Keywords: edema disease / glycolipids / lipid rafts / mass spectrometry / TLC immunostaining

Transcription

1 Glycobiology vol. 23 no. 6 pp , 2013 doi: /glycob/cwt013 Advance Access publication on February 21, 2013 Expression of Shiga toxin 2e glycosphingolipid receptors of primary porcine brain endothelial cells and toxin-mediated breakdown of the blood brain barrier Iris Meisen 2,3,, Regina Rosenbrück 2,, Hans-Joachim Galla 4, Sabine Hüwel 4, Ivan U Kouzel 2, Michael Mormann 2, Helge Karch 2, and Johannes Müthing 1,2,3 2 Institute for Hygiene, Robert-Koch-Str. 41; 3 Interdisciplinary Center for Clinical Research (IZKF); and 4 Institute of Biochemistry, University of Münster, D Münster, Germany Received on December 3, 2012; revised on February 15, 2013; accepted on February 18, 2013 Shiga toxin (Stx) 2e, released by certain Stx-producing Escherichia coli, is presently the best characterized virulence factor responsible for pig edema disease, which is characterized by hemorrhagic lesions, neurological disorders and often fatal outcomes. Although Stx2e-mediated brain vascular injury is the key event in development of neurologic signs, the glycosphingolipid (GSL) receptors of Stx2e and toxin-mediated impairment of pig brain endothelial cells have not been investigated so far. Here, we report on the detailed structural characterization of Stx2e receptors globotriaosylceramide (Gb3Cer) and globotetraosylceramide (Gb4Cer), which make up the major neutral GSLs in primary porcine brain capillary endothelial cells (PBCECs). Various Gb3Cer and Gb4Cer lipoforms harboring sphingenine (d18:1) or sphinganine (d18:0) and mostly a long-chain fatty acid (C20 C24) were detected. A notable batch-to-batch heterogeneity of primary endothelial cells was observed regarding the extent of ceramide hydroxylation of Gb3Cer or Gb4Cer species. Gb3Cer, Gb4Cer and sphingomyelin preferentially distribute to detergent-resistant membrane fractions and can be considered lipid raft markers in PBCECs. Moreover, we employed an in vitro model of the blood brain barrier (BBB), which exhibited strong cytotoxic effects of Stx2e on the endothelial monolayer and a rapid collapse of the BBB. These data strongly suggest the involvement of Stx2e in cerebral vascular damage with resultant neurological disturbance characteristic of edema disease. Keywords: edema disease / glycolipids / lipid rafts / mass spectrometry / TLC immunostaining 1 To whom correspondence should be addressed: Tel: ; Fax: ; jm@uni-muenster.de Both are first authors. Introduction Shiga toxin-producing Escherichia coli (STEC) are zoonotic food- and waterborne enteric pathogens that elicit systemic vascular damage diseases in humans and animals (Bielaszewska and Karch 2005; Mainil and Daube 2005; Tarr et al. 2005; Karmali et al. 2010; Melton-Celsa et al. 2012). In humans, STEC infections primarily cause watery and bloody diarrhea, and in severe cases patients develop hemorrhagic colitis, life-threatening hemolytic uremic syndrome (HUS) and central nervous system impairment (Karch et al. 2005; Obata 2010; Obrig 2010; Zoja et al. 2010; Meuth et al. 2012; Obrig and Karpman 2012). STEC of human isolates produce mostly Stx1a and/or Stx2a (in previous publications imprecisely designated as Stx1 and Stx2, respectively; for exact nomenclature of Stx subtypes refer to Scheutz et al. 2012); among them Stx2a is more potent as a cause of severe HUS and cerebrovascular dysfunction (Bielaszewska et al. 2011; Mellmann et al. 2011). Pigs are known to develop edema disease through STEC producing Stx2e, which represents the most frequent Stx2 subtype in porcine feces (Fratamico et al. 2004). Edema disease is a common cause of illness and death loss in pigs during the first 2 weeks after weaning (Moxley 2000). The edema disease is characterized clinically by neurologic signs such as ataxia, incoordination and recumbency and/or subcutaneous and submucosal edema in various tissues (Matise et al. 2000). Vascular damage with resultant infarction and malacia in the brain stem is the main cause of death in affected pigs. Clinical signs and lesions are largely the result of Stx2e, which causes impairment of porcine brain capillary endothelial cells (PBCECs) (Moxley 2000), suggesting weakening and leakage of the endothelial blood brain barrier (BBB), which might explain severe brain tissue lesions (Matise et al. 2000) during the disease. However, the precise mechanism of Stx2e-mediated interaction with endothelial target cells of the brain is only poorly understood. All Stxs have a common AB 5 structure composed of a single A subunit and five identical B subunits (Lingwood 1996; Beddoe et al. 2010). The enzymatically active A subunit inhibits protein biosynthesis at ribosomal level (Sandvig et al. 2010) and the pentameric B subunit binds to glycosphingolipids (GSLs) of the globo-series (Smith et al. 2004). GSLs are composed of a highly variable oligosaccharide moiety and a double-tailed ceramide lipid anchor The Author Published by Oxford University Press. All rights reserved. For permissions, please journals.permissions@oup.com 745

2 I Meisen et al. (sphingosine and fatty acid) (Lochnit et al. 2001; Levery 2005; Merrill 2011), which both have a functional impact on membrane microdomain formation (Prinetti et al. 2009; Gupta and Surolia 2010) and thus in GSL-Stx interaction events (Smith et al. 2006; Römer et al. 2007). According to the present knowledge, ceramide heterogeneity of globotriaosylceramide (Gb3Cer) molecules and their organization in lipid rafts may affect the toxin s pathogenicity, since association of GSLs with detergent-insoluble membrane microdomains seems as a functional requirement to bind and sort AB 5 toxins backward to the intracellular targets (Falguières et al. 2001; Lencer and Saslowsky 2005; Lingwood et al. 2010). After endocytosis, Stxs are transported by a retrograde pathway to the Golgi apparatus and the endoplasmic reticulum, before the enzymatically active moiety of the A subunit is translocated to the cytosol, where the toxins inhibit protein biosynthesis by inactivating ribosomes (Sandvig et al. 1992, 2010; Johannes and Römer 2010; Bergan et al. 2012; Spooner and Lord 2012). The initial step of toxin-mediated injury is binding of the B subunits of Stx1a and Stx2a subtype to their cell-surface receptor Gb3Cer/CD77 (Müthing et al. 2009; Lingwood et al. 2010; Bauwens et al. 2011). Stx1a recognizes Gb3Cer