were treated with neither VT nor adaGb₃ analogues. B, effect of adaGb₃ analogues on CHO/Jurkat cell viability. CHO or Jurkat cells were incubated with 0, 10, 20, or 50 μm adaGb₃ (closed circles), OHEtadaGb₃ (triangles), carboxyadaGb₃ (squares), urea-adaGb₃ (diamonds), or adabisGb₃ (open circles) at 37 °C. Cell viability was monitored after 72 h and expressed as a percentage of non-treated control cells.

In contrast to the other adaGb₃ derivatives, treatment of Jurkat or CHO cells with adabisGb₃ had no subsequent effect on VT1 or VT2 sensitivity. AdabisGb3 treated cells remained resistant to VT1/VT2.

At high dosage, some adaGb₃ analogues were toxic to Jurkat cells (Fig. 3B). A concentration of >20 μm adaGb₃ or urea-adaGb₃ was toxic, whereas OHEtadaGb₃, carboxyadaGb₃, and adabisGb₃ were non-toxic at all doses. No analog showed cytotoxicity to CHO cells.

VT1/VT2 Binding to Gb₃ Analog-treated CHO/Jurkat Cells

AdaGb₃-treated CHO or Jurkat cells showed significant cell surface binding of fluorescent VT1 or VT2 at 4 °C (Fig. 4A). Cell membrane labeling was punctate, particularly for CHO cells. No VT binding to adabisGb₃-treated CHO or Jurkat cells was detected.

FIGURE 4.

A, binding of VT1 or VT2 to adaGb₃- or adabisGb₃-treated Gb₃-negative cells. CHO cells were treated with 20 μm adaGb₃ (left) or 50 μm adabisGb₃ (right), and Jurkat cells were treated with 10 μm adaGb₃ (left) or 50 μm adabisGb₃ (right). Then 4 μg/ml Alexa488-VT1B or Texas Red-VT2 was added at 4 °C. After 1 h, cells were washed and fixed. Stained cells were viewed by confocal microscopy. B and C, TLC of total lipids from adaGb₃ or adabisGb₃-treated cells. Glycolipids were isolated from CHO and Jurkat cells, separated by TLC, and visualized with orcinol (left). VT binding to the lipids was detected by TLC overlay with VT1 B-subunit (right). B, lane 1, adaGb₃; lane 2, adabisGb₃; lane 3, total lipids of untreated CHO cells; lane 4, total lipids of adaGb₃-treated CHO cells; lane 5, total lipids of adabisGb₃-treated CHO cells. C, lane 1, adaGb₃; lane 2, adabisGb₃; lane 3, total lipids of untreated Jurkat cells; lane 4, total lipids of adaGb₃-treated Jurkat cells; lane 5, total lipids of adabisGb₃-treated Jurkat cells. S, standard glycolipid mixture (from the top, glucosylceramide, galactosylceramide, lactosylceramide, Gb₃, Gb₄, and Gb₅). D and E, trafficking of VT1 to the early endosome, Golgi, and ER in Vero cells compared with adaGb₃-treated or OHEtadaGb₃-treated CHO/Jurkat cells. VT1 was internalized at 37 °C for 10 min (D), 1 h (D and E), or 6 h (E) as indicated. After fixation and permeabilization, VT1 was detected with mAb PH-1/Alexa488 anti-mouse IgG (for EEA1 colocalization) or polyclonal anti-VT1B/Alexa488 anti-rabbit IgG (for RAb6 and calnexin colocalization). The early endosome marker EEA1 (D), Golgi marker Rab6 (E), or ER marker calnexin (E) was detected with Alexa546-labeled anti-goat or Alexa594 anti-rabbit antibodies (red). Fluorescently stained cells were viewed with a confocal microscope. Toxin colocalization (arrowheads) with organelle markers was quantitated (supplemental Table 1).

Cellular Uptake of AdaGb₃ or AdabisGb₃

To determine whether adabisGb₃ is incorporated into the target cell membrane but becomes receptor-inactive, we extracted total lipids from adaGb₃ or adabisGb₃-treated CHO (Fig. 4B) and Jurkat cells (Fig. 4C) and detected the analogues by VT1 binding in a TLC overlay assay. AdaGb₃ was incorporated into both Jurkat and CHO cells, whereas adabisGb₃ was not detected in the extracts of adabisGb₃-treated cells. These results show adaGb₃ inserted into the cell membrane of Gb₃-negative cells to provide a pseudoreceptor for VT1 and VT2. In contrast, adabisGb₃ does not incorporate into the cell membrane of Gb₃-negative cells.

