Targeted Disruption of Gb3/CD77 Synthase Gene Resulted in the Complete Deletion of Globo-series Glycosphingolipids and Loss of Sensitivity to Verotoxins*

To examine whether globotriaosylceramide (Gb3/CD77) is a receptor for verotoxins (VTs) in vivo, sensitivity of Gb3/CD77 synthase null mutant mice to VT-2 and VT-1 was analyzed. Although wild-type mice died after administration of 0.02 μg of VT-2 or 1.0 μg of VT-1, the mutant mice showed no reaction to doses as much as 100 times that administered to wild types. Expression analysis of Gb3/CD77 in mouse tissues with antibody revealed that low, but definite, levels of Gb3/CD77 were expressed in the microvascular endothelial cells of the brain cortex and pia mater and in renal tubular capillaries. Corresponding to the Gb3/CD77 expression, tissue damage with edema, congestion, and cytopathic changes was observed, indicating that Gb3/CD77 (and its derivatives) exclusively function as a receptor for VTs in vivo. The lethal kinetics were similar regardless of lipopolysaccharide elimination in VT preparation, suggesting that basal Gb3/CD77 levels are sufficient for lethal effects of VTs.

VT-producing E. coli are considered to be causative agents of hemorrhagic colitis (7), and infections have often been associated with hemolytic uremic syndrome (HUS), which has been associated with acute renal failure, thrombocytopenia, and microangiopathic hemolytic anemia (8). HUS is also frequently associated with disruption of the central nervous system (9) and can be lethal.
Gb3/CD77 was reported to bind specifically to the ␤-subunit of VTs. Consequently, tissue damage by VTs is considered to depend on Gb3 expression in individual tissues (10). However, the following issues remain to be clarified, specifically, whether Gb3 is the only specific receptor for VTs in vivo and whether clinical signs of HUS are dependent upon Gb3 expression.
In the present study, we established knock-out mice for the Gb3/ CD77 synthase gene (11) and analyzed their sensitivity to VTs as well as Gb3/CD77 expression in mouse tissues. We clearly demonstrated that Gb3/CD77 and/or its derivatives are exclusive receptors in vivo and that they mediate the tissue damage and pathological features caused by VTs. These results suggest that therapeutic approaches directed at disrupting the function of Gb3/CD77 could be an effective method for protecting against disorders related to pathogenic E. coli infection.

EXPERIMENTAL PROCEDURES
Generation of Knock-out Mice-A targeting vector was constructed using a 10.5-kb mouse genomic fragment including the ␣1,4Gal-T gene. Part of the ␣1,4Gal-T gene (nucleotides 68 -615 in the open reading frame) was replaced with a neo-resistant gene (Fig. 1A). The targeting vector was linearized and transfected into embryonic stem cells, and G418-resistant clones were isolated. Homologous recombination was confirmed by dual Southern blotting. Chimeric mice were generated by microinjection of the embryonic stem clones into embryos at the 8-cell stage and were mated with C57BL/6 mice to generate heterozygotes. Genotyping was performed by PCR using genomic DNA isolated from mouse tails and amplified with primers: P1, 5Ј-ACG ACC TCT GGA CTT GCA AGA ACT GTT TGA-3Ј; P2, 5Ј-AAG GCA CCG TTG AGG ACA TAG CGG GAT-3Ј; and P3, 5Ј-GCC TGC TTG CCG AAT ATC ATG GTG GAA AAT-3Ј (Fig. 1C). The Nagoya University Committee on Animal Research approved all experimental procedures.
Northern Blotting-Total RNA was prepared from mouse tissues using TRIzol TM (Invitrogen). Total isolated RNAs were then separated using 1.25% agarose gels and hybridized with 32 P-labeled probes. The cDNA fragment corresponding to nucleotides 90 -637 in the open reading frame of the mouse ␣1,4Gal-T gene (GenBank accession number AA682117) was used as a probe.
