Toxicity and Immunogenicity of a Verotoxin 1 Mutant with Reduced

INFECTION AND IMMUNITY,0019-9567/97/$04.0010

June 1997, p. 2019–2028 Vol. 65, No. 6

Copyright © 1997, American Society for Microbiology

Toxicity and Immunogenicity of a Verotoxin 1 Mutant with ReducedGlobotriaosylceramide Receptor Binding in Rabbits

DARRIN J. BAST,1,2 JAMES L. BRUNTON,1,2,3 MOHAMED A. KARMALI,2,4

AND SUSAN E. RICHARDSON2,4*

Samuel Lunenfeld Research Institute, Mount Sinai Hospital,1 Department of Microbiology, The Toronto Hospital,3

The Research Institute and the Division of Microbiology, Department of Paediatric Laboratory Medicine,The Hospital for Sick Children,4 and Department of Microbiology,

The University of Toronto,2 Toronto, Ontario, Canada

Received 22 November 1996/Returned for modification 31 January 1997/Accepted 10 March 1997

The verotoxins (VT1 and VT2), produced by strains of enterohemorrhagic Escherichia coli, have beenimplicated in the pathogenesis of hemorrhagic colitis and the hemolytic uremic syndrome. To better under-stand the role of globotriaosylceramide (Gb3) receptor binding by the verotoxins in disease production, weexamined the clinicopathologic effects of an intravenously (i.v.) administered verotoxin 1 mutant holotoxin(Phe30Ala) in rabbits. The substitution of alanine for phenylalanine 30 in the VT1 B subunit has been shownpreviously to reduce both Gb3 binding affinity and capacity in vitro. This reduction in receptor bindingcorresponded to a 105-fold reduction in the toxic activity of VT1 on a Vero cell monolayer. In this study, purified125I-labeled Phe30Ala was administered i.v. to rabbits to determine its specific distribution in rabbit tissues.In contrast to the rapid elimination of i.v. administered 125I-VT1 from the bloodstream, 125I-Phe30Ala had a52-fold-longer half-life in serum and failed to localize preferentially in the gastrointestinal tract and centralnervous system (CNS). Rabbits challenged with Phe30Ala at a dose equivalent to 10 times the 50% lethal dose(LD50) of VT1 showed no visible clinical symptoms typical of VT effect after 7 days. Administration of Phe30Alaat a dose equivalent to 100 times the LD50 of VT1, however, caused both clinical and histopathologic featuresindistinguishable from VT1 toxemia in rabbits, although the onset of symptoms was delayed. Rabbits wereimmunized with Phe30Ala and challenged i.v. with either 125I-VT1 or 125I-VT2. The specific uptake of 125I-VT1in the gastrointestinal tract and CNS was totally inhibited in Phe30Ala immune rabbits. Only a partialdecrease in target organ uptake was observed in Phe30Ala immune rabbits challenged with 125I-VT2. From thisstudy, we conclude that Gb3 binding is responsible for target organ localization of VT1 and disease productionin the rabbit. The ability of Phe30Ala to induce both strong antibody and protective responses against VT1suggests that VT mutants with reduced receptor binding properties may be useful in vaccine strategies. Afurther reduction in the toxicity of Phe30Ala would be required for its use as a natural toxoid to protect againsthuman verotoxigenic E. coli infections.

Verotoxigenic Escherichia coli (VTEC) are associated withsporadic cases and outbreaks of both hemorrhagic colitis (14,22) and the hemolytic-uremic syndrome (8, 9). The virulenceof these strains is determined to a large extent by the produc-tion of potent cytotoxins termed Shiga-like toxins or verotoxins(VTs) (27). The verotoxins (VT1, VT2, and VT2c) are A:B5subunit toxins that are structurally related yet antigenicallydistinct (4, 15, 25). The B subunit (35 kDa) is a pentamer (25),responsible for the binding of the toxin to its eukaryotic glo-boseries glycosphingolipid receptor globotriaosylceramide(Gb3) (11, 31). The enzymatically active A subunit (32 kDa)inhibits eukaryotic protein synthesis by catalytically inactivat-ing the 60S ribosomal subunit (3).

Several animal models, including the mouse (32), pig (13),dog (7), and rabbit (20, 24, 33), have been used to study theclinicopathologic effects of the verotoxins. Following intrave-nously (i.v.) administered injections of purified VT1, the rabbithas been shown to develop nonbloody diarrhea and progres-sive flaccid paralysis (20, 33). These clinical features are hy-pothesized to be a result of binding of the toxin to Gb3 recep-tors in the microvasculature of the gastrointestinal tract and

central nervous system (CNS), with subsequent endothelial cellinjury and microvascular thrombosis in the affected organs.These features are preventable by inducing immunity to VT1(20). The vascular lesions observed in the rabbit following i.v.VT1 administration are similar to those observed in the kid-neys and other organs of hemolytic-uremic syndrome patients(20, 21). Based on these studies, the pathological features ofHUS are thought to be the result of toxin-mediated damage tothe vascular endothelial cells of the renal glomeruli and othermicrovasculature via specific Gb3 receptor binding. Thus, therabbit appears to be a suitable model for studying the interac-tion of the VTs with endothelial cells in target blood vesselsand for examining the efficacy of toxin-derived vaccines.