and also globotetraosylceramide (Gb4Cer) to a considerable extent, whereas Stx2a preferentially binds to Gb3Cer and to a lesser extent to Gb4Cer as recently shown in a comparative investigation for a collection of Stx1a- and Stx2a-containing supernatants from STEC isolates of human origin (Müthing et al. 2012). Gb4Cer has been previously shown as a principal receptor for Stx2e (DeGrandis et al. 1989), and moreover, to bind (in addition to Gb3Cer and Gb4Cer) to Forssman GSL that constitutes a GalNAcα3-elongated Gb4Cer structure conferring upon Stx2e a unique promiscuous recognition feature (Müthing et al. 2012). So far, no data are available on the GSL receptors of Stx2e expressed by pig endothelial cells and the knowledge about Stx2e-mediated cellular damage of the pig brain microvasculature is poor. We, therefore, address these issues in this report and present the first investigation on detailed structural characterization of Gb3Cer and Gb4Cer species identified as Stx2e receptors of primary PBCECs, their assembly in detergentresistant membranes (DRMs) and Stx2e-mediated cellular breakdown of the BBB using a highly appropriate cell culture model based on the measurement of the transendothelial electrical resistance (TEER). Results Owing to the limited number of primary PBCECs prepared from pig brains, which could be propagated only for one passage in tissue culture flasks, we were unable to perform all the following investigations with one and the same batch of cells. Therefore, we employed in this study PBCECs from three individual batches, i.e., derived from three different pigs of the same race. Isolated total GSLs from PBCECs of Batch 1 (small batch) were entirely spent for antibody-mediated thinlayer chromatography (TLC) overlay detection of Gb3Cer and Gb4Cer combined with structural characterization by electrospray ionization quadrupole time-of-flight mass spectrometry 746 (ESI-Q-TOF MS) (Figures 3 and 4). Batch 2 (large batch) was employed for comparative antibody and Stx2e TLC overlay assays (Figures 1 and 2) as well as for the preparation of DRMs (Figure 5). Batch 3 (small batch) was used in the BBB cell culture model to determine Stx2e-mediated cytotoxicity (Figures 6 and 7). Detection of Stx2e receptors in the neutral GSL fraction of PBCECs using anti-gb3cer and anti-gb4cer antibodies Neutral GSLs of PBCECs (Batch 2) were separated by TLC along with GSL reference mixture 1, composed of Lc2Cer, Gb3Cer and Gb4Cer, and GSL reference mixture 2, Fig. 1. Orcinol stain of neutral GSLs from PBCECs (A) and antibody-mediated detection of Gb3Cer (B) and Gb4Cer (C) in the neutral GSL preparation of PBCECs. GSL amounts are equivalent to cells (A, lane b) and cells (B and C, lane b). Neutral GSL reference mixture 1 containing mostly Gb4Cer and lower quantities of Gb3Cer and Lc2Cer corresponds to 10 µg (A, lane a), 2 µg (B, lane a) and 0.2 µg (C, lane a). Neutral GSL reference mixture 2 comprising equimolar amounts of Gb3Cer, Gb4Cer and Forssman GSL corresponds to 5 µg (A, lane c) and 0.4 µg (B and C, lane c).

3 GSL receptors and Stx2e-mediated injury of endothelial cells Fig. 2. Stx2e-mediated detection of Gb3Cer and Gb4Cer but not of Forssman GSL in the neutral GSL preparation of PBCECs. (A) GSL amounts equivalent to and cells were applied for the orcinol stain (lane a) and the Stx2e overlay assays (lanes b and c), respectively. Stx2e (h) and Stx2e (p) variants from human (lane b) and porcine (lane c) STEC isolates were used (for details of the two Stx2e variants, refer to the paragraph Stx2e-containing supernatants and controls of the Materials and methods section). (B) GSL amounts of 5 and 0.4 µg of reference mixture 2 (harboring equimolar quantities of Gb3Cer, Gb4Cer and Forssman GSL) were applied for the orcinol stain and for the anti-forssman GSL antibody overlay assay in lanes a and b, respectively. A GSL amount equivalent to PBCECs was used for the anti-forssman GSL antibody overlay assay in lane c. containing equimolar quantities of Gb3Cer, Gb4Cer and Forssman GSL. The orcinol stain of the neutral GSLs of PBCECs suggested the presence of Gb3Cer and Gb4Cer (Figure 1A, lane b) as deduced from their chromatographic behavior in comparison with Gb3Cer and Gb4Cer of reference mixture 1 (Figure 1A, lane a) and reference mixture 2 (Figure 1A, lane c). TLC overlay assays with anti-gb3cer (Figure 1B, lane b) and anti-gb4cer antibody (Figure 1C, lane b) confirmed the expression of Gb3Cer and Gb4Cer, respectively, additionally verified by parallel immunostains of Gb3Cer (Figure 1B, lanes a and c) and Gb4Cer (Figure 1C, lanes a and c) in the reference mixtures. Detection of Stx2e receptors in the neutral GSL fraction of PBCECs using Stx2e variants In order to detect to what extent Stx2e binds to PBCEC-derived Gb3Cer and/or Gb4Cer, TLC overlay assays were performed with neutral GSLs (Batch 2) using Stx2e-O101-VUB-EH60 from a human STEC isolate and Stx2e-O139-S123G from a pig STEC isolate (Figure 2A). For the sake of simplification, the former isolate of human origin is named Stx2e (h) and the latter isolate from porcine source is denoted by Stx2e (p) in the following text, whereby h and p stand for human and porcine origin, respectively. The Stx2e overlay assays were performed with an anti-stx2 antibody in conjunction with a secondary alkaline phosphatase (AP)-labeled anti-mouse IgG antibody and 5-bromo-4-chloro-3-indolyl phosphate (BCIP) as the substrate (see paragraph Antibodies in the Materials and methods section). In comparison with the orcinol stain (Figure 2A, lane a), the TLC overlay assays using Stx2e (h) and Stx2e ( p) clearly indicate binding of both Stx2e variants to Gb3Cer and Gb4Cer as shown in lanes b and c of Figure 2A, respectively. A remarkably stronger interaction with Gb4Cer than with Gb3Cer could be observed for both Stx2e variants, indicated by a relatively higher color intensity of the BCIP-positive Stx2e-immunostained Gb4Cer versus Gb3Cer bands when compared with the orcinol stain (Figure 2A, lane a). It has been recently shown that Stx2e binds in addition to Gb3Cer and Gb4Cer to Forssman GSL (Müthing et al. 2012), which represents a GalNAcα3-elongated Gb4Cer structure and separates beneath Gb4Cer. Therefore, we employed a highly specific and sensitive monoclonal anti-forssman GSL antibody to reconsider the presence of Forssman GSL in the neutral GSL preparation (Batch 2) of PBCECs (Figure 2B). The orcinol stain of