Nonspecific Binding of Clostridium difficile Toxin A to Murine Immunoglobulins Occurs via the Fab Component

Abstract

Clostridium difficile toxin A binds nonspecifically to a mouse monoclonal antibody (MAb) immunoglobulin G3 λ chain [IgG3(λ)], through the Fab component. This binding, which is retained even after boiling the MAb, is temperature dependent, with more toxin bound at 4 than 37°C (P = 0.0024). The nonspecific binding was decreased by incubation of the IgG3 λ MAb with α- or β-galactosidase (P = 0.0001 and 0.029, respectively), indicating that toxin A binds to a carbohydrate moiety on the Fab. However, binding was not blocked by the Bandeiraea simplicifolia lectin BS-1, indicating that a terminal α-galactose may not be involved. Binding was also not affected by competitive assays with Lewis X antigen. The dependence on carbohydrate moieties in nonspecific binding was also shown for two other MAbs, IgA(κ) and IgM(λ), with demonstration of a significant reduction in binding with α-galactosidase (P = 0.0001 and 0.0002, respectively) but not β-galactosidase (P = 0.27 and 0.25, respectively).

Clostridium difficile is the major etiological agent of antibiotic-associated pseudomembranous colitis and of many cases of antibiotic-associated diarrhea (2, 9, 10, 17). The disease results from exposure to, and colonization by, C. difficile and production of two major toxins, toxins A and B (1, 3). Both toxins are able to bind nonspecifically to many murine monoclonal antibodies (MAbs) raised against antigens other than toxin A (19), although binding is greater to toxin A (19). None of the MAbs that were examined neutralized or precipitated the biological activity of toxin A, leading the authors to conclude that toxin A binding did not occur via a true immune reaction. Toxin A also binds to human para-proteins (unpublished data). This nonspecific binding is in keeping with the ability of a number of proteins from a variety of gram-positive (8, 14, 16, 23) and gram-negative (4, 7, 15, 20, 26, 27) microorganisms to interact nonimmunologically with immunoglobulins (Igs) (26). To date, all of these proteins have been shown to bind to the Fc domain of the antibody molecule. We undertook to determine which component of the Ig bound nonspecifically to toxin A and to identify the nature of the interaction.

MATERIALS AND METHODS

Reagents.

Commercially purified mouse IgG3(λ) (Y5606), IgM(λ) (MOPC 104E), and IgA(κ) (TEPC 15), goat anti-mouse IgG conjugated to alkaline phosphatase, and Bandeiraea simplicifolia BS-1 isolectin B4 were purchased from Sigma. A specific MAb to C. difficile toxin A, PCG-4 [IgG2a(κ)] was a gift from D. Lyerly (18). Coffee bean α-galactosidase and Escherichia coli β-galactosidase were purchased from Boehringer Mannheim.

Preparation of toxin A.

C. difficile VPI 10463 was incubated for 4 days at 37°C in dialysis flasks, and the resulting toxin A was purified to homogeneity from the culture filtrate, as described in detail previously (13).

Preparation of Fab and Fc fragments.

Mouse IgG3(λ) (1 mg) was digested with immobilized papain, and the resulting Fab and Fc fragments were separated on a protein A column (ImmunoPure Fab preparation kit; Pierce Chemical Co., Rockford, Ill.) as described by the manufacturers. Both fragments were dialyzed against phosphate-buffered saline (PBS) (pH 7.4) overnight at 4°C and then concentrated to a final volume of 1 ml with concentrators (Vivapore Science Ltd., Lincoln, United Kingdom). Fragment purity was determined by the ability to bind protein A. Fab or Fc fragments (10 μg/ml) were incubated on plates coated with protein A (2.5 μg/ml). By using the enzyme-linked immunosorbent assay (ELISA) protocol described below, optical densities at 405 nm (OD₄₀₅) of 0.198 ± 0.009, 0.472 ± 0.012, and 0.199 ± 0.0005 were obtained for Fab, Fc, and conjugate controls, respectively, indicating that Fc fragments were absent from the Fab preparation.

ELISA determination of nonspecific binding of MAb to toxin A.

