Proteolytic Activation of Cholera Toxin and Escherichia coli Labile Toxin by Entry into Host Epithelial Cells

Cholera and Escherichia coliheat-labile toxins (CT and LT) require proteolysis of a peptide loop connecting two major domains of their enzymatic A subunits for maximal activity (termed “nicking”). To test whether host intestinal epithelial cells may supply the necessary protease, recombinant rCT and rLT and a protease-resistant mutant CTR192H were prepared. Toxin action was assessed as a Cl− secretory response (Isc) elicited from monolayers of polarized human epithelial T84 cells. When applied to apical cell surfaces, wild type toxins elicited a brisk increase in Isc (80 μA/cm2). Isc was reduced 2-fold, however, when toxins were applied to basolateral membranes. Pretreatment of wild type toxins with trypsin in vitro restored the “basolateral” secretory responses to “apical” levels. Toxin entry into T84 cells via apical but not basolateral membranes led to nicking of the A subunit by a serine-type protease. T84 cells, however, did not nick CTR192H, and the secretory response elicited by CTR192H remained attenuated even when applied to apical membranes. Thus, T84 cells express a serine-type protease(s) fully sufficient for activating the A subunits of CT and LT. The protease, however, is only accessible for activation when the toxin enters the cell via the apical membrane.

Over the last several years, we have used the human intestinal T84 cell line to examine the cell biology of Vibrio cholerae and Escherichia coli heat-labile toxins on polarized epithelial cells. Both toxins (cholera toxin, CT 1 , and labile toxin, LT) are structurally, immunologically, and functionally nearly identi-cal (1)(2)(3). They account for the massive secretory diarrhea seen in infections caused by these microorganisms (4 -6).
CT and LT consist of five identical B subunits that bind ganglioside G M1 at the cell surface and a single A subunit comprised of two functional domains termed the A 1 and A 2 peptides (1,2). The A 1 peptide exhibits the toxin's ADP-ribosyltransferase activity, which is necessary for signal transduction. The A 2 peptide tethers the A and B subunits together and contains the endoplasmic reticulum-targeting motif K(R)DEL at its COOH terminus. Enzymatic activity of the A subunit is latent. For full ADP-ribosyltransferase activity, the peptide loop connecting the A 1 and A 2 peptides must be proteolytically cleaved at residue Arg-192 (7). After proteolytic cleavage (termed "nicking"), the A 1 and A 2 peptides remain covalently associated via a single disulfide bond. This bond is likely to be reduced when the A 1 peptide translocates across the membrane to the cytosolic membrane surface (8). Translocation is necessary for the A 1 peptide to gain access to its substrate the heterotrimeric GTPase G s␣ (9). Toxin-induced ADP-ribosylation of G s␣ activates adenylate cyclase and raises intracellular cAMP levels, which in intestinal crypt epithelial cells elicits a Cl Ϫ secretory response, the primary transport event responsible for secretory diarrhea (10).
In nature, both toxins make initial contact with the intestinal cell apical membrane but must gain access to adenylate cyclase on the cytoplasmic surface of the basolateral membrane. This process takes 30 -40 min. We have obtained evidence that this "lag phase" corresponds to the time required for CT to enter the cell via apical endosomes and move to its site of action on the basolateral membrane by transcytosis (11). The intracellular site where the A subunit translocates across the membrane, however, remains undefined, but available evidence indicates that this may occur as the toxin-G M1 complex (or toxin alone) moves retrograde through Golgi cisternae or into the endoplasmic reticulum (12)(13)(14)(15).
Several un-nicked bacterial toxins including anthrax protective antigen, diphtheria toxin, Pseudomonas enterotoxin, and possibly shiga toxin, require activation by endogenous proteases of the host cell (in these cases furin) to elicit a biologic response (16). These toxins also require endocytosis for biologic activity and may encounter furin within endosomes, trans-Golgi, or possibly at the membrane surface of the host cell. As many Vibrios secrete their own proteases (17) and both CT and LT act within the gastrointestinal tract, it has largely been assumed that V. cholerae activates its own toxin, and proteases in the gut lumen activate LT. However, the very first events in the pathogenesis of diarrhea due to both V. cholerae and E. coli likely entail bacterial adhesion to the enterocyte surface, as evidenced by the nature of identified invasion factors (type IV pilus, surface glycoproteins, and inner membrane regulatory proteins such as ToxR, (18,19)). Thus, in vivo, both CT and LT may bind to the intestinal cell apical membrane immediately after release from the microbe.