reference mixture 2, containing equimolar quantities of Gb3Cer, Gb4Cer and Forssman GSL, is depicted in lane a and the corresponding positive control TLC binding assay performed with the anti-forssman GSL antibody is displayed in lane b of Figure 2B. As deduced from the TLC overlay assay shown in lane c of Figure 2B, PBCECs do not express this pentaglycosylceramide, in agreement with failure of detection of the Forssman GSL in the Stx2e overlay assays (Figure 2A, lanes b and c). Mass spectrometric identification of Gb3Cer and Gb4Cer lipoforms Structures of antibody-detected Gb3Cer and Gb4Cer species of the GSL preparation of PBCECs (Batch 1) were identified by ESI-Q-TOF MS. For this purpose, the silica gel of immunopositive Gb3Cer bands and Gb4Cer bands was scraped off the plate and the extracted GSLs were submitted to MS analysis using the positive ion mode (Figures 3 and 4). Gb3Cer lipoforms. To give evidence of putative Gb3Cer structures, we initially performed MS 1 analysis followed by collision-induced dissociation (CID) of selected precursor ions hereafter referred to as MS 2. The overview MS 1 spectrum of all detected [M + Na] + ions obtained from the extracted scratched silica gel of the framed immunopositive area (see inset of Figure 3A) revealed huge diversity of Gb3Cer species and suggests considerable ceramide variability of the detected Gb3Cer variants. The predominant [M + Na] + ions at m/z were tentatively assigned as hydroxylated Gb3Cer (d18:0, C22:0-OH) and those at / correspond to predicted Gb3Cer (d18:0, C24:1-OH/C24:0-OH) structures, although alternative structures such as Gb3Cer (d18:0, C25:0) cannot be excluded (for list of detected major and minor [M + Na] + ions and proposed structures refer to Table I). The less abundant Gb3Cer ions with m/z values of / can be attributed to Gb3Cer (d18:1, C24:1/C24:0), and those at m/z were assigned to hydroxylated Gb3Cer (d18:0, C20:0-OH). By reason of diminished TLC mobility upon introduction of an OH-group, the hydroxylated Gb3Cer species are supposed to accumulate in the more intensively immunostained lower band (see inset of Figure 3A), whereas the aforementioned Gb3Cer species carrying Cer (d18:1, C24:1/C24:0) and minor Gb3Cer (d18:1, C22:0) with m/z at may separate in the less intensive upper band. The very minor ions with m/z values of were assigned to Gb3Cer (d18:1, C16:0), which separates together with the hydroxylated Gb3Cer variants in the lower band of the immunopositive double band. Based on the data obtained from MS 1 analysis, the proposed structures were corroborated by CID after selection of the desired precursor [M + Na] + ions. MS 2 spectra of [M + Na] + ions at m/z and are shown as representative 747

4 I Meisen et al. Fig. 3. ESI-Q-TOF MS 1 and MS 2 spectra of antibody-detected Gb3Cer species from PBCECs. The MS 1 spectrum (A) was obtained from the silica gel extract of the TLC overlay assay, which was performed with a GSL aliquot corresponding to cells as shown in Figure 1B (lane b). The framed dotted rectangle of the inset portrays the immunostained double band of Gb3Cer, from which the silica gel was scraped off. A survey of the m/z values of the [M + Na] + ions detected in the MS 1 spectrum and the corresponding proposed Gb3Cer lipoforms with differing ceramide lipid anchors are provided in Table I. The asterisk marks a Gb3Cer species for 748 examples in Figure 3B and C, respectively, together with the auxiliary fragmentation scheme of Gb3Cer (d18:1, C24:0), which is portrayed in Figure 3D following the nomenclature of Domon and Costello (1988a, b). The MS 2 spectrum of the species with m/z (Figure 3B) reveals a full series of Y-type ions (Y 2 ions at m/z , Y 1 ions at m/z and Y 0 ions at m/z ) indicating the loss of three hexose moieties and a full series of B- and C-type ions. Additionally, W ions at m/z , which are diagnostic for the presence of sphingosine (4-sphingenine, d18:1), give rise to the unambiguous assignment of this species as Gb3Cer (d18:1, C24:0). The MS 2 spectrum of the Gb3Cer variant at m/z , tentatively assigned as Gb3Cer (d18:0, C20:0-OH), is depicted in Figure 3C. Again, complete series of complementary Y- and B-type ions are detected, accompanied by few C- and Z-type ions. The ceramide moiety is represented by Y 0 and Z 0 ions at m/z and , respectively. Even though no further fragmentation within the lipid part was detected, the presence of dihydrosphingosine (sphinganine, d18:0) and the hydroxylated fatty acid C20:0-OH within the ceramide moiety is the sole reasonable assumption. Gb4Cer lipoforms. The MS 1 spectrum derived from the silica gel extract of the Gb4Cer positive double band is shown in Figure 4A as marked by a dotted rectangle in the inset. In accordance with an intensely stained upper band, Gb4Cer variants with Cer (d18:1, C22:0) at m/z , Cer (d18:1, C24:1) at m/z and Cer (d18:1, C24:0) at m/z are present as species of highest abundance. A synopsis of all Gb4Cer species detected in PBCECs is provided in Table II together with their proposed structures. The structural characterization by MS 2 experiments is exemplified by the CID of [M + Na] + ions at m/z (Figure 4B) and (Figure 4C) and explained by the auxiliary fragmentation scheme of Gb4Cer (d18:1, C16:0) shown in Figure 4D. The loss of the terminal HexNAc moiety is indicated by Y 3 ions at m/z (Figure 4B) and Y 2,Y 1 and Y 0 ions at m/z , and are generated by the additional loss of one to three hexose moieties, the latter representing the lipid portion Cer (d18:1, C24:1). In addition, a full series of B-type ions (B 1 B 4 ) and internal fragment ions, marked with asterisks in the MS 2 spectrum, are observed. The MS 2 spectrum of the Gb4Cer variant at m/z is depicted in Figure 4C. Though this species is only present as minor component (see Figure 4A), a clear-cut spectrum comprising complete series of complementary Y-type and B-type accompanied by few C-type, Z-type and internal fragment ions (marked by asterisks) was obtained giving rise to the structural assignment as Gb4Cer (d18:1, C16:0). which only one of three possible lipoforms is assigned in the MS 1 spectrum (see Table I). The MS 2 spectra of proposed Gb3Cer (d18:1, C24:0) at m/z (B) and of predicted Gb3Cer (d18:0, C20:0-OH) at m/z (C) were obtained by CID of the corresponding precursor ions and are exemplarily shown. (D) Molecular structure and explanatory fragmentation scheme of Gb3Cer (d18:1, C24:0).

