Mouse Toxicity and Cytokine Release by Verotoxin 1 B Subunit Mutants

Abstract

The crystal structure of the verotoxin 1 (VT1) B subunit complexed with a globotriaosylceramide (Gb₃) analogue showed the presence of three receptor binding sites per monomer. We wished to study the effects of altering the three sites, singly or in combination, on animal toxicity and cytokine induction in vitro. We found that while the site 1 and 2 mutants were modestly (two- to sevenfold) reduced in their ability to cause disease in BALB/c mice, the site 3 mutant, W34A, was as toxic as VT1. However, all the double-mutant proteins, irrespective of which two sites were mutated, exhibited approximately a 100-fold reduction in their 50% lethal doses for mice. These results suggest that multivalent receptor binding is important in vivo and that all three binding sites make a similar contribution to the latter process. The triple-mutant holotoxin, F30A G62T W34A, administered intraperitoneally without adjuvant, stimulated a strong antibody response in BALB/c mice, and the immune sera neutralized the activity of VT1 in vitro. Induction of tumor neurosis factor alpha release from differentiated human monocytes (THP-1 cells) was relatively impaired for site 1 and site 2 but not site 3 mutants, suggesting an auxiliary role for the latter site in mediation of cytokine release in vitro. Cytotoxicity assays on undifferentiated THP-1 cells have also demonstrated the importance of sites 1 and 2 and the relatively small role played by site 3 in causing cell death. These data suggest an association between the cytotoxicity of the protein and its ability to induce cytokine release.

Verotoxin-producing Escherichia coli (VTEC) strains are recognized as the etiological agents of a number of human diseases, including hemorrhagic colitis (19, 26) and the hemolytic uremic syndrome (HUS) (10). Much evidence suggests that verotoxins (VTs; also known as Shiga-like toxins) produced by VTEC are directly involved in the genesis of the vasculopathies of both of these conditions (24, 25). VTs belong to the A:B5 family of bacterial toxins and consist of an enzymatic A subunit (32 kDa) and a homopentamer of B subunits (7.5 kDa each) (8, 29). In the case of VTs, the A subunit possesses N-glycosidase activity which catalyzes the depurination of the 28S rRNA component of the eukaryotic ribosome, leading to the inhibition of protein synthesis (7). The B subunit pentamer mediates binding to the toxin receptor, the globo-series glycosphingolipid globotriaosylceramide (Gb₃) (14, 35).

The structure of the cocrystal of the VT1 B pentamer and the Pk trisaccharide, a Gb₃ analogue, revealed three Pk binding sites on each VT1 B monomer (12). Site 1 involved a hydrophobic stacking interaction between F30 and Gal-β of the Pk trisaccharide and hydrogen bonds with a number of residues, including D17, E28, and T21. Site 2 involved hydrogen bonds with R33, N32, and G62 and hydrophobic interactions with F30 and A56. Site 3 involved a hydrophobic stacking of W34 on Gal-β of the Pk moiety and hydrogen bonding with D18 of the adjacent monomer. Computer modeling studies predicted the presence of two Gb₃ binding sites on the VT1 B subunit: site 1, a cleft defined by the solvent-exposed face of F30 and an aspartate loop (D16 to D18), and site 2, a crevice formed by the back side of the same F30, a glycine loop (G60 to G62) and residues N32 and R33 (17, 18). Site 1 was very similar to that demonstrated by crystallography, while the details of site 2 differed significantly, although it was located in the same area. The third site identified in the cocrystal, termed site 3, had not been predicted by modeling. Studies of VT1 B mutants in our laboratory have demonstrated that sites 1 and 2 function as the principal sites responsible for high-avidity binding to Gb₃ in the membrane context and that both are important in mediating cell toxicity (3, 6; A. M. Soltyk, C. R. MacKenzie, W. M. Wolski, T. Hirama, and J. L. Brunton, submitted for publication). In contrast, site 3 was shown to contribute significantly to receptor binding but appeared to play a relatively minor role in inducing cytotoxicity in vitro (3; Soltyk et al., submitted).

It has been proposed that two synergistic signals may play a role in the pathogenesis of VT-mediated disease: systemic VTs and elevated levels of proinflammatory cytokines, such as tumor neurosis factor alpha (TNF-α) and interleukin-1β released in response to the toxins (31). The latter have been shown to sensitize endothelial cells to VTs by upregulating the cell membrane expression of Gb₃ (33, 34). In addition, clinical studies have shown that HUS patients have elevated levels of these proinflammatory cytokines in plasma as well as in urine (11, 15). Recently, it was demonstrated that, in contrast to the VT1 B subunit and anti-Gb₃ monoclonal antibody, only the VT1 holotoxin was capable of inducing cytokine release from THP-1 cells, a human monocytic cell line (28). Subsequently, Yamasaki et al. showed that cytokine release induction was dependent on the enzymatically active A subunit (37).