We have recently demonstrated that the substitution of ala-nine for phenylalanine 30 in the VT1 B subunit (Phe30Ala)resulted in a significant reduction in toxin binding to Gb3 (2).Furthermore, the cytotoxicity of Phe30Ala was reduced by105-fold compared with that of the wild-type toxin when ex-amined on Vero cell monolayers (2). The Phe30Ala B subunitwas crystallized, and its crystal structure was found to be iden-tical to that of the wild-type B subunit with the exception of thephenyl ring at position 30 (2). We concluded that the dramaticreduction in in vitro cytotoxic activity by Phe30Ala was a resultof the reduction in Gb3 binding, due to the loss of the phenylring at position 30 and not to the disruption of the overallsecondary structure of the molecule.

* Corresponding author. Mailing address: Division of Microbiology,Department of Paediatric Laboratory Medicine, The Hospital for SickChildren, 555 University Ave., Toronto, Ontario, Canada M5G 1X8.Phone: (416) 813-5992. Fax: (416) 813-5993.

2019

on February 11, 2018 by guest

http://iai.asm.org/

Dow

nloaded from

http://iai.asm.org/

To better understand the role of Gb3 receptor binding by theverotoxins in disease production, we examined the in vivobiological effects of Phe30Ala by comparing the results of thesystemic administration of Phe30Ala and wild-type VT1 inrabbits. We hypothesized that Phe30Ala might be useful as anatural toxoid to protect against VT-mediated disease due toits retention of the secondary structure of the wild-type VT1 Bsubunit and its ability to associate with the A subunit. Wetherefore examined its efficacy and toxicity as a natural toxoidin rabbits.

(Presented in part at the 96th Annual Meeting of the Amer-ican Society for Microbiology,-New Orleans, La., May 1996.)

MATERIALS AND METHODS

Bacterial strains. Plasmid pJLB128 encodes the wild-type VT1 holotoxin geneand was expressed in E. coli JM101 (19, 29). Plasmid pJLB128(F30A) encodesthe Phe30Ala holotoxin gene and was expressed in E. coli JM101 (2).

Purification of VT1 and Phe30Ala holotoxins. By using polymyxin B (19), VT1and Phe30Ala holotoxins were extracted from the periplasm of the overproduc-ing E. coli strains described above. The holotoxins were purified by chromatog-raphy on hydroxylapatite (Bio-Rad, Richmond, Calif.), chromatofocusing (Phar-macia, Uppsala, Sweden), and affinity chromatography on a Cibachron bluecolumn (Bio-Rad) as previously described (17). The purity of the protein prep-arations was determined by reducing sodium dodecyl sulfate-polyacrylamide gelelectrophoresis with 16% Tricene gels. The protein concentrations in purifiedsamples were determined by the Pierce bicinchoninic acid protein assay (Pierce,Rockford, Illinois). All toxin preparations used in animal injections tested neg-ative in the Limulus amoebocyte lysate assay (E- toxate; Sigma Chemical Co., St.Louis, Mo.) at a detection level of 0.05 to 0.1 endotoxin unit/ml.

Iodination of VT1 and Phe30Ala holotoxins. VT1, Phe30Ala, and VT2 holo-toxins were radiolabelled with Na125I (Amersham)- and Iodogen (Pierce)-coatedtubes as previously described (5). The radiolabelled toxins were tested for Gb3binding by the microtitre plate assay (5) to ensure that labeling procedure hadnot altered the previously observed in vitro receptor binding pattern of all threetoxins.

Intravenous administration of radiolabelled toxins. All procedures performedon animals were conducted in accordance with the principles put forth by theAnimal Care Committee of the Hospital for Sick Children, Toronto, Canada.Saline suspensions of 125I-VT1 and 125I-Phe30Ala, at doses equivalent to 10times the previously reported 50% lethal dose (LD50) of VT1 (4 mg) (20), wereadministered intravenously (i.v.) via the marginal ear vein to three and five NewZealand White male rabbits (2 kg each) (Reimens Fur Ranches), respectively.Samples of blood (1 ml) were taken from the artery of the ear at defined intervalsimmediately following injection. The total amount of circulating radiolabelledtoxin in the blood was quantitated by monitoring for radioactivity on a Beckmangamma counter and multiplying the radioactivity per milliliter by the estimatedtotal blood volume (55 ml/kg of body weight). The half-lives of both 125I-VT1 and125I-Phe30Ala in serum were determined. At 2 h postchallenge, the level ofradioactivity remaining in the bloodstream of 125I-VT1-injected rabbits was 2%of that initially administered. These rabbits were then sacrificed by a lethalinjection of 0.5 ml of pentobarbital (Euthanyl [240 mg/ml]; MTC Pharmaceuti-cals, Cambridge, Ontario, Canada) per kg. The five rabbits given injections of125I-Phe30Ala were sacrificed 10 h after the initial injection to allow the radio-activity in the bloodstream to reach 2% of the initial amount injected.

Following sacrifice, an autopsy was performed and individual organs wereremoved and analyzed for radioactivity. Samples (1 to 3 g) of the followingtissues and organs were obtained: colon, cecum, small intestine, stomach, liver,skeletal muscle, heart, lung, kidney, spleen, spinal cord, and brain.

Clinicopathologic studies with Phe30Ala challenged rabbits. Five male NewZealand White rabbits (2.0 kg each) were challenged by i.v. injections ofPhe30Ala holotoxin at a dose of 4 mg/rabbit (equivalent to 103 the LD50 forVT1). Three control rabbits were challenged i.v. with wild-type VT1 at the samedose. All injections were administered via the marginal ear vein.

