Neutral glycosphingolipids in human blood: a precise mass spectrometry analysis with special reference to lipoprotein-associated Shiga toxin receptors.

Shiga toxin (Stx)-producing Escherichia coli are the leading cause of hemorrhagic colitis and life-threatening extraintestinal complications in humans. Stx1 and Stx2 are transferred by yet to be delineated mechanisms from the intestine to the circulation where they injure microvascular endothelial cells. The resulting vascular lesions cause renal failure and brain damage. Because lipoproteins are potential carriers of Stx through the circulation, we investigated human lipoprotein-associated neutral glycosphingolipids (GSLs) with emphasis on high (globotriaosylceramide) and low (globotetraosylceramide) affinity Stx-receptors. TLC overlay employing Stx1, Stx2, and anti-GSL antibodies demonstrated preferential distribution of globo-series GSLs to very low- and low-density lipoproteins compared with minor association with high-density lipoproteins. Electrospray ionization quadrupole time-of-flight mass spectrometry portrayed C24:0/C24:1 and C16:0 as the major fatty acid of the ceramide moieties of Stx-receptors carrying nonvarying d18:1 sphingosine. This structural heterogeneity was also found in precursor lactosylceramide, glucosylceramide, and galactosylceramide, the last showing an exceptionally high degree of hydroxylated C24 fatty acids. Our findings provide the basis for exploring the functional role of lipoprotein-associated Stx-receptors in human blood.


Human plasma
Frozen pooled fresh plasma of healthy donors of blood group A was obtained from the Heart and Diabetes Centre Northrhine-Westphalia (Bad Oeynhausen, Germany; charge 0426 127710 1). The plasma (protein concentration 69.4 mg/ml) was thawed, fi ltrated through a sterile fi lter (pore size 0.2 µm), and aliquots of 360 ml and 100 ml were immediately used for the preparation of lipoproteins and the extraction of GSLs, respectively (see below).

Preparation of lipoprotein fractions from human plasma
VLDL (fraction I, d = 0.94 to 1.006 g/ml), LDL (fraction II, d = 1.019 to 1.063 g/ml), and HDL (fraction III, d = 1.125 to 1.210 g/ml) were isolated from 360 ml of human plasma by discontinuous KBr gradients as described by Havel et al. ( 38 ). Fraction I comprises chylomicrons, is enriched in VLDL, and contains in addition intermediate-density lipoproteins (IDL), thereafter named "VLDL". Due to their high content of LDLs and HDLs, fractions II and III were designated as "LDL" and "HDL", respectively. The lipoprotein fractions were dialyzed against 0.3 mmol/l Tris-HCl, 0.14 mol/l NaCl, 1.0 mmol/l EDTA, pH 7.2, and stored under exclusion of light at 4°C for no longer than 7 days. The protein concentration in lipoprotein fractions [VLDL (1.3 mg/ ml), LDL (6.4 mg/ml), and HDL (11.4 mg/ml)] were determined by the method of Lowry et al. ( 39 ).

Isolation and purifi cation of neutral GSLs from human plasma
One hundred milliliters of pooled plasma were dialyzed against deionized water followed by lyophilization. The residue was extracted twice with chloroform/methanol/water (30/60/8, each by vol.). After rotary evaporation, the extract was resolved in 100 ml of aqueous 1 N NaOH and incubated for 2 h at 37°C in order to saponify coextracted phospholipids and triacylglycerides. After neutralization with 10 N HCl followed by dialysis and lyophilization, neutral GSLs were separated from gangliosides by anion-exchange column chromatogrpahy employing DEAE-Sepharose CL-6B (GE Healthcare, München, Germany) in the acetate form and subsequently purifi ed by silica gel 60 (particle size 0.040 to 0.063 mm, Merck, Darmstadt, Germany) chromatography according to standard procedures (40)(41)(42). The neutral GSL fraction was adjusted to 2 ml of chloroform/methanol (2/1, v/v) whereby 1 µl corresponds to 50 µl of plasma (equivalent to 3.47 mg of plasma protein).