Given the central role that VTs are believed to play in the pathogenic processes of both hemorrhagic colitis and HUS, immunization of humans against these virulence factors could protect against the systemic complications of VTEC infection, as has already been shown for edema disease of swine (9). Although the B subunit of VT1 generates strong antibody responses (5, 27), holotoxin toxoids may be superior to vaccines incorporating only the B subunits of VTs, as they provide protection against both homologous and heterologus toxins (4).

In this study, we investigated the roles played by the three glycolipid binding sites of the VT1 B subunit in inducing disease in the animal host and stimulating cytokine release from monocytes in vitro. We also assessed the immunogenicity of the triple-mutant protein, F30A G62T W34A, and the ability of the induced humoral response to neutralize the activity of VT1.

Toxicity of VT1 mutants in BALB/c mice.

Wild-type and mutant holotoxins used in this study were purified from periplasmic extracts of E. coli JM101 transformed with the plasmid pJLB128 (22) encoding the VT1 holotoxin and from derivative plasmids carrying mutations in the B cistron of VT1 that were designed to specifically disrupt the three Gb₃ binding sites. The mutant holotoxins were F30A (site 1 [6]), G62T, G62A, and A56Y (site 2 [3; Soltyk et al., submitted]), and W34A (site 3 [3]); the double mutants were F30A W34A, G62T W34A, and F30A G62T; and the triple mutant was F30A G62T W34A (Soltyk et al., submitted). Crystal structures of several mutants complexed with the Pk trisaccharide analogue have shown that the G62T mutant can interact with Pk at site 1 but not site 2 while the F30A mutant does not interact with Pk at site 1 (13). The holotoxins were purified as previously described (20). The purified holotoxins were resuspended in phosphate-buffered saline, purged of endotoxin with polymyxin B resin (Bio-Rad Laboratories, Hercules, Calif.), and tested for residual endotoxin contamination by the Limulus amebocyte lysate assay (Associates of Cape Cod, Inc., Falmouth, Mass.). The toxin preparations were found to contain, on average, less than 1 pg of endotoxin per 1 μg of a given holotoxin. Groups of five 4- to 6-week-old female mice of the inbred BALB/c strain (Charles River Laboratories, Wilmington, Mass.) were injected intraperitoneally with serial dilutions of the wild-type and mutant holotoxins and observed for the development of hind limb paralysis and death. To check for the lethality of the residual endotoxin contamination of the protein samples, five mice were injected with boiled toxins (100°C, 1 h) at the highest toxin dose used in the 50% lethal dose (LD₅₀) studies. Control mice for the LD₅₀ studies were injected with phosphate-buffered saline by the protocol outlined above. The experimental procedures carried out on the mice were in accordance with the principles of the Animal Care Committee of Mount Sinai Hospital, Toronto, Canada. The LD₅₀s (Table 1) were calculated according to the method described by Reed and Muench (23). For F30A and A56Y, the LD₅₀s fell between 1 and 2 μg (at 1 μg and lower doses, 100% of the mice survived; at 2 μg and higher doses, 100% of the mice died). The LD₅₀s for F30A W34A, G62T W34A, F30A G62T, and F30A G62T W34A fell between 50 to 100 μg (for all these toxins at 50 μg and lower doses, 100% of the mice survived; at 100 μg and higher doses, 100% of the mice died). Ranges of LD₅₀s are provided for the toxins above, since the determination of precise LD₅₀s would involve sacrificing more animals at doses between the twofold doses, which yielded an all-or-none response. Control mice injected with boiled toxins showed 100% survival.

TABLE 1.