The rabbits were observed twice daily for evidence of the following symptomsand any other changes not observed in the control animals: ruffled fur, lethargy,decreased water and food consumption, mild to severe watery diarrhea with orwithout blood or mucus, and forelimb and/or hindlimb paralysis. The symptomslisted are those previously observed in a similar study conducted by Richardsonet al. (20).

Rabbits showing visible signs of severe clinical disease (diarrhea and/or limbparalysis) were sacrificed immediately by i.v. injection of 0.5 ml of pentobarbitalper kg. Rabbits showing no signs of clinical disease were sacrificed by lethalinjection of pentobarbitol 7 days postinjection. At autopsy the organs wereexamined for gross pathological abnormalities, and representative tissue speci-mens were taken from the brain, spinal cord, and cecum, as well as any otherobviously involved organs. Tissues for histological examination were preserved in10% buffered formalin immediately after removal. The same experiment was

then performed at doses equivalent to 100 times the previously reported LD50 ofVT1 (40 mg/rabbit).

Histopathologic analysis of tissue specimens. Formalin-preserved tissue spec-imens were embedded in paraffin and sectioned into 8- to 10-mm-thick sections.All sections were stained with hematoxylin and eosin and examined by lightmicroscopy. Each tissue section was examined by one observer who was unawareof the treatment status and clinical details. Abnormalities were scored on thebasis of the presence or absence of edema, hemorrhage, infarction, and throm-botic vasculopathy on a scale of 0 to 31, representing absent, mild, moderate,and severe involvement, respectively, as previously described (20). Scores foreach of six target tissues (cecum, anterior brain, midbrain, cerebellum, cervicalspinal cord, and lumbar spinal cord) were determined and a mean score per testanimal calculated.

Immunization of rabbits with Phe30Ala and challenge with 125I-VT1 and125I-VT2. Purified Phe30Ala was used to immunize six specific-pathogen-freeNew Zealand White rabbits. The rabbits were initially immunized with 4 mg ofPhe30Ala dissolved in 100 ml of phosphate-buffered saline and emulsified with anequal volume of Freund’s complete adjuvant (Sigma). On days 10 and 20 fol-lowing the initial immunization, the rabbits were boosted with an additional 4 mgof Phe30Ala dissolved in 100 ml of phosphate-buffered saline and emulsified withan equal volume of Freund’s incomplete adjuvant (Sigma). All immunizationswere administered subcutaneously on the back of the animal. Sera were collectedby a marginal ear vein bleed 14 days following the last immunization. Threerabbits were then challenged with 125I-VT1, and three were challenged with125I-VT2, all at doses equivalent to 10 times the LD50 of VT1.

ELISA and neutralization assays. Anti-VT1, anti-Phe30Ala, and anti-VT2immunoglobulin G antibodies were assayed by a standard antibody enzyme-linked immunosorbent assay (ELISA) optimized for use with VT1, Phe30Ala,and VT2 holotoxins as the coated test antigens (1a). The ability of anti-Phe30Alaantisera to inhibit the cytotoxic effects of both VT1 and VT2 toward Vero cellswas determined by an in vitro neutralization assay (9). Serial twofold dilutions ofantisera were preincubated with 0.003 ng of purified VT1 or VT2, which is 10times the measured 50% cytotoxic dose. The serum neutralization titer wasdefined as the highest serial twofold dilution of serum at which 50% of the cellswere killed by the effects of the toxin. Antibody titers were converted to loga-rithmic values [log2 (x) and log5 (x) for neutralizing and ELISA titers respec-tively, where x is the reciprocal of the serum dilution).

Statistical analysis. Standard errors of the mean were determined for allvalues in the figures. The one-tailed Student t test was used to determine signif-icance. Statistical analyses were done with the functions contained within thePrimer of Biostatistics (4a).

RESULTS

Elimination of 125I-VT1 and 125I-Phe30Ala from the blood-stream of immunologically naive rabbits. i.v. administered125I-VT1 was rapidly eliminated from the bloodstream of im-munologically naive rabbits (Fig. 1). A mean half-life in serumof approximately 2 min was determined. 125I-Phe30Ala waseliminated much more slowly from the bloodstream, with a

FIG. 1. Levels of 125I-VT1 (■) and 125I-Phe30Ala (E) in serum of immuno-logically naive New Zealand White rabbits. Each value is the mean value atindividual time intervals for 125I-VT1 (n 5 3) and 125I-Phe30Ala (n 5 5) rep-resented as the percentage of total radioactivity administered. Error bars repre-sent standard errors of the means.

2020 BAST ET AL. INFECT. IMMUN.


http://iai.asm.org/

Dow

nloaded from

http://iai.asm.org/

half-life in serum of 104 min. At 2 h postinjection, the amountsof 125I-VT1 and 125I-Phe30Ala remaining in the bloodstreamwere 2 and 48%, respectively, of the initial amounts adminis-tered (Fig. 1). All rabbits were sacrificed for localization stud-ies when the level of radioactivity in the bloodstream wasapproximately 2% of the amount initially injected, so that theresidual blood remaining in tissues would not interfere with theassessment of tissue localization, particularly for highly vascu-larized tissues. Rabbits that were given 125I-VT1 were sacri-ficed at 2 h postinjection. The rabbits given 125I-Phe30Ala werenot sacrificed until 10 h postinjection due to the much slowerdecay of the mutant.

