Inhibition of T-cell Response by Escherichia coli Heat-Labile Enterotoxin-Treated Epithelial Cells

Abstract

Escherichia coli heat-labile enterotoxin (LT) is an extensively studied adjuvant of mucosal responses. Nevertheless, its mode of action as an adjuvant remains incompletely understood. In this study, we describe a simplified in vitro model with which to look at some aspects of immunoregulation by LT. The interaction of LT with the apical surface of a monolayer of CaCo-2 epithelial cells induces the release of a soluble factor which inhibits the antigen-induced release of interleukin-2 by T cells cultured at the basolateral side of the cells. The release of this factor requires the ADP-ribosylating activity of LT since the isolated B subunit, as well as an enzymatically silent LT mutant, loses biological activity in this model. The inhibitory activity is likely to be due to prostaglandin release, since it is blocked by indomethacin. The contribution of LT-induced prostaglandin release to the complex immunoregulatory activity of LT is discussed.

Escherichia coli heat-labile enterotoxin (LT) and the closely related Vibrio cholerae toxin are potent adjuvants of mucosal immune responses, and there is considerable interest in their potential use for the development of novel oral (or nasal) vaccines. To this end, much work has been directed at trying to identify the molecular features of the toxins which are responsible for their adjuvant properties and dissociate these elements of molecular structure from those responsible for the inherent toxicity of LT and cholera toxin (CT), which limits their use in vaccine formulations (reviewed in references 16 and 17).

LT and CT are highly homologous molecules, both comprising two subunits: the enzymatically active A subunit and a pentameric B subunit. The latter is responsible for cellular attachment of the toxin molecule to the cell membrane via interaction with the GM₁-ganglioside (GM₁) molecule, a glycosphingolipid found ubiquitously on the cell surface of mammalian cells, as well as on other surface glycoconjugates. After effective attachment of the toxin to the cell surface, the A subunit is internalized and undergoes selective proteolysis and reduction, generating two fragments—the A1 and A2 peptides. The A1 peptide interacts with and is activated by 20-kDa GTP binding proteins (known as ADP-ribosylation factors), which play a major role in the regulation of the cellular cytoskeleton. Finally, the activated A1 molecule is translocated to the cell membrane, where it catalyzes the irreversible transfer of ADP-ribose from NAD to the α subunit of G_s and possibly other G proteins located on the plasma membrane. This results in the irreversible activation of adenylate cyclase and a consequent increase in intracellular levels of cyclic AMP (cAMP). The toxicity of the molecules is generally believed to result from this abnormal accumulation of cAMP, which modulates ion and water transport across the gut epithelium, resulting in a characteristic diarrhea.

A large number of mutant forms of both LT and CT have now been generated to probe the structure and/or function characteristics of the molecules. These include those of recombinant B subunit alone, which have no associated toxicity and, of course, no enzymatic activity. There are also a number of mutants within the active site of the A1 subunit which decrease or abolish its enzymatic ADP-ribosylating activity and other mutants in which the cleavage of the A subunit to form A1 and A2 is substantially inhibited. All these mutants have been tested in a wide variety of in vivo immunization systems in an effort to understand the basis for the molecules' adjuvanticity (17). Intensive studies have highlighted some of the characteristics of LT and CT that contributed to adjuvanticity, but much remains to be discovered about the mechanisms of action. Significant differences in adjuvanticity may occur depending on the route of administration, the dose, and the host species (2, 8, 13, 18). The nasal route, for example, requires much less antigen and also much less adjuvant than the oral route, and it has been relatively easy to demonstrate adjuvant activity of the enzymatically inactive AB complexes or even B holotoxin.

In this study, we looked at the behavior of LT and its mutants in an in vitro model described previously (11). This model allowed us to look at the action of LT in a context simpler than that offered by in vivo experiments but sufficiently complex to allow us to examine the action of the toxin on the interaction among three key cellular components of the mucosal immune system: a polarized epithelial cell, a T cell, and an antigen-presenting cell. In vivo, the polarized epithelial monolayer provides an important barrier between enterotoxins released in the gut and immune cells within mucosal lymphoid tissues or within the epithelium. However, the epithelium acts not only as a passive physical barrier between gut contents and the internal milieu but is increasingly recognized as playing an active part in the immune response itself, releasing defensins, cytokines, and inflammatory mediators and participating in specific molecular interactions with underlying hemopoetic cells (reviewed in reference 9).