Intracellular VT1 Trafficking in AdaGb₃- or OHEtadaGb₃-reconstituted Cells

Confocal microscopy was used to compare the intracellular trafficking of fluorescent VT1 bound to endogenous Gb₃ in Vero cells and adaGb₃ inserted into CHO or Jurkat cells. After 10 min at 37 °C, a large fraction of internalized VT1 colocalized with the early endosome marker EEA1 in both Vero cells and adaGb₃-treated CHO/Jurkat cells (Fig. 4D). The VT1 also colocalized with transferrin in the “reconstituted” cells at this time (not shown). After 1 h at 37 °C in Vero cells, some of the internalized VT1 overlaps with the Golgi marker, Rab6 (Fig. 4E, arrowheads), and some remained colocalized with EEA1 (Fig. 4D). In adaGb₃- or OHEtadaGb₃-treated CHO/Jurkat cells, Rab6 coincidence with internalized VT1 was insignificant compared with Vero cells (Fig. 4E), although VT1 staining was much less overall. VT1 did not colocalize with the lysosomal marker, Lamp-2, at any time (supplemental Fig. 1). After 6 h, most VT1 colocalized with the ER marker calnexin in Vero cells (Fig. 4E). In contrast, most of VT1 was lost and rarely overlapped with calnexin in adaGb₃- or OHEtadaGb₃₃-treated CHO/Jurkat cells (Fig. 4E). Essentially the same results were found for OHEtadaGb₃-treated CHO and Jurkat cells (Fig. 4, D and E) (and for cells treated with carboxyl or urea-adaGb₃. Cells treated with adaSGC (sulfatide) did not bind VT (supplemental Fig. 2). Thus, these Gb₃ analogues mediate VT1 internalization to early endosomes, but trafficking to Golgi/ER, as seen in Gb₃-expressing, VT1/VT2-sensitive Vero cells, is not detectable. The differential targeting (quantitated in supplemental Table 1) provides an explanation for the reduced efficacy of adaGb₃, compared with natural Gb₃, to mediate VT cytotoxicity.

AdaGb₃ Treatment of VT-sensitive Vero Cells Subverts Endogenous Gb₃-mediated VT1 and VT2 Retrograde Transport

To assess any relationship between these “exogenous” versus “endogenous” trafficking pathways, Gb₃-expressing Vero cells were also “reconstituted” with adaGb₃ or OHEtadaGb₃. Cells were treated with VT1 or VT2, and cell viability was monitored after 72 h (Fig. 5A). Vero cell VT1 sensitivity was reduced 10-fold after adaGb₃ treatment, but little effect on VT2 sensitivity was seen. OHEtadaGb₃ treatment of Vero cells had significantly less effect on VT1 sensitivity (Fig. 5A).

FIGURE 5.

A, toxicity of VT1/VT2 to adaGb₃- or OHEtadaGb₃-treated Vero cells. Vero cells were incubated with 0, 10, 20, or 50 μm adaGb₃ or OHEtadaGb₃ at 4 °C for 1 h. Cells were treated with 10-fold serially diluted VT1 or VT2 and incubated at 37 °C. Cell viability was monitored after 72 h and expressed as a percentage of control cells, which were treated with neither VT nor adaGb₃ analogues. B, staining of non-treated, adaGb₃-treated, or OHEtadaGb₃-treated Vero cells with fluorescent VT1 or VT2. Vero cells were treated with or without adaGb₃ or OHEtadaGb₃. Then Alexa488-VT1B or Texas Red-VT2 was bound on ice for 1 h. Cells were washed, fixed, and viewed with a confocal microscope. Bar, 10 μm. C, trafficking of VT1 in adaGb₃ or OHEtadaGb₃-treated Vero cells. VT1 was internalized at 37 °C for 10 min, 1 h, or 6 h in adaGb₃- or hydroxyethyl adaGb₃-treated Vero cells. VT1, EEA1, Rab6, and calnexin were localized as described in the legend to Fig. 4. D, trafficking of VT2 to the Golgi and ER in non-treated, adaGb₃-treated, or hydroxyethyl adaGb₃-treated Vero cells. Bound VT2 was internalized at 37 °C for 1 or 6 h. Cells were fixed, permeabilized, and labeled with anti-Rab6 (Golgi) or anti-calnexin (ER). Fluorescently stained cells were viewed with a confocal microscope. (Texas Red-VT2 is pseudocolored green, and organelle markers detected with anti-rabbit-Alexa488 colored red, for ease of comparison). E, VT2-induced vacuolation in non-treated, adaGb₃-treated, or OHEtadaGb₃-treated Vero cells. VT2 was bound and internalized at 37 °C for 6 h. VT2 (green) and calnexin (red) were detected as described for D. Fluorescently stained cells were viewed with a confocal microscope. Arrowhead, vacuoles; bar, 10 μm. Toxin colocalization with organelle markers was quantitated (supplemental Table 2).