Wells of Nunc Maxisorp C96 microtiter plates (Life Technologies Ltd., Paisley, United Kingdom) were coated with 5 μg of toxin A per ml in 0.05 M carbonate buffer (pH 9.6). The plates were incubated overnight at 4°C and then washed three times with PBS containing 0.1% Tween 20 (PBST). The plates were blocked with 2% bovine serum albumin (Sigma) in PBST for 1 h at 22°C. The MAbs or IgG3 λ fragments (50 μl) in 1% bovine serum albumin–PBST were added to the toxin-coated wells, and the plates were incubated for 2 h at 37°C for complete MAb or overnight at 37°C for IgG3(λ) MAb fragments. The plates were washed three times with PBST, and 50 μl of goat anti-mouse alkaline phosphatase-conjugated IgG in 1% BSA–PBST was added to each well. Goat anti-mouse IgG-alkaline phosphatase was used to detect binding to IgG as well as IgA and IgM since it binds to all three Ig classes (19). After 1 h at 37°C, the plates were washed three times with PBST. Soluble alkaline phosphatase (p-nitrophenyl phosphate) substrate (50 μl; Sigma) was added to each well, the plates were incubated at room temperature, and the optical density at 405 nm (OD₄₀₅) was monitored hourly for 3 h. The highest OD₄₀₅ value obtained with wells that did not contain primary antibody was taken as the background value.

Effect of the reaction temperature on binding.

To determine if nonspecific binding of toxin A to the IgG3(λ) MAb was temperature dependent, 50 μl of MAb (40 μg/ml) was incubated on toxin-coated plates at either 4 or 37°C for 2 h. To detect the complex, goat anti-mouse IgG-alkaline phosphatase conjugate was added and allowed to react for 2 h at 37°C.

Effect of boiling the MAb on binding.

To determine the effect of boiling the IgG3(λ) MAb on binding to toxin A, MAb-coated plates instead of toxin A-coated plates had to be used since the goat anti-mouse conjugate did not react with boiled MAb. MAb (40 μg/ml) in bicarbonate buffer (pH 9.6) was boiled for 10 min at 100°C, and 50-μl portions were incubated in microtiter wells overnight at 4°C. After blocking (2% BSA–PBST for 1 h at 22°C), 50 μl of toxin A (5 μg/ml) or diluent (1% BSA–PBST) was added to each of the wells, which were incubated for 2 h at 37°C. Binding was detected with a 1:1,000 dilution of PCG-4 (1 mg/ml), a specific MAb against toxin A, followed by a 1:1,000 dilution of goat anti-mouse IgG alkaline phosphatase conjugate, with incubation for 1 h at 37°C after each step. The OD₄₀₅ was recorded after 1 h at room temperature.

Effect of α- or β-galactosidase treatment of the MAbs.

To determine whether α- or β-galactosidase or their combination could reduce the binding of the MAbs to toxin A, the ELISA procedure with toxin A-coated plates (described above) was used. MAbs (50 μg/ml) in 1% BSA–PBST (pH 6.0) were preincubated with 1.5 U of α- or β-galactosidase per ml at 22°C for 1 h for all three MAbs, and for 24 h to determine the effect of prolonged incubation on the IgG3(λ) MAb only, before being added to a microtiter plate as the primary antibody. Wells containing only 1% BSA–PBST (conjugate controls), both treated with α- or β-galactosidase and untreated, served as negative controls to determine the background absorbance. The specific antibody for toxin A (PCG-4) at 2 μg/ml in 1% BSA–PBST was added to toxin A-coated plates as a positive control.

To test the effect of α-galactosidase on toxin A (as opposed to the MAb) and its subsequent binding of the MAb, an ELISA plate was coated with 50 μl of toxin A (5 μg/ml) at 4°C overnight. The plates were blocked before addition of α-galactosidase (1.5 U/ml) or diluent and holding at room temperature overnight. The IgG3(λ) MAb was then added at 20 μg/ml and incubated for 2 h at 37°C.

Effect of α-galactosidase treatment on nonspecific binding of the IgG3(λ) fragments.

The procedure to determine the effect of α-galactosidase on binding of the IgG3(λ) fragments was as above, except that the Fab and Fc fragments of IgG3(λ) were used instead of whole MAb. Treatment with α-galactosidase only was performed since MAb fragments were only available in small quantities, and this was shown to be most effective for whole MAb.

Effect of B. simplicifolia lectin on binding.

The ability of B. simplicifolia lectin, which is specific for terminal α-linked galactose (12), to block the binding of the IgG3(λ) MAb to toxin A was investigated by ELISA. A dilution range of 5 to 200 μg of lectin per ml in 1% BSA–PBST was preincubated with the MAb (40 μg/ml) for 2 h at 4°C, and 50 μl of each mixture was incubated in toxin-coated wells for 2 h at 37°C. Lectin-only negative controls were used to determine background values.