Tissue localization of 125I-VT1 and 125I-Phe30Ala in immu-nologically naive rabbits. The tissue localization of 125I-VT1administered to immunologically naive New Zealand Whiterabbits was highly organ specific (Fig. 2a). Maximal accumu-

lation of 125I-VT1 (counts per minute [cpm] per milligram oftissue per total cpm injected) was found in the cecum, colon,small intestine, stomach, spinal cord, and brain (listed in de-scending order of cpm per milligram per total cpm). Otherorgans examined (spleen, lung, heart, liver, kidney, and skele-tal muscle) showed negligible 125I-VT1 accumulation. In con-trast, those rabbits given injections of 125I-Phe30Ala had mark-edly reduced accumulation in target organs. In particular,there was no sequestration of toxin in the gastrointestinal tractand CNS at 10 h postinjection (Fig. 2a). The difference be-tween the means of the uptake of radioactivity per milligram oftissue in 125I-VT1 and 125I-Phe30Ala was statistically differentfor the cecum, colon, small intestine, spinal cord, and brain(P , 0.01).

Tissue localization results for both 125I-VT1- and 125I-Phe30Ala-treated rabbits were corrected for differences in or-gan blood flow by expressing the data in cpm per 100 g of tissueper specific organ blood flow (milliliters per minute per 100 g)(20) per total toxin dose (cpm) (Fig. 2b). When the resultswere corrected for blood flow, the colon now had the largestaccumulation of 125I-VT1 followed by the cecum, spinal cord,small intestine, stomach, and brain. Again, the accumulation of125I-Phe30Ala was significantly lower than that of 125I-VT1 inthe colon, cecum, spinal cord, small intestine, and brain(P,0.01).

Challenge with VT1 and Phe30Ala at doses equivalent to 10times the LD50 of VT1. The three control animals receivingVT1 holotoxin at a dose of 10 times the previously reportedLD50 of wild-type VT1 (20) exhibited the typical clinical fea-tures of VT1 intoxication at a mean of 17 h postchallenge;these included general malaise followed by severe paralysisand/or severe diarrhea. Histopathologic analysis showed allthree rabbits to have evidence of VT effect, with mild to severegastrointestinal tract and CNS involvement which was indistin-guishable from that observed in previous studies with VT1(20). In the gastrointestinal tract, mucosal and submucosaledema, hemorrhage, and microvascular angiopathy were ob-served (Fig. 3B). CNS lesions were characterized by focal hem-orrhage and infarction, with thrombotic vasculopathy mostprominent in the grey matter of the spinal cord (Fig. 4B).Lesions were similar but more focal and less severe in the brainsections and more prominent in the cerebellum compared toanterior brain sections.

In contrast, the five rabbits challenged with Phe30Alashowed no clinical signs consistent with VT intoxication over aperiod of 7 days of observation. The gross pathologic findingswere also normal. However, mild to moderate histopathologicabnormalities consistent with VT effect were observed in tworabbits, in the gastrointestinal tract (Fig. 3C) and CNS in onerabbit and in the CNS only (Fig. 4C) in the other. The otherthree rabbits showed minimal yet equivocal changes which mayhave been early VT effect or artifacts. The mean histopatho-logic score of the Phe30Ala-treated rabbits was significantlylower than that of the rabbits treated with VT1 (P , 0.05).

Challenge with VT1 and Phe30Ala at doses equivalent to 100times the LD50 of VT1. Clinical symptoms typical of VT effectwere observed in all three rabbits challenged with Phe30Ala ata dose equivalent to 100 times the LD50 of VT1; one hadsevere CNS symptoms, and two had moderate diarrhea with-out CNS symptoms. Typical symptoms also occurred in thewild-type-VT1-challenged control, although they occurred 7 hpostchallenge as compared to 40 h in the Phe30Ala-challengedrabbits. At autopsy, mild cecal edema was evident in all threerabbits, although the cecal contents were adherent to the cecalmucosa as in unchallenged animals. Histopathologic testingshowed that of the two rabbits with diarrhea, one had moder-

FIG. 2. (a) Tissue localization of 125I-VT1 at 2 h postinjection (n 5 3) (solidbars) and 125I-Phe30Ala at 10 h postinjection (n 5 5) (hatched bars) in immu-nologically naive rabbits. Values are expressed as a fraction of the total radio-activity administered. The values were significantly lower only in the colon,cecum, spinal cord, small intestine, and brain for animals challenged withPhe30Ala as compared with those challenged with wild-type VT1 toxin (P ,0.01). (b) Values are expressed as means of the total radioactivity administeredand adjusted to account for the variability of blood flow according to organ site(21) (milliliters per minute per 100 g of body weight). P , 0.01 for all valuesobtained with Phe30Ala-challenged versus VT1-challenged rabbits except forskeletal muscle, lung, liver, heart, kidney, spleen, and stomach. Error bars rep-resent the standard errors of the means. Abbreviations: sm. int., small intestine;sp. cord, spinal cord; sk. mus., skeletal muscle.

VOL. 65, 1997 IN VIVO TOXICITY OF A VEROTOXIN 1 MUTANT IN THE RABBIT 2021


http://iai.asm.org/

Dow

nloaded from

http://iai.asm.org/

FIG. 3. Hematoxylin-and-eosin-stained sections of cecum from an untreated rabbit (A), a rabbit given an i.v. injection of 103 the LD50 of VT1 (B), a rabbit givenan i.v. injection of Phe30Ala at a dose equivalent to 103 the LD50 of VT1 (C), and a rabbit given an i.v. injection of Phe30Ala at a dose equivalent to 1003 the LD50of VT1. (A) Normal mucosal (c) and submucosal (d) histologic findings. (B and D) Severe mucosal and submucosal edema (e) and hemorrhage (f). (C) Mild tomoderate mucosal (h) and submucosal (g) edema. Magnificatioon, 3116.

