Shiga toxin glycosphingolipid receptors of Vero-B4 kidney epithelial cells and their membrane microdomain lipid environment1[S]

Shiga toxins (Stxs) are produced by enterohemorrhagic Escherichia coli (EHEC), which cause human infections with an often fatal outcome. Vero cell lines, derived from African green monkey kidney, represent the gold standard for determining the cytotoxic effects of Stxs. Despite their global use, knowledge about the exact structures of the Stx receptor glycosphingolipids (GSLs) and their assembly in lipid rafts is poor. Here we present a comprehensive structural analysis of Stx receptor GSLs and their distribution to detergent-resistant membranes (DRMs), which were prepared from Vero-B4 cells and used as lipid raft equivalents. We identified globotriaosylceramide (Gb3Cer) and globotetraosylceramide (Gb4Cer) as the GSL receptors for Stx1a, Stx2a, and Stx2e subtypes using TLC overlay detection combined with MS. The uncommon Stx receptor, globopentaosylceramide (Gb5Cer, Galβ3GalNAcβ3Galα4Galβ4Glcβ1Cer), which was specifically recognized (in addition to Gb3Cer and Gb4Cer) by Stx2e, was fully structurally characterized. Lipoforms of Stx receptor GSLs were found to mainly harbor ceramide moieties composed of sphingosine (d18:1) and C24:0/C24:1 or C16:0 fatty acid. Moreover, co-occurrence with lipid raft markers, SM and cholesterol, in DRMs suggested GSL association with membrane microdomains. This study provides the basis for further exploring the functional impact of lipid raft-associated Stx receptors for toxin-mediated injury of Vero-B4 cells.

toxin or Stxs from pathogenic E. coli ( 35,56,57 ). Cholesterol-rich lipid rafts make them relatively resistant to solubilization by nonionic detergents, allowing for the isolation of detergent-resistant membranes (DRMs) from low buoyant density fractions after sucrose density ultracentrifugation ( 58 ). Although a matter of debate ( 59 ), compositional analysis of lipid rafts is largely obtained from DRMs as the ruling method to assigning lipid raft-association ( 60 ).
Here we present the fi rst comprehensive investigation of Vero cells regarding identifi cation and structural characterization of globo-series Stx receptor GSLs for various Stx subtypes, their molecular assembly with phospholipids, and cholesterol in membrane microdomains using DRMs and Stx-mediated cytotoxicity. This study might be helpful to further our understanding of Vero cell sensitivity to various Stx subtypes and the functional role of lipid rafts in kidney epithelial cells.

Cultivation of Vero-B4 cells
Vero-B4 cells were obtained from the Leibniz Institute Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbH (DSMZ, German Collection of Microorganisms and Cell Cultures, Braunschweig, Germany; DSMZ number ACC 33 ). The cell line, which was established from the kidney of a normal adult African green monkey ( Cercopithecus aethiops ) in Japan ( 1 ), was grown in MEM culture medium with nonessential amino acids and 1 mM sodium pyruvate (Lonza, Verviers, Belgium), MEM vitamin mix (Lonza, Wakersville, MD) plus 5% FCS (PAA, Pasching, Austria) in a humidifi ed atmosphere at 37°C with 5% CO 2 . The cells grow as adherent-elongated fi broblast-like cells in monolayers. Cultures were routinely passaged every 2 to 3 days using 0.25% trypsin-EDTA (Invitrogen, Karlsruhe, Germany; catalog number 25200) before cells became confl uent. Appropriate cell production for the isolation of preparative amounts of GSLs from total cells (see Purifi cation of neutral GSLs from Vero-B4 cells below) and for the preparation of sucrose density gradient fractions (see Preparation of sucrose density gradient fractions: DRMs and nonDRM fractions below) was performed in 175 cm 2 tissue culture fl asks (Greiner Bio-One, Frickenhausen, Germany), as previously described for endothelial cells (61)(62)(63).