Isolation of GSLs from lipoprotein preparations
Three types of lipoprotein fractions were prepared according to their density (see "Preparation of lipoprotein fractions from human plasma"). Aliquots of VLDL, corresponding to 4 mg of lipoprotein, and aliquots of LDL and HDL, equivalent to 20 mg of lipoprotein, respectively, were dialyzed against deionized water and dried by rotary evaporation. After saponifi cation of nonGSL contaminants with 1 N NaOH for 1 h at 37°C and neutralization with 10 N HCl, samples were lyophilized. GSLs were extracted from the residues with chloroform/methanol/water (30/60/8, v/v/v) under sonication. The extracts were centrifuged and the supernatants dried by rotary evaporation. The residues were redissolved in chloroform/methanol (2/1, v/v) and adjusted to concentrations whereby 1 µl of extract corresponds to 1 µg of lipoprotein.

Stx1 and Stx2
Stx1 and Stx2 were purifi ed from E. coli C600(H19J) carrying the stx 1 gene from E. coli O26:H11 strain H19 ( 43 ) and from E. coli endothelial cells ( 17 ). Stxs are internalized and undergo retrograde transport through the Golgi apparatus to the endoplasmic reticulum ( 18 ). After translocation into the cytosol, the enzymatically active A1 fragment exerts its toxic function due to specifi c depurination of adenosine at a highly conserved loop of the 60S rRNA, causing death of the target cell through inhibition of protein synthesis ( 19,20 ). Vascular damage caused by Stx-producing Escherichia coli (STEC), the major causes of hemorrhagic colitis and the life-threatening hemolytic uremic syndrome (HUS), is largely mediated by Stxs, which particularly injure endothelial cells in the kidney, the brain, and other organs and also participate in thrombotic mechanisms ( 15,17,(21)(22)(23). After being released by the infecting STEC in the intestine, Stx is translocated across the gut into circulation ( 24 ) and transported to endothelial cells. Importantly, Gb3Cer has not been detected in human gastroepithelial cells ( 25 ), and the mechanism of toxin translocation across the intestinal barrier remains an enigma. Although the role of polymorphonuclear leukocytes as a Stx carrier has been indicated ( 26,27 ), the mechanism of toxin delivery is still a matter of debate. Interestingly, lipoproteins can bind GSLs, including Gb3Cer ( 28 ), and it seems possible that Stx could be cotransported, bound in a piggyback fashion with dietary lipoproteins, from the lumen of the intestine to the circulation ( 29 ). However, knowledge about the structural diversity of GSLs in lipoproteins is rather poor and little is known about their functional role in human blood. Though the existence of neutral GSLs in lipoproteins is well known ( 30,31 ), the composition of GSLs, being minor constituents of lipoproteins, and their fi ne structure have generally drawn little attention ( 32,33 ). This prompted us to perform a compositional analysis of neutral GSLs in human blood with special reference to monohexosylceramides and Stx receptors and their association with lipoproteins. Notably, glucosylceramide (GlcCer) has been reported to modulate the Stx-mediated cytotoxic effect ( 34 ) and may directly contribute to venous thrombosis ( 35 ), and galactosylceramide (GalCer) has been identifi ed as a (co)receptor of type 1 human immunodeficiency virus ( 36,37 ).

Extraction of GSLs from silica gel of TLC plates
The silica gel of primulin-, antibody-, and Stx-detected GSL bands was extracted with chloroform/methanol/water (30/60/8, C600(933W) carrying the stx 2 gene from E. coli O157:H7 strain EDL933 ( 44 ), respectively. Briefl y, bacteria were disintegrated by sonication and subjected to extraction with polymyxin B (3000 U/ ml) (Sigma-Aldrich Chemie GmbH, Taufkirchen, Germany). After centrifugation, the supernatant containing crude Stx was concentrated by ultrafi ltration and subjected to chromatography on hydroxylapatite column followed by chromatofocusing ( 45,46 ). The column fractions containing Stx were pooled, dialyzed, and aliquots were stored at Ϫ 70°C until use. Purity of the Stx1 and Stx2 preparations was monitored by SDS-PAGE, and the structural integrity of both toxins was checked by peptide mapping employing MS ( 13 ).

Anti-GSL, anti-Stx1, and anti-Stx2 antibodies
All polyclonal rabbit and chicken antibodies were produced according to the method of Kasai et al. ( 47 ). The preparation and specifi cities of the antibodies against various neutral GSLs have been reported in previous publications. Structures of GSLs and the antibody-relevant references are listed in Table 1 . Mouse IgG monoclonal antibodies 109/4-E9b and 135/6-B9 (Sifi n, Berlin, Germany) were used for the detection of Stx1 and Stx2, respectively.