2022


http://iai.asm.org/

Dow

nloaded from

http://iai.asm.org/

FIG. 3—Continued.



http://iai.asm.org/

Dow

nloaded from

http://iai.asm.org/

FIG. 4. Hematoxylin-and-eosin-stained sections of cervical spinal cord from an untreated rabbit (A), a rabbit given an i.v. injection of 103 the LD50 of VT1 (B),a rabbit given an i.v. injection of Phe30Ala at a dose equivalent to 103 the LD50 of VT1 (C), and a rabbit given an i.v. injection of Phe30Ala at a dose equivalent to1003 the LD50 of VT1. (A) Normal tissue histologic findings. (B and D) Severe hemorrhage (c). (C) Focal petechial hemorrhage (g). (D) Areas of extensive infarction(edema [d] and loss of neural tissue), with signs of motor neuron necrosis (e) and thrombosed blood vessels (f). Magnification, 3116 (A) and 3279 (B to D).

2024


http://iai.asm.org/

Dow

nloaded from

http://iai.asm.org/

FIG. 4—Continued.



http://iai.asm.org/

Dow

nloaded from

http://iai.asm.org/

ate histopathologic changes (Fig. 3D) while the other had mildor equivocal changes, as did the animal without diarrhea whichhad only CNS symptoms. CNS histopathologic testing showedsevere involvement (Fig. 4D) in the affected rabbit and absentto mild involvement in the other two rabbits. The findings weretypical of those observed in animals with VT1 intoxication inboth the gastrointestinal tract and the CNS. The histopatho-logic score of the rabbits challenged with Phe30Ala at a doseequivalent to 100 times the LD50 of VT1 was not significantlydifferent from those challenged with wild-type VT1 at a dose of10 times the LD50 (P 5 0.498).

Tissue distribution of 125I-VT1 and 125I-VT2 in rabbits im-munized with Phe30Ala. In contrast to the organ-specific lo-calization of wild-type 125I-VT1 in the gastrointestinal tractand CNS, its distribution in rabbits immunized againstPhe30Ala was significantly different (Fig. 5). The liver andspleen showed the greatest accumulation of 125I-VT1, withnegligible amounts measured in the gastrointestinal tract andCNS. The difference between the means of the uptake ofradioactivity per milligram of tissue in immunologically naiveand Phe30Ala immune rabbits was statistically significant forall tissues except the skeletal muscle, lung, heart, and kidney(P , 0.01). To examine the ability of Phe30Ala-raised anti-bodies to prevent specific tissue binding by the heterologousverotoxin VT2, both Phe30Ala immunized and immunologi-cally naive rabbits were given i.v. injections of 125I-VT2. 125I-VT2 localized predominantly in the cecum, colon, and smallintestine of rabbits not immunized with Phe30Ala (Fig. 6).Phe30Ala immune rabbits showed a partial reduction in thespecific distribution of 125I-VT2 in the cecum, colon, and smallintestine. The difference between the means of the uptake ofradioactivity per milligram of tissue in immunologically naiveand Phe30Ala immune rabbits was statistically significant foronly the colon and small intestine (P , 0.05) (Fig. 6). Further-more, a significant increase in the amount of 125I-VT2 wasnoted in the liver (P , 0.05) but not the spleen of rabbitsimmunized with Phe30Ala.

Immunogenicity of the Phe30Ala holotoxin. Sera collectedfrom Phe30Ala-immune rabbits immediately before the ad-ministration of radiolabelled toxins were subjected to titer de-termination for both cross-reactive and neutralizing antibodiesto wild-type VT1, Phe30Ala, and VT2 holotoxins. When ana-

lyzed by ELISA, the reciprocal geometric mean titers of specificantibody to VT1, Phe30Ala, and VT2 were 8.15 (1:500,000), 8.15(1:500,000), and 5.72 (1:10,000), respectively. Phe30Ala-raisedantiserum also neutralized the toxic effects of VT1 in the invitro Vero cell assay but was not able to neutralize the effectsof VT2. The reciprocal geometric mean titer of neutralizingantibodies determined against 10 times the 50% cytotoxic doseof VT1 was 11.30 1 0.35 (1:2519) (n 5 6).

DISCUSSION

The specific localization of i.v. administered VT1 to thegastrointestinal tract and CNS of the rabbit (20) correlateswith the presence of Gb3 in these tissues (33). Furthermore,the localization of VT1 to endothelial cells of the gastrointes-tinal tract and CNS microvasculature correlates with the his-tologic changes observed in these tissues (20). Based on theseobservations and the assumption that Gb3 is present on micro-vascular endothelial cells of both tissues, it is hypothesized thatthe specific binding of VT1 to Gb3 on vascular endothelium ofthe gastrointestinal tract and CNS of the rabbit leads to thevasculopathy observed in these tissues following i.v. VT1 chal-lenge. To our knowledge, this study is the first direct in vivodemonstration that Gb3 binding is responsible for target organlocalization of VT1 and disease production in the rabbit.