Our aim in the present study was to examine whether the enterocyte itself may proteolytically activate the nascent A subunits of cholera or E. coli heat-labile toxins. As before, we utilized the human intestinal cell line T84 to model the interaction between toxin and intestinal epithelial cell. Our results show that a serine protease(s) endogenous to the apical membrane or apical endocytic compartment of T84 cells is sufficient to activate fully nascent CT or LT. Proteolytic activation, however, is not apparent when toxin enters the cell via the basolateral membrane, and this is rate-limiting for signal transduction. These data provide evidence that in vivo epithelial cells of the intestine (the physiologic target cell of CT in nature) may activate the nascent A subunits of CT or LT after toxin entry into apical endosomes or after binding apical receptors.
For electrophysiological studies, confluent monolayers on Transwell inserts were transferred to HBSS and preincubated with apical or basolateral CT at 4°C for 30 min prior to shifting to fresh HBSS at 37°C. Measurements of short circuit current and resistance were performed with 0.33-cm 2 monolayers, and biochemical studies were performed with 5-cm 2 monolayers as described previously (20 -22).
Preparation of Recombinant Wild Type Toxins (rCT and rLT)-E. coli XL1-Blue [recA 1 lac Ϫ endA 1 gyrA96 thi hsdR17 supE44 relA1 {FЈ proAB lacIq lacZ⌬M15 Tn10}] (23) or TX1 harboring plasmids encoding wild type (wt) CT or LT were inoculated from stocks into 200 ml of Luria-Bertani medium (L-broth) (24) supplemented with 200 g/ml ampicillin and 10 g/ml tetracycline. The cells were grown at 37°C on a rotary shaker. When the culture had reached an optical density of 0.2 (600 nm), isopropyl-1-thio-␤-D-galactopyranoside (0.5 mM final) was added, and the cells were cultured for an additional 2 h. The cells were then collected by centrifugation (6,000 rpm, 4°C for 15 min) and washed in 4 ml of ice-cold phosphate-buffered saline (140 mM NaCl, 10 mM Na 2 HPO 3 /NaH 2 PO 3 , pH 7.2) and then resuspended in 8 ml of ice-cold 0.3 M sucrose, 0.1 M Na 2 HPO 3 /NaH 2 PO 3 , pH 7.6. EDTA (5 mM) and lysozyme (10 g/ml) were added and the cells kept on ice for an additional 15 min with occasional agitation. The treated cells were centrifuged (12,000 rpm at 4°C for 5 min), and the supernatant containing the periplasmic extract was removed, dialyzed overnight against two changes of HBSS (1:1,000 v/v), separated into 20 -50-l aliquots, flash frozen, and stored at Ϫ80°C until required. Periplasmic extracts prepared from control E. coli strain XL1-Blue not harboring toxin clones or harboring clones containing wt B subunits alone were electrically silent when applied to T84 cell monolayers (i.e. they did not elicit a secretory response or affect monolayer resistance). Periplasmic extracts of wt rCT and rLT were characterized in a previous publication (25).