5 GSL receptors and Stx2e-mediated injury of endothelial cells Fig. 4. ESI-Q-TOF MS 1 and MS 2 spectra of antibody-detected Gb4Cer species from PBCECs. The MS 1 spectrum (A) was obtained from the silica gel extract of the TLC overlay assay, which was performed with a GSL aliquot corresponding to cells as shown in Figure 1C (lane b). The framed dotted rectangle of the inset portrays the immunostained double band of Gb4Cer, from which the silica gel was scraped off. A survey of the m/z values of the [M + Na] + ions detected in the MS 1 spectrum and the corresponding proposed Gb4Cer lipoforms with differing ceramide lipid anchors are provided in Table II. The MS 2 spectra of proposed Gb4Cer Distribution of Gb3Cer, Gb4Cer and phospholipids to sucrose gradient fractions of PBCECs Nine sucrose gradient fractions, which can be further subdivided into DRM fractions (F1 F3), intermediate fractions (F4 F6), bottom fractions (F7 and F8) and the sediment (S) fraction (F9), were prepared from PBCECs (Batch 2) using 1% Triton X-100. Out of the sucrose gradient fractions (1.5 ml each) two-third (1.0 ml each) were submitted to TLC overlay assay analysis for the detection of Gb3Cer and Gb4Cer (see TLC overlay assays of Figures 1 and 2, which were performed with GSLs from the same batch of cells) and one-third (0.5 ml) was used to determine the phospholipid content (Figure 5). The TLC immunodetection of Gb3Cer displayed preference in distribution to DRM fractions as shown in Figure 5A. Relative densitometric quantification gave 37% for canonical DRM fraction F2 and 17 and 7% for the adjacent fractions F1 and F3, respectively, amounting in total for 61% with clear evidence for preferential DRM association of Gb3Cer. The remaining Gb3Cer quantities were detected in the intermediate fractions F4 F6 (14%), the bottom fractions F7 and F8 (22%) and the sediment fraction F9 (3%). Thus, the majority of Gb3Cer molecules floated in the DRM fractions, which are equivalent to the membrane liquid ordered phase, and residual 36% are made up of soluble Gb3Cer, which distributed more or less equally to the non-drm fractions F4 F8 representing the membrane liquid disordered phase. Negligible Gb3Cer amounts (3%) were found in the sediment fraction. An even more pronounced gradient distribution pattern was observed for Gb4Cer with predominant occurrence in DRM fraction F2 (51%) accompanied by fractions F1 and F3 with 18 and 2%, respectively (Figure 5B). Thus, 71% of Gb4Cer floated in the DRM fractions, whereas the residual 29% constituted soluble Gb4Cer in the non-drm fractions F4 F8. No Gb4Cer was detectable in the sediment fraction. In conclusion, these data suggest that both Gb3Cer and Gb4Cer reside preferentially, but not exclusively, in the DRM fractions. Next, the phospholipids of PBCECs were investigated with respect to their gradient separation (Figure 5C). The molybdenum blue stain of the phospholipids of individual gradient fractions demonstrated preponderant occurrence of sphingomyelin (SM) in the classical DRM fraction F2 (45%) and neighboring fractions F1 (14%) and F3 (8%). Far smaller amounts were detectable in the intermediate fractions F4 F6 (18%) and the bottom fractions F7 and F8 (15%). Thus, SM could be identified as a true microdomain marker for PBCECs through its predominance in DRM fractions F1 F3 (summed 67%) and more or less equal distribution of the remaining SM to the non-drm fractions over the entire gradient (Figure 5C). On the other hand, phosphatidylcholine (PC) exhibited a reverse distribution due to preferential occurrence in the bottom fractions F7 and F8 (65%) and minute appearance in the DRM fractions F1 F3 (19%) as well as the intermediate fractions F4 F6 (15%) rendering PC a true non-drm (d18:1, C24:1) at m/z (B) and of predicted Gb4Cer (d18:1, C16:0) at m/z (C) were obtained by CID of the corresponding precursor ions and are exemplarily shown. Peaks marked by asterisks represent fragment ions generated by internal cleavages. (D) Molecular structure and explanatory fragmentation scheme of Gb4Cer (d18:1, C16:0). 749

6 I Meisen et al. Table I. Monoisotopic m/z values and proposed ceramide composition of the Gb3Cer species detected in PBCECs by ESI MS coupled with TLC immunostain Gb3Cer a [M + Na] + m/z Ceramide d18:1, C16: d18:0, C20:0-OH d18:1, C22: d18:0, C22:0-OH d18:1, C24: d18:1, C24: d18:0, C24:1-OH; d18:1, C24:0-OH; d18:0, C25: d18:0, C24:0-OH a Gb3Cer lipoforms were identified in extracts of immunostained bands (Figure 1B, lane b). The MS 1 spectrum of Gb3Cer species is shown in Figure 3A and MS 2 spectra of predicted Gb3Cer (d18:1, C24:0) and Gb3Cer (d18:0, C20:0-OH) are exemplarily shown in Figure 3B and C, respectively. Major lipoforms are typed in bold print. Table II. Monoisotopic m/z values and proposed ceramide composition of the Gb4Cer species detected in PBCECs by ESI MS coupled with TLC immunostain Gb4Cer a [M + Na] + m/z Ceramide d18:1, C16: d18:1, C20: d18:1, C22: d18:1, C23: d18:1, C24: d18:1, C24: d18:1, C24:1-OH; d18:1, C25:0; d18:0, C25: d18:1, C24:0-OH; d18:0, C24:1-OH; d18:0, C25: d18:1, C26: d18:1, C26:0 a Gb4Cer lipoforms were identified in extracts of immunostained bands (Figure 1C, lane b). The MS 1 spectrum of Gb4Cer species is shown in Figure 4A and MS 2 spectra of predicted Gb4Cer (d18:1, C24:1) and Gb4Cer (d18:1, C16:0) are exemplarily shown in Figure 4B and C, respectively. Major lipoforms are typed in bold print. marker of PBCECs. The same holds true for phosphatidylethanolamine (PE), which was only detectable in the F7 bottom fraction (Figure 5C). In summary, the preferential distribution of Gb3Cer, Gb4Cer and SM to DRM fractions evidenced their colocalization in the membrane liquid ordered phase and identified these three membrane constituents as potential lipid raft markers, whereas PC and PE represent typical molecules of the membrane liquid disordered phase and can be considered nonlipid raft markers in PBCECs. Mass spectrometric identification of Gb3Cer and Gb4Cer lipoforms of DRM fraction F2 Mass spectrometric analysis of canonical DRM fraction F2 (Batch 2) revealed almost identical lipoforms of Gb3Cer (Supplementary data, Figure S1A) but different lipoforms