A previously characterized VT1 B-subunit mutant (Phe30Ala)(2) was used in this study to better demonstrate the significance ofGb3 receptor binding in disease production in the rabbit. Thesubstitution of Ala for Phe30 in the VT1 B subunit caused a4-fold reduction in Gb3 binding affinity and a 10-fold reductionin the number of receptors recognized (2). This is consistentwith the 52-fold-longer intravascular half-life and the signifi-cant reduction in target organ accumulation by 125I-Phe30Alacompared to 125I-VT1 in the rabbit in this study. This resultstrongly supports the hypothesis that Gb3 receptor recognitionand binding affinity play an important role in the intravascularclearance of VT1 and its accumulation in specific organs in theimmunologically naive rabbit (20, 33).

Although the accumulation in the target organs of 125I-Phe30Ala (at a dose equivalent to 103 the LD50 of VT1) wassignificantly reduced compared to that of wild-type 125IVT1,mild VT-related histologic changes in the target organs were

FIG. 5. Localization of 125I-VT1 in tissues of naive (n 5 5) (solid bars) andPhe30Ala-immunized (n 5 3) (hatched bars) rabbits. P , 0.01 for all values fornaive versus immunized rabbits except for data on the skeletal muscle, lung,heart, and kidney. Error bars represent the standard errors of the means. Rabbitswere challenged with 125I-VT1 at a dose equivalent to 103 the LD50 of VT1.

FIG. 6. Localization of 125I-VT2 in tissues of naive (n 5 3) (solid bars) andPhe30Ala-immunized (n 5 3) (hatched bars) rabbits. P , 0.05 for all values fornaive versus immunized rabbits for data on the colon, small intestine, and liveronly. Error bars represent the standard errors of the means. Rabbits werechallenged with 125I-VT2 at a dose equivalent to 103 the LD50 of VT1.



http://iai.asm.org/

Dow

nloaded from

http://iai.asm.org/

observed in rabbits challenged with nonradiolabelled Phe30Ala atequivalent doses. However, the challenge of rabbits withPhe30Ala at 10-fold-higher doses (equivalent to 1003 the LD50of VT1) resulted in both clinical and pathologic effects indis-tinguishable from those in VT1-treated rabbits, although theonset of symptoms was delayed. This suggests that while ahighly significant reduction in 125I-Phe30Ala accumulation intarget tissues and lack of clinical symptomatology was demon-strated at 103 the LD50 of VT1, at the higher dose (1003 theLD50 of VT1) there was still enough residual binding to exerta significant biological effect.

The apparent 10-fold reduction in the biological effects inrabbits of Phe30Ala compared to wild-type VT1 is significantlyless than the 105-fold reduction in in vitro cytotoxicity on aVero cell monolayer (2). This most probably reflects the dif-ferences between the effect of the toxin on a cell monolayerand in a complex biological organ system in vivo. Several ex-planations may account for this discrepancy. It is conceivablethat the residual binding by Phe30Ala induces host cytokineproduction. This in turn may sensitize the vascular endothe-lium to the effects of the toxin (26) by increasing the synthesisand membrane expression of Gb3, as has been previously dem-onstrated with wild-type VT1 in vitro (12, 18, 28, 30). In recentstudies, a correlation between the cytotoxic ability of Phe30Alaand the total Gb3 concentration in various sensitive cell lineshas been demonstrated. As the total Gb3 content increases, so,too, does the ability of Phe30Ala to cause cell death (1). Sec-ond, Phe30Ala may bind preferentially to fatty acyl isotypes ofGb3 which are not present in Vero cells but are present in theendothelium of the target microvascular beds of the gastroin-testinal tract and CNS of the rabbit. Modifications in the fattyacid composition of the ceramide moiety of Gb3 may alter theconformation or surface accessibility of the carbohydrate moi-ety necessary for toxin binding (10, 16, 23). Binding ofPhe30Ala to different Gb3 isoforms may result in a distinctmechanism of endocytosis, intracellular routing, and translo-cation of the toxin necessary for cytotoxic action. Further stud-ies are required to confirm both these hypotheses.

It has been shown that immunity to either VT1 or VT2prevents the accumulation of both toxins in the gastrointestinaltract and CNS of rabbits (1b) despite the lack of cross-reactiv-ity of anti-VT1 antisera against the cytotoxic effects of VT2 inthe in vitro neutralization assay (1b, 6). Thus, we proceeded toexamine the efficacy of Phe30Ala as a natural toxoid in pro-tecting against challenges with VT1 and VT2 in the rabbitmodel. The inhibition of accumulation of 125I-VT1 in the gas-trointestinal tract and CNS of Phe30Ala-immunized rabbits isconsistent with the strong anti-VT1 IgG antibody responseelicited in these animals, as measured by both ELISA and theVT1 neutralization assay. While no anti-VT2 in vitro neutral-izing ability was detected, a partial inhibition of 125I-VT2 ac-cumulation in the gastrointestinal tract was demonstrated,which correlated with reduced anti-VT2 antibody ELISA titers(compared to anti-VT1 antibody ELISA titers). This may bedue to the elimination of an important epitope in Phe30Alathat is necessary for VT2 clearance in vivo. While our studydemonstrates that immunity to Phe30Ala reduces the accumu-lation of toxin in target tissues following an i.v. VT challenge,it is important to note that it does not measure biologicalprotection. It is possible, though, that in a natural infectionstate where the toxin challenge would be much lower, immu-nity to Phe30Ala provides full protection against both VT1 andVT2 mediated disease.