Cell cytotoxicity assay
Indirect determination of cell viability was performed using the crystal violet assay, as previously described (61)(62)(63). Briefl y, Vero-B4 cells were grown until subconfl uence in tissue culture fl asks (Greiner Bio-One), trypsinized and seeded in 100 l volumes in 96-well tissue culture plates (Corning Inc., Corning, NY) (initial cell seeding density of 4 × 10 3 cells/well). One hundred microliters of Stx-containing supernatants from E. coli cultures (see Stx1a, Stx2a, Stx2e, anti-Stx, anti-GSL, and secondary antibodies below) were added in various dilutions in cell culture medium to cell culture plate wells (fi nal volume of 200 l) and incubated for 48 h (37°C, 5% CO 2 ). Stx-free cell culture medium served as a control. Incubation was stopped by removal of the diluted supernatants. Remaining adherent cells were fi xed with formalin, stained with crystal violet, and densitometrically quantifi ed (61)(62)(63). Results represent the mean ± SD of quadruplicate determinations and are portrayed as percentage values of untreated control cells. synonymously named as "verotoxins" or "verocytotoxins" ( 28,29 ) owing to their cytotoxic capacity toward Vero cells, preferentially target microvascular endothelial cells of the human kidneys and the brain (30)(31)(32) that follow gastrointestinal infection and culminate as renal insuffi ciency and an often fatal outcome ( 26,31,33 ). Stxs are composed of a single 30 kDa A subunit and a pentamer of noncovalently attached identical 7 kDa B subunits ( 34 ). The B pentamer binds to cell surface-exposed glycosphingolipids (GSLs) of the globo-series and is therefore dependent on these lipids for transport into the cells ( 35 ). After retrograde routing of the holotoxin to the endoplasmic reticulum and cleavage of the A subunit, the A 1 fragment halts eukaryotic protein biosynthesis by inactivating ribosomes leading to cell death ( 36,37 ). The Stx subtypes, Stx1a, Stx2a, and Stx2e [for nomenclature of Stx subtypes refer to Scheutz et al. ( 38 )], which have so far been investigated in more detail, can be distinguished with regard to their exact binding specifi city ( 39 ). Variants of the Stx1a subtype have been shown to exhibit preferential and moderate binding toward globotriaosylceramide (Gb3Cer, Gal ␣ 4Gal ␤ 4Glc ␤ 1Cer) and globotetraosylceramide (Gb4Cer, GalNAc ␤ 3Gal ␣ 4Gal ␤ 4Glc ␤ 1Cer), respectively. Stx2a variants prefer Gb3Cer and exhibit only marginal binding toward Gb4Cer, while Stx2e binds, in addition to Gb3Cer and a preference for Gb4Cer, also to the Forssman GSL, which represents a pentahexosylceramide with GalNAc ␣ 3GalNAc ␤ 3Gal ␣ 4Gal ␤ 4Glc ␤ 1Cer structure ( 39 ), indicating promiscuous binding of Stx2e to globo-series-related GSLs with elongated Gb3Cer/Gb4Cer core structures.
GSLs belong to the structurally very diverse group of sphingolipids (40)(41)(42) and have long been considered mysterious and puzzling riddles, beginning with the initial naming of "sphingosin" in the 1880s by the German physician and biochemist J. L. W. Thudichum due to their enigmatic ("Sphinx-like") properties ( 43,44 ). Their biosynthesis is rather complex ( 42 ) and they play important roles in the formation of the nervous system, where they are especially abundant (45)(46)(47). "Sphingolipidomic" analysis is now becoming feasible, at least for ceramides and other bioactive sphingolipid backbones, using MS ( 48 ), including novel MALDI MS imaging techniques ( 49 ). GSLs are believed to be involved in many cellular events such as traffi cking, signaling, and cellular interaction ( 50 ), as well as infl ammation ( 51 ). They function as important components in keratinocytes, for instance, where preservation of the skin barrier depends on GSL biosynthesis, and in enterocytes of the small intestine, where GSLs are involved in endocytosis and vesicular transport ( 50 ). In particular, they participate in the formation of lipid rafts, which are considered to be functional nanoscale membrane domains enriched in cholesterol and sphingolipids, characteristic of the external leafl et of cell membranes in particular ( 52 ). Although the understanding of the structure of lipid rafts in living cells is quite limited, they are suggested to be involved in a great variety of cellular functions and biological events (53)(54)(55). Plasma membrane organization in microdomains seems to be a functional requirement for binding and effi cient internalization of pathogens and toxins, such as Vibrio cholerae (30/9/25/18/6, each by volume) and stained with molybdenum blue Dittmer-Lester reagent ( 73,74 ). Cholesterol was stained upon TLC separation in chloroform/acetone (96/4, v/v) with manganese(II)chloride ( 62,75 ). Neutral GSLs were separated in chloroform/methanol/water (120/70/17, each by volume) and stained with orcinol ( 76 ).

Densitometry of phospholipids, cholesterol, GSLs, and TLC immunostained GSLs
All TLC-separated lipids were quantifi ed densitometrically by means of a CD60 scanner (Desaga, Heidelberg, Germany, software ProQuant ® , version 1.06.000) in refl ectance mode with a light beam slit of 0.02 × 4 mm. Molybdenum blue-colored phospholipid, dark brownish cholesterol, and deep purpled orcinol bands were scanned at wavelengths of = 700 nm, = 365 nm, and = 544 nm, respectively, and immunopositive GSL bands were densitometrically quantifi ed at a wavelength of = 630 nm ( 62,70,78 ). All scans were performed in triplicate from three independent experiments. Linear results for relative percentage values and concentrations of lipids were obtained for colored lipid bands as optical density values between ‫ف‬ 0.1 and ‫ف‬ 1.

