Shiga Toxins 1 and 2 Translocate Differently across Polarized Intestinal Epithelial Cells

Transcription

1 INFECTION AND IMMUNITY, Dec. 1999, p Vol. 67, No /99/$ Copyright 1999, American Society for Microbiology. All Rights Reserved. Shiga Toxins 1 and 2 Translocate Differently across Polarized Intestinal Epithelial Cells BRYAN P. HURLEY, 1 MARY JACEWICZ, 1 C. M. THORPE, 1 LISA L. LINCICOME, 1 ANDREW J. KING, 2 GERALD T. KEUSCH, 1 AND DAVID W. K. ACHESON 1 * Division of Geographic Medicine and Infectious Disease 1 and Division of Nephrology, 2 New England Medical Center, Boston, Massachusetts Received 27 May 1999/Returned for modification 14 July 1999/Accepted 17 September 1999 Shiga toxin-producing Escherichia coli (STEC) is an important food-borne pathogen that causes hemolyticuremic syndrome. Following ingestion, STEC cells colonize the intestine and produce Shiga toxins (Stx), which appear to translocate across the intestinal epithelium and subsequently reach sensitive endothelial cell beds. STEC cells produce one or both of two major toxins, Stx1 and Stx2. Stx2-producing STEC is more often associated with disease for reasons as yet undetermined. In this study, we used polarized intestinal epithelial cells grown on permeable filters as a model to compare Stx1 and Stx2 movement across the intestinal epithelium. We have previously shown that biologically active Stx1 is able to translocate across cell monolayers in an energy-dependent, saturable manner. This study demonstrates that biologically active Stx2 is also capable of movement across the epithelium without affecting barrier function, but significantly less Stx2 crossed monolayers than Stx1. Chilling the monolayers to 4 C reduced the amount of Stx1 and Stx2 movement by 200-fold and 20-fold respectively. Stx1 movement was clearly directional, favoring an apical-to-basolateral translocation, whereas Stx2 movement was not. Colchicine reduced Stx1, but not Stx2, translocation. Monensin reduced the translocation of both toxins, but the effect was more pronounced with Stx1. Brefeldin A had no effect on either toxin. Excess unlabeled Stx1 blocks the movement of 125 I-Stx1. Excess Stx2 failed to have any effect on Stx1 movement. Our data suggests that, despite the many common physical and biochemical properties of the two toxins, they appear to be crossing the epithelial cell barrier by different pathways. There are several different types of Escherichia coli that cause disease in humans: enteroaggregative, enteropathogenic, enterotoxigenic, and enteroinvasive E. coli, as well as Shiga toxin-producing E. coli (STEC). STEC is also known as verotoxic E. coli or enterohemorrhagic E. coli (24). Of the various types of pathogenic E. coli, STEC causes the greatest degree of morbidity (24, 28). STEC infection has been associated with hemorrhagic colitis (HC), where severe damage to intestinal tissue results in grossly bloody diarrhea, and with hemolyticuremic syndrome (HUS), characterized by acute renal failure, thrombocytopenia, and hemolytic anemia (13, 28, 30). Most individuals infected with STEC who develop diarrhea and/or bloody diarrhea will recover from the infection without further complication. However, 5 to 10% of patients, primarily children and the elderly, develop potentially life-threatening systemic complications such as HUS or thrombotic thrombocytopenic purpura (13, 18, 28, 30). One of the most important differences between STEC and the other diarrheagenic types of E. coli is the production of cytotoxins known as Shiga toxins (24). Shiga toxins are generally thought to be responsible for the thrombotic microangiopathy seen in multiple sites, which is the histopathological hallmark of STEC-related disease (28). The precise sequence of events that leads to HC and/or HUS is unclear. STEC cells are usually either ingested from a contaminated food or water source or transmitted from an infected individual. They colonize the colon and produce Shiga toxins in * Corresponding author. Mailing address: Division of Geographic Medicine and Infectious Diseases, New England Medical Center, Box 041, 750 Washington St., Boston, MA Phone: (617) Fax: (617) david.acheson@es.nemc.org. Present address: Fogarty International Center, National Institutes of Health, Bethesda, MD. the intestinal lumen. It is thought that Shiga toxins are then able to cross the epithelial barrier and presumably enter the bloodstream, targeting the endothelia of susceptible tissues in multiple sites resulting in intestinal as well as systemic