The objective of our study was to see if this model would identify any specific immune modulatory interactions which were activated by LT in epithelial cells and which might contribute to its in vivo adjuvant activity. The study highlighted one specific immune modulatory pathway activated by LT which has received relatively little attention but may play an important role in altering the balance of immune responses in vivo. Of course, this in vitro model has intrinsic limitations in the extent to which it replicates the in vivo environment. Further studies combining both in vivo and in vitro approaches of CT and LT will be required to fully understand the activity of these pleiotropic immune modulators.

MATERIALS AND METHODS

Reagents.

Wild-type LT, LT mutants LTK63 (Ser→Lys) and LTR72 (Ala→Arg), and the recombinant B subunit of LT were a kind gift of Mariagrazia Pizza, IRIS, Siena, Italy. They were isolated from the periplasm of recombinant E. coli and characterized as previously described (4). Lipopolysaccharide (LPS) (Salmonella enterica serovar Typhimurium), ovalbumin grade V, GM₁, indomethacin, and forskolin were all obtained from Sigma (Poole, Dorset, United Kingdom). Hen egg white lysozyme was obtained from Roche Diagnostics (formerly Boehringer Mannheim, Lewes, United Kingdom). Peptides constituting amino acids 46 to 61 of lysozyme and 323 to 339 of ovalbumin were synthesized by the Imperial Cancer Research Fund peptide synthesis facility (London, United Kingdom).

Cell culture.

CaCo-2 epithelial cells were maintained in stock culture as previously described (11). For Transwell filter insert cultures (12-mm diameter, 3-μm pore size; Costar UK Ltd., High Wycombe, Bucks, United Kingdom), cells were seeded at approximately 10⁵ cells/filter with a change of medium every two to three days. The epithelial cells were allowed to grow to confluence for 10 to 14 days, when they differentiate to form a polarized “tight” epithelium as previously described in detail (11).

T-cell hybridomas IC5.1 (recognizing hen egg lysozyme amino acids 46 to 61 on I-A^k) and DO11.10 (recognizing ovalbumin amino acids 323 to 339 on I-A^d) were cultured in RPMI 1640 medium and Dulbecco's modified Eagle's medium, respectively, supplemented with 10% fetal calf serum (Life Technologies, Paisley, United Kingdom), 2 mM glutamine, 100 U of penicillin/ml, and 100 μg of streptomycin/ml.

Toxin treatment of CaCo-2 cells and collection of supernatants.

Differentiated CaCo-2 cell monolayers grown on Transwells were used for experiments after 10 to 14 days of culture. Under these conditions, the apical microvillar membrane of CaCo-2, which corresponds to the surface of the epithelium in contact with the gut contents in vivo, was exposed only to the medium within the upper chamber of the Transwells, while the basolateral membrane was in contact with the lower chamber. LT was added to the upper culture chamber (the apical surface) and incubated for 3 h at 37°C. Control monolayers were incubated with medium only or LPS as indicated. In some experiments, indomethacin (2 μg/ml) was added to both apical and basolateral chambers for 30 min prior to incubation with LT and also during the antigen presentation assay. For experiments using GM₁ blocking, LT was incubated with a five times molar excess of GM₁ for 30 min at 37°C before being added to the cell monolayers.

After incubation with toxins, all cell monolayers were carefully washed with MEM on both sides with a 5-min interval between washes. CaCo-2 cell monolayers were then used for the coculture assays (see below) or incubated with complete medium, and 24 h later, supernatant aliquots from apical and basolateral compartments were collected, centrifuged, and stored at −20°C until required.

In some experiments, forskolin (1 or 10 μM) was added to CaCo-2 cells for 18 h at 37°C. After incubation, cell monolayers were carefully washed with minimal essential medium (MEM), and fresh complete medium was added to the cultures. Supernatant aliquots from apical and basolateral compartments were collected after 24 h, centrifuged, and kept at −20°C until required.