ESI MS
Native and permethylated GSLs, as well as phospholipids, were analyzed by nanoESI MS using a SYNAPT G2-S mass spectrometer (Waters, Manchester, UK) equipped with a Z-spray source. The source settings were: temperature 80°C, capillary voltage 0.8 kV, sampling cone voltage 20 V, and offset voltage 50 V. For fragmentation by low energy collision-induced dissociation (CID), the lipid precursor ions were selected in the quadrupole analyzer and ion mobility separation was applied (wave velocity 700-800 m/s, wave height 40 V, nitrogen gas fl ow rate 90 ml/min, helium gas fl ow rate 180 ml/min). Subsequent fragmentation was performed in the transfer cell with a collision gas (Ar) fl ow rate of 2.0 ml/min and collision energies up to 100 eV (E lab ). Individual GSLs and phospholipids were detected as singly charged monosodiated [M+Na] + and/or protonated [M+H] + species and structures were deduced from CID spectra. Purifi ed neutral GSLs from Vero-B4 cells, GSL extracts from immunostained Gb5Cer, and lipid extracts from density gradient fractions F2 and F7 were dried under a stream of nitrogen, dissolved in methanol, and analyzed in the positive ion sensitivity mode.

Statistics
Nonparametric statistical analysis was performed in the R v3.2.0 computing environment ( 81 ). Spearman's rank correlation coeffi cient r S was applied to determine the strength of association between phospholipids, cholesterol, and GSLs in DRM and nonDRM fractions obtained from Vero-B4 cells (see Preparation of sucrose density gradient fractions: DRMs and nonDRM fractions above). Ranks were allocated to measured values with rank 1 corresponding to the highest value, rank 2 to the second highest value, etc., and rank 8 to the lowest value. All tests were two-tailed and the r S and respective P values were calculated with R software, whereby r S values were considered signifi cant at adjusted P values <0.01.

RESULTS
The orcinol stain of TLC-separated neutral GSLs of Vero-B4 cells ( Fig. 1 , Vero) suggests the presence of globo-series GSLs with Gb3Cer as the dominant GSL when compared with reference R1, composed of equimolar concentrations of Gb3Cer, Gb4Cer, and the Forssman GSL, and reference R2 of neutral GSLs from human erythrocytes, containing mostly Gb4Cer and lesser amounts of Gb3Cer. In Vero-B4 cells, Gb3Cer makes up about 52% of total neutral GSLs, as determined on the average of three independent preparations by densitometric quantifi cation (see exemplary scan in Fig. 1 ). Gb4Cer amounts to 9% and postulated pentahexosylceramide, with supposed Gb5Cer structure, appears with 12% in the chromatogram of Vero-B4 cells.

Preparation of sucrose density gradient fractions: DRMs and nonDRM fractions
Sucrose density gradient fractions were prepared according to Brown and Rose ( 80 ), as previously described ( 62,69,70 ) with minor modifi cations. The following detergent extraction was done on ice with prechilled solutions and centrifugations were performed at 4°C. Briefl y, 1 × 10 8 confl uent grown cells of each approach were disrupted in lysis buffer, cell debris was removed by gentle centrifugation (400 g ), and separation of membranes from cytosol was performed by ultracentrifugation (150,000 g ) of the supernatant. The membrane sediment was thoroughly solubilized in 1 ml of 1% Triton X-100 buffer and mixed with 1 ml of 85% sucrose. This solution (42.5% sucrose) was then overlayed with a discontinuous sucrose bottom-to-top gradient of 30% and 5% sucrose. Three top DRM and DRM-associated fractions (F1 to F3) were separated from the bulk of nonDRM fractions (F4 to F8), which were further subdivided into intermediate (F4 to F6) and bottom fractions (F7 and F8) by ultracentrifugation (200,000 g ), whereby the light DRMs fl oated to the interface between 5% and 30% sucrose in the density gradient. Eight fractions of 1.5 ml volume each were collected and used for lipid analysis (see next section).

Isolation of phospholipids, cholesterol, and GSLs from sucrose density gradient fractions
For phospholipid analysis, 0.375 ml (one quarter) of each sucrose gradient fraction was dialyzed at 4°C for 3 days against distilled water and freeze dried. Total lipids were solubilized under short sonication in chloroform/methanol (2/1, v/v) and adjusted to defi ned volumes corresponding to 1 × 10 5 cells/ l. For cholesterol and GSL analysis, 0.75 ml of each sucrose gradient fraction (two quarters) was adjusted to 1 N NaOH, incubated for 1 h at 37°C under gentle rotation to saponify glycerophospholipids and triglycerides, and afterwards neutralized with HCl. The samples were dialyzed at 4°C for 3 days against distilled water, freeze-dried, and then taken up in chloroform/methanol (2/1, v/v) corresponding to 1 × 10 5 cells/ l. Phospholipids, cholesterol, and immunostained GSLs were semi-quantifi ed by densitometric scanning (see Densitometry of phospholipids, cholesterol, GSLs, and TLC immunostained GSLs above). The remaining 0.375 ml aliquot of the sucrose gradient fractions served as reserve.