ESI Q-TOF-MS
The extracted GSL samples were analyzed in positive ion mode by electrospray ionization quadrupole time-of-fl ight mass spectrometry (ESI Q-TOF-MS 1 ) and low-energy collision-induceddissociation (CID) MS 2 using a Q-TOF mass spectrometer (Micromass, Manchester, UK). After selecting the precursor ions of interest with the fi rst quadrupole, CID was performed to obtain fragment ions enabling sequence analysis [for further details see Meisen et al. ( 54,60 ) and Schweppe et al. ( 58 )]. The nomenclature introduced by Domon and Costello ( 61, 62 ) and Adams and Ann ( 63 ) was used for the assignment of the fragment ions.

TLC immunodetection of human plasma neutral GSLs
The initial identifi cation of neutral GSL structures, deduced from their chromatographic behavior in comparison to refer-  Table 2 . fragment ions obtained by internal cleavage of the ceramide portion clearly evidenced the occurrence of h24:0 and the concomitant appearance of sphingosine (d18:1) as indicated by specifi c diagnostic ions.

Antibody-detected neutral GSLs
Next, we investigated the GSL extracts from TLC immunostained Lc2Cer, Gb3Cer, Gb4Cer, and nLc4Cer double bands in detail by ESI Q-TOF-MS. The major [M+Na] + ions identifi ed in the ESI Q-TOF-MS 1 spectra are listed in Table 3 . Examples of MS 1 spectra including the corresponding insets of anti-Lc2Cer, anti-Gb3Cer, anti-Gb4Cer, and anti-nLc4Cer antibody stained GSLs, where the silica gel bands scraped off the TLC plate are indicated by a dotted rectangle, are provided in Fig. 4A , B, C, and D , respectively, for lower-band GSLs of the respective doublets carrying short-chain fatty acids C14 to C18. The main species in the spectra correspond to [M+Na] + ions of neutral GSLs with saturated C16:0 fatty acids, which may be fl anked by the corresponding but low abundant protonated [M+H] + ions as shown in the Lc2Cer and Gb3Cer spectra in Fig. 4A and B , respectively. In addition to these major species, minor ions indicating GSL species with C14:0 and C18:0 fatty acids were detected as well. None of them, including the dominant C16:0 fatty acid carrying species, showed any modifi cation such as unsaturation or hydroxylation of the acyl chains. The neutral GSLs of the immunostained double bands harboring long-chain C20 to C24 fatty acids are shown in supplementary Fig. I and the major monosodiated ions acquired from the MS 1 spectra are listed in Table 3 . As shown for the GSLs with short-chain fatty acids ( Fig. 4 )  Interestingly, GSL heterogeneity owing to variable acylation of the sphingosine (d18:1) moieties with both monounsaturated and saturated C22 to C24 fatty acids was observed for the prevalent tri-and tetrahexosylceramides, whereas Lc2Cer was found to harbor exclusively monounsaturated C24:1 fatty acid.