AdaGb₃ treatment of Vero cells altered the VT1 and VT2 cell surface staining pattern (Fig. 5B). Gb₃-expressing VT-sensitive Vero cells showed punctate cell surface binding, as observed previously (12). However, Vero cells treated with adaGb₃ showed a more uniform cell surface VT1/VT2 labeling pattern, as if adaGb₃ had served to fuse previously separate cell surface Gb₃ domains. In contrast, OHEtadaGb₃-treated Vero cells retained the punctate cell surface binding of non-treated Vero cells (Fig. 5B). Quantitation of cell surface binding showed that adaGb₃ treatment reduced the amount of Alexa488-VT1B bound by 20–30% (supplemental Fig. 3), consistent with loss of a GSL clustering component in membrane binding affinity (50).

The intracellular trafficking of VT1 in non-treated (Fig. 4, D and E) and in adaGb₃- or OHEtadaGb₃-treated Vero cells (Fig. 5C) was then compared. After 10 min at 37 °C, a large portion of internalized VT1 colocalized with early endosomal EEA1 in all cases. In non-treated Vero cells after 1 h at 37 °C, some internalized VT1 colocalized with the Golgi marker, Rab6 (Fig. 4E, arrowheads), and some remained with EEA1 (Fig. 4D). However, in adaGb₃-treated Vero cells after 1 h at 37 °C, little VT1 colocalized with EEA1, and VT1 coincidence with Rab6 (Fig. 5C) was far less than in non-treated Vero cells. In contrast, in OHEtadaGb₃-treated Vero cells, a large portion of internalized VT1 remained colocalized with EEA1 at 1 h (Fig. 5C) as in non-treated Vero cells. After 6 h, most VT1 colocalized with the ER marker calnexin in non-treated Vero cells (Fig. 4E). However, VT1 rarely overlapped with calnexin in adaGb₃-treated cells. In contrast, VT1/calnexin overlap at 6 h was retained in OHEtadaGb₃-treated Vero cells (Fig. 5C) although reduced compared with in non-treated Vero cells.

Because adaGb₃ was less effective to reduce VT2, as compared with VT1, Vero cell sensitivity (Fig. 5A), the intracellular trafficking of VT2 was compared in Vero and adaGb₃-treated Vero cells (Fig. 5D). After 1 h at 37 °C, some internalized VT2 was overlapping with the Golgi marker Rab6, in non-treated Vero cells (Fig. 5C, arrowheads). In adaGb₃-treated Vero cells, Rab6 coincidence with VT2 was barely detectable at 1 h, but in OHEtadaGb₃-treated Vero cells, internalized VT2 colocalized with Rab6 as for non-treated Vero cells. After 6 h of culture, significant VT2 colocalization with the ER marker calnexin in non-treated Vero cells was found. In contrast, VT2 showed no overlap with calnexin in adaGb₃-treated Vero cells at this time. However, for OHEtadaGb₃-treated cells, VT2/calnexin overlap (Fig. 5C) was similar to that in non-treated Vero cells after 6 h.

Thus, adaGb₃ changed the intracellular VT1 and VT2 trafficking in Gb₃-expressing cells. Initial endosomal entry was retained, but subsequent Golgi/ER traffic was compromised. OHEtadaGb₃ treatment had little obvious effect.

As previously reported (12), intracellular perinuclear vacuoles were detected in some VT2-treated Vero cells (Fig. 5E, arrowheads). VT2 vacuolation was retained for adaGb₃- or OHEtadaGb₃-treated Vero cells, but VT2 staining of these vesicles was lost (Fig. 5, D and E).