Permethylation of neutral GSLs
An aliquot of neutral GSLs corresponding to 6 × 10 6 Vero-B4 cells was evaporated under a stream of nitrogen and resuspended in 500 l of dimethyl sulfoxide for 30 min. Ground sodium hydroxide pellets and 500 l of methyl iodide were added, followed by centrifugation at 1,000 g for 30 min. The supernatant was concentrated to a volume of 500 l under a stream of nitrogen and permethylated GSLs were purifi ed by Sephadex-LH-20 (GE Healthcare Bio-Sciences AB, Uppsala, Sweden) chromatography. For this purpose, the column was equilibrated with chloroform/methanol (1/1, v/v) and GSL derivatives were eluted with chloroform/methanol (1/1, v/v) in seven fractions of 3 ml each. The eluates were dried under a stream of nitrogen and submitted to mass spectrometric analysis. viability when employing the lowest supernatant dilution (1:2 4 ) ( Fig. 3 ). Thus, half-maximal cell killing was obtained with extremely diluted Stx1a-and Stx2a-containing supernatants, with Stx1a showing a stronger effect than Stx2a; whereas, the Stx2e-containing supernatant was almost inactive. This indicates a much higher susceptibility of Vero-B4 cells toward Stx1a and Stx2a than Stx2e.

Globo-series Stx-receptors Gb3Cer and Gb4Cer from Vero-B4 cells
For characterization of structural details of preliminarily identifi ed Gb3Cer and Gb4Cer Stx receptors, GSLs were eluted from the silica gel of immunostained bands (see Fig. 2A ) and the methanolic extracts were submitted to ESI MS using the positive ion mode.
Gb3Cer species. The structural characterization of Gb3Cer species by MS 1 analysis (an overview mass spectrum is shown in supplementary Fig. 1) followed by CID Stx2a (from E. coli O26:H11, strain 2574/97), and Stx2e (from E. coli O139:K82, strain S123G) in conjunction with respective anti-Stx1 or anti-Stx2 antibody using the TLC overlay assay ( Fig. 2B ). Stx1a and Stx2a were found to bind to prevalent Gb3Cer and did not bind to minute Gb4Cer expressed by Vero-B4 cells (see orcinol stain in Fig. 2A ). On the other hand, Stx2e recognized Gb4Cer and putative Gb5Cer ( Fig. 2B ) in addition to Gb3Cer, whereby relative binding intensities toward Gb4Cer and proposed Gb5Cer were higher in comparison to Gb3Cer (see scan in Fig. 2B ), as could be deduced from the orcinol stain of neutral GSLs of Vero-B4 cells (see scan in Fig. 1 ). The appearance of upper and lower bands of antibody-and Stxdetected Gb3Cer, Gb4Cer, and putative Gb5Cer suggests considerable ceramide heterogeneity, harboring GSLs with long chain C24 fatty acid in the upper band and those with short chain C16 fatty acid in the lower band. Their structures will be determined in detail by MS below.

Cytotoxicity of Stx1a, Stx2a, and Stx2e toward Vero-B4 cells
Subconfl uent grown Vero-B4 cells were exposed to various dilutions of Stx-containing bacterial cell culture supernatants, and cell viability was determined in comparison to controls without toxin using the crystal violet assay. Cell viability decreased in a concentration-dependent manner when applying increasing Stx concentrations ranging from 8,192-fold (2 Ϫ 13) to 16-fold (2 Ϫ 4) diluted supernatants ( Fig. 3 ). The supernatants of EHEC producing Stxs of Stx1a and Stx2a subtype exerted substantial cell damage showing ‫ف‬ 50% loss in cell viability when applying supernatant of highest dilution (1:2 13 ). Stx2e-containing supernatant exerted only low cell killing, reaching ‫ف‬ 20% loss in  explanatory fragmentation scheme in Fig. 4B , and a further CID spectrum of Gb3Cer (d18:1, C24:1) is provided in supplementary Fig. 2A.