Antigen presentation assays. (i) Coculture assay.

After treatment with the toxin as described above, complete tissue culture medium was added to the upper (apical) compartment of the Transwells. Then, 2 × 10⁶ murine spleen cells (as a source of antigen-presenting cells, irradiated with 3,000 rads from an X-ray source to prevent proliferation), 2 × 10⁵ antigen-specific T-hybridoma cells, and antigen were added to the lower chamber. Controls included CaCo-2 cells only; CaCo-2, spleen, and T cells in the absence of antigen; and cultures in the absence of CaCo-2 cells. After 24 h, supernatants were collected and analyzed for interleukin 2 (IL-2) content.

(ii) Supernatant test.

In these experiments, antigen presentation assays were performed in the absence of the CaCo-2 monolayer. An aliquot of the supernatant from CaCo-2 cell monolayers treated as above was tested in an antigen presentation assay that was set up in 96-well flat-bottomed tissue culture plates. All supernatants were assayed in triplicate. Irradiated murine spleen cells (5 × 10⁵/well) were incubated with T-cell hybridomas (5 × 10⁴/well) in the presence of the specific antigen. After 24 h of incubation, supernatants were collected and assayed for IL-2 content.

(iii) IL-2 assay.

The IL-2 content of the culture supernatants was determined by testing a series of dilutions of culture supernatants using the IL-2-dependent cell line (CTLL) proliferation assay as described previously (11). A standard curve was obtained by incubating CTLL in the presence of recombinant IL-2 (Prepotech, London, United Kingdom). In order to exclude a direct effect of LT or a CaCo-2-secreted product on the CTLL cells, a standard curve was set up, containing supernatant derived from LT-treated and nontreated CaCo-2 cell monolayers. No difference in the proliferative response of CTLL cells was observed.

In selected experiments, IL-2 concentrations were confirmed using an IL-2 enzyme-linked immunosorbent assay (R & D, Abingdon, United Kingdom), performed according to the manufacturer's instructions.

RESULTS

Indirect inhibition of T-cell response by LT.

In our in vitro model, LT was added to the apical surface of CaCo-2 cells in order to approximate the situation in vivo where LT initially comes in contact with epithelial cells, rather than have it interacting with T cells or antigen-presenting cells. We therefore tested if the exposure of epithelial cells to LT could have any subsequent, indirect effect on antigen-specific T-cell responses.

Differentiated CaCo-2 cell monolayers were treated with LT (2 μg/ml), washed extensively to remove residual LT, and then cocultured with IC5.1 T cells, mouse spleen cells, and lysozyme or lysozyme peptide. As shown in Fig. 1A and B, the antigen-specific release of IL-2 was significantly inhibited in those cultures where CaCo-2 cells had been pretreated with LT, compared with the results for untreated cultures. Treatment of CaCo-2 cells with as little as 50 ng of LT/ml gave similar levels of inhibition, although longer incubation times (18 h) were required. The inhibitory effect was not specific to the IC5.1 hybridoma, since LT also indirectly inhibited the activation of the ovalbumin-specific T-cell hybridoma DO11.10 (data not shown).

FIG. 1.

LT induces the release of an immunosuppressive factor by CaCo-2 monolayers. The upper chamber (apical surface) of differentiated CaCo-2 monolayer cultures was replaced by medium or medium containing LT (2 μg/ml), and the cultures were incubated at 37°C for 3 h. Residual LT was then removed by repeated washing, and fresh culture medium was added to the upper chamber. (A and B) IC5.1 T cells and antigen-presenting cells were then added to the lower chamber with 200 μg of lysozyme/ml (A) and 2 μg of lysozyme peptide/ml (B) as described in Materials and Methods. IL-2 release into the lower chamber of these cocultures was measured after 24 h. (C and D) CaCo-2 cultures were incubated for a further 24 h. Supernatants from the lower chamber were then collected and subsequently tested for their ability to inhibit the release of IL-2 by IC5.1 T cells stimulated with 200 μg of lysozyme/ml (C) or 2 μg of lysozyme peptide/ml (D). Supernatants were diluted 1:2 (black columns) or 1:50 (white columns) before addition to the IC5.1 cultures. The IL-2 results are the mean of triplicate cultures ± standard error of the mean (SEM) from one representative experiment out of four. The IL-2 concentration in the absence of antigen was <2 pg/ml. IL-2 release by IC5.1 cells in the absence of CaCo-2 cells was 300 (±50) pg/ml (lysozyme) and 630 (±43) pg/ml (lysozyme peptide). ∗, P < 0.01.