BFA, which prevents Golgi-dependent retrograde traffic, protects cells from VT1 (51). To confirm that intracellular VT trafficking in adaGb₃-treated cells mediates toxicity without Golgi access, the effects of BFA on VT-induced cytotoxicity in adaGb₃-treated CHO/Jurkat cells (Fig. 6A) and Vero cells (Fig. 6B) were compared. BFA was virtually ineffective to prevent VT1/VT2 killing of adaGb₃-treated CHO cells and provided minimal protection to adaGb₃-treated Jurkat cells (Fig. 6A). In contrast, BFA completely protected Vero cells against VT1/VT2 (Fig. 6B). However, for adaGb₃-treated Vero cells, VT1/VT2 cytotoxicity became BFA-resistant. These data indicate that intracellular VT traffic and toxicity in adaGb₃-treated CHO, Jurkat, and Vero cells is Golgi-independent.

FIGURE 6.

Effect of BFA on VT1- and VT2-induced cell killing. Cells were incubated ± 0.5 μg/ml BFA for 30 min prior to the VT addition. VT1 was added at 10 μg/ml for adaGb₃-treated CHO/Jurkat cells (A), 0.1 ng/ml for non-treated Vero cells (B, top), and 100 ng/ml for adaGb₃-treated Vero cells (B, bottom). VT2 was added at 10 μg/ml for adaGb₃-treated CHO/Jurkat cells, 1 ng/ml for non-treated Vero cells, and 100 ng/ml for adaGb₃-treated Vero cells. Cell viability was monitored after 21.5 h and expressed as a percentage of control cells in the absence of VT.

AdaGb₃ Analog Inhibition of VT Cytotoxicity

Previously, we reported that adaGb₃ competed with Gb₃ for VT1 binding in receptor ELISA (44) and was effective to prevent the Vero cell binding of both VT1 and VT2 (12, 46). We therefore compared the efficacy of other adaGb₃ analogues for protection of Vero cells from VT binding.

Increasing concentrations of adaGb₃, hydroxylethyladaGb₃, carboxyadaGb₃, or adabisGb₃ were preincubated with VT1/VT2 and tested for reduction of Vero cell cytotoxicity (Fig. 7). Of the “monomer” species, only adaGb₃ showed significant protection against VT1 and, more effectively, VT2. CarboxyadaGb₃ and hydroxylethyladaGb₃ showed no protection. AdabisGb₃ inhibited VT1/VT2 Vero cell cytotoxicity to a greater extent than adaGb₃. Cytotoxicity of VT1/VT2 preincubated with 50 μm adalbisGb₃ was reduced 150–250-fold. AdaGb₃ had no effect on Vero cell susceptibility to ricin, which also undergoes Golgi/ER retrograde transport (52); thus, overall transport to the Golgi and ER was not blocked (supplemental Fig. 4).

FIGURE 7.

Inhibition of verotoxin cytotoxicity. Increasing concentrations of adaGb₃, carboxyadaGb₃, OHEtadaGb₃, or adabisGb₃ were premixed with VT1 or VT2 dilutions at 37 °C for 1 h and then added to Vero cells for 1 h. Cells were then washed and incubated at 37 °C for 72 h. Live cells were stained with crystal violet, and viability was plotted as a percentage of untreated Vero cells. From this, the CD₅₀ values of VT1 (left)/VT2 (right) preincubated with adaGb₃ analogues were determined and plotted as a function of analog concentration.

Unlike Sphingolipids, AdaGb₃ Does Not Stabilize Membrane Cholesterol in Phospholipid Liposomes

Because cholesterol is central to intracellular membrane GSL traffic, the ability of Gb₃ and adaGb₃ to interact with cholesterol was compared by their ability to induce resistance to MβCD cholesterol extraction from a model phospholipid membrane (48, 53). Approximately 40% of the [³H]cholesterol in PC liposomes was extracted by 0.25 mm MβCD (Fig. 8) from the liposomal pellet. Inclusion of SPM or Gb₃ within the liposomes significantly reduced the extracted cholesterol to 10 and 15%, respectively, indicating stabilization of the cholesterol within the membrane. In contrast, inclusion of adaGb₃ consistently increased cholesterol susceptibility to MβCD extraction to 50%. Inclusion of adaGb₃ together with SPM had no effect on SPM-cholesterol stabilization, but adaGb₃ reduced the stabilizing effect of Gb₃ on cholesterol by ∼30%, indicating that membrane adaGb₃ interfered with the interaction between Gb₃ and cholesterol. Thus, in contrast to Gb₃ and SPM, adaGb₃ destabilizes rather than stabilizes membrane cholesterol and partially reverses Gb₃-cholesterol stabilization. This effect could explain the lack of adaGb₃ Golgi/ER trafficking and the adaGb₃ modulation of VT1-bound Gb₃ intracellular trafficking observed.