To determine the binding of the lectin to the d-galactose of the Lewis X antigen of rabbit erythrocytes, a hemagglutination assay was used. Lectin was diluted to 10, 20, and 100 μg/ml in PBS. Then 50 μl of each dilution, 50 μl of C. difficile VPI 10463 culture filtrate (positive control), or 50 μl of PBS (negative control) was added to V-bottom wells of a microtiter tray (Scientific Laboratory Supplies, Nottingham, United Kingdom), 100 μl of rabbit erythrocytes was added to each well, and the plate was incubated at 4°C for 1 h before hemagglutination was determined.

Competitive binding of Lewis X antigen.

A 1:1 ratio of the IgG3(λ) MAb to Lewis X antigen (Oxford Glycosystems) (50 μl) was incubated in toxin A-coated wells at 4°C for 2 h, and 50 μl of anti-mouse IgG alkaline phosphatase conjugate was added to each well after the plates were washed with ice-cold PBST.

For all ELISAs, the OD₄₀₅ readings were recorded as mean ± standard error for a minimum of three separate experiments undertaken at different times, and each was conducted in triplicate. The significance of binding was determined from comparison with the conjugate control.

RESULTS

All three mouse MAbs significantly bound nonspecifically to toxin A (Table 1) (P < 0.0001 for all the IgG, IgA, and IgM MAbs). The IgG3(λ) MAb was used to confirm reproducibility. Analysis of the results of 15 separate experiments (i.e., 45 separate readings) over a period of 1 year showed nonspecific binding in all cases, yielding a mean OD₄₀₅ of 0.485 ± 0.025 (P < 0.0001).

TABLE 1.

Effect of α- and β-galactosidase on binding of IgG3(λ), IgM(λ), and IgA(κ) to toxin A^a

Treatment	OD405 (mean ± SE) (% binding reduction) for:
	IgG3(λ)	IgM(λ)	IgA(κ)
None	0.384 ± 0.039	0.502 ± 0.036	0.752 ± 0.059
	0.327 ± 0.032b
α-Galactosidase	0.179 ± 0.009 (53.4)	0.326 ± 0.022 (35.1)	0.443 ± 0.029 (41.1)
	0.132 ± 0.008b (59.7)b
β-Galactosidase	0.282 ± 0.017 (26.6)	0.454 ± 0.035 (9.6)	0.664 ± 0.049 (11.7)
	0.165 ± 0.019b (49.6)b
Both	0.191 ± 0.010 (50.3)	0.327 ± 0.023 (34.9)	0.434 ± 0.029 (42.3)
	0.103 ± 0.006b (68.5)b

^aMAbs (50 μg/ml) were incubated for 1 h at 22°C or for 24 h [IgG3(λ) only] with 1.5 U of enzyme per ml and incubated for 2 h at 37°C on toxin-coated plates. The OD₄₀₅ was measured after 1 h. The 1-h incubation values represents the mean of six assays with triplicate wells. The conjugate control values are as follows: α-galactosidase treated, 0.067 ± 0.002; β-galactosidase treated, 0.069 ± 0.002; α- and β-galactosidase treated, 0.066 ± 0.0018; untreated conjugate control, 0.069 ± 0.002. 

^b24 h; mean ± SE for three assays for prolonged incubation of IgG3(λ). 

The nonspecific binding of toxin A to the IgG3(λ) MAb (40 μg/ml) was temperature dependent (Table 2), with more binding at 4 than 37°C (P = 0.0024). Boiled MAb (40 μg/ml) also nonspecifically bound toxin A (Table 2) (P = 0.0014). It was possible to use murine MAbs for both capture and detection because the boiled capture MAb is not recognized by the goat anti-mouse antibody. The binding of the IgG3(λ) Fab and Fc fragments was then investigated (Table 2). The Fab fragments bound nonspecifically to toxin A (P = 0.0001), whereas the Fc fragments did not (P = 0.41) compared with the conjugate control.

TABLE 2.

Effect of incubation temperature, boiling, or fragmentation on MAb binding to toxin A^a

Test	OD405 (mean ± SE)b
Incubation temp (°C)
4°C	0.800 ± 0.021
37°C control	0.679 ± 0.016
Boiling
Boiled MAb	0.653 ± 0.012
Conjugate control	0.081 ± 0.002
Fragmentation
Fc	0.081 ± 0.002
Fab	0.125 ± 0.005
Whole MAb	0.479 ± 0.009
Conjugate control	0.084 ± 0.007

^aMAb (40 μg/ml) was added to toxin A (5 μg/ml)-coated plates except for the boiled-MAb assay, where toxin A was added to MAb-coated plates. Fragments of IgG3(λ) were incubated for 24 h at 37°C on toxin A-coated plates. Goat anti-mouse alkaline phosphatase was balanced (i.e., appropriate concentrations were used to give equivalent absorbance per microgram of fragment to account for differing binding to Fab or Fc). 