dysfunction (28). Members of the family of Shiga toxins produced by E. coli are structurally and functionally related to the Shiga toxin from Shigella dysenteriae type 1 (2). There are two major members of the Shiga toxin family, Shiga toxin 1 (Stx1), which is essentially identical to the Shiga toxin produced by Shigella, and Shiga toxin 2 (Stx2), which has only 56% homology with Stx1 and which is immunologically distinct (2). Variants of Stx2 termed Stx2c, Stx2d, and Stx2e are produced by certain STEC strains, but no variants of Stx1 have been described (22). All Shiga toxin family members are made up of one A subunit exhibiting an RNA-glycohydrolase activity and five B subunits responsible for toxin binding (2). Stx1, Stx2, Stx2c, and Stx2d bind to the well-characterized glycolipid receptor Gb3 present on certain eukaryotic cells (2). The toxins are endocytosed and transported in a retrograde manner to the Golgi apparatus, followed by passage into the endoplasmic reticulum (ER). The toxins then gain access to the cytoplasm, where they interact with the ribosomes acting as a glycohydrolase, cleaving a specific adenine, the result being inhibition of protein synthesis (10, 31). Although Stx1 and Stx2 are structurally and functionally similar, several clinical studies have reported that STEC producing only Stx2 or Stx2 plus Stx1 is more frequently associated with disease than is STEC producing only Stx1 (6, 23, 26, 28, 32, 35). One comprehensive study of an international collection of STEC, representing numerous serotypes, concluded that STEC cells containing the Stx2 gene were five times more likely to be associated with severe disease than STEC cells of the same serotype that did not have the Stx2 gene (6). Exper- 6670

2 VOL. 67, 1999 TRANSLOCATION OF Stx1 AND Stx2 ACROSS EPITHELIUM 6671 iments with mice and gnotobiotic pigs also suggest that Stx2 may be more pathogenically relevant (11, 34, 37). Stx2 has been shown to be 400-fold more toxic to mice than Stx1 (34). Despite the evidence suggesting that the risk of developing systemic complications is more frequently associated with Stx2- producing STEC infections, very little is known about differences between Stx1 and Stx2 that provides insight into this observed clinical phenomenon. It is speculated that STEC strains may be capable of producing more Stx2 than Stx1 in the gut lumen and that carriage of the Stx2 gene may be accompanied by some yet-undiscovered pathogenic factor that is lacking in strains that produce Stx1 only (28). Unlike Shigella spp., STEC strains are not invasive and are thought to be restricted to the lumen of the gut, where Shiga toxins are produced, from which they can be recovered in the stools of infected patients (1, 4, 28). Our laboratory is interested in understanding how Shiga toxins penetrate the intestinal epithelial cell barrier and gain access to the underlying tissue. We have previously described a tissue culture model to investigate the movement of Stx1 across polarized intestinal epithelial cells and have shown that biologically active Stx1 was able to cross both T84 and CaCo2A intestinal epithelial cell lines without destroying the cells or the tight junctions (4). Stx1 translocation appeared to be both energy dependent and saturable (4). The aim of the present study was to investigate whether the increased association of disease with Stx2-producing strains might be explained by differences in the movement across intestinal epithelial cell monolayers between Stx2 and Stx1 (6, 23, 26). We examined the movement of both toxins under several conditions, including the use of drugs that affect various cellular processes. The results from these studies indicate that there are indeed significant differences in the movement of the two toxins. (This work was presented at the 99th General Meeting of the American Society for Microbiology, 1999.) MATERIALS AND METHODS Tissue culture. Cells used in this study include the human intestinal epithelial cell lines CaCo2A and HCT-8, as well as the African green monkey kidney epithelial Vero cell line. All cell lines were grown at 37 C in 5% CO 2 with 100 U of penicillin and streptomycin per ml and 10% heat-inactivated fetal bovine serum (Gibco/BRL). CaCo2A cells were grown in Dulbecco modified Eagle medium (high glucose) 25 mm HEPES; HCT-8 cells were grown in RPMI mm HEPES; Vero cells were grown in modified McCoy s 5a media. Vero and HCT-8 cells were obtained from the American Type Culture Collection (Manassas, Va.), and CaCo2A cells were obtained from