FIGURE 8.

GSLs/adaGSL stabilize/destabilize cholesterol within PC liposomes. The effect of inclusion of SPM, Gb₃, or adaGb₃ (alone and in combination) on [³H]cholesterol availability to MβCD extraction from cholesterol/PC liposomes was determined. The percentage of cholesterol extracted by PBS (open bars) or 0.25 mm MβCD (gray bars) after 1 h at room temperature is shown. As expected, inclusion of the sphingolipid SPM or Gb₃ increased resistance to cholesterol extraction by MβCD, but adaGb₃ showed a reverse effect. Moreover the inclusion of adaGb₃ together with Gb₃, but not SPM, partially reversed the stabilizing effect on liposomal cholesterol. Error bars, S.D.

AdaGb₃ Does Not Alter Internalization and Retrograde Traffic of CT

To address whether the effect of adaGb₃ on native Gb₃-mediated VT intracellular traffic might be in any way selective, we examined the intracellular retrograde traffic of GM1-bound Cy3-CTB in adaGb₃-treated HEK-293 cells. VT1 and CT preferentially bind different Vero cell subsets during the cell cycle (54), making comparison of differential trafficking in a single cell difficult. CHO cells do not express GM1 (55) (or Gb₃), and cell suspension cultures (Jurkat) are inconvenient to study intracellular traffic. We therefore treated HEK-293 cells (Gb₃−ve, GM1+ve) with adaGb₃ and monitored the cell binding and internalization of VT1B and CTB (Fig. 9). The cell surface binding of CTB to HEK-293 cells at 4 °C was largely unaffected by adaGb₃ treatment (Fig. 9). VTB bound the HEK-293 cell surface only after adaGb₃ treatment and colocalized extensively with CTB at 4 °C (Fig. 9). Warming to 37 °C induced plasma membrane-bound CTB internalization to the same juxtanuclear Golgi structures in both control and adaGb₃-treated cells (Fig. 9). VTB, however, was internalized into punctate intracellular vesicles, for the most part, distinct from CTB-labeled Golgi (Fig. 9). VTB containing vesicles were in the Golgi area (as defined by CTB) but remained separate from CTB. Thus, cell surface-colocalized GM1-bound CTB and adaGb₃-bound VTB are differentially trafficked to separate structures within the cell, such that GM1-CTB Golgi retrograde traffic is retained, whereas adaGb₃-VT1 is trafficked to an alternative destination within the same cells.

FIGURE 9.

AdaGb₃ insertion and VT trafficking do not perturb the intracellular traffic of cholera toxin. AdaGb₃ was inserted into HEK-293 cells, and the simultaneous binding and internalization of Alexa488-VTB and Cy3-CTB were assessed. Top panels, binding at 4 °C; bottom panels, detection after 1 h of internalization at 37 °C. DAPI nuclear staining is shown for VT1 B-treated cells without adaGb₃. Bar, 10 μm. Only plasma membrane staining is seen at 4 °C. AdaGb₃-bound VTB and GM1-bound CTB show significant cell surface overlap. At 37 °C, Cy3-CTB is internalized to juxtanuclear Golgi in both control and adaGb₃-treated cells. In contrast, Alexa488-VTB is internalized to punctate vesicles distinct from CTB-labeled Golgi in adaGb₃-treated cells.

DISCUSSION

The binding of the VT family of AB₅ subunit toxins to their receptor GSL, globotriaosyl ceramide, is of interest for many reasons. First, the VT B subunit pentamer binding to Gb₃ provides the basis for renal glomerular endothelial cell targeting following systemic verotoxemia and therefore plays a central role in the pathology of HUS (17, 56), which remains a life-threatening complication of gastrointestinal verotoxin-producing E. coli infection, an ever increasing threat in the developed world (9). Second, verotoxin binding to cell surface Gb₃ provides an index of the complex manner in which cell surface GSLs can be presented within a bilayer for ligand recognition and is thus a probe for GSL membrane organization (57). Third, Gb₃ and verotoxin internalization and intracellular traffic provide a probe of the molecular basis of retrograde transport to the ER (58). Fourth, Gb₃ is up-regulated in many human tumor cells, and thus verotoxin itself (59, 60, 61) or the B subunit pentamer coupled to cytotoxic drugs (62, 63) offers new antineoplastic approaches (64). In this area also, as with HUS, endothelial cells within the neovasculature express Gb₃ and are VT-sensitive (65, 66). Last, Gb₃ expression is a key risk factor for HIV susceptibility (67), and aglycone modulation of gp120-Gb₃ binding is similar to that of VT1 (16).