The reductions in toxicity observed for the mutant toxins in BALB/c mice and on THP-1 cells^a

Toxin	Mutated site(s)	LD50 (μg)	Time to death (days)b	CD50 (ng/ml) on THP1 cellsb
VT1		0.7	2.5 ± 0.3	0.042 ± 0.005
W34A	3	0.5	2.6 ± 0.2	0.5 ± 0.0
A56Y	2	1–2	3.3 ± 0.1	20 ± 6.9
G62A	2	3.0	3.3 ± 0.2	11,000 ± 3,400
G62T	2	5.8	3.4 ± 0.2	50,000 ± 0.0
F30A	1	1–2	3.3 ± 0.1	8,500 ± 2,300
F30A G62T	1, 2	50–100	7.6 ± 0.7	ND
F30A W34A	1, 3	50–100	8.0 ± 1.0	12,000 ± 3,200
G62T W34A	2, 3	50–100	9.1 ± 0.8	50,000 ± 2,100
F30A G62T W34A	1, 2, 3	50–100	8.1 ± 1.1	ND

^aThe results shown for the THP-1-cell cytotoxicity assay are those of one representative experiment done in triplicate. Two other independent experiments showed similar results. 

^bData are means ± standard deviation. ND, not determined. 

Work in our laboratory has shown the importance of high-affinity Gb₃ binding sites 1 and 2 and the relatively small contribution of low-affinity Gb₃ binding site 3 in mediating cell death in vitro (Table 1) (3, 6; Soltyk et al., submitted). Thus, the high animal toxicity of the site 3 mutant, W34A, both with respect to its LD₅₀ and time elapsed to the onset of disease symptoms and death (the latter being identical to that of VT1), is consistent with the in vitro findings. These results show that the presence of functional high-affinity sites 1 and 2 is sufficient for the toxin to mediate VT1-like levels of toxicity in mice. Surprisingly, however, site 1 (F30A) and site 2 (G62A and G62T) mutants also exhibited relatively high toxicity levels in vivo, even though their 50% cytotoxic doses (CD₅₀s) were 5 to 6 logs higher than those for VT1 (Table 1). The A56Y site 2 mutation is believed to block site 2 less effectively than the G62 mutations, resulting in a relatively more toxic protein both in vivo and in vitro (Table 1).

All the double-mutant proteins (F30A W34A, G62T W34A, and F30A G62T) exhibited a significant decrease in toxicity in comparison to their constituent single mutants. While the mutation of any one site resulted in no reduction (W34A) or reductions of <1 order of magnitude in mouse lethality (F30A, A56Y, G62A, and G62T), a mutation of any two sites caused an approximately 100-fold decrease in toxicity relative to that of VT1. This is particularly striking for the F30A W34A and G62T W34A mutants, which in vitro exhibited no statistically significant (P > 0.05) reductions in cytotoxicity relative to the single mutants F30A and G62T, respectively (Table 1). The reductions in animal toxicity caused by adding the W34A mutation to a preexisting site 1 or 2 mutation are comparable to the reductions in binding to Gb₃-containing liposomes demonstrated by surface plasmon resonance (Soltyk et al., submitted). A plausible explanation for these observations is that maintaining multivalency of receptor-toxin interactions is critical to achieve high-avidity binding in the membrane context which is required for cellular internalization. Thus, with any two sites disrupted, the one remaining Gb₃ binding site per B subunit is insufficient to support high levels of receptor engagement. This results in a protein possessing only residual lethality similar to that of the triple-mutant toxin, in which all three sites are disrupted. The requirement for multivalent binding in vivo is consistent with the results of VT1 mutants binding to Gb₃ in liposomes (Soltyk et al., submitted).

Another observation emerging from our studies is that the mutant toxins are much more toxic in vivo than in vitro relative to the toxicity of VT1. One reason for this may be the complexity of an animal system versus a simple cell monolayer. Specifically, host response to the toxins by cytokine release, resulting ultimately in target-cell sensitization by means of increased Gb₃ expression on the cells of susceptible tissues, could in part be responsible for this phenomenon (15, 31, 32). Another plausible explanation for this increased toxicity of the mutants of VT1 relative to that of the wild-type protein could be that, in mice, there exists a Gb₃-independent mechanism facilitating the intoxication of renal tubular cells, the site of pathology in the murine model of VT-mediated disease (30, 36). The latter might involve tubular reabsorption of filtered proteins, including VTs, by a nonspecific, receptor-independent process of pinocytosis (16).

For all the mutants except for W34A, the time elapsed to the onset of disease development and death was delayed relative to that of VT1 (Table 1). The mechanism underlying this phenomenon may consist of two components. First, the reduced level of receptor recognition exhibited by the mutants (3; Soltyk et al., submitted) may lead to slow, reduced uptake in the susceptible tissues. This was in fact observed for the F30A mutant in the rabbit model of verotoxemia (1). The serum half-life of the mutant was close to 50-fold longer than that of VT1, and the mutant also failed to preferentially localize to the Gb₃-rich tissues targeted by the wild-type protein. Also, mutants exhibiting reduced Gb₃ recognition are internalized by Vero cells to a lesser extent than the wild-type protein (Soltyk et al., submitted). The second component of the mechanism responsible for the observed delay may be that the mutant holotoxins are less efficient than the wild-type protein at stimulating cytokine release. We have shown that this is indeed the case in vitro (Fig. 1). This reduced cytokine release could lead to less target cell sensitization, and thus it would take longer for the disease symptoms to develop.