This study provides the first direct in vivo observation thatGb3 recognition and binding by VT1 are important for intra-vascular toxin clearance, organ targeting, and disease produc-

tion in the rabbit. Furthermore, the results of this study dem-onstrate that VT mutants with reduced receptor binding, suchas Phe30Ala, may provide sufficient immunity to protectagainst VT-associated diseases. However, despite its markedlyreduced in vitro cytotoxicity, in vitro Gb3 binding ability, and invivo target organ accumulation, Phe30Ala is too toxic to beuseful as a natural toxoid, especially for humans. The construc-tion of VT1 mutant derivatives with additional B-subunit mu-tations may demonstrate acceptable toxicity profiles for use asnatural human toxoids.

ACKNOWLEDGMENTS

This work was supported by grants PG1123 and MT13071 from theMedical Research Council of Canada. D.J.B. is the recipient of ascholarship from the Searle Pharmaceutical Company and a scholar-ship from the Ontario Graduate Scholarship Program of the Ministryof Education and Training of Ontario.

We thank Bonnie Welsh and Gerry Kent for their assistance andhelpful advice. We also thank Cliff Lingwood for supplying purifiedVT2.

REFERENCES

1. Bast, D. Unpublished data.1a.Bast, D. J., et al. Unpublished data.1b.Bielaszewska, M., M. A. Karmali, and M. Petric. 1994. Localization of

verocytotoxin (VT2) and antigenic cross-reactivity of VT1 and VT2 in therabbit model, p. 249–252. In M. A. Karmali and A. G. Goglio (ed.), Recentadvances in Verocytotoxin-producing Escherichia coli infections. ElsevierBiomedical Press, Amsterdam, The Netherlands.

2. Clarke, C., D. J. Bast, A. Sharp, P. M. St. Hilaire, R. Agha, P. E. Stein, E. J.Toone, R. J. Read, and J. L. Brunton. 1995. Phenylalanine 30 plays animportant role in receptor binding of verotoxin-1. Mol. Microbiol. 19:891–899.

3. Endo, Y., K. Tsurgi, T. Yutsudo, Y. Takeda, T. Ogasawara, and K. Igarashi.1988. Site of action of a Vero toxin (VT2) from Escherichia coli O157:H7 andof Shiga toxin on eukaryotic ribosomes. Eur. J. Biochem. 171:45–50.

4. Frazer, M., M. Chernai, Y. Kozlov, and M. James. 1994. Crystal structure ofthe holotoxin from Shigella dysenteriae at 2.5 Å resolution. Nat. Struct. Biol.1:59–64.

4a.Glantz, S. A. 1992. Primer of biostatistics, 3rd ed. McGraw-Hill, Inc., NewYork, N.Y.

5. Head, S. C., M. A. Karmali, and C. A. Lingwood. 1991. Preparation of VT1and VT2 hybrid toxins form their purified dissociated subunits. Evidence forB subunit modulation of A subunit function. J. Biol. Chem. 266:3617–3621.

6. Head, S. C., M. A. Karmali, M. E. Roscoe, M. Petric, N. A. Strockbine, andI. K. Wachsmuth. 1988. Serological differences between verocytotoxin 2 andShiga-like toxin II. Lancet ii:751.

7. Hertzke, D., L. Cowan, P. Schoning, and B. Fenwick. 1995. Glomerularultrastructural lesions of idiopathic cutaneous and renal glomerular vascu-lopathy of greyhounds. Vet. Pathol. 32:451–459.

8. Hofmann, S. L. 1993. Southwestern International Medicine Conference.Shiga-like toxins in hemolytic uremic syndrome and thrombotic thrombocy-topenic purpura. Am. J. Med. Sci. 306:398–406.

9. Karmali, M. A., M. Petric, C. Lim, R. Cheung, and G. Arbus. 1985. Theassociation between idiopathic hemolytic uremic syndrome and infection byverotoxin-producing Escherichia coli. J. Infect. Dis. 151:775–782.

10. Kiarash, A., B. Boyd, and C. Lingwood. 1994. Glycosphingolipid receptorfunction is modified by fatty acid content. Verotoxin 1 and Verotoxin 2cpreferentially recognize different globotriaosyl ceramide fatty acid homo-logues. J. Biol. Chem. 269:11138–11146.

11. Lingwood, C., H. Law, S. Richardson, M. Petric, J. L. Brunton, S. DeGran-dis, and M. A. Karmali. 1987. Glycolipid binding of purified and recombi-nant Escherichia coli produced verotoxin in vitro. J. Biol. Chem. 262:8834–8839.

12. Louise, C. B., and T. G. Ogrig. 1991. Shiga toxin-associated hemolytic-uremic syndrome: combined cytotoxic effects of Shiga toxin, interleukin-1b,and tumor necrosis factor alpha on human vascular endothelial cells in vitro.Infect. Immun. 59:4173–4179.

13. MacLeod, D. L., and C. L. Gyles. 1991. Immunization of pigs with purifiedShiga-like toxin II variant toxoid. Vet. Microbiol. 29:309–318.

14. O’Brien, A. D., V. Tesh, A. Donohue-Rolfe, M. Jackson, S. Olsnes, K. Sand-vig, A. Lindberg, and G. T. Keusch. 1992. Shiga toxin: biochemistry, genetics,and mode of action and role in pathogenesis. Curr. Top. Microbiol. Immu-nol. 180:65–94.