Membrane GSL organization and its role in intracellular vesicular traffic are poorly understood but are of high potential significance (68–70). The amphipathic GSL analogues we have made, which in part retain the receptor function of membrane GSLs (71), provide new insight into these processes and the means to alter cellular GSL metabolism selectively (72). We now show that adaGSLs have an immediate effect on plasma membrane GSL receptor function and intracellular traffic. Our results, summarized in Scheme 1, include several novel observations as described below.

SCHEME 1.

1–3, endogenous Gb₃ within cell surface lipid rafts mediates VT internalization (1), endosomal transport to Golgi-associated vesicle (2), and retrograde transport to Golgi and thence ER (3). 4, plasma membrane adaGb₃ can mediate VT internalization to early endosomes without further retrograde transport. 5, A-subunit may be released into cytosol. 6, toxin-adaGb₃ complex may be recycled and lost from the cell surface. Some VT may undergo lysosomal degradation. 7, mixing of endogenous and adaGb₃ alters Gb₃ organization to disburse Gb₃ from raft restriction. In combination, Gb₃ and adaGb₃ mediate VT internalization to endosomes (5) or Golgi-associated vesicles (3), but retrograde transport (8) to Golgi and ER does not occur.

Induction of VT1/VT2 Cell Sensitivity

We synthesized a series of modified adaGb₃ species and found preferential VT2 (cf. VT1) binding. These Gb₃ mimics incorporated into the plasma membrane of receptor negative cells to induce cell VT1/VT2 cytotoxicity. This is the first report in which a Gb₃ derivative has been shown capable of this function and opens the potential to make any cell sensitive to VT cytopathology. Significantly, in such “pseudoreceptor”-reconstituted cells, the toxin-receptor complex was internalized to endosomes but did not mediate Golgi/ER retrograde transport, as for endogenous Gb₃-mediated VT traffic (Scheme 1). Despite the different functional groups present, this was seen for all adaGb₃ species. Prolonged association with EEA1 vesicles was seen for OHEtadaGb₃ but not other adaGb₃-treated cells. Lack of Golgi/ER targeting suggests that A-subunit cytosolic translocation from endosomes mediates the induced toxicity (51, 73). The internalized toxin was less long-lived compared with that within Vero cells. (70–100% versus 30–50% loss in 1–6 h; see supplemental Table 2). This may indicate proteolysis or, more likely, loss due to recycling from endosomes to the cell surface (Scheme 1). This VT loss from endosomes may contribute to the lack of Golgi/ER VT detection. We have not observed any adaGSL breakdown within the time frame of the present experiments.

The intracellular transport of adaGb₃ is distinct from exogenous BODIPY and NBD-GSL analogues, which readily traffic from the cell surface to the Golgi (31). The more planar structure of these fluorescent substituents may permit a cholesterol interaction. This clearly shows that the lipid structure of membrane GSLs can provide differential intracellular membrane addresses for exogenous (and, by inference, endogenous) GSL species. AdaGSLs may be defective in lateral membrane “connectivity” (74).

AdaGb₃-reconstituted CHO cells were significantly less sensitive to VT1/VT2 than similarly reconstituted Jurkat cells. This indicates that properties in addition to receptor status regulate cytotoxicity.

AdabisGb₃ Does Not Induce VT1/VT2 Cell Sensitivity

Our second novel observation is that adabisGb₃, in which two lyso-Gb₃s are coupled to a single adamantane frame, does not incorporate into the cell membrane. This lack of membrane partitioning of adabisGb₃ is of significance and must be a structure-related property. Although the hydrophobicity of adabisGb₃ is significantly reduced compared with adaGb₃, gangliosides of greater polarity are readily taken up into the membranes of cultured cells (75). It is possible that the 1–3 coupling to the adamantane frame positions the sphingosine tails in a skewed orientation relative to one another, and as such, the non-parallel alkyl chains may be unable to insert and stack in a lamellar bilayer to prevent plasma membrane incorporation. Nevertheless, in solution, adabisGb₃ can bind to VT1/VT2 tightly to function as an extracellular inhibitor of VT1/VT2 cell binding. Furthermore, adabisGb₃ itself is not toxic to Jurkat cells, whereas high concentrations of adaGb₃ can be toxic. This is probably a function of the lack of cell membrane insertion of adabisGb₃. Thus, adabisGb₃ provides a potential basis for protection against verotoxemia.