Enzyme Assay-The enzyme activity of ␣1,4Gal-T was measured as described previously (11), and membrane fractions were prepared from kidneys.
VT-1 and VT-2 Preparation-A clinically isolated E. coli O157 H7 strain was used for production of VTs as reported (14). The Luria broth culture supernatant was precipitated with 60% saturated ammonium sulfate at 4°C. The precipitate was dissolved in phosphate-buffered saline and used after dialysis against phosphate-buffered saline. Contaminating lipopolysaccharides (LPS) in this preparation (10.62 mg/ml) were eliminated by Detoxi-Gel TM (Pierce); the LPS concentration was Ͻ 0.001% (final 99.52 ng/ml), but VT-2 concentration was Ͼ93% after the purification. VT-2 concentration was determined by the VTEC-RPLA TM kit (Denka Seiken, Tokyo). LPS concentration in the crude VT-2 preparation was determined by the ENDOSPESY TM kit (Seikagaku Corp., Tokyo).
Histology and Immunohistochemistry-For pathological analysis, tissues from 10 -15-week-old mice before and after treatment with VT-2 were fixed with 3.7% formalin in phosphate-buffered saline and embedded in paraffin. The sections were stained with hematoxylin-eosin. For immunohistochemical analysis, tissues were frozen in liquid nitrogen and 7-m sections were prepared on a cryostat (Leica) and fixed with ice-cold acetone for 15 min at Ϫ20°C. After blocking with 0.05% H 2 O 2 and 10% normal goat serum, cryosections were incubated with mAb-38. 13 (1:50) or an anti-CD31 mAb-390 (1:50) (eBioscience), and the antibody bindings were detected with Histofine Simple Stain Mouse Ϫ/Ϫ, homozygote. E, synthetic pathway of globoseries glycolipids. ␣1,4Gal-T is a key enzyme in the synthesis of the globo-series glycolipids. F, the products of the enzyme assay using UDP-[ 14 C]galactose and LacCer were analyzed by TLC/ autoradiography (left). TLC of neutral glycolipids from kidney tissues with a solvent system chloroform/methanol/water (60:35:8) was visualized by orcinol-H 2 SO 4 . Neutral glycolipids from B erythrocytes were used as a standard (St). TLC immunostaining of kidney neutral glycolipids with mAb 38.13 (right). CMH, ceramide monohexoside; CDH, ceramide dihexoside; Gb4, globotetraosylceramide; Gb5, galactose ␤3-Gb4.
MAX PO TM (Rat) (Nichirei, Tokyo) and 3,3Ј-diaminobenzidine-tetrahydrochloride (Dojin, Kumamoto, Japan) as a substrate. Nuclei were stained with hematoxylin. For negative controls, immunohistochemistry was carried out using non-relevant antibodies with the same isotypes. To determine the expression sites of Gb3 in tubules, we carried out immunohistochemistry using fluorescence-conjugated second antibodies and performed our analysis with a Fluoview FV500 TM confocal laser microscope (Olympus, Tokyo). Antibody binding to Gb3 and CD31 was detected with anti-rat-IgM fluorescein isothiocyanate (ICN Pharmaceuticals, Aurora, OH) and anti-rat-IgG-Alexa-488 (Molecular Probe, Invitrogen), respectively. GM1 expression in tissues was detected with CTB-Alexa-555 (1:100) (Molecular Probe, Invitrogen).
Measurement of TNF-␣ and IL-1␤-After injection of VT-2, mouse serum was obtained and used for enzyme-linked immunosorbent assay using Mouse TNF-␣ or IL-1␤ immunoassay TM kits (Biosource International, Camarillo, CA).

RESULTS
Generation of ␣1,4 Galactosyltransferase Knock-out Mice-After transfection of the targeting vector (Fig. 1A) and G418 selection, the obtained clones were Southern blotted. Of the 12 candidate homologous recombinants selected from the first screening, 9 clones were identified in the second screening (data not shown). Chimeric mice were generated by microinjection of three isolated clones, and heterozygotes were mated with each other to generate homozygotes. The results of Southern blotting and PCR genotyping are shown in Fig. 1, B and C. Northern blotting revealed that ␣1,4Gal-T mRNA was completely absent (Fig. 1D).