Stx-detected neutral GSLs
The MS 1 spectra obtained from the silica gel extracts of Stx1/anti-Stx1 and Stx2/anti-Stx2 stained Gb3Cer and Gb4Cer variants were identical to those acquired from the anti-Gb3Cer and anti-Gb4Cer antibody detected species. MS 1 data of Stx1/anti-Stx1 and Stx2/anti-Stx2 detected [M+Na] + ions are summarized in Table 3 in addition to those obtained by antibody identifi cation. The Stx1 and Stx2 positive TLC overlay assay bands and in particular the monosodiated ions of the MS 1 spectra acquired from the Stx1 stains served as starting point for the fi nal confi rmation of Stx-binding Gb3Cer and Gb4Cer species using CID tandem MS. The complete structural characterization was performed for the main species assigned to Gb3Cer and Gb4Cer and is exemplarily shown in the following sections for Gb3Cer and Gb4Cer carrying sphingosine (d18:1) as the sole and constant amino alcohol and a variable fatty in the TLC binding assay of Fig. 1 . The ganglio-series neutral GSLs Gg3Cer and Gg4Cer were not detected in the neutral GSL fraction of human plasma (data not shown).
The proposed structures, and particularly the assessment of hydroxylated species of GlcCer and GalCer, were further verifi ed by MS 2 analysis. The MS 2 spectrum, together with the corresponding fragmentation scheme of GalCer (d18:1, h24:0), is shown as a representative example of a hydroxylated monohexosylceramide in Fig. 3 . The The neutral GSLs GlcCer and GalCer from human plasma were separated as borate complexes by TLC (see insets of Fig. 2 ) and isolated by preparative TLC using primulin as nondestructive dye. Monosodiated ions were recorded in the positive ion mode as shown in Fig. 2 . Major monohexosylceramides are printed in boldface.
The complete structural characterization of the Stx1 ligands was exemplifi ed for the main species assigned to Gb3Cer (d18:1, C24:1) by use of low energy CID MS 2 as depicted in Fig. 5B together with the fragmentation scheme in Fig. 5C . The fragment ions originating from the monosodiated precursor ions at m/z 1156.75 are assigned according to the nomenclature of Domon and Costello ( 61,62 ) and Adams and Ann ( 63 ) and gave rise to the complete structure ( Table 4 ). A full series of Y-type ions at m/z 994.70 (Y 2 ), 832.68 (Y 1 ), and 670.63 (Y 0 ) was obtained, indicating the sequential loss of three hexose moieties from Gb3Cer and evidencing the Gb3Cer (d18:1, C24:0) structure, which was further supported by the appearance of high abundant B 2 and B 3 as well as C 2 and C 3 ions and additional 0,2 A 2 and 0,2 A 3 ions, the latter two of them generated by ring cleavages ( Fig. 5B ). The detection of N II -ions at m/z 264.29 formed by loss of the N-linked fatty acid of the ceramide moiety and cleavage at the reducing end of the oligosaccharide (Z 0 -type ions in addition to Y 0 -type ions) clearly indicates the presence of 4-sphingenine (d18:1) and rules out 4-sphinganine (d18:0) as the longchain aminodiol constituent of the ceramide moiety.

Stx-detected Gb4Cer species
The prevalent monosodiated molecular ions detected in the MS 1

Stx-detected Gb3Cer species
The     Fig. 6A . Full structural characterization of these low-affi nity Stx1 ligands was performed with the proposed Gb4Cer (d18:1, C24:1) species employing low energy CID MS 2 as demonstrated in Fig. 6B in conjunction with the fragmentation scheme depicted in Fig. 6C . The fragment ions originated from the monosodiated precursor ions at m/z 1359.87 exhibited a full series of Y-type ions starting with Y 3 at m/z 1156.79, indicating the loss of a HexNAc moiety from the nonreducing end, followed by the sequential cleavage of three hexose moieties, resulting in Y 2 ions at m/z 994.72, Y 1 ions at m/z 832.68, and ending up with the Y 0 fragment ions at m/z 670.62, representing the ceramide moiety carrying C24:1 fatty acid (see synopsis of fragment ions in Table 4 ). Additionally, a complete series of B-type ions and sporadic C-and Z-type ions were detected, thus enabling, together with the N II -ions at m/z 264.27 generated by loss of the N-linked fatty acid of the ceramide moiety and cleavage at the reducing end of the oligosaccharide (Z 0 -type ions in addition to Y 0 -type ions), an unambiguous structural assignment of Gb4Cer (d18:1, C24:1).
The  Table 4 .