Distribution of Stx receptors Gb3Cer, Gb4Cer, and Gb5Cer and lyso-PC to sucrose density gradient fractions
Antibody-mediated detection of Stx receptors in sucrose density gradient fractions exhibited clear DRM preference for the major GSL Gb3Cer of Vero-B4 cells, as exemplarily for terminally 1-3-linked galactose in Fig. 6C . Almost full sets of Y-and B-type ions accompanied by some Z-and C-type ions are detected, thus again giving rise to the complete monosaccharide sequence ( Fig. 6A, B ). Secondary fragmentation of B-type ions generated from permethylated glycoconjugates by cleavage of the glycosidic bond of a HexNAc preferentially eliminates the substituent in position 3 of the amino sugar ( 82 ). In the present case the B 2 ion (GalMe 4 -GalNAcMe 3 + ) at m/z 464.24 exclusively eliminates GalMe 4 -OH yielding the very stable B 2 /Z 4 ion at m/z 228.12 ( Fig. 6A, C ), thus, proving unambiguously the 1-3 linkage of the terminal galactose.
To sum up, the Stx2e-mediated detection of a pentahexosylceramide that was identifi ed as Gb5Cer by means of combined immunostain, galactosidase treatment, and MS of native and permethylated Gb5Cer unequivocally substantiated this pentahexosylceramide as a reliable and specifi c receptor for Stx2e.

Distribution of cholesterol, SM, and PC to sucrose density gradient fractions
Eight top-down fractions were obtained from sucrose gradient fractions after ultracentrifugation and denoted as DRM fractions (F1 to F3) and nonDRM fractions (F4 to F8), which were further subdivided into intermediate fractions (F4 to F6) and bottom fractions (F7 and F8). Gradient fractions from three independent cell preparations of Vero-B4 cells were produced using 1% Triton X-100 as the detergent.
Cholesterol. Cholesterol concentrations of gradient fractions were determined as nanograms per 1 × 10 6 cells and contents of SM and PC are portrayed as relative amounts within the eight fractions. Figure 7A shows the mean values of cholesterol and Fig. 7B and Fig. 7C the relative distribution of SM and PC, respectively, to the DRM (top) and nonDRM (intermediate and bottom) fractions from three cell preparations as bar diagrams (for listed data refer to Table 2 ). The canonical DRM fraction F2 harbored the highest concentration of cholesterol (averaged 602 ng/10 6 cells) as expected ( Fig. 7A ). Considerable amounts of cholesterol were detected in the bottom fractions F7 (224 ng/10 6 cells) and F8 (223 ng/10 6 cells) and DRM-associated fraction F3 (213 ng/10 6 cells), whereas intermediate fractions F4 to F6 each contained 113, 108, and 126 ng/10 6 cells, respectively, and top fraction F1 exhibited the lowest cholesterol content of 80 ng/10 6 cells. In summary, 53% of cholesterol distributed to DRMs with preference to classical DRM fraction F2 (35.6%), indicating its lipid raft association, and 47% scattered to nonDRM fractions ( Table 2 ).
SM and PC. SM could be identifi ed as a true lipid raft marker due to its extremely high content in canonical DRM fraction F2 (86.1%), accompanied by 12.2% in the adjacent F3 fraction ( Fig. 7B , Table 2 ), which amount to 98.3% of the DRMs. PC shows a similar trend with 55.8% in DRM fraction F2 and 10.8% in the adjacent F3 fraction, amounting to a summed average of 67.2% being DRM  Table 2 . structure with relative amounts of 64 and 28% (summed 92%) in nonDRM fractions F7 and F8, respectively. These single-tailed polar phospholipids were subsequently confi rmed by ESI MS (see next paragraph).