the GRASP Tissue Culture Core, New England Medical Center. Polycarbonate filter inserts (0.33-cm 2,5- m-pore-size Transwell membranes; Costar, Cambridge, Mass.) were placed in 24-well or 12-well plates and were precoated with rat tail collagen prior to being seeded with either CaCo2A or HCT-8 cells. The cells were fed every 2 days, and experiments were conducted 8 to 10 days postseeding. For inverted monolayers, where cells are grown on the underside of the filter, cells were added to filters placed upside down in a 12-well plate and allowed to attach overnight. The next day, the filter membranes were flipped back upright in a 24-well plate. This serves to reverse the polarity of the cells, with the basolateral surface facing up and the apical surface facing down (27). Experimental procedures. Purified Stx1 and Stx2 (8) were used at 1 g/ml unless otherwise indicated. [ 3 H]inulin (New England Nuclear, Boston, Mass.) was used at 2.5 Ci/ml. Horseradish peroxidase (HRP; Sigma Chemical Co., St. Louis, Mo.) was used at 500 g/ml. These molecules were diluted in the appropriate culture medium and added to either the upper chamber (100 l) or the lower chamber (600 l) of the Transwell membrane. These volumes were used in accordance with the manufacturer s instructions to avoid the effects of hydrostatic pressure. Following the addition of the various agents to the upper or lower chamber according to the study protocol, the plates were incubated at 37 C in 5% CO 2 for 24 h unless otherwise noted. Samples were then collected from the lower or upper chamber and frozen at 70 C prior to further processing (4). Electrical resistance measurements across CaCo2A and HCT-8 monolayers grown on filter membranes for 8 to 10 days were taken prior to each experiment with a Millicell-ERS resistance reader (Millipore Corporation, Bedford, Mass.) to insure that the monolayers demonstrated intact tight junctions and thus formed a functional barrier (27). After 8 days, measurements were 750 to 950 /cm 2 for CaCo2A monolayers and 1,000 to 1,200 /cm 2 for HCT-8 cells. Toxin, [ 3 H]inulin, and HRP measurement. Two assay systems were used for the measurement of Stx1 and Stx2. First, biological activity was measured by quantifying the inhibition of incorporation of [ 3 H]leucine into the protein as previously described (8). Bioactivity was converted to amount of toxin by comparison with a control dose-response curve run concurrently. In circumstances where other reagents interfered with the bioassay, Stx1 and Stx2 were quantified as amounts of antigen by a monoclonal antibody capture enzyme-linked immunosorbent assay (ELISA) (3). [ 3 H]inulin concentration was determined by adding 10 l of sample to 4 ml of liquid scintillation fluid and measuring the counts per minute in a scintillation counter. HRP was measured in terms of activity as described by Hecht et al. (14). The amounts of toxin, [ 3 H]inulin, and HRP in the starting material as well as in the medium recovered from the opposite chamber following translocation were measured. Values are reported as the amount of translocated as a percentage of the amount added. Effect of drugs. All drugs (colchicine, monensin, and brefeldin A) were purchased from Sigma Chemical Co.). To initiate the experiment, medium in both chambers was replaced with medium containing various concentrations of colchicine (10 9 to 10 3 M), monensin (10 8 to 10 4 M), or brefeldin A (0.1 to 500 ng/ml) and the chambers were held for1hat37 C in 5% CO 2. Then the medium in the top chamber was replaced with medium containing the same concentration of drug with Stx1 or Stx2 and [ 3 H]inulin. Samples were collected from the bottom chamber at 24 h as described above. The ELISA rather than the bioactivity assay was used to quantify toxin in these experiments because the drugs interfered with the bioactivity assay. Toxin binding. CaCo2A monolayers on 96-well plates or Transwell membranes were cooled to 4 C and then exposed to 125 I-Stx1 or 125 I-Stx2 for 1 h followed by washing as described previously (17). 125 I-Stx1 and 125 I-Stx2 were labeled with Bolton-Hunter reagent in accordance with the manufacturer s instructions (ICN Pharmaceuticals Inc. Irvine, Calif.). Alternatively, 125 I-Stx1 was labeled with chloramine-t (17). Data was subjected to Scatchard analysis to determine the binding affinity and the number of binding sites per cell for each toxin. Competitive toxin movement experiments. 125 I-Stx1 (0.5 g/ml) was added to the apical chambers of inverted CaCo2A monolayers with a