^bValues represent mean ± SE of three assays of triplicate wells. 

Treatment of the IgG3(λ) MAb (50 μg/ml) with α-galactosidase and β-galactosidase (1.5 U/ml) produced a significant decrease in nonspecific binding of the MAb to toxin A (Table 1) (P = 0.0001 and 0.029, respectively) and reductions in binding of 53.4 and 26.6%, respectively. Incubation with both enzymes (Table 1) resulted in a 50.3% reduction in binding (P = 0.0002). Treatment of the IgM(λ) and IgA(κ) MAbs with α-galactosidase produced decreases in nonspecific binding to toxin A similar to those seen for IgG3(λ) (Table 1). Comparison of α-galactosidase- or β-galactosidase-treated MAbs to untreated MAbs gave P = 0.0002 and 0.25 for IgM(λ) and P = 0.0001 and 0.27 for IgA(κ), respectively. Treatment of toxin A with α-galactosidase before addition of the MAb did not affect binding (OD₄₀₅ = 0.368 ± 0.007 for untreated toxin and 0.336 ± 0.028 for treated toxin [P > 0.05]). Prolonged preincubation (24 h) of the enzymes, either alone or in combination, with the IgG3(λ) MAb produced reductions in binding of 59.7, 49.6, and 68.5%, respectively (Table 1). A specific MAb for toxin A, PCG-4 (2 μg/ml), was used as a positive control. Neither α-galactosidase nor β-galactosidase treatment reduced the degree of binding to toxin A, with α-galactosidase-treated PCG-4 giving an OD₄₀₅ of 2.549 ± 0.029 and β-galactosidase-treated PCG-4 giving an OD₄₀₅ of 2.577 ± 0.025 compared to a value of 2.587 ± 0.022 for untreated PCG-4 after 15 min at room temperature (P = 0.24 and 0.71 respectively).

Treatment of IgG3(λ) Fab fragments with α-galactosidase resulted in a significant reduction of nonspecific binding to toxin A with values of 0.098 ± 0.002 compared to 0.125 ± 0.005 for untreated fragments (P = 0.0046). However, B. simplicifolia lectin did not block binding of the MAb to toxin A (Table 3), even though it agglutinated rabbit erythrocytes, indicating that it recognizes the d-galactose of Lewis X antigen. Equally Lewis X antigen did not significantly decrease the binding of the MAb to toxin A. OD₄₀₅ values of 0.30 ± 0.004 for the MAb only and 0.259 ± 0.016 for MAb incubated with Lewis X antigen were obtained (P = 0.071).

TABLE 3.

Effect of B. simplicifolia lectin on the binding of IgG3(λ) to toxin A^a

Amt of lectin (μg/ml)	OD405 (mean ± SE) forb:
	MAb	Conjugate control
0	0.609 ± 0.002	0.114 ± 0.004
5	0.604 ± 0.004	0.070 ± 0.006
10	0.691 ± 0.024	0.076 ± 0.011
50	0.683 ± 0.007	0.114 ± 0.021
100	0.834 ± 0.005	0.242 ± 0.006
150	0.861 ± 0.018	0.242 ± 0.023

^aIgG3(λ) (40 μg/ml) and lectin dilutions were incubated at 4°C for 2 h and then at 37°C for 2 h on toxin-coated plates. 

^bMean OD₄₀₅ ± SE of triplicate determinations after 1 h. 