The inhibitory activity of CaCo-2 cells treated with LT was mediated by a stable, soluble mediator, since supernatants from CaCo-2 cells treated with LT also resulted in significant dose-dependent inhibition when added to separate cultures of IC5.1, spleen cells, and antigen (Fig. 1C and D). Supernatants could be stored at −20°C with little loss of inhibitory activity.

The activity of LT is inhibited by GM₁.

The B-unit pentamer of LT binds directly to GM₁, and cell surface-bound GM₁ therefore acts as the receptor for LT on CaCo-2 apical cell membranes (6). To confirm that the activity of LT in this model was indeed a result of binding to the LT receptor, the toxin was preincubated with excess soluble GM₁, which can competitively block subsequent binding to the cell-associated form. As shown in Fig. 2, the inhibitory effects of LT were blocked by preincubation of the toxin with excess GM₁. In contrast, the addition of GM₁ after LT had already interacted with the CaCo-2 cells had no effect on IC5.1 activation.

FIG. 2.

GM₁ blocks the ability of LT to induce the release of an inhibitory factor by CaCo-2 cells. CaCo-2 cultures were treated for 3 h with LT and then cocultured with IC5.1, antigen-presenting cells, and lysozyme (200 μg/ml) as described. In some cultures GM₁ (in a 5:1 molar ratio) was added either together with LT (LT + GM₁), after LT was removed (LT / GM₁), or in the absence of LT (GM₁). Results show the IL-2 release (mean and SEM from triplicate cultures) by IC5.1 cells as a percentage of that in CaCo-2 cells cultured in medium only. Absolute IL-2 levels corresponding to 100% were 400 (±30) pg/ml. IL-2 concentration in the absence of antigen was less than 2 pg/ml. The figure shows the results of one of three replicate experiments. ∗, P < 0.01.

Enzymatic activity of LT is required for inhibitory activity.

In order to identify which features of LT structure were required to induce the release of inhibitory activity by CaCo-2 cells, we tested several mutant forms of LT in the assay. As shown in Fig. 3, the LT B-subunit holotoxin had no activity. Similarly, LTK63, in which serine at position 63 has been changed to lysine (which results in a total abrogation of ADP-ribosylating activity), exhibits no activity in this assay. In contrast, LTR72, in which alanine at position 72 has been mutated to arginine, retains approximately 0.6% of enzymatic activity (5) as well as the ability to induce the release of the inhibitory factor by CaCo-2 cells. LPS (1 μg/ml), at a concentration well in excess of the level that was present in the LT preparations, had no effect in this model.

FIG. 3.

Immune modulation of CaCo-2 cells by LT is related to LT ADP-ribosylating activity. LT, LT holotoxin, LTR72, LTK63 (all at 2 μg/ml), and LPS (1 μg/ml) were added to the apical surface of CaCo-2 cultures for 3 h, and supernatants were collected from the lower chamber of the Transwells as described. DO11-10 T-hybridoma cells were cocultured with antigen-presenting cells and ovalbumin peptide (1 μg/ml) in the presence of these CaCo-2 supernatants diluted 1:2 (solid bars) or 1:5 (dotted bars). Results show IL-2 release by DO11-10 (mean ± SEM for triplicate cultures) after 24 h. IL-2 release from similar DO11-10 cultures without any CaCo-2 supernatants was equivalent to that from samples containing supernatants from CaCo-2 cultured in the absence of LT (Medium). IL-2 release in the absence of antigen was less than 2 pg/ml. The figure shows the results of one of three replicate experiments. ∗, P < 0.025.