Loss of Gb3 Synthase Activity and Globo-series Glycolipids in ␣1,4Gal-T Ϫ/Ϫ Mice-We examined Gb3 synthase activity and globoseries glycolipids (Fig. 1F). Gb3 synthase activity in membrane fractions from kidney was reduced by ϳ50% in ␣1,4Gal-T ϩ/Ϫ mice compared with the wild type, with no activity observed in ␣1,4Gal-T Ϫ/Ϫ mice. TLC analysis revealed that the globo-series glycolipids were completely deleted in ␣1,4Gal-T Ϫ/Ϫ mice; ceramide monohexoside and ceramide dihexoside were observed to increase. Neutral glycolipids in mouse kidney were similar to those reported previously (15), and TLC immunostaining confirmed that ␣1,4Gal-T Ϫ/Ϫ mice contained no Gb3. These mutant mice were born at ratios that followed Mendelian inheritance and grew up without apparent abnormalities over a year. Routine examinations revealed no apparent blood, serum, urine, or fecal disorders (data not shown).
Immune Tissues and Antibody Levels in Gb3 Null Mutant Mice-A number of studies reported that a subset of immature B-cells expressed Gb3/CD77 and that they underwent apoptosis under certain conditions (3). These findings indicate that this antigen is a functional molecule involved in B-cell differentiation. In this study, routine immunological analyses, such as counting the numbers of spleen cells and thymocytes, measurement of immunoglobulin levels, and ratios of lymphocyte populations or T-cell subsets, were performed and no apparent abnormalities were found ( Table 1).
Sensitivity of Mutant Mice to VTs-Although a number of studies indicated that Gb3/CD77 was a receptor for VTs (16) and its possible roles in HUS, direct evidence for Gb3/CD77 being a target for VTs in vivo has never been demonstrated. To clarify whether tissue damage depends on Gb3 expression in vivo, responses of mice to VTs were examined. As expected, wild-type mice exhibited convulsions and shivering within 48 h and died 48 -120 h after injection of more than 0.02 g of VT-2 or 1 g of VT-1. Conversely, ␣1,4Gal-T Ϫ/Ϫ mice exhibited no abnormal behavior and survived even after the injection of 2.0 g of VT-2 or VT-1 ( Table 2). Histological analysis of VT-2-treated mouse tissues revealed tissue damage in the brain cortex, proximal tubules, glomerulus, and liver in wild-type mice after VT-2 injection. Pathological changes in the brain cortex were observed from 24 h after VT-2 injection. In the cortex, extravasated erythrocytes, swelling of the endothelial cells (Fig. 2B, arrows), and edematous changes in cortical gray matter were observed at 48 h (Fig. 2, C-H). In the kidney, apparent changes were observed in proximal tubules and some glomeruli. A number of proximal tubules were observed to have sloughed epithelial cells with eosinophilic cytoplasm and contained eosinophilic fluid (Fig.  2, J and L, arrows), indicating cell loss due to necrosis as reported previously (17,18). Several glomeruli were associated with segmental endocapillary congestion without fibrin and platelet thrombi and diminished numbers of endothelial cells (Fig. 2, N, arrows, and P). These changes were observed 24 h after VT-2 injection, increasing in extent to almost all glomeruli at 48 h. In the liver, microsteatosis of hepatocytes was observed 48 h after injection (data not shown). No morphological changes were observed in the ␣1,4Gal-T Ϫ/Ϫ mice.