DISCUSSION

MAbs are powerful tools that can be used in trying to identify functional epitopes of protein molecules. Many researchers raised MAbs to toxins A and B of C. difficile for structure-function studies. One of the earliest of these, PCG-4 [1gG2a(κ)] raised to toxin A (18), was shown to recognize particular repeat units at the C-terminal end of toxin A (5) and to inhibit enterotoxic activity (18). Only one of the MAbs raised by these researchers cross-reacted with toxin B. Since the toxins were also shown to be antigenically distinct (18, 21), the authors concluded that the toxins share one or a few epitopes (18). However, others reported up to as many as 60 cross-reacting MAbs (6, 24), causing Lyerly et al. to propose that the toxins may bind MAbs nonspecifically (19). This was shown to be the case (19). We had also independently noted an apparent similar phenomenon with human monoclonal para-proteins (unpublished data), as well as observing that a negative in-house control “nonsense” MAb (a MAb specific for herpes simplex virus type 1), consistently reacted with toxin A (unpublished data). The results of the study reported here clearly show that the apparent nonspecific interaction of one of these MAbs is due to interaction with the Fab component rather than with the Fc component. This was a surprising observation, particularly since all nonspecific interactions of bacterial antigens with Igs described to date have occurred via the Fc component (see, e.g., references 7, 8, 14–16, 20, 23, 26, and 27). There were therefore no precedents from which to infer the nature of this interaction. Preliminary characterization demonstrated that the effect exhibited by the IgG3(λ) MAb was resistant to boiling raising the possibility of involvement of a carbohydrate. This possibility is supported by the fact that Igs are glycoproteins consisting of 4 to 18% carbohydrate (11). In general, the carbohydrate is found only in the secretory component, the J chain, and the constant regions of the heavy chains and, with few exceptions, is not found associated with the light chains or the variable regions of the heavy chains (11). The presence of carbohydrate on Igs and the known ability of toxin A to bind to certain carbohydrates (25) raised the possibility that toxin A was binding to the Fab component of MAb IgG3(λ) via carbohydrate localized on the constant domain of the heavy chain (C_H1). Since the binding of toxin A to the trisaccharide Galα1-3 Galαβ1-4 GlacNAc and to the carbohydrate antigens designated I, X, and Y is temperature dependent, the temperature dependency of binding to IgG3(λ) MAb was investigated. There was significantly more binding at 4 than at 37°C, supporting the possibility that toxin A bound to Fab through carbohydrate. The fact that pretreatment of the IgG3(λ) MAb with α- or β-galactosidase significantly reduced binding to the toxin and the demonstration that the reduction was not due to an effect of the galactosidases on the toxin further support the involvement of carbohydrate. Both galactosidases gave reductions in binding when used alone, although α-galactosidase always had a greater effect, and their combination reduced binding to a similar degree to that of α-galactosidase alone. This difference between α- and β-galactosidase was less pronounced following prolonged incubation with the enzymes. These observations are consistent with the involvement of carbohydrates bearing terminal α-galactose and terminal β-galactose. Further, α-galactosidase treatment of Fab (β-galactosidase was not tested) reduced binding. Interestingly, only α-galactosidase treatment of IgA and IgM gave reductions in binding, suggesting the possibility that only α- galactose residues are involved in binding for at least these two MAbs.

There is evidence that binding of toxin A to carbohydrate is dependent on the type 2 core (Galβ1-4 GlcNAc), with a branch either on or immediately adjacent to this core (25). Such an antigen is Lewis X, which is also present on the secretory component of Ig (21). However, based on the observations of the IgG3(λ) MAb, it is unlikely that Lewis X antigen was involved in the nonspecific binding of toxin A to the Fab fragment, since the binding could not be significantly inhibited by Lewis X antigen when tested at the optimum temperature of 4°C. However, these observations do not rule out the involvement of other toxin A binding carbohydrates with terminal α-galactose.

What is more difficult to explain is the reduction in binding by α-galactosidase but the inability of B. simplicifolia lectin to prevent binding. It is possible to speculate that two separate receptors are not involved, since the combined inhibition of the α- and β-galactosidase is not equal to the sum of their individual inhibitory activities, even after prolonged exposure to the enzymes. Since type 2 cores in unbranched molecules do not bind (25), it is possible that there is a single receptor with a terminal type 2 core and a branch(es) with a terminal α- galactose. Blockage of the α-galactose would not affect the ability of the branch to sterically hinder changes in the conformation of the core, whereas removal of the α-galactose could do so.

It is known from the work of others that the repeats at the C-terminal end of toxin A are involved in carbohydrate binding (5, 22). If this is the case and binding to the Fab fragment occurs via carbohydrate, the C-terminal end of toxin A alone should also have this activity. Preliminary evidence shows that this is the case, with the C-terminal peptide behaving like the whole toxin but none of five different overlapping N-terminal fragments binding the Fab fragment (4a). The significance of the ability of toxin A to bind nonspecifically with Igs by interaction with the Fab component is unknown. It may promote interaction with mucus due to the presence of secretory IgA, which contains more carbohydrate than other Igs do (11). A practical outcome of our findings is the indication that MAbs raised to toxin A should be retested following treatment with α- and β-galactosidase to determine the specificity of the interaction. To our knowledge, this is the first observation of a bacterial antigen binding nonspecifically to the Fab component of Igs. The extent to which other microorganisms can do this and the significance of such binding should be explored.