Preparation of Recombinant Protease-resistant CT Variant (CTR192H)-A plasmid encoding CTR192H (pMGJ6705) was made by oligonucleotide-directed mutation of a wt CT clone (pMGJ67) and will be described elsewhere. 2 CT holotoxin made by this clone contains A190G and R192H substitutions in CT A subunit. rCTR192H was made by overnight induction of 500-ml mid-log phase cultures (A 600 nm ϭ 0.6 -0.8) of E. coli TX1 carrying the recombinant plasmid, grown in Luria broth at 30°C using 200 M isopropyl-1-thio-␤-D-galactopyranoside. Cells were harvested by centrifugation (6,000 ϫ g, 10 min at 4°C), washed in 10 ml of TEN (50 mM Tris, pH 7.4, 1 mM EDTA, 0.2 M NaCl), resuspended in 25 ml of TEN containing 1 mg/ml polymyxin B, and incubated at 37°C for 10 min. This extract was applied to a galactose affinity resin (Pierce) at room temperature, washed with 20 ml of TEN, and eluted with 1 M galactose in TEN (26). Pooled fractions were dialyzed extensively against HBSS and stored at Ϫ80°C. Analysis of protein content in these pooled fractions by SDS-PAGE (reducing conditions) showed single protein bands at Ϸ 28 and 11 kDa (corresponding to A and B subunits) when stained by Coomassie Blue.
Assessment of Recombinant Toxin Concentrations-To assay toxin concentrations an enzyme-linked immunosorbent assay was used. Periplasmic extracts containing recombinant toxins were applied to 96-well microtiter plates coated with ganglioside G M1 as described previously (27). After 30 or 60 min at 37°C, the plates were washed with phosphate-buffered saline, pH 7.4, and the presence of CT or LT bound to G M1 was assayed by routine techniques using rabbit polyclonal antiserum raised against either CT A subunit or CT B subunit (1:500 in phosphate-buffered saline) (11, 12) (both cross-react with LT subunits), or mouse monoclonal against LT B subunit (28). Apparent toxin concentrations were confirmed by SDS-PAGE and Western blot using serial dilutions of CT or LT B subunit as standards.
Trypsin Treatment of Recombinant CT and LT in Vitro-Periplasmic extracts (1 ml final volume) containing recombinant CT or LT (15 nmol) were incubated with 0.2-2 mg/ml trypsin at 37°C for 30 min as modified from methods described previously (7,29,30). The reaction was stopped by adding 200 mg/ml soybean trypsin inhibitor at 4°C. Nicking was assessed structurally by SDS-PAGE and Western blot and functionally by increase in efficiency of toxin-induced Cl Ϫ secretion after applying the nicked toxin to basolateral reservoirs of T84 cell monolayers. Proteolytic damage to the recombinant toxin was assessed functionally as a reduction in potency of toxin-induced Cl Ϫ secretion after applying the nicked toxin to apical reservoirs of T84 cell monolayers.
Proteolytic Nicking of rCT and CTR192H by T84 Cells in Situ-Monolayers were incubated with apical or basolateral rCT or CTR192H (20 nM) at 4°C for 15 min before transferring to fresh buffer at 37°C or at 4°C for the indicated times (up to 180 min). Nicking of the CT A subunit was assessed as a 5-kDa shift in molecular mass after immunoprecipitation, SDS-PAGE, and Western blot as described below.
To identify the "class" of protease likely responsible for nicking, we utilized inhibitors of the proteolytic reactions catalyzed as described in (31). Monolayers were preincubated with the specified protease inhibitors or buffer alone (containing the carrier dimethyl sulfoxide, Me 2 SO) at 4°C for 30 min before adding 20 nM rCT or CTR192H. The following protease inhibitors were used to determine "functionally" the class of protease involved (based on reaction catalyzed). Serine-type peptidases were diisopropyl fluorophosphate (DFP, 2.5 M), phenylmethylsulfonyl fluoride (PMSF, 175 M from 35 mM stock in ethanol), and 3,4-dichloroisocoumarin (3,4-DCI, 1 mM from 100 mM stock in Me 2 SO). Cysteine-type peptidases were leupeptin (1 M, from 1 mM stock in water) and trans-epoxysuccinyl-L-leucylamido(4-guanidino)butane (E-64, 10 M from 100 mM stock in water). Metallopeptidases were EDTA (5 mM), and 1,10-phenanthroline (1 mM from 100 mM stock in Me 2 SO). The aspartic-type peptidase was pepstatin (1 mM from 100 mM in Me 2 SO).