Mass spectrometric characterization of DRM-and nonDRM-associated phospholipids
ESI MS 1 profi les were produced from classical DRM fraction F2 and bottom fraction F7 to identify those phospholipids that specifi cally distribute to DRM or nonDRM fractions, respectively, obtained from Vero-B4 cells. The particular aim was to structurally elucidate the microdomain-associated marker, SM, in DRM fraction F2 and to unravel the structures of those single-tailed phospholipids that appear as nonDRM markers in the F7 sucrose density gradient fraction. Supposed and TLC "pre-identifi ed" lyso-PC was detectable in the MS 1 spectrum through ions at m/z 496.33, 522.34, and 544.33, depicting protonated lyso-PC (16:0), protonated lyso-PC (18:1), and monosodiated lyso-PC (18:1), respectively, as minor but unique phospholipid species in shown for one approach ( Fig. 8A ). Relative quantities of 55% distribute to DRM fraction F2, which is fl anked by F3 with 13% and sum up together with F1 to 68% of DRM associates. The calculation of average Gb3Cer distribution of three independent preparations revealed 67.3% of Gb-3Cer in DRM fractions F1 to F3 (grouped), indicating its preponderant DRM association ( Table 3 ). The remaining average Gb3Cer quantities were detected in the intermediate fractions F4 to F6 (grouped 13.2%) and in the bottom fractions F7 and F8 (grouped 19.5%), with clear evidence for preferential, but not exclusive, microdomain association of this GSL.
With regard to DRM association, the same holds true for less abundant Gb4Cer and Gb5Cer, as shown for the gradient analysis of the same approach ( Fig. 8B ). Even more clearly, Gb4Cer distributes with 70% to F2 and 6% to F3, summed up (including F1) to 76% with DRM association, and Gb5Cer was detected with 92% in F2 and 2% in F3, indicating (including F1) 94% DRM association. The calculation of average Gb4Cer and Gb5Cer distribution of three independent preparations revealed 77.5% of Gb-4Cer and 91.0% of Gb5Cer in DRM fractions F1 to F3 (grouped), indicating their preferred DRM association ( Table 3 ). Remnant average Gb4Cer and Gb5Cer quantities were detected in the intermediate fractions F4 to F6 (grouped 3.4 and 1.7%, respectively) and in the bottom fractions F7 and F8 (grouped 19.1 and 7.3%, respectively), showing again unequivocal preference, but not exclusive microdomain association, of these two GSLs. Collectively, data on Gb3Cer, Gb4Cer, and Gb5Cer suggest them as lipid raft candidates according to their preferred occurrence in the F1 to F3 DRM fractions.
As an example of a phospholipid profi le of a gradient preparation, the molybdenum blue-stained chromatogram of the same approach is shown in Fig. 8C , which demonstrates preponderant PC and uppermost SM DRM association with DRM fraction F2 (see also Fig. 7 , Table 2 ). Searching for a typical nonDRM marker, a faint band of a highly enriched phospholipid that separates below PC in the two bottom fractions indicated the presence of a lyso-PC Average values of triplicate TLC densitometric determinations of three separate experiments of each analyzed lipid in respective sucrose density gradient fractions are listed (see Fig. 7 ). Fractions were grouped into DRM (F1 to F3) and nonDRM fractions (F4 to F8). The nonDRM fractions were further subdivided into intermediate (F4 to F6) and bottom fractions (F7 and F8). For statistical analysis using Spearman's correlation coeffi cient r S , ranks were assigned to each fraction and P values were adjusted to P р 0.01. Rank 1 corresponds to the highest, rank 2 to the second highest, etc., and rank 8 to the lowest value. Percentages describe the relative content of each lipid in fractions F1 to F8 or in the grouped fractions F1 to F3, F4 to F6, and F7 and F8, summarized to 100% for single and grouped values, respectively. PC ( Table 2 ) and the neutral GSLs, Gb3Cer, Gb4Cer, and Gb5Cer ( Table 3 ). The highest percentage value received rank number 1, the second highest value rank number 2, etc. until fraction 8, which corresponds to the lowest value. The calculated rank correlation coeffi cient of r S = 0.875 ( P = 4.419 × 10 Ϫ 3 ) for the canonical lipid raft marker SM versus Gb3Cer suggests association of Gb3Cer membrane microdomains, as expected from the gradient distribution pattern of SM ( Fig. 7B ) and Gb3Cer ( Fig. 8A ). Similar DRM association was determined for Gb4Cer (r S = 0.786, P = 2.063 × 10 Ϫ 2 ) and Gb5Cer (r S = 0.875, P = 4.419 × 10 Ϫ 3 ) (see Fig. 8B ) when compared with SM distribution. Microdomain association could also be evidenced for direct comparison of Gb3Cer and Gb4Cer content, as well as Gb-3Cer and Gb5Cer content, by receiving an identical r s value of 0.976 ( P = 3.968 × 10 Ϫ 4 ), which indicates an equivalent membrane distribution of the Stx receptors. In addition, partition of both SM and Gb3Cer in comparison to the cholesterol patterns (see Fig. 7A ) with given r S values of 0.736 ( P = 3.751 × 10 Ϫ 2 ) and 0.929 ( P = 2.232 × 10

DISCUSSION
Stxs, also referred to as verotoxins, were fi rst described as a novel cytotoxic activity against Vero cells ( 11,12 ). Since this discovery, numerous studies have shown the functional impact of globo-series GSLs as Stx-specifi c receptors and their key role regarding toxin-mediated cytotoxicity in Vero cells ( 14,15,18,20,22,23,83 ). TLC analysis and corresponding overlay binding assays revealed the presence of Gb3Cer, the major Stx receptor, and Gb-4Cer, the less effi cient receptor, in Vero cells. Heterogeneity regarding TLC separation was observed in several previous studies and orcinol as well as Stx-detected bands of receptor GSLs suggested ceramide heterogeneity owing to fatty acids with varying acyl chain lengths ( 14,15,20,23 ). We provide here the full structural elucidation of Stx receptors of Vero-B4 cells and identifi ed Gb3Cer (d18:1, C24:0/C24:1) and Gb3Cer (d18:1, C16:0) as the prevalent Gb3Cer species (see Table 1 ), as well as similar variability for the corresponding prevalent Gb4Cer species. Minor quantities of GSLs harbored Cer (d18:1, C22:0) and Cer (d18:1, C18:0) as the lipid anchor. This receptor repertoire closely resembles those of human brain and glomerular microvascular endothelial cells ( 62,78 ), which represent the bottom fraction F7 (highlighted by grayed boxes in the mass spectrum shown in Fig. 9B ). Thus, the presence of lyso-PC in the bottom fractions indicates the most prominent difference when compared with its lack in DRM fraction F2, and lyso-PC can be considered as a reliable nonDRM marker that represents a characteristic phospholipid species of the liquid disordered membrane phase of Vero-B4 cells.