FIG. 1.

Net TNF-α release induced by purified VT1 and its mutants from a human monocytic cell line. TPA-differentiated THP-1 cells were incubated in media containing the indicated concentrations of VT1 and its mutants for 12 h at 37°C. The culture supernatants were then assayed for the presence of TNF-α by ELISA. The data shown are the means ± standard deviation of three independent experiments. The release mediated by A56Y at 10 μg/ml was not statistically significant (P > 0.05). The background release from cells incubated in medium only was 5.3 ± 0.58 pg/ml. Student's t test was used to determine statistical significance.

Immunogenicity of the F30A G62T W34A holotoxin.

Five mice were injected intraperitoneally with 50 μg of the biologically active triple-mutant holotoxin, followed by boosters with the same dose 14 days later; 14 days after the booster injections, mice were bled by cardiac puncture. Control mice were injected with phosphate-buffered saline. Anti-F30A G62T W34A-specific antibodies and anti-VT1-specific antibodies were detected using a solid-phase enzyme-linked immunosorbent assay (ELISA) as previously described (2). Briefly, each well of an Immulon 1 plate (Dynatech Laboratories, Chantilly, Va.) was coated with 400 ng of F30A G62T W34A and VT1 holotoxins dissolved in 100 μl of the carbonate-bicarbonate buffer (pH 9.6). The coated wells were subsequently blocked with 2% (wt/vol) bovine serum albumin (Sigma Chemical Co., Oakville, Ontario, Canada), incubated with serial dilutions of sera, peroxidase-conjugated goat anti-mouse immunoglobulin G secondary antibody (Bio-Rad Laboratories), and developed with o-phenylenediamine dihydrochloride (Sigma Chemical Co.) for 30 min. The chromogen produced was measured by determining the optical density at 490 nm (OD₄₉₀) on an MR600 ELISA plate reader (Dynatech Laboratories). The antibody titer was defined as the highest serial dilution of serum at which the OD₄₉₀ was 2 standard deviations above the mean OD₄₉₀ of the negative-control sera at a 1:100 dilution (2). The ability of anti-F30A G62T W34A holotoxin antisera to inhibit the cytotoxic effects of VT1 toward Vero cells (a kind gift of C. Lingwood, The Hospital for Sick Children, Toronto, Canada) was determined in an in vitro neutralization assay (10). Equal volumes of serial twofold dilutions of sera and toxin at 0.03 ng/ml (10 times the measured CD₅₀ for VT1) were preincubated for 1 h and then added to the wells of a 96-well plate. The serum neutralization titer was defined as the highest serial dilution of serum at which 50% of the cells were killed by VT1. Antibody titers were converted to logarithmic values (log₅ x and log₂ x for ELISA and neutralizing titers, respectively, where x equals the reciprocal of the serum dilution) for calculation of geometric means and standard deviations. Mice injected with phosphate-buffered saline produced no anti-toxin antibodies.

Our results show that the F30A G62T W34A holotoxin produced a strong anti-toxin antibody response in mice, even without the use of any form of adjuvant. Importantly, the immune sera reacted equally well with both the triple-mutant and the wild-type holotoxins in a solid-phase ELISA assay. The reciprocal geometric mean (± standard deviation) titers of specific antibody to F30A G62T W34A and VT1 holotoxins were 8.29 ± 0.21 (serum dilution, 1:625,000) for both proteins. The sera also neutralized the cytotoxic activity of VT1 on Vero cells. The reciprocal geometric mean titer of neutralizing antibodies determined against 10 times the CD₅₀ of VT1 was 9.93 ± 0.42 (1:975) (n = 5). The advantage of using holotoxin molecules as vaccines is that, unlike B subunit vaccine formulations, they would be expected to provide protection against both homologous and heterologous toxins (VT2) in vivo (4). In addition, Bast et al. showed that the antibody response to B subunit presented as holotoxin is major histocompatibility complex independent (in the haplotypes tested), while the response to the B subunit alone is major histocompatibility complex dependent (2). Their data were compatible with the thesis that the A subunit contained T-helper epitopes. However, even the disruption of all three Gb₃ binding sites of VT1 (F30A G62T W34A) resulted in a holotoxin which, despite an 8-log reduction in toxicity in vitro (Soltyk et al., submitted), mediated relatively high levels of toxicity in vivo. Consequently, additional mutations, most likely in the active site of the A subunit, would have to be made to produce a safe natural toxoid.