Lipoprotein-associated neutral GSLs
VLDL, LDL, and HDL fractions harboring very low-, low-, and high-density lipoproteins, respectively, were isolated and separated by density gradient ultracentrifugation from the same batch of pooled human plasma, which was used for the structural characterization of individual GSLs from human plasma as described above. Total lipids including GSLs were extracted from lipoprotein fractions and coextracted phospholipids and triglycerides saponifi ed by alkaline treatment. GSL aliquots corresponding to 40 µg of lipoproteins  Table 3 . Unknown ions, which appeared in the spectrum of minor nLc4Cer (D) and originated most likely from the silica gel extraction procedure, are marked with asterisks and were not further analyzed. be observed in the HDL fraction (upper four panels of Fig.  7 ). Stx1 and Stx2 clearly bound to the individual Gb3Cer species showing binding preference to the lower of the two Gb3Cer bands that harbors the Gb3Cer species with the shorter fatty acids, but did not recognize the minute quantities of Gb4Cer species in the lipoprotein fractions, which were also only faintly detectable in the neutral GSL fraction derived from human plasma (lower two panels of Fig. 7 ). The presence of Gb4Cer, however, could be shown in the lipoprotein fractions and the human plasma sample with the anti-Gb4Cer antibody. Importantly, nonlipoprotein containing fractions, e.g., the intermediate albumin-containing fraction obtained from the ultracentrifugation gradient, did not contain any GSLs, indicating strict association of all neutral GSLs with lipoproteins in human blood.
in case of Lc2Cer and nLc4Cer detection, 60 µg using anti-Gb3Cer and anti-Gb4Cer antibody, and 120 µg employing Stx1 and Stx2 were separated on TLC plates. Neutral GSLs were detected with the same set of anti-GSL antibodies and Stx1 and Stx2 preparations as described above ( Table 1 ). The upper four panels of Fig. 7 show the TLC immunodetection assays of lipoprotein-associated neutral GSLs with the polyclonal antibodies against Lc2Cer, Gb3Cer, Gb4Cer, and nLc4Cer, and the two lower panels, those TLC overlay assays performed with Stx1/anti-Stx1 and Stx2/anti-Stx2. As being obvious from the antibody assays, all neutral GSLs, which were comparatively analyzed in lipoprotein fractions corresponding to identical protein amounts, preferentially distributed to the VLDL and LDL fractions calculated as percentage values, whereas only low binding intensities could  Table  4 . C: Fragmentation scheme of Gb3Cer (d18:1, C24:1). The nomenclature introduced by Domon and Costello ( 61, 62 ) and Adams and Ann ( 63 ) was used for the assignment of the fragment ions.
ranging from C16 to C24 acyl chains. As deduced from the MS 1 spectrum, GlcCer (d18:1, C24:1/C24:0) and GlcCer (d18:1, C22:0) were the most prominent variants. In contrast to this, the fatty acid profi le of GalCer ranged from C18 to C26 and ions corresponding to hydroxylated GalCer (d18:1, h24:1/h24:0) showed the highest abundance in the MS 1 spectrum in accordance to the TLC staining intensities. Thus, GalCer with h24:1 and h24:0 acyl chains were the predominant species, whereas only extremely low abundant ions indicative for h24:0 and h24:1 fatty acid appeared in the GlcCer spectrum. Interestingly, the occurrence of GalCer and GalCer-derivatives containing hydroxyl C18 and C24 fatty N -acylation is a distinctive feature of vertebrate (including human) myelin from both the central and peripheral nervous system ( 65, 66 , and references therein). Thus, it is tempting to speculate about myelin GalCer as a possible indication of plasma source.
With respect to the development of novel technologies, both GalCer and GlcCer monohexosylceramide molecular species can now be discriminated by ESI MS/MS ( 67 ) exploiting the differential fragmentation pattern of GSL using chlorine or Li + adducts ( 68,69 ). Furthermore, Zama et al. ( 70 ) recently published a novel HPLC-based technology to simultaneously analyze and quantify GalCer and GlcCer using o-phthalaldehyde derivatives prepared with sphingolipid ceramide N -deacylase. Due to the biological importance, for example, of GalCer as (co)receptor of HIV-1 gp120 and gp41 ( 36,37 ) or GlcCer exhibiting thrombosis modulatory potential ( 35 ) and auxiliary function in Stx-mediated cytotoxicity ( 34 ), those techniques might be useful for the determination of GlcCer and GalCer levels in various biological samples such as target tissues or body fl uids.
Human serum contains a number of lipid protein complexes that can be isolated on the basis of their different fl otation densities and classifi ed as VLDL, LDL, and HDL ( 38 ). The protein content of such particles varies from 5% to 60%, and all contain a mixture of neutral lipids, phospholipids, and GSLs organized to form a stable particle. GSLs are present on the surface of all human serum lipoprotein particles where they make up minor constituents We further evaluated the relative quantitative distribution of neutral GSLs among lipoprotein fractions by TLC scanning densitometry of orcinol-stained bands. Interestingly, Gb3Cer exhibited a similar relative distribution in the lipoprotein fractions compared with human plasma (see supplementary Table I), whereas Lc2Cer was found considerably enriched in the VLDL fraction accompanied by reduced relative content of monohexosylceramides. In addition, a relative decrease of Gb4Cer to half of the plasma value was detected in the VLDL and LDL fractions. This heterogeneity will be addressed in future investigations on the distribution of GSLs within lipoprotein fractions from individual probands.