Statistics of lipid distribution in DRM and nonDRM fractions of Vero-B4 cells
The distribution of membrane lipids to DRM and non-DRM fractions was investigated more precisely by statistical nonparametric analysis based on Spearman's rank correlation coeffi cient r S . Ranks were assigned to percentage values of the gradient fractions of cholesterol, SM, and Gradient fractions F1 to F8 were assigned as described in Fig. 7 . TLC immuno-overlay assays of separated GSL extracts of fractions F1 to F8, equivalent to 1 × 10 6 cells, were performed with anti-Gb3Cer antibody for detection of Gb-3Cer (A) and with anti-Gb4Cer antibody for detection of Gb4Cer and (via antibody cross-reaction) Gb5Cer, the latter marked with arrow heads (B). Amounts of 2 g and 0.2 g of neutral GSLs from human erythrocytes (R2) were used as positive control in the anti-Gb3Cer (A) and the anti-Gb4Cer (B) overlay assay, respectively. The phospholipids, PC, SM, and lyso-PC (C), were detected in lipid extracts corresponding to 1 × 10 6 cells after TLC separation with bromophenol blue and matched to a phospholipid reference mixture (R3), whereby SM and lyso-PC are highlighted with arrow heads as specifi c markers of DRMs and nonDRMs, respectively. GSL and phospholipid chromatographies were done in triplicate with material from three separate experiments (data summarized in Tables 2, 3 ) and representative runs are shown.
DRM and nonDRM preparations is still the method of choice to investigate presumed lipid raft association ( 60 ). Canonical lipid raft markers identifi ed in Vero-B4 cells were cholesterol, SM, and GSLs, whereby PC was also found to preferentially occur in classical DRM fraction F2 of the sucrose density gradient. Notably, the density of Gb-3Cer distribution within the lipid environment in DRMs has been shown to contribute to Stx resistance of mutant Vero cells, which were shown to harbor a higher content the major targets in human EHEC infections ( 32,84 ). Anyway, the biological function of this ceramide variability is largely unknown, although differential binding of Stxs to such Gb3Cer lipoforms may indicate a signifi cant role in the eventual pathogenic outcome ( 85 ). However, this GSL heterogeneity, which appears as GSL doublets on the TLC plate, has been previously observed in TLC identifi cation of Gb3Cer and Gb4Cer species detected in Vero cells ( 15 ) and a green monkey kidney-derived cell line, termed GMK AH-1 ( 86 ). In addition to Gb3Cer and Gb4Cer, a novel pentaglycosylceramide with Gb5Cer structure was proposed and evidenced by the authors by means of MS combined with gas chromatography ( 86 ). Such Gb5Cer structure was also proposed to be expressed by Vero cells and identifi ed by TLC immunostaining with Gb5Cerspecifi c monoclonal antibody D579 ( 15 ). Furthermore, proposed Gb5Cer was specifi cally recognized (in addition to Gb3Cer and Gb4Cer) by the Shiga-like toxin (SLT) of subtype SLT-IIvp, which corresponds to Stx2 of subtype 2e ( 15 ). This is in agreement with observed Stx2e-binding toward fully structurally characterized Gb5Cer of Vero-B4 cells and failure of binding of Stx1a and Stx2a subtypes to this structure, as shown in our study. Collectively, we provide entire structural information on Stx receptors Gb3Cer, Gb4Cer, and Gb5Cer and their distribution to DRMs, as well as their association with lipid rafts and lipid (phospholipid and cholesterol) environment in those supramolecular structures. Furthermore, our data also suggest a common presence of globo-series GSLs, including Gb5Cer, in various Vero cell lines.
GSLs participate as driving forces in the formation of membrane microdomains and exert specifi c functions related to their physicochemical characteristics through GSL-protein interactions in those domains (87)(88)(89)(90)(91). GSLs preferentially distribute to phase-separated microdomains in the outer leafl et of the plasma membrane, which are suggested to operate as attachment platforms for host pathogens and their toxins on epithelial cells ( 57,92,93 ) to gain entry into cells and retrograde routing to intracellular targets in the cytosol ( 35,56,94 ). The analysis of  Average values of triplicate TLC densitometric determinations of three separate experiments of each analyzed GSL in respective sucrose density gradient fractions are listed (see Fig. 8 ). Fractions were grouped into DRM (F1 to F3) and nonDRM fractions (F4 to F8). The nonDRM fractions were further subdivided into intermediate (F4 to F6) and bottom fractions (F7 and F8). For statistical analysis using Spearman's correlation coeffi cient r S , ranks were assigned to each fraction and P values were adjusted to P р 0.01. Rank 1 corresponds to the highest, rank 2 to the second highest, etc., and rank 8 to the lowest value. Percentages describe the relative content of each GSL in fractions F1 to F8 or in the grouped fractions F1 to F3, F4 to F6, and F7 and F8, summarized to 100% for