Gb3 Expression in Damaged Tissue-To investigate the involvement of Gb3 in tissue damage, we used immunohistochemistry to analyze

Gb3/CD77 Knockout Resulted in Verotoxin Resistance
Gb3 expression in the tissues of wild-type mice in which damage occurred. In the brain, Gb3 was expressed on the endothelial cells in the cortex (Fig. 3, B, D, and E, arrows) and the pia mater (Fig. 3C, arrowheads), with the intensity of antibody staining increasing slightly after VT-2 treatment (Fig. 3H). These staining patterns were very similar to that observed for CD31 (Fig. 3, F and G, white arrowheads), a marker of endothelial cells, indicating that Gb3 was expressed in some of the endothelial cells of the cortex. In the kidney, Gb3 was strongly expressed in the basilar regions of the proximal tubules (Fig. 3, J and L, arrowheads), with the intensity of antibody staining being slightly enhanced   APRIL 14, 2006 • VOLUME 281 • NUMBER 15

Gb3/CD77 Knockout Resulted in Verotoxin Resistance
after VT-2 treatment (Fig. 3N). However, no such expression was observed in the glomerulus at all (Fig. 3, J and N, arrows). The regions that were positively stained in the tubules were considered to be capillaries, because they overlapped partly with the expression sites of CD31 (Fig. 3, K and M, arrowheads). Gb3 expression was undetectable in liver even after the VT-2 stimulus (data not shown).
The Effects of LPS-VT preparations contain high amounts of LPS. Because it has been reported that LPS induced Gb3 expression via the induction of inflammatory cytokines (19), we analyzed serum levels of TNF-␣ and IL-1␤ after VT-2 administration (Fig. 4, A and B). Both cytokines showed increased serum levels. Real-time PCR also showed up-regulation of ␣1,4Gal-T mRNA in various tissues (Fig. 4C) as well as an increased intensity in antibody staining. However, no apparent changes were observed in the survival times of mice administered 0.2 g of VT-2 (contaminated with LPS) or 0.2 g of purified-VT-2 (LPS-free) (Fig. 5).

DISCUSSION
As expected, targeted disruption of the Gb3 synthase gene resulted in complete loss of the mRNA, Gb3 synthase activity, and all glycolipids belonging to the globo-series in the tissues examined. These results indicated that the targeted gene is the only gene coding Gb3 synthase. This mutant mouse line appeared useful for functional analysis of globoseries glycolipids in vivo and also for the expression analysis of glycolipids that are synthesized through the synthesis of Gb3 in murine tissues and organs.
As shown in Table 1, there are no clear abnormalities in lymphocyte populations and subsets in spleen and thymus. Immunoglobulin levels were also equivalent between wild type and the null mutants. However, implication of Gb3/CD77 in the immune responses to various antigens remains to be investigated. Although brainiac gene is involved in the synthesis of glycolipids in Drosophila, the finding that null mutant mice showed no definite morphological or functional changes during their development was surprising (20), and it is likely that the remaining glycolipid structures might compensate the functions of globo-series glycolipids.
A number of studies have indicated that Gb3 and its derivatives may be involved in various biological events such as clonal selection and differentiation of B lymphocytes, apoptosis, bacterial adhesion, viral invasion, and many other physiological processes. Therefore, we expect that these mutant mice possess potential defects particularly in the humoral immunity, although they are not overt at this moment. However, extensive studies and long-term observation are needed to elucidate the biological importance of Gb3 and its derivatives, and we are now further investigating these mutant mice for physiological defects.
Numerous studies showing that Gb3 is a receptor for VTs have been conducted using cultured cells, and it has been proposed that other carbohydrate-containing molecules such as N-glycans (4) or other alternative receptors (21) may also be able to receive VT signals. The findings of this study revealed that mice lacking functional Gb3 synthase gene  To analyze the effects of LPS in the crude VT-2 preparation, mice were treated with crude and purified VT-2 (LPS-free). Filled and open circles represent crude or purified (LPS-free) VT-2 (0.2 g) preparations, respectively. The crude preparations contained 10 g of LPS. n ϭ 12 (LPS-free VT-2) or 6 (others). Data were analyzed with the Student's unpaired t-test. No significant differences could be found in the results between crude and LPS-free VT-2 preparations. appear to be completely insensitive to VTs, with neither clinical signs nor pathological changes found in the null mutant mice. These results clearly indicated that Gb3/CD77 (including its derivatives) is the only structure that plays a role as a receptor for VTs. It was reported that the P1 antigen might also be a receptor for VTs (4). Recently, it was shown that the P1 antigen is also a product of Gb3/CD77 synthase (22) and consequently the null mutant mouse is expected to lack the P1 antigen. However, because no cells expressing only the P1 antigen (not Gb3/ CD77) are currently available, it seems very difficult to confirm that P1 acts as a receptor for VTs. It can therefore be concluded that only Gb3/ CD77 synthase products act as receptors for VTs.