Immunoprecipitation, SDS-PAGE, and Western Blotting-Monolayers were incubated with apical or basolateral rCT or CTR192H (20 nM) at 4°C for 15 min before transferring to 37°C for the indicated times (to allow endocytosis and entry of toxin into the cell). Control monolayers were kept at 4°C for the duration of the experiment (90 or 120 min). Little or no endocytosis occurs at 4°C. After incubations at 37°C were complete, further cellular processing of internalized toxins was quenched by returning the monolayers to 4°C. Individual monolayers were then removed intact on their filter supports, immersed in 0.6 ml of 0.5% SDS, 20 mM Tris, 150 mM NaCl, 5 mM EDTA, 20 mM triethanolamine, 0.18 mM PMSF, and 20 g/ml chymostatin, and heated at 100°C for 3 min. The cell lysate was diluted with Triton X-100 to form a mixed micelle buffer (0.25% SDS, 2% Triton X-100, final). DNA was sheared by vortex at 4°C for 30 min and removed by centrifugation (12,000 ϫ g for 10 min) in the presence of Sepharose CL-4 (Pharmacia Biotech Inc.) until clear. Cell lysates were precleared again by a 30-min incubation with protein A-Sepharose (Pierce). This procedure solubilizes all cellular proteins and proteins associated with the cell monolayer including the entire fraction of toxin bound to the cell surface or internalized via endocytosis (11,12,25).
After cell lysis and solubilization of all cellular and cell-associated proteins, CT A and B subunits were immunoprecipitated using polyclonal antibodies raised against denatured toxin subunits eluted from 2 M. G. Jobling and R. K. Holmes, submitted for publication. SDS-PAGE gels (11). Antisera against toxin subunits were first covalently coupled to protein A beads using dimethyl pimelimidate (20 mM final). Immune complexes formed overnight at 4°C were recovered by centrifugation (3,000 ϫ g, for 5 min) and washed five times with fresh mixed micelle buffer and finally once with buffer containing the same salts and 8% sucrose. Immunoprecipitated proteins were released from the beads by heating in 100 l of Laemmli sample buffer at 100°C for 3 min and analyzed by SDS-PAGE on 10 -20% acrylamide gradient gels (reducing conditions), transfer to 0.25-m nitrocellulose, and Western blotting with polyclonal antibody raised against CT B or A subunits (1:3,000 dilution serum). Blots were developed using affinity-purified goat anti-rabbit horseradish peroxidase-conjugated secondary antibody (1:2,500, Sigma) and enhanced chemiluminescence (Amersham). Immunoblots were initially blocked with 5% nonfat dry milk in 50 mM Tris, 200 mM NaCl, pH 7.4.

RESULTS
Proteolytic Activation of rLT and rCT by Host Epithelia: Role of Cell Polarity-In recent studies utilizing recombinant CT and LT (25), we noticed that the time courses of Cl Ϫ secretion elicited by basolaterally applied toxins (20 nM) were slower and significantly less robust than we had previously observed with toxin preparations purified from V. cholerae supernatants. In contrast, when these same recombinant toxins were applied to apical membranes, there was little or no delay in the anticipated time course of the secretory response. In all past experiments (11,12,21,25,32), we consistently found that toxin preparations purified from V. cholerae supernatants elicited Cl Ϫ secretion with a shorter lag phase and more rapid rate of onset when applied to basolateral rather than apical surfaces of the same cells (Fig. 1, panel A). Unmistakably, the faster rate of signal transduction elicited by recombinant toxins (rCT or rLT) entering the cell through the apical endosome was altogether opposite to that observed previously in earlier studies (Fig. 1, panels B and C). As preparations of native CT purified from V. cholerae supernatants are proteolytically nicked and preparations of recombinant toxins expressed in E. coli XL1-Blue are not, we hypothesized that when applied to basolateral surfaces of T84 cell monolayers the attenuated time courses of signal transduction elicited by recombinant rCT or rLT were due to the absence of or a delay in A subunit activation. In vitro, the enzymatic activities of both CT and LT are enhanced 10fold by proteolytic cleavage (activation) of the A subunit (7).