AdaGb₃ Compromises Endogenous Gb₃-mediated VT1 Retrograde Transport

Our third observation is that adaGb₃ can subvert the natural retrograde transport of VT1/VT2 bound to endogenous cellular Gb₃. This property is dependent on the lipid structure of the Gb₃ mimic; OHEtadaGb₃ did not have this effect. In addition, the effect was selective, in that GM1-CTB intracellular traffic was virtually unaffected. AdaGb₃ plasma membrane incorporation altered the surface distribution of VT1 and VT2 overall, generating a more uniform cell surface receptor (Gb₃ + adaGb₃) distribution. This implies cooperation between the membrane-incorporated adaGb₃ and non-uniformly distributed Gb₃ (Scheme 1). The subsequent retrograde transport of VT1/VT2 from endosomes to Golgi and hence to ER was largely circumvented in adaGb₃-treated Vero cells. Consistent with the lack of retrograde Golgi/ER transport, VT1/VT2 cytotoxicity for adaGb₃-treated cells became insensitive to BFA protection. VT1 (but not VT2) cytotoxicity was significantly reduced for adaGb₃-treated Vero cells, suggesting more effective cytosolic VT2 A-subunit translocation from endosomes.

Lipid Structural Dependence

The early intracellular transport of VT-bound adaGb₃ in Gb₃-negative cells is similar to endogenous Gb₃-bound VT, in that the toxin is rapidly targeted to early endosomes. We did not observe any VT/adaGb₃ colocalization with Lamp-2, indicating a fate other than lysosomal degradation. For Gb₃-positive plasma membranes into which adaGb₃ is incorporated, VT1 and VT2 must simultaneously bind both endogenous Gb₃ and incorporated adaGb₃. The VTB subunit pentamer will probably bind five Gb₃ molecules (76) to induce membrane curvature by compaction (77). Membrane adaGb₃ has a larger molecular area and is more resistant to compression than Gb₃ (25), and inclusion within this toxin-GSL membrane complex could compromise compaction/membrane curvature. The non-uniform Vero cell distribution of Gb₃, as detected by VT1 or VT2 binding, was rendered more uniform after adaGb₃ incorporation, consistent with a lack of clustering (77). This redistribution was not seen after OHEtadaGb₃ incorporation and correlates with protection against VT cytotoxicity by adaGb₃ but not OHEtadaGb₃.

Subsequent internalization was similar for adaGb₃-treated CHO or Jurkat cells and, initially, for untreated Vero cells, in that early endosomes were targeted, but in adaGb₃-treated cells, the later retrograde transport to Golgi/ER was compromised. Thus, the adaGb₃ internalization and trafficking route dominated that of endogenous Gb₃. This was not observed for OHEtadaGb₃-treated cells.

Retrograde transport overall was not affected because Vero cell susceptibility to ricin and Golgi traffic of cholera toxin/GM1 in HEK-293 cells were unaffected by adaGb₃. Our previous studies showed partial cell surface colocalization but the separate internalization of VT1 and CT (78). Internalization is mediated through both clathrin-dependent and clathrin-independent mechanisms, both of which access Golgi/ER retrograde transport (27). Although VT2 bound adaGb₃ in preference to native Gb₃ by TLC overlay, adaGb₃ did not have a greater effect on VT2 compared with VT1 intracellular traffic, indicating that membrane organization rather than binding per se is the key factor.

AdaGb₃ Does Not Interact with Cholesterol

The endosome to TGN transport of VT is compromised by clathrin blockade (79, 80). Cholesterol depletion can reduce both clathrin-dependent (81) and caveolin-dependent (82) internalization and can block actin-dependent endosome-TGN CT (83) and endosome-TGN ricin retrograde transit (84). GSL-cholesterol interaction is key to the formation of liquid-ordered domains in model membranes (85) and increased order in cell membranes (74). Aberrant cholesterol and GSL retrograde transport (and metabolism) are intimately connected in sphingolipid storage diseases (86).

AdaGb₃ was unable to stabilize liposomal membrane cholesterol and partially reduced cholesterol stabilization by Gb₃; these novel biophysical properties may explain the effect of adaGb₃ on intracellular VT routing. If the transition of VT1/VT2-bound Gb₃ between endosomes and TGN is cholesterol-dependent, adaGb₃-bound VT should transit ineffectively. Similarly, if the punctate cell surface Gb₃ distribution were cholesterol-dependent, a more uniform distribution for adaGb₃ would be expected.