These results suggest that the enzymatic activity of LT was essential for its biological function in this assay. Since a major downstream effect of ADP-ribosylation by LT is the activation of adenylate cyclase activity (7) and a consequent rise in intracellular cAMP concentration, we also tested if the inhibitory activity would be released in response to direct up-regulation of intracellular cAMP. As shown in Fig. 4, treatment of CaCo-2 monolayers with forskolin, an agent which increases adenylate cyclase activity directly, also induced the release of a soluble factor which inhibited the activation of DO11.10 cells.

FIG. 4.

Forskolin induces the release of an inhibitory factor from CaCo-2 cultures. CaCo-2 cells were cultured in the presence or absence of forskolin (1 or 10 μM), and supernatants were collected as described. DO11-10 T-hybridoma cells were cocultured with antigen-presenting cells and ovalbumin peptide (1 μg/ml) in the presence of these CaCo-2 supernatants diluted 1:2 (solid bars) or 1:50 (dotted bars). The results show IL-2 release by DO11-10 (mean ± SEM for triplicate cultures) after 24 h. IL-2 release from DO11-10 cells stimulated in the absence of CaCo-2 supernatants was equivalent to that from cells containing supernatants from CaCo-2 cultured in the absence of LT (column 0). IL-2 release in the absence of antigen was less than 2 pg/ml. The figure shows the results of one of two replicate experiments. ∗, P < 0.01.

Indomethacin abrogates the effects of LT.

One family of well-known inhibitors of IL-2 release by T cells is prostaglandins (12, 14), and the release of prostaglandins is also known to be triggered by elevated intracellular cAMP. To determine whether the effect of LT was associated with the secretion of prostaglandins, indomethacin, a highly selective inhibitor of cycloxygenase 1 and 2 activity (15) and hence prostaglandin release, was added to CaCo-2 cell monolayers during incubation with LT. As shown in Fig. 5, indomethacin completely blocked the ability of LT to stimulate the release of inhibitory activity by CaCo-2 cells.

FIG. 5.

Indomethacin inhibits the release (induced by LT) of the immune modulating factor from CaCo-2 cells. CaCo-2 cells were cultured in the presence of indomethacin (2 μg/ml) for 30 min, and then LT (2 μg/ml) was added for a further 3 h as described. Both indomethacin and LT were removed by extensive washing, and the CaCo-2 cells were then cocultured with IC5.1 T-hybridoma cells, antigen-presenting cells, and either intact lysozyme (200 μg/ml) (A) or lysozyme peptide 46 to 61 (2 μg/ml) (B). Results show IL-2 release (mean and SEM from triplicate cultures) by IC5.1 cells as a percentage of that in CaCo-2 cells cultured in medium alone. Absolute IL-2 levels corresponding to 100% were 420 (±20) pg/ml for lysozyme, and 630 (±43) pg/ml for lysozyme peptide. The IL-2 concentration in the absence of antigen was less than 2 pg/ml. The figure shows the results of one of three replicate experiments.

DISCUSSION

This study was undertaken to see if it was possible to identify any major regulatory interactions among epithelial cells, T cells, and antigen-presenting cells in response to LT and to see whether such interactions might clarify the mode of action of LT as an adjuvant of mucosal immunity. The model exhibits both the advantages and inherent disadvantages of in vitro models in general. Thus, it simplifies the experimental system by identifying three cellular components and isolating them from the other cellular components found in vivo. In this way, it is easier to identify single molecular interactions, but other equally important interactions may of course be lost. The specific in vitro model also retains the vectorial nature of the epithelial barrier by using CaCo-2 cells, which, although from a transformed cell line, retain the ability to differentiate in vitro into a polarized “tight” epithelial monolayer. In vivo as in vitro, adjuvant must cross this epithelial barrier before it can gain direct access to the antigen-presenting cells and T cells, which initiate mucosal immune responses.

This model unexpectedly demonstrated that LT induces the release of a factor(s) by the CaCo-2 cells which was inhibitory for antigen-dependent T-cell activation (as measured here by the release of IL-2). Inhibition was equally apparent for responses to intact protein antigen and antigen peptide, suggesting that the target of the factor was the antigen presentation and T-cell activation step rather than the ability of the antigen-presenting cell to process the antigen for subsequent presentation.