To test this idea, we proteolytically nicked recombinant rCT and rLT in vitro by pretreatment with low dose trypsin. Pretreatment with trypsin resulted in nicking of Ͼ20% of total A subunit (as assessed by SDS-PAGE) without attenuating ADPribosyltransferase activity (assessed by the ability of nicked toxins to elicit a secretory response when applied to apical surfaces of T84 cells). Fig. 2 shows the time course of Cl Ϫ secretion elicited by recombinant rCT (20 nM)and rLT (40 nM), nicked or not nicked by pretreatment with trypsin in vitro. When applied apically, toxins not pretreated with trypsin elicited a brisk secretory response from T84 cells (upper panels). When applied to basolateral membranes, however, the secretory responses elicited by these recombinant toxins were clearly attenuated (lower panels). Signal transduction by toxin entering the cell through the basolateral membrane displayed a longer lag phase and slower rate of onset compared with the time courses of Cl Ϫ secretion elicited by the same toxin preparations applied apically. Pretreatment of rCT or rLT with trypsin in vitro, however, resulted in a faster time course of signal transduction elicited by the basolaterally applied toxins, presumably due to nicking of the rA subunits (lower panels, ϩTrypsin). Trypsin had no effect on the time course of apically applied toxins (upper panels, ϩTrypsin). These data (summarized in Table I) show that proteolytic activation of CT or LT is rate-limiting for signal transduction when toxin enters the cell via the basolateral (but not apical) cell surface. As such, these results suggest that entry of nascent CT or LT into the apical endosome leads to cleavage of the A subunit at position Arg-192, but this does not occur when toxin enters the cell via the basolateral endocytic pathway.
Nicking of CT A subunits Occurs during Toxin Entry via the Apical but Not Basolateral Cell Surface-To examine whether T84 cells could activate nascent CT by proteolytic nicking of the A subunit, we again utilized recombinant toxins. rCT was applied to apical (Fig. 3, upper panel) or basolateral (Fig. 3, lower panel) reservoirs of T84 monolayers at 4°C and then shifted to 37°C or kept at 4°C for the indicated times. The entire fraction of toxin associated with the monolayers was then solubilized by lysing the monolayers in 0.5% SDS at 100°C as described under "Experimental Procedures." Toxin subunits were concentrated by immunoprecipitation, and nick- ing was assessed as a shift in apparent molecular mass of the A subunit from 28 to 22 kDa by SDS-PAGE and Western blot. Fig. 3 shows the results of these studies. The first lane from the left in the upper panel shows that when toxin was applied apically, incubations at 4°C led to proteolytic nicking of the A subunit. At 37°C (third through seventh lanes in the upper panel), nicking was apparent by 15 min and progressed steadily over time. When applied to basolateral cell surfaces, however, nicking of the A subunit was not apparent even after a 2-h incubation (lower panel). Proteolysis of as little as 0.5% of total toxin bound to T84 cell monolayers can be detected by Western blot (assessed by serial dilution's of toxin bound at 4°C). Thus, toxin entry into the cell via apical endosomes resulted in cleavage of the A subunit into A 1 and A 2 peptides. Nicking was time-dependent and appeared able to occur at the cell surface as evidenced by proteolysis of toxin bound to apical cell surface receptors at 4°C.
To determine the class of protease responsible for this cleavage (as defined by reaction catalyzed; 31), we preincubated T84 cell monolayers with serine, cysteine, aspartic, and metalloprotease inhibitors before applying recombinant CT. As shown in Fig. 4, A and B

TABLE I Effect of trypsin treatment on kinetics of Cl Ϫ secretion elicited by recombinant LT or CT entering the cell via apical or basolateral cell surfaces
The Cl Ϫ secretory response was assessed as a short circuit current (Isc). In vitro activation of either recombinant toxin by trypsin treatment resulted in a significant acceleration of signal transduction when CT or LT entered the cell from the basolateral but not apical membrane. Mean Ϯ S.E. N represents the number of independent experiments with each experiment comprised of two independent samples/group. For wt rLT and nicked rLT applied apically, calculated means were based on N ϭ 3 and N ϭ 2, respectively. For wt rLT and nicked rLT applied basolaterally, calculated means were based on N ϭ 5 and N ϭ 3, respectively.