15. Olsnes, S., R. Reisbig, and K. Eiklid. 1981. Subunit structure of Shigellacytotoxin. J. Biol. Chem. 256:8732–8738.

16. Pellizzari, A., H. Pang, and C. Lingwood. 1992. Binding of verocytotoxin 1 to



http://iai.asm.org/

Dow

nloaded from

http://iai.asm.org/

its receptor is influenced by differences in receptor fatty acid content. Bio-chemistry 31:1363–1370.

17. Petric, M., M. A. Karmali, S. E. Richardson, and R. Cheung. 1987. Purifi-cation and biological properties of Escherichia coli verocytoxin. FEMS Mi-crobiol. Lett. 41:63–68.

18. Ramegowda, B., and V. L. Tesh. 1996. Differentiation-associated toxin re-ceptor modulation, cytokine production, and sensitivity to Shiga-like toxinsin human monocytes and monocytic cell lines. Infect. Immun. 64:1173–1180.

19. Ramotar, K., B. Boyd, G. Tyrrell, J. Gariepy, C. Lingwood, and J. L. Brun-ton. 1990. Characterization of Shiga-like toxin 1B purified from overproduc-ing clones of the SLT-1 B cistron. Biochem. J. 272:805–811.

20. Richardson, S., T. Rotman, V. Jay, C. Smith, L. Becker, M. Petric, N.Olivieri, and M. Karmali. 1992. Experimental verocytotoxemia in rabbits.Infect. Immun. 60:4154–4167.

21. Richardson, S. E., M. A. Karmali, L. E. Becker, and C. R. Smith. 1988. Thehistopathology of the hemolytic uremic syndrome (HUS) associated withverocytotoxin-producing Escherichia coli (VTEC) infections. Hum. Pathol.19:1102–1108.

22. Riley, L., R. Remis, S. Helgerson, H. McGee, J. Wells, B. Davis, R. Herbert,E. Olcott, L. Johnson, N. Hagrett, P. Blake, and M. Cohen. 1983. Hemor-rhagic colitis associated with a rare Escherichia coli serotype. N. Engl.J. Med. 308:681–685.

23. Sandvig, K. M. Ryd, O. Garred, E. Schweda, P. K. Holm, and B. van Deurs.1994. Retrograde transport from the Golgi complex to the ER of both Shigatoxin and the non-toxic Shiga-B-fragment is regulated by butyric acid andcAMP. J. Cell Biol. 126:53–64.

24. Sjogren, R., R. Neill, D. Rachmilewitz, and D. Fritz. 1994. Role of Shiga-liketoxin I in bacterial enteritis: comparison between isogenic strains induced inrabbits. Gastroenterology 106:306–317.

25. Stein, P. E., A. Boodhoo, G. J. Tyrrell, J. L. Brunton, and R. J. Read. 1992.Crystal structure of the cell binding B oligomer of verotoxin-1 from Esche-

richia coli. Nature (London) 355:748–750.26. Tesh, V. L., B. Ramegowda, and J. E. Samuel. 1994. Purified Shiga-like toxins

induce expression of proinflammatory cytokines from murine peritonealmacrophages. Infect. Immun. 62:5085–5094.

27. Tesh, V. L., and A. D. O’Brien. 1991. The pathogenic mechanisms of Shiga-toxin and the Shiga-like toxin. Mol. Microbiol. 5:1817–1822.

28. Tesh, V. L., J. E. Samuel, L. P. Perera, J. B. Sharefkin, and A. D. O’Brien.1991. Evaluation of the role of Shiga and Shiga-like toxins in mediatingdirect damage to human vascular endothelial cells. J. Infect. Dis. 164:344–352.

29. Tyrrell, G. J., K. Ramotar, B. Toye, B. Boyd, C. A. Lingwood, and J. L.Brunton. 1992. Alteration of the carbohydrate binding specificity of vero-toxins from Gala1-4Gal to GalNAcb1-3Gala1-4Gal and vice versa by site-directed mutagenesis of the binding subunit. Proc. Natl. Acad. Sci. USA89:524–528.

30. Van der Kar, N. C. A. J., L. A. H. Monnens, M. A. Karmali, and V. W. M. vanHinsbergh. 1992. Tumor necrosis factor and interleukin-1 induce expressionof the verocytotoxin receptor globotriaosylceramide on human endothelialcells: implications for the pathogenesis of the hemolytic uremic syndrome.Blood 80:2755–2764.

31. Waddell, T., S. Head, M. Petric, A. Cohen, and C. A. Lingwood. 1988.Globotriaosyl ceramide is specifically recognized by the Escherichia coliverocytotoxin 2. Biochem. Biophys. Res. Commun. 152:674–679.

32. Wadolkowski, E. A., J. A. Burris, and A. D. O’Brien. 1990. Mouse model forcolonization and disease caused by enterohemorrhagic Escherichia coliO157:H7 Infect. Immun. 58:2438–2445.

33. Zoja, C., D. Corna, C. Farina, G. Sacchi, C. Lingwood, M. Doyle, V. V.Padhye, M. Abbate, and G. Remmuzzi. 1992. Tissue localization of verotoxinglycolipid receptor determines clinical expression of the disease in rabbitschallenged with verotoxin-1. J. Lab. Clin. Med. 120:229–238.

Editor: A. O’Brien



http://iai.asm.org/

Dow

nloaded from

http://iai.asm.org/

Documents

Toxicity and Immunogenicity of a Verotoxin 1 Mutant with Reduced