single and grouped values, respectively. routes via retrograde vesicular transport through the Golgi apparatus to the endoplasmic reticulum, from which they enter the cytosol ( 94,99 ). Although the initial mechanisms of GSL-toxin interaction are now partly understood, much less is known about the release of toxins from pathogenic enterobacteriaceae into the gut and their interaction with intestinal epithelial cells. Evidence has increased, for example, for V. cholerae toxin and the Stx of the E. coli O104:H4 outbreak strain, that released toxins are associated with outer membrane vesicles, which are shed by gram-negative bacteria during growth ( 100,101 ). The capability of gram-negative bacteria to release lipid vesicles has been recently reported for EHEC-derived hemolysin, which is delivered to the intestine in a free and a vesicular form ( 102 ). However, Stxs entrapped in outer membrane vesicles and binding as well as internalization of toxinloaded vesicles by human intestinal epithelial cells suggest a novel mechanism to deliver pathogenic cargoes and injure host cells ( 101 ). Thus, these novel fi ndings on extracellular transport of AB 5 toxins in lipid vesicles nicely show an unexpected analogy when compared with the intracellular vesicular transportation and retro-translocation of GSL-bound toxins through the Golgi apparatus and the endoplasmic reticulum ( 35 ). However, the potential functional impact of globo-series GSL receptors of Stxs, exposed on the cell surface of target epithelial cells, remains obscure in the process of vesicle internalization at this stage of research.
of Gb3Cer than wild-type Stx-sensitive cells ( 95 ), indicating that not only the presence of Gb3Cer, but also Gb3Cer density in lipid rafts, is important for Stx binding and most likely for Stx-mediated host cell intoxication. Cholesterol has been proposed to play a key role in this GSL recognition process in which communication is controlled through the collective behavior of lipids, where cholesterol modulates GSL conformation and receptor activity ( 96 ). In particular, cholesterol seems to mask Stx receptor GSLs in lipid rafts of Vero cells and may impinge many other GSL receptor functions ( 97 ). As a further marker of plasma membrane microdomains, SM may have important impact in cell-signaling through its structural role in lipid rafts ( 98 ).
The detailed mass spectrometric investigation of DRMand nonDRM-associated phospholipids revealed, besides exclusive distribution of SM (sphingolipid) to the liquidordered phase, specifi c enrichment of monounsaturated PC (glycerophospholipid) species PC (36:1), PC (34:1), and PC (32:1) and saturated PC (30:0) in classical DRM fraction F2; whereas doubly unsaturated pendants PC (36:2), PC (34:2), and PC (32:2) were overwhelmingly found in the nonDRM bottom F7 fraction, representing the liquid-disordered phase. This distribution fi ts to the common assumption that phospholipids with saturated fatty acids may "prefer" the liquid-ordered phase and "like to distribute" to those microdomains; whereas additional double bonds in fatty acids may disturb or hinder tight packaging of PC glycerophospholipids in lipid rafts. Interestingly, detected PE species, PE (36:2), PE (34:1), and PE (34:2), were only found in the F7 bottom fraction, independent of degree of fatty acid unsaturation, and represent further nonDRM markers, which may preferentially distribute to the liquid-disordered phase. In addition to those differences between DRM and nonDRM fractions, we detected, by means of detailed MS, single-tailed phospholipids, namely lyso-PC (18:1) and lyso-PC (16:0), as typical nonDRM markers in the F7 bottom sucrose density gradient fraction of Vero-B4 cells. Interestingly, both lyso-PC species have been previously identifi ed as non-lipid raft markers in leukocyte-derived Raji (B cell) and THP-1 (monocyte) cell lines ( 70 ). Thus, the identifi ed nonDRM markers may act as DRM opponents in the liquid-disordered phase of the plasma membrane of Vero-B4 cells, whereby their functional role requires further investigations to deepen our understanding of the complexity of the plasma membrane and, in particular, the involvement of lipid raft-associated GSLs in traffi cking of Stxs to subcellular targets . This study might be helpful to further our understanding of Vero cell sensitivity to various Stx subtypes and the functional role of lipid rafts in kidney epithelial cells.
Our knowledge has continuously increased with regard to elucidation of fundamental biological mechanisms such as the interaction of bacterial AB 5 toxins, e.g., V. cholerae toxin or Stxs from pathogenic E. coli , with GSL receptors exposed on the surface of susceptible cells, followed by internalization and retrograde transportation to intracellular target structures ( 36,37,56 ). Upon binding and delivery to early endosomes, GSL-bound toxins follow diverse intracellular