Toxin
ϪTrypsin ϩTrypsin nanthroline (1 mM) or by the cysteine peptidase inhibitor leupeptin (1 M). The aspartic-type protease inhibitor pepstatin (1 g/ml) and the cysteine peptidase inhibitor E-64 (1 mM) displayed incomplete inhibition. Both pepstatin and E-64 may interfere with reactions catalyzed by proteases of other classes, notably those that may exhibit significant thiol dependence or those maximally active at low pH and dependent on carboxyl groups (31).
When taken together, these data provide evidence that the nascent A subunits of CT and LT are likely activated by a serine protease(s) endogenous to the apical membrane or apical endosome, or both, of human intestinal T84 cells. In contrast, entry into T84 cells via the basolateral endocytic pathway did not result in detectable nicking of the A subunit (Fig. 3, lower  panel). Nevertheless, recombinant toxins applied to basolateral cell surfaces elicited an attenuated but clearly detectable secretory response (Figs. 1 and 2).
Protease-resistant CT Variant-To confirm our interpretation of these results, we prepared recombinant rCT replacing Arg-192 with histidine (CTR192H). This mutation inactivates the trypsin nicking site connecting the A 1 and A 2 peptides of the A subunit. Fig. 5 shows that CTR192H was not nicked by T84 cells, even after 2-h incubation at 37°C. CTR192H, however, was still able to elicit a secretory response (Fig. 6). When applied basolaterally, the secretory response elicited by CTR192H was similar to that of wt but not nicked rCT (Fig. 6, lower panel, and Table II). In contrast, when applied to apical cell surfaces, the time course of Cl Ϫ secretion elicited by wt rCT was accelerated (presumably due to nicking), but the response elicited by mutant CTR192H remained dramatically attenuated (Fig. 6, upper panel, and Table II). These data confirm our findings that intestinal T84 epithelial cells proteolytically activate nascent rCT when the toxin enters the cell through the physiologically relevant apical membrane. DISCUSSION The results of these studies show that nascent rCT (and presumably rLT) can be proteolytically activated by a serine protease(s) endogenous to the apical membrane or apical endosome (or both) of human intestinal T84 cells. In contrast, toxin entry into the cell via basolateral endosomes did not result in nicking of either rCT or rLT A subunit, and this appears to be rate-limiting for signal transduction. These data provide evidence that in vivo, after colonization of the intestine by V. cholerae or E. coli, the process of toxin binding and entry into the host intestinal cell via the physiologically relevant apical membrane may be fully sufficient for activation of the nascent enterotoxins secreted by these microbes.
Although full activity of both CT and LT requires proteolytic cleavage of the exposed loop connecting the A 1 and A 2 fragments of their respective A subunits (residues 187-199), the amino acid sequences of these functionally conserved regions lack any real sequence homology, with the exception of Arg at position 192 (1,33). The importance of this residue in toxin action is confirmed by replacing Arg-192 with His in recombinant toxin CTR192H. Neither trypsin treatment in vitro nor proteases endogenous to T84 cells in situ were able to cleave the A subunit of CTR192H into A 1 and A 2 fragments, and this was correlated with a clear decrease in the ability of mutant CTR192H to induce Cl Ϫ secretion. Similar results confirming the importance of Arg-192 in LT were obtained by replacing Arg-192 with Gly as described recently by Grant et al. (34) and Dickinson and Clements (35).
The functionally defined T84 cell protease(s) responsible for cleavage of the A subunit displayed complete sensitivity to both DFP and 3,4-DCI and likely represents a membrane-associated serine protease (31). E-64 (1 mM) also inhibited nicking, though not completely. The related cysteine peptidase inhibitor leupeptin, however, did not. The aspartic-type protease inhibitor pepstatin appeared to have an effect on nicking. As such, these data provide some evidence that the apically located serinetype protease(s) responsible for toxin activation may exhibit significant thiol or pH dependence, or both (such as that found for the carboxypeptidase C family, prolyl oligopeptidase, and a subset of the subtilisin family) (31).