DISCUSSION
The purpose of this comprehensive study employing MS combined with TLC overlay assay was to explore the molecular diversity of different neutral GSLs from human blood, particularly of those which are supposed to have a functional role in human disease. Mono-, di-, tri-, and tetraosylceramides were exhaustively analyzed by ESI Q-TOF-MS 1 and MS 2 with emphasis on so far unresolved monohexosylceramide composition and neutral GSLs of the globo-series with binding potential to both Stx1 and Stx2.
The monohexosylceramides GlcCer and GalCer were found to constitute approximately one third among the neutral GSLs in human plasma with a relative content of 53% GlcCer and 47% GalCer within the monohexosylceramide fraction. Thus, GalCer represents one of the major neutral GSLs in human plasma that has not been detected (most likely due to failure of separation from GlcCer) in previous investigations ( 30,31,64 ). For the discrimination of both types of monohexosylceramides, which are indistinguishable using conventional TLC, we reactivated an "old-fashioned" but simple method for separation of GlcCer and GalCer as borate complexes in alkaline solvent, followed by preparative TLC and MS 1 and MS 2 analysis. Their fi ne characterization by ESI Q-TOF-MS 2 revealed GlcCer species with variable fatty acid chain lengths discrepancy with previous quantitative investigations by Dawson et al. ( 30 ), who showed roughly similar amounts in VLDL and HDL fractions, whereas considerably lower GSL amounts were detected by us in the HDL fraction. One possible explanation might be that Dawson and colleagues investigated lipoprotein fractions from two Caucasian donors whereas we analyzed lipoproteins from pooled plasma of unknown ethnic origin. Notably, our data were obtained from TLC overlays and expressed on the basis of antibody and Stx staining intensities related to defi ned milligrams of plasma protein or micrograms of lipoproteins. Direct ESI Q-TOF-MS 1 and MS 2 analysis of the Stxdetected Gb3Cer and Gb4Cer variants demonstrated a broad ceramide heterogeneity of the individual GSL species that ranged from C16 to C24 fatty acid chain length with invariable sphingosine (=4-sphingenine). Furthermore, our study demonstrates almost equal ratios of longchain C24:1/C24:0 versus short-chain C16:0 fatty acids in the phospholipid bilayer surrounding the hydrophobic cores of lipoproteins ( 30 ). The content of serum lipoprotein fractions may vary depending on the nutritional or hormonal stage of the donors ( 71 ) and the GSL content of lipoproteins (so far shown for GlcCer) may increase during infection and infl ammation ( 72,73 ). Previous quantitative determination of neutral GSLs revealed 0.6, 0.4, 0.2, and 0.15 µmole/dl in normal human serum for mono-, di-, tri-, and tetraosylceramides, respectively ( 30 ). We detected all of the neutral GSLs found in human plasma in each of the three major lipoprotein classes analyzed and no GSLs were detected in the lipoprotein-free fractions obtained by ultracentrifugation. This coincides with several investigations of other research groups, for example, with a study on the GlcCer distribution in human plasma lipoproteins where no GlcCer was detectable in the lipoprotein-defi cient plasma ( 64 ). Our data revealed predominant presence of GSLs in VLDL and LDL being in some  Table 3 . B: MS 2 spectrum of Stx1-bound Gb4Cer (d18:1, C24:1). The precursor ions at m/z 1359.87 were selected from the MS 1 spectrum and the position of the analyzed Gb4Cer-species is marked with an arrow in the inset. Type of fragment ions and corresponding m/z -values are listed in Table 4 . Asterisks indicate minor fragment ions that originate from internal cleavages of the oligosaccharide. C: Fragmentation scheme of Gb4Cer (d18:1, C24:1). The nomenclature introduced by Domon and Costello ( 61,62 ) and Adams and Ann ( 63 ) was used for the assignment of the fragment ions. proteins from these two organs and their peripheral catabolism determine lipoprotein levels that are normally present in blood ( 71,74 ). Enterocytes, which line the small intestine, assemble chylomicrons from triglycerides, phospholipids, and