of Gb4Cer (Supplementary data, Figure S1B) regarding the extent of ceramide hydroxylation in comparison with the 750 Fig. 5. Distribution of Gb3Cer (A), Gb4Cer (B) and phospholipids (C) to sucrose density gradient fractions of PBCECs. Gradient fractions (F1 F9) were numbered from top (F1) to bottom (F8); F9 was obtained by solubilization of the sediment (S). Fractions F1 F3 represent the DRM fractions and fractions F4 F9 the non-drm fractions. Lipid extracts were separated by TLC and immunostained Gb3Cer (A), immunostained Gb4Cer (B) and molybdenum blue-stained phospholipids PC and SM (C) were quantified by densitometric scanning. The percentage values of the relative content of Gb3Cer, Gb4Cer and the phospholipids PC and SM in the respective fractions are given below each panel [F1 F9 (%)]. GSL (A and B) and phospholipid amounts (C) of the gradient fractions correspond to cells and cells, respectively. GSL amounts of 2 µg (A) and 0.2 μg (B) of reference mixture 1 (R) containing mostly Gb4Cer and lower quantities of Gb3Cer and Lc2Cer were coseparated as positive controls; 60 µg of a well-defined phospholipid reference mixture (R) was applied for assignment of phospholipids in the gradient fractions (C). Phospholipid references from top to bottom: CL, cardiolipin; PG, phosphatidylglycerol; PE, phosphatidylethanolamine; PA, phosphatidic acid; PC, phosphatidylcholine; SM, sphingomyelin. spectra obtained from the GSL preparation of total PBCECs (Batch 1). However, despite the conspicuous differences in the ceramide moieties of Gb4Cer, indicating a batch-to-batch heterogeneity of this Stx2e receptor, the expression of Gb3Cer and Gb4Cer represents a stable feature of primary PBCECs from different cell culture batches. Due to the preferential distribution of both Stx2e receptors to DRM fractions only low

7 GSL receptors and Stx2e-mediated injury of endothelial cells quantities of Gb3Cer (Figure 5A) and Gb4Cer (Figure 5B) were present in the non-drm fraction F7. These amounts were not sufficient to obtain precise mass spectra of fraction F7. However, some ions of very low abundance suggest almost identical Gb3Cer and Gb4Cer species when compared with DRM fraction F2 (data not shown). Stx2e-mediated destruction of PBCECs that form the BBB To investigate Stx2e-caused damage of the BBB, we employed a well-established cell culture model consisting of primary PBCECs (Batch 3) cultured on Transwell filter inserts (Franke et al. 2000; Matthes et al. 2011). This in vitro model based on endothelial cell monolayers closely resembles the in vivo situation of the microvasculature of the porcine brain. The bioelectric properties determined by impedance spectroscopy provide quick access to cytotoxic properties of a wide range of effector molecules such as bacterial toxins, including Stxs. Characterization of the BBB model. PBCECs were cultivated on filters in a two-chamber system, in which the endothelial cell monolayer separates the apical from the basolateral compartment representing the blood and brain side of the capillary endothelium, respectively (Figure 6A). The characteristic feature of the porcine endothelial cell monolayer (Figure 6B) is its high TEER of >1000 Ω cm 2 and capacitances per unit area ranging from 0.45 to 0.60 µf cm 2, which closely resemble the physiological conditions (Franke et al. 2000). The barrier properties can be maintained for at least 24 h as demonstrated by time-dependent continuous relative TEER measurements after apical administration of serum-free (sf) chemically defined medium, which was used as the assay medium (Figure 6C). Relative TEER measurements ranging between 70 and 90% upon apical exposure of the cells to sf medium indicate a stable cell monolayer that fulfills the barrier function (Figure 6C, open triangles). The control assay with Luria-Bertani (LB) medium diluted 1:32 in sf medium hints at tolerance of PBCECs toward sf medium containing small volumes of LB medium as evidenced by TEER values between 85 and 90% (Figure 6C, filled triangles). Next, the supernatants obtained from nontoxic laboratory E. coli K12 and E. coli C600 strains, which were propagated in LB medium and do not produce any toxins, were administered in small volumes to sf medium (1:32 dilutions). Sf medium spiked with bacterial supernatants did not significantly affect the endothelium and indicates maintenance of the cell barrier function as shown by a constant relative TEER of 60% over the whole time span of 24 h (Figure 6C, filled and open circles). Stx2e-mediated breakdown of the BBB. In order to detect to what extent Stx2e harms the integrity of the BBB, 1:32- diluted bacterial cell culture supernatants of Stx2e-O101- VUB-EH60 (human STEC isolate) and Stx2e-O139-S123G (pig STEC isolate), designated as Stx2e (h) and Stx2e (p) (see paragraph Detection of Stx2e receptors in the neutral GSL fraction of PBCECs using Stx2e variants in the Results section above), were administered to PBCEC monolayers to characterize Stx2e-mediated changes of the relative TEER. As demonstrated in Figure 7A, apical application of Stx2e (h) (Figure 7A, filled triangles) and Stx2e ( p) supernatants Fig. 6. Endothelial cell culture model of the BBB. (A) Sketch of the cell culture model. PBCECs grow as monolayer on filter inserts and separate an apical (blood side) and basolateral compartment (brain side). (B) Phase contrast image of confluent grown PBCECs. Wells with TEER values of 1000 Ω cm 2 and capacitances per unit area between 0.45 and 0.60 µf cm 2 revealing a confluent PBCEC monolayer were employed for Stx2e treatments and controls without toxin. (C) Examples of continuous monitoring of the relative TEER of PBCECs for 24 h after apical application of sf pure chemically defined medium (assay medium) and LB medium as well as Stx-free E. coli K12 and E. coli C600 cell culture supernatants. The latter three control experiments were conducted with 1:32 dilutions of LB medium and supernatants (E. coli K12 and E. coli C600) in sf medium. This dilution corresponds to the dilution of Stx2e-containing supernatants used in the cell culture model (see Figure 7). (Figure 7A, filled circles) exerted a strong cytotoxic cellular effect resulting in a 50% reduction of the TEER after 2.5 and 2.2 h exposure, respectively, and an entire breakdown of the 751