The possibility that the inhibitory effect was due to the presence of some contaminant within the toxin preparations, especially the presence of low amounts of LPS, was addressed. Direct measurement showed negligible amounts of LPS in the preparations (1 ng/ml or less); in contrast, 1 μg of LPS/ml, when added directly to the CaCo-2 cells, did not induce the release of any inhibitory factor. Furthermore, the LT activity was completely blocked by preincubation with excess GM₁, which binds the B-subunit pentamer of LT and prevents the toxin from interacting with its cellular receptor on CaCo-2 cells. GM₁ was, however, unable to suppress the inhibitory activity if added after LT had already bound to CaCo-2 cells. Thus we ruled out the possibility that inhibition was due to residual LT remaining within the CaCo-2 monolayers after washing and then acting directly on the T cells or antigen-presenting cells themselves. This finding was important, since a direct effect of CT (and CT B holotoxin) on both T cells (3, 14, 19) and antigen-presenting cells (1) has been described previously. The third piece of evidence arguing for a specific activity of LT was the comparison between LT mutants: although these were all isolated and purified in a very similar way from recombinant bacteria, only some forms were able to induce the release of inhibitory activity.

The comparison of the LT mutants also suggested that the ability to induce inhibitory activity was absolutely dependent on retaining some ADP-ribosylating activity, since both the B subunit alone and LTK63 were completely without biological activity. However, LTR72, which retains about 0.6% of the wild-type enzyme activity (5), retained its ability to induce the release of inhibitor in this model. These results further suggest that the mode of action of LT in inducing the release of the inhibitory molecules is via ADP-ribosylation of G proteins and a consequent increase in intracellular cAMP. This hypothesis was confirmed by showing a parallel release of inhibitory activity by CaCo-2 cells in response to forskolin, a pharmacological agent which activates adenylate cyclase and increases cAMP concentrations directly.

Increased cAMP levels can activate cycloxygenase enzymes and hence stimulate release of prostaglandins in many cell types. Indeed, there is one report of prostaglandin E₂ release in ileal loop explant cultures in response to CT (15), although the cell type responsible for the prostaglandin synthesis in this system was not identified. Since prostaglandins are known to block IL-2 production by T cells, it seemed a reasonable prediction that the biological effect documented in this study was also mediated by prostaglandin release; this prediction was confirmed by the reversal of inhibitory activity by pretreatment of CaCo-2 cells with indomethacin, a potent inhibitor of cycloxygenases.

It is at first sight difficult to reconcile the ability of LT to induce the release of an inhibitory factor from CaCo-2 cells with the known adjuvant activity of LT as seen in vivo. Indeed, other factors must be in operation in vivo, since both the LTK63 mutant and, to a lesser extent, even the B subunit holotoxin have adjuvant activity in vivo but have no ability to induce the release of prostaglandins in vitro. Nevertheless, prostaglandins are immunomodulatory, as well as immunoinhibitory, mediators—in particular, they have been shown to selectively block the release of IL-2 and gamma interferon (TH1 cytokines), while not affecting the release of IL-4 (a TH2 cytokine) (12, 14). They may also regulate the phenotype of antigen-presenting dendritic cells to induce T cells to produce TH2 cytokines in preference to TH1 cytokines (10). In the context of an adjuvant, therefore, the release of prostaglandins could play a role in the bias of the immune response away from TH1- towards TH2-type cytokines, which has been documented in a number of systems following pretreatment with LT or CT (13, 18).

In conclusion, this study identifies one particular molecular interaction resulting from LT action on epithelial cells, which may play a role in the overall complex immunoregulatory action of this molecule. Further extension of the in vitro model, for example, using T cells capable of secreting both TH1 and TH2 cytokines or using immature or mature dendritic cells as antigen-presenting cells, may help to identify other important molecular connections. Ultimately, the biological significance of any such molecular pathways identified in such simplified in vitro models will have to be evaluated within the context of the complicated interactions which occur during mucosal immunization in vivo.