cholesterol derived from the diet and then secrete the nascent lipoprotein via the lymph into the blood ( 75 ) whereas the liver synthesizes and assembles VLDL. Upon secretion, VLDL circulates and its triglyceride core is a substrate for lipoprotein lipase, an enzyme that resides on the luminal surface of the capillary endothelium ( 71 ). This leads to the formation of VLDL remnants, which can then be rapidly cleared by the liver or can continue to be processed to become LDL. GSLs are transported in plasma on lipoproteins and are taken up and metabolized by cells through an LDL receptor-mediated pathway. However, only sparse data are available about the biological and functional role of lipoprotein-associated GSLs. A few early studies have evidenced the reversible exchange of GSLs between HDL and LDL ( 76 ) and, moreover, that cholera toxin insensitive cells can acquire a functional cholera toxin receptor ganglioside GM1 following culture in the presence of serum lipoproteins ( 77 ). Due to the strict association of Gb3Cer with lipoproteins in human blood, it is tempting to speculate on the functional role of lipoprotein-associated Gb3Cer in human blood as transporter of Stx from the intestine into the circulation. We thus speculate that this way, Stx might reach the microvascular endothelial cells of the kidney and the brain, which represent the major targets of Stx in the clinical onset of HUS.
Thus, despite the hypothetical functional role of GSLs, our study provides useful information for future investigations of the transfer of neutral GSLs among lipoproteins and from lipoproteins to cells. Furthermore, the analysis of the molecular details of Stx-receptors and their association with lipoproteins in patients suffering from Stx-mediated diseases such as HUS may help us to better understand the involvement of GSLs in the interaction with Stxs and to inspire us to develop preventive and therapeutic measures for Stx-mediated diseases ( 78 ).
among the Lc2Cer and globo-series species. The exception was nLc4Cer, which exhibited a preponderance of C16:0 fatty acid in the ceramide portion. Interestingly, hydroxylation as a characteristic feature of the C24:1 and C24:0 fatty acids in the GalCer species did not appear in the Stx-receptors Gb3Cer and Gb4Cer and was found to be restricted to GalCer. This fi nding is in line with the compositional analysis of GlcCer (the precursor of Lc2Cer, Gb3Cer, and Gb4Cer), indicating almost no hydroxylated C24 acyl chains corresponding to the failure of this modifi cation in Gb3Cer and Gb4Cer. Although GSLs are intimately associated with serum lipoproteins, the mode of association and the implication of fatty acid heterogeneity in this association remain undetermined.
The liver and intestine essentially account for all circulating plasma lipoproteins. Rates of synthesis of lipo- Fig. 7. TLC immunodetection of lipoprotein-associated individual neutral GSLs from human plasma. GSL-aliquots of extracts from VLDL, LDL, and HDL lipoprotein-fractions and from human (hu) plasma were separated by TLC and analyzed with polyclonal antibodies (pAb) against Lc2Cer, Gb3Cer, Gb4Cer, and nLc4Cer in comparison to Stx1/anti-Stx1-antibody and Stx2/anti-Stx2-antibody. Bound anti-GSL and anti-Stx-antibodies were visualized with alkaline phosphatase conjugated secondary antibodies and BCIP as a substrate. Aliquots applied for anti-Lc2Cer and anti-nLc4Cer immunostains correspond to 12.2 mg and those used for anti-Gb3Cer, anti-Gb4Cer, and the Stx1 and Stx2 overlay assays are equivalent to 18.0 mg of human plasma proteins. GSL aliquots of lipoprotein extracts equal the amounts of 40 µg in the anti-Lc2Cer and anti-nLc4Cer, 60 µg in the anti-Gb3Cer and anti-Gb4Cer, and 120 µg of lipoproteins in the Stx1-and Stx2-overlay assays. The densitometrically determined relative distributions of GSLs between the different lipoproteins were calculated as percentage values and are presented for the antibody-detected GSLs as arabic numerals. The vertical white lines indicate areas of noncontiguous lanes assembled. The GSL structures and backgound information on the employed antibodies and Stxs are provided in Table 1 .