The molecular identity of the protease responsible for nicking the CT A subunit remains unknown. Although intestinal epithelia in vivo express at least two endoproteases and numerous peptidases on their lumenal membranes (36), the activities of cell surface or endosomal proteases in human T84 cells have not been systematically examined. T84 cells do express the serine protease tissue-kallikrein (37,38), together with a novel serpin protease inhibitor kalistatin (39). These data raise the possibility that T84 monolayers may activate nascent CT and LT by utilizing tissue-kallikrein. T84 cells also express furin, as evidenced by the ability of T84 cells to activate, process, and respond to edema factor/protective antigen of nascent anthrax toxin. 3 Nascent anthrax toxin must be cleaved by furin to elicit a response from intact cells. However, the expression of furin by T84 cells cannot explain our results as neither CT nor LT contains the RXXR amino acid motif required for furin cleavage (16). Whatever the molecular identity, this functionally defined protease must display a remarkable degree of apical polarization as little or no activity can be detected when CT or LT enters the cell from basolateral membranes.
The results of these studies also show that proteolytic nicking of the A subunit may not be necessary for toxin action, as both CTR192H and wt but un-nicked rCT (i.e. basolaterally applied) elicit a Cl Ϫ secretory response from T84 monolayers. Similar findings were reported by Grant and co-workers for a nearly identical mutation in LT tested on Chinese hamster ovary and nonpolarized Caco-2 cells (34). It has also been well described that the intact A subunit displays clear (though attenuated) enzymatic activity in vitro (7). Nevertheless, it remains possible that after entry into basolateral or apical endosomes, both wt and mutant CTR192H may be nicked at position 192 or at an alternative site(s), in very small amounts not detectable by Western blot, by the same or another protease. When taken together, however, these data provide evidence that in vivo protease-resistant variants such as CTR192H may continue to display the ability to elicit secretory diarrhea though with attenuated potency.
Since the A subunits of both CT and LT maintain extensive and stable noncovalent interactions with the B pentamer (2, 7), the ability of un-nicked CTR192H to elicit a Cl Ϫ secretory response raises the distinct possibility that domain A 1 of CTR192H may translocate across the membrane and exhibit enzymatic activity in the cytosol while tethered via the A 2 domain to the B subunit on the contralateral (lumenal) membrane surface. If this is correct, it is also possible that the A 1 peptide of nicked CT or LT may not fully dissociate from the A 2 peptide and B subunit after entry into the cell. If so, this would fit nicely with our previous observations that in polarized cells signal transduction by CT is not diffusion-limited even after translocation of the A subunit (21) and that both the pentameric B subunit and translocated A subunit appear to travel together across the cell en route to the basolateral membrane (11). Alternatively, it remains possible that the A 1 peptide of both mutant and wt toxin may dissociate from the A 2 peptide/B subunit-G M1 complex after a small (but not measurable) amount of nicking by the same or another protease as discussed above or that proteolytic cleavage of the A subunit is not required for complete dissociation of A subunit from the B pentamer. In support of this possibility, the entire CT A subunit including the A 2 peptide appears to separate from the pentameric B subunit after toxin entry into Vero cells (13,14).
In summary, these studies show that in nature the host intestinal cell may engage V. cholerae or E. coli in a form of molecular interaction by supplying the necessary protease to 3 W. I. Lencer and R. J. Collier, unpublished results.  6. Time courses of Cl ؊ secretion elicited by wt rCT or CTR192H (1 nM) applied to apical or basolateral surfaces of T84 cell monolayers. Concentrations of 1 nM were used in these studies as they more closely approximate ED 50 , and bioactivities of rCT were thus more sensitive to proteolytic activation. activate the nascent A subunits of cholera or E. coli heat-labile toxins. As modeled by the T84 cell system, a protease(s) endogenous to the enterocyte apical membrane, or endosome, or both is fully sufficient to nick the A subunits of CT and LT after they are released from the microbe and bind to the cell surface. Soluble proteases produced by V. cholerae or endogenous to the gut lumen may have little or no effect on the activation of these enterotoxins in vivo.