Activity of B subunit mutant toxins on THP-1 cells.

THP-1 cells, a human monocytic cell line (a kind gift of O. Rotstein, The Toronto General Hospital, Toronto, Canada), were used to assess the toxicity and cytokine release induction by VT1 and its mutants. For undifferentiated THP-1-cell cytotoxicity assays, THP-1 cells grown in RPMI 1640 (Gibco BRL, Bethesda, Md.) containing 10% fetal calf serum (Sigma Chemical Co.) were seeded in 96-well plates, and serial 10-fold dilutions of lipopolysaccharide-purged holotoxins were added to the wells in triplicate. After 72 h of incubation, the assays were developed using an MTT cell viability assay kit (Promega, Madison, Wis.). Wells with no added toxin were used to calculate percent survival. For TNF-α release assays, THP-1 cells were differentiated for 65 h in the presence of 60 nM 12-O-tetradecanoylphorbol-13-acetate (TPA) (Sigma Chemical Co.) in complete RPMI 1640 containing 10% fetal calf serum. Following differentiation (assessed by flow cytometry analysis of cell surface expression of CD11b and morphological changes), the cells were first incubated for 24 h in complete RPMI 1640 containing no TPA and then for 12 h with serial dilutions of endotoxin-purged wild-type and mutant holotoxins. The supernatants were collected and assayed for the presence of immunoreactive TNF-α by using a human TNF-α ELISA kit (Medicorp, Inc., Montreal, Quebec, Canada). Boiled (100°C, 1 h) toxins served as controls for cytokine induction due to the residual endotoxin contamination of the protein samples. The Student t test was used to determine statistical significance for the toxicity and cytokine release assays using the functions contained within Primer of Biostatistics (McGraw-Hill). The CD₅₀s for undifferentiated THP-1 cells are presented in Table 1. The CD₅₀ for VT1 on TPA-differentiated THP-1 cells was 500 ± 150 (mean ± standard deviation) ng/ml (72-h incubation period). Differentiated THP-1 cells released TNF-α upon exposure to the toxins. At 10 μg of toxin/ml, F30A, G62T, F30A W34A, D17E W34A, and F30A G62T produced no net TNF-α release (Fig. 1). The cytokine release induced by the wild-type and W34A proteins at both the toxin doses was not significantly different (P > 0.05); VT1 and W34A also produced equivalent amounts of TNF-α at 1 and 5 μg of toxin per ml (data not shown). The induction mediated by A56Y at 10 μg/ml did not differ significantly from background release from cells incubated in medium only (P > 0.05). The above cytokine induction was toxin dependent, since the equivalent amounts of boiled holotoxins produced no net TNF-α production. Undifferentiated THP-1 cells did not release TNF-α upon exposure to VT1 (results not shown).

Our results indicate that while site 1 (F30A) and site 2 (G62A and G62T) mutants were severely impaired in their ability to cause THP-1-cell death, the site 3 mutant, W34A, exhibited relatively little reduction in that respect (Table 1). These findings closely mirror the results of cytotoxicity assays performed with these mutant proteins on other cell lines, such as the Gb₃-rich Vero cells, and thus confirm the importance of the high-affinity Gb₃ binding sites 1 and 2 in toxicity mediation (3; Soltyk et al., submitted). Interestingly, the mutant proteins exhibited much smaller reductions in their ability to induce TNF-α release than they did in their ability to mediate cell killing. However, in keeping with the cytotoxicity assay results, site 1 (F30A) and site 2 (A56Y, G62A, and G62T) mutants were relatively impaired in their ability to elicit cytokine release whereas the site 3 mutant, W34A, was as effective an inducer as VT1 at the doses tested. These results imply that toxin uptake mediated by Gb₃ binding through sites 1 and 2 is important for both toxicity and cytokine mediation, with site 3 playing an auxiliary role in those processes. Thus, these data suggest an association between the cytotoxicity of the protein and its ability to induce cytokines, an idea which was recently proposed by Yamasaki et al., who showed that the ability of an A subunit VT1 mutant to induce cytokines, at both the mRNA and protein levels was directly related to the mutant's toxicity (37).