8 I Meisen et al. Fig. 7. Loss of BBB integrity upon exposure of PBCECs to Stx2e-containing STEC supernatants. Stx2e (h) and Stx2e (p) from human and porcine STEC isolates, respectively, were administered to confluent grown PBCECs from the apical or basolateral side, and the effects are depicted as relative TEER (A) and capacitance measurements (B) for 24 h. Stx2e supernatants were applied as 1:32 dilution in sf medium (see Figure 6). Calculated standard deviations 5% on average and were omitted from the graph for reasons of simplification. BBB after 7.7 h incubation observed for both toxins with the result of an approximate TEER of < 2%. The complementary capacitance values corroborated the TEER measurements as being obvious from the sigmoidal curve of the capacitance, which started to increase after 6.8 and 6.2 h exposure from an average value of 0.5 µf cm 2 reaching a plateau of 11.8 µf cm 2 and 10.9 µf cm 2 after 14 h exposure in the case of Stx2e (h) (Figure 7B, filled triangles) and Stx2e ( p) (Figure 7B, filled circles), respectively. The TEER time course of Stx2e (h) and Stx2 ( p) upon basolateral administration exhibited basically a similarly strong cell damaging picture when compared with the apical application mode. However, a 50% reduction of the TEER was determined after an exposure time of 3.5 h for Stx2e (h) (Figure 7A, open triangles) and 2.8 h treatment using Stx2e ( p) (Figure 7A, open circles) indicating a slight parallel shift toward marginally diminished cytotoxicity of the two basolaterally applied Stx2e variants. The total breakdown of the endothelium occurred at 10.4 and 9.0 h for Stx2e (h) and Stx2e ( p), respectively, resulting in a minute delay in reaching a TEER value of < 2% when compared with apical toxin 752 delivery. The corresponding capacitance data confirmed the TEER values deduced from a matchable delay to reach a plateau of 11.9 µf cm 2 after 14.8 h and 12.4 µf cm 2 after 14.0 h treatment with Stx2e (h) (Figure 7B, open triangles) and Stx2e ( p) (Figure 7B, open circles), respectively. The application of 1:16-diluted Stx2e supernatants showed a slightly steeper drop down and applying 1:64-diluted Stx2e supernatants revealed a notably retarded decline of relative TEER measurements portrayed in Supplementary data, Figures S2 and S3, respectively. In conclusion, highly diluted Stx2e-containing supernatants obtained from STEC strains of human and porcine origin exhibited strong cytotoxic effects on primary PBCECs and a rapid collapse of the BBB irrespective of the side of application. In other words, both the apical (luminal) and the basolateral (tissue) sides of primary PBCECs are sensitive to the action of Stx2e, giving evidence for expression of globoseries receptor GSLs in the plasma membrane on both sides of this type of endothelial cells. Lipopolysaccharide of E. coli does not infringe BBB integrity. In order to answer the question whether lipopolysaccharides (LPS), which are common constituents of the supernatants of E. coli cell cultures, could impact on the BBB integrity, control experiments were performed using LPS solutions with defined concentrations of 10 up to 200 ng ml 1. Notably, they did not show any harmful effect on PBCECs at these relevant concentrations, thereby excluding interference and/or involvement of E. coli endotoxin in the abrupt crash of the BBB, which was observed for Stx2e-containing supernatants that may harbor LPS in the analyzed range of concentrations (Supplementary data, Figure S4). These experiments confirmed the results obtained by employing supernatants from E. coli K12 and E. coli C600 laboratory strains being unable to produce virulence factors due to limited genetic endowment. Discussion Stx2e is the key virulence factor involved in the pathogenesis of porcine edema disease. Endothelial cell necrosis in the brain is an early event in the pathogenesis of edema disease caused by Stx2e-producing STEC strains and is most probably due to the direct effect of Stx2e responsible for the morphological changes observed in the arterioles of infected pigs (Matise et al and cited references). Because data regarding the Stx2e receptors of swine cerebral endothelial cells are the missing link that could explain the direct effect of Stx2e on the brain target cells, we addressed this issue in the present study and identified Stx2e-binding Gb3Cer and Gb4Cer as the prevalent neutral GSLs in primary PBCECs. The detection of Gb3Cer and Gb4Cer in PBCECs is of particular interest in the context of a previous study, in which both GSLs were detected in the GSL extracts of several swine tissues, but not in the brain (Boyd et al. 1993). Although Gb3Cer and Gb4Cer receptors have not been demonstrated in the brain, Stx2e binds to small arteries and capillaries in this tissue, and lesions occur in the brain of pigs after injection of Stx2e (MacLeod et al. 1991; Boyd et al. 1993; Waddell et al. 1998). Our in-depth structural analysis of Gb3Cer and Gb4Cer

9 GSL receptors and Stx2e-mediated injury of endothelial cells derived from PBCECs revealed various lipoforms exhibiting an unusual high extent of ceramide hydroxylation. This particularly applies to the dominant Gb3Cer species detected in both primary PBCEC batches carrying a long-chain C20:0, C22:0 or C24:1/24:0 fatty acid (see Table I and Supplementary data, Table SI). Remarkably, hydroxylated ceramide of Gb4Cer was detected in minor Gb4Cer variants derived from Batch 1 carrying a C24:1/C24:0 fatty acid (see Table II), but was identified in major Gb4Cer variants of Batch 2 carrying a long-chain C20:0, C22:0 or C24:1/24:0 fatty acid (see Supplementary data, Table SII). This modification and the relatively high content of C24:0/C24:1 versus C16:0 fatty acid harboring Gb3Cer and Gb4Cer species illustrate the most prominent differences of PBCECs in comparison with human endothelial cells such as human umbilical vein endothelial cells (Müthing et al. 1999; Okuda et al. 2010), human brain microcvascular endothelial cells (HBMECs) (Schweppe et al. 2008; Betz et al. 2011) and human