As for gangliosides, it is well known that GT1b/GD1b are receptors for botulinum toxins and/or tetanus toxins (23), and this has been confirmed using complex ganglioside-lacking mice (24), which were relatively resistant to the injected toxins. However, the fact that the mice finally died suggested that coreceptors for individual toxins were present (23). Conversely, VT-injected null mutant mice survived for more than six months without exhibiting any disorders. Consequently, the finding that Gb3/CD77 behaved as the sole receptor for VTs was thus relatively unique.
Because mutant mice were observed to be perfect negative controls in immunohistochemistry, it was possible to distinguish minimal expression of Gb3/CD77 from the background and confirm that the main targets of VTs in tissue damage were Gb3/CD77 on endothelial cells in restricted regions (10,25). There have been numerous studies on the expression of Gb3/CD77 in human (16,(25)(26)(27)(28) and mouse (17,18) tissues, with some ambiguity observed in the precise localization thereof. In our study, definite expression of Gb3/CD77 was observed in the microvascular endothelial cells of the brain cortex and pia mater and the capillary vascular endothelial cells surrounding renal tubules and glomeruli, as well as lumenal side of the proximal tubules. Actually, pathological changes observed shortly after VT administration occurred in these regions, suggesting that initial VT target sites are endothelial cells expressing Gb3/CD77. Pathological changes detected at relatively later phases in the glomeruli and liver appeared to be because of circulation disturbances caused by primary disruption of the endothelial cells. Interestingly, damage in liver was observed immediately before death, suggesting that the liver damage was not a direct consequence of VT-2 but rather that it occurred secondary to systemic damage.
Implication of vascular endothelial cells in the brain damage under HUS has been also been considered (9,29). Colocalization of Gb3/CD77 and CD31 and accordance of their expression sites and pathological changes in brain cortex in VT-injected mice strongly supported the idea that endothelial cells are the primary targeting sites of VTs in the brain. These findings should explain the neurological signs and unconsciousness episodes observed in HUS.
Because it was reported that LPS and inflammatory cytokines induced Gb3/CD77 expression on some endothelial cells (19), it may be possible that they potentiate VT cytopathology (30). However, the effects of LPS on VT toxicity depend on when it is administered (31). In addition, the mouse strains used and tissue origins of endothelial cells are also critical for the assessment of LPS effects (10,31). In our results, the ␣1,4Gal-T gene might be induced by LPS via cytokine induction, but it scarcely affected the sensitivity to VTs. Specifically, the roles of LPS/ cytokines in inducing Gb3/CD77 expression may not be a major cause underlying tissue damage in vivo. Consequently, basal levels of Gb3/ CD77 expressed on vascular endothelial cells in non-treated mice should be sufficient to cause tissue damage and induce lethal toxicity. One explanation for the lack of any observable difference between crude and LPS-eliminated VT-2 preparations in our experiment was that the injected VT2 may have been cleared from the blood before sufficient Gb3/CD77 was induced by LPS.
The data obtained with the null mutant mice are useful as they clearly infer the presence of targets for protection against the toxic effects of VTs. Although apoptotic pathways mediated by Gb3/CD77 may be heterogenous (32), efforts to combat HUS should focus on inhibiting the interaction between VTs and Gb3/CD77.