In adaGb₃-treated cells, VT1/VT2 cell surface distribution was altered, but internalization was similar. Thus, the cholesterol-dependent “decision” to undergo Golgi/ER retrograde transport may be taken at the cell surface.

GSL-cholesterol interaction can promote (25) or prevent ligand-GSL binding (21). This may be a function of GSL fatty acid content (16, 20), GSL/cholesterol ratio (20), and membrane curvature (21). The carbohydrate conformation of GSLs is changed when in complex with cholesterol (20). Hydrogen bonding of the sterol OH and adjacent GSL anomeric oxygen “bends” the carbohydrate from a membrane perpendicular to parallel orientation (18, 20, 87), restricts exogenous ligand binding (21), and shields the sterol from water interaction by an “umbrella” effect (87). In model membranes, the GSL glycan thickness is an inverse function of the cholesterol concentration, from perpendicular (thickest) to parallel (thinnest) (20), suggesting a range of intermediate cholesterol-dependent carbohydrate conformations. Nine potential membrane GSL carbohydrate conformations have been modeled (88).

The lack of cholesterol interaction we show predicts adaGSLs to be resistant to this masking effect, and such analogues might therefore have a binding advantage over natural GSL species in cholesterol-containing membranes. This could explain the dominant trafficking effect of adaGb₃ in Gb₃-expressing cells and the ability of adamantyl monohexosyl ceramides to modulate GSL metabolism (72). The lack of OHEtadaGb₃ efficacy could be consistent with this mechanism. The GSL conformational change induced by cholesterol is mimicked in GSLs containing 2-hydroxy fatty acids (18). Hydrogen bonding between the sugar and fatty acid OH can similarly bend the carbohydrate (89). The OH of OHEtadaGb₃ might similarly mediate such a carbohydrate conformational change, and the potential ligand binding advantage would be lost.

OHEtadaGb₃ induced VT susceptibility in Gb₃-negative cells. In OHEtadaGb₃-treated (or carboxyadaGb₃- or urea-adaGb₃-treated), Gb₃-negative cells, intracellular VT trafficking is similar to that of adaGb₃-treated cells. This suggested that the incorporated adaGb₃ analogues define a shared intracellular retrograde routing, which is different from natural Gb₃ traffic. The differential VT cell sensitivity may due to binding differences between adaGb₃ analogues. AdaGb₃ is strongly bound by VT1/VT2. VT1-adaGb₃ binding was similar to VT1-Gb₃ binding, and VT2-adaGb₃ binding was ∼4-fold greater than VT2-Gb₃ binding. On the other hand, VT1/OHEtadaGb₃ binding was significantly less than VT1/adaGb₃ or Gb₃. VT2/OHEtadaGb₃ binding was similar to VT2/Gb₃ binding, significantly weaker than VT2/adaGb₃ binding. These binding differences are consistent with an OHEtadaGb₃ conformational restriction that could explain the VT sensitivity, cell surface staining, and intracellular trafficking observed in adaGb₃-treated versus OHEtadaGb₃-treated Vero cells. The punctate VT1/VT2 Vero cell surface staining of Gb₃-containing plasma membrane domains that we observed at 4 °C, as previously (12), was replaced in adaGb₃-treated Vero cells by a more uniform membrane staining pattern. Membrane-incorporated adaGb₃ may intercalate to disperse such domains by reducing cholesterol interaction and “fill the gaps” between domains (Scheme 1). This may reroute intracellular traffic similar to non-raft, as compared with raft Gb₃ (22).

We found that the VT2-induced vacuolation we previously reported for a subpopulation of Vero cells (12) was retained for adaGb₃-treated or OHEtadaGb₃-treated Vero cells. Thus, this vacuolation response is independent of VT2 retrograde traffic to the Golgi/ER, which may relate to the increased clinical severity of VT2. In Vero cells, VT2 was present in the limiting membrane of the VT2-induced vacuoles. However, for adaGb₃-treated or OHEtadaGb₃-treated Vero cells, VT2 was not detected in the vacuolar membrane, indicating that the vacuoles arise from a signaling mechanism rather than direct effect of toxin membrane Gb₃ binding.

In conclusion, we show the novel, lipid-dependent, pseudoreceptor function of adaGb₃ mimics in receptor negative cells and their structure-dependent domination over native intracellular Gb₃-dependent but not GM1-dependent traffic. This may be mediated by the lack of cholesterol association of adaGb3 mimics and their ability to preferentially reduce membrane Gb₃-cholesterol interaction.