glomerular microvascular endothelial cells (GMVECs) (Betz et al. 2012) as recently reviewed by Bauwens et al. (2013). Although expression of Gb3Cer (and most likely also of Gb4Cer) and accessibility of lipid-bound oligosaccharides of the globo-series on the cell surface are essential for initiation of Stx-mediated cellular injury, the knowledge of the functional impact of the ceramide moiety and, in particular, its contribution to efficient binding of Stx to endothelial cells and subsequent internalization is still rather poor. Thus, it remains tempting to speculate whether the abnormally high content of hydroxylated Gb3Cer or the reduced amount of hydroxylated Gb4Cer observed in one of the two batches of primary PBCECs, from which GSLs were isolated, may play a functional role in Stx2e-mediated cell death and fatal cerebral complications in the course of pig edema disease. Gb3Cer and Gb4Cer have also been determined in previous studies as the major neutral GSLs in porcine aorta endothelial cells (Bouhours et al. 1996). In another approach, substantial susceptibility to the cytotoxic action of Stx2e, measured by the neutral red cytotoxicity assay, was observed for porcine aorta endothelial cells indicating relationship between cell susceptibility and content of Gb4Cer (Valdivieso-Garcia et al. 1996). Although their exact lipoforms have not been established at that time, both Gb3Cer and Gb4Cer can be likely considered to carry mainly long (C24) and short chain fatty acids (C16) according to their classical TLC separation in double bands, respectively (Bouhours et al. 1996), as previously observed for many other endothelial cell types. Since pigs express several human xenoantigenic GSL determinants, some studies have focussed in the past on the identification of xenoantigens like Galα3Gal-epitopes in swine tissues including endothelial cells (Bouhours et al. 1996; Vanhove et al. 1998; Koma et al. 2000; Leonardsson et al. 2004). Extended neolacto-series penta- and heptaglycosylceramides with terminal Galα3Gal-determinant (xenoreactive epitope) have been identified in porcine aorta endothelial cells (Bouhours et al. 1996), but simple isogb3cer and isogb4cer with Galα3Galβ4Glc- and GalNAcβ4Galα3Galβ4Glc-structure, respectively, which have been determined in human thymus by permethylation and multiple ion trap mass spectrometry (Li, Teneberg et al. 2008, Li, Zhou et al. 2008), have so far not been identified in porcine endothelial cells neither derived from aorta (Bouhours et al. 1996; Valdivieso-Garcia et al. 1996) nor from the brain as reported in this study analyzing primary PBCECs. We detected Gb3Cer and Gb4Cer by means of highly specific antibodies and Stx2e, which to the best of our knowledge do not bind to the isoforms of Gb3Cer and Gb4Cer. However, although the presence of isogb3cer and/or isogb4cer in PBCECs cannot be excluded, it is suggested that they (if present) should not have any functional impact on Stx2e-mediated cytotoxicity, because binding of Stxs to GSLs carrying a terminal or an internal Galα3Gal-sequence has never been described so far. A pronounced association of Stx2e receptors Gb3Cer and Gb4Cer with DRMs has been determined for primary PBCECs in our study, suggesting preferential assembly of Stx2e receptors in membrane lipid raft microdomains. The compositional analysis of such domains has been obtained from DRMs (London and Brown 2000) as the ruling method for assigning lipid raft-association (Lingwood and Simons 2007). The organization of GSLs in lipid rafts is deemed the prerequisite for efficient binding and subsequent subcellular retrograde transport for various carbohydrate-binding toxins, including Stxs to intracellular targets as demonstrated in several studies using nonendothelial cell lines (Fantini et al. 2000; Falguières et al. 2001, 2006; Kovbasnjuk et al. 2001; Fantini 2003; Lencer and Saslowsky 2005; Smith et al. 2006; Lingwood et al. 2010). The association of GSLs with lipid rafts in endothelial cells and its functional impact on endothelial cell injury has been mostly ignored so far and, thus, remains largely elusive. The exception is the work of Warnier et al. (2006) that nicely demonstrated requirement of lipid rafts for the formation of receptor-stx complexes and retrograde transport in primary GMVECs. The preferential DRM distribution of Gb3Cer and Gb4Cer in PBCECs as demonstrated in this work hint at lipid raft-association in PBCECs and was quite similar to HBMECs (Betz et al. 2011). The only deviation compared with this canonical association has been recently reported for immortalized GMVECs, which exhibited preferential distribution of Gb3Cer and Gb4Cer to non-drm fractions (Betz et al. 2012). This unexpected finding contrasts with the current opinion regarding the imperative requirement of DRM association for Stx-mediated cellular impairment. Thus, the unknown mechanisms underlying differences in Stx-mediated deterioration of endothelial cells still remain largely unexplored. The cerebral microvascular endothelium, which mainly exerts barrier function, protects the brain parenchyma by restricting, for instance, toxic substances and guarantees an undisturbed homeostasis of the central nervous system, which is a prerequisite for neuronal function. Damage to brain microvascular endothelial cells caused by Stx2e is the key pathomechanism underlying neurological impairment, and vascular damage in the brain stem is the main cause of death in pigs suffering from edema disease (Mainil 1999; Matise et al. 2000; Moxley 2000; Obata 2010). The experiments presented here were performed with a well-established in vitro two-chamber cell culture model based on primary PBCECs isolated from porcine brain vessels (Hoheisel et al. 1998). This BBB model is perfectly suited to determination of Stx2e-caused cellular deterioration, because the tight layer of PBCECs closely reflects the in vivo situation where cells do 753

10 I Meisen et al. not proliferate (Franke et al. 1999) and, moreover, this approach is superior to many other assays allowing analysis of label-free living cells and monitoring disturbances and disintegration of a cell monolayer by real-time TEER measurements. In order to gain a deeper insight into the dynamics of Stx2e-challenged PBCECs, we applied one Stx2e variant originating from a human and another variant descending from a porcine isolate. The reaction of PBCECs observed for both variants was a rapid and dramatic decrease of barrier integrity, which started immediately upon exposure of the two Stx2e variants from the apical side and 1 h later after basolateral administration. This little delay is most plausibly due to the fact that the toxins have to pass the filter inserts (on which cells are growing) when applied basolaterally. Thus, a 50% decline of relative TEER was observed after 2.5 and 3.5 h upon apical and basolateral application, respectively, indicating a 1 h parallel shift as well as a delayed start of action of Stx2e from the brain side. The steepness of TEER reduction curves was similar for Stx2e (h) and Stx2e (p) when administered from the vascular and from the brain side, and basically the same regime of BBB destruction was determined for both Stx2e variants. With regard to the effect of Stx2e-mediated cellular damage, vascular ultrastructural investigation of brain sections from swine infected with STEC suggested that the Stx-induced arteriolar lesions were primarily necrotic (Matise et al. 2000). Hence, it is tempting to interpret the rapid decline of the TEER as the initiation process that may end up in severe endothelial damage as a result of necrotic processes as shown by the in vivo data of Matise et al. (2000). Edema disease usually occurs 1 2 weeks postweaning, suggesting protection of suckling piglets against infections of Stx2e-producing STEC strains presumably via compounds present in milk. Although control of edema disease is mainly based on antimicrobial therapy, use of antibiotics in swine production increases the risk of the development of antibiotic-resistance of porcine STEC and is becoming a major concern. An alternative strategy is the immunization of piglets (Bosworth et al. 1996; Johansen et al. 1997) or of breast-feeding sows, since maternal immunity against Stx2e toxoid can be transmitted from the vaccinated sows to the piglets via the colostrum (Oanh et al. 2012). However, future concepts should also consider the anti-adhesive concept as prophesied by Nathan Sharon in his review entitled Safe as mother s milk: carbohydrates as future anti-adhesion drugs for bacterial diseases that anti-adhesive oligosaccharides will in the near future join the arsenal of drugs for the therapy of bacterial diseases (Sharon and Ofek 2000). Such toxin blockers could be applied for prevention and cutting the risk of infection (Sharon 2006). However, the manufacture of soluble oligosaccharide-based Stx2e competitors aimed at amelioration or prevention of STEC-mediated infections in pigs remains challenging, because the major obstacle lies in the production of Stx2e neutralizers at a reasonable price. Materials and methods Porcine endothelial cells and cell cultivation Primary porcine brain capillary endothelial cells (PBCECs) were isolated from the brains of 6-month old pigs as 754 described in detail by Franke et al. (2000) and stored in the gas phase over liquid nitrogen. Cells were thawed and propagated for one passage in 75 cm 2 collagen-precoated cell culture tissue flasks (Nunc, Langenselbold, Germany) using culture medium composed of Medium M199 with Earle s salts (Biochrom KG, Berlin, Germany) supplemented with 10% (v/ v) newborn calf serum (Cat. No. S0125, Biochrom), 0.7 mm glutamine and 200 U ml 1 of penicillin and 200 µg ml 1 of streptomycin, respectively, in a humidified 5% CO 2 atmosphere at 37 C. Cells from confluently grown tissue culture flasks were used for subsequent GSL and phospholipid analysis (see next paragraph). Cell proliferation and morphology were microscopically controlled by use of an Axiovert 40 C microscope (Carl Zeiss, Göttingen, Germany) in phase contrast. Images were taken with a compact PowerShot G10 camera (Canon, Inc., Tokio, Japan) and processed with the software Axiovision Rel. 4.8 (Carl Zeiss). Extraction of phospholipids and isolation of GSLs from total PBCECs Lipids were extracted from 10 tissue culture flasks as described (Betz et al. 2011). Briefly, harvested cells were taken up in 30 ml of phosphate-buffered saline (PBS) whereof 1/3 (10 ml) was used for total lipid extraction and 2/3 (20 ml) was used for DRM preparation (see chapter Preparation of DRMs below). After centrifugation, total lipids were extracted sequently from the cell sediment with chloroform/methanol in different mixing ratios (Betz et al. 2011). The combined extracts were dried under rotary evaporation and redissolved in 20 ml of chloroform/methanol (2/1, v/v). Forty percent (8 ml) of the extract were used directly for phospholipid analysis (see chapter High-performance TLC of phospholipids and GSLs below) and the remaining 60% (12 ml) of the extract were applied to isolation of GSLs. For this purpose, the extract was dried and phospholipids were removed by saponification as described (Betz et al. 2011). Finally, the GSL containing sample was redissolved in a definite volume of chloroform/ methanol (2/1, v/v) corresponding to cells µl 1. Preparation of DRMs DRMs were prepared according to Brown and Rose (1992) with minor modifications as described previously (Betz et al. 2011, 2012). Briefly, membranes from confluent grown cells were disintegrated by cell lysis and cell debris was removed by low-speed centrifugation. Separation of the membranes from the cytosol was performed by ultracentrifugation of the supernatant. After resuspension of the membrane sediment in 1 ml of 1% Triton X-100 (w/v) in TNE buffer, the solution was overlayed with a discontinuous sucrose bottom-to-top gradient (for details refer to Betz et al. 2011). Eight fractions of 1.5 ml volume each were collected from top to bottom and the sediment was dissolved in 1.5 ml of PBS (fraction 9). Extraction of phospholipids and isolation of GSLs from DRM and non-drm fractions of PBCECs For comparative phospholipid analysis of DRM and non-drm fractions, 0.5 ml (one-third) of each sucrose gradient fraction was dialyzed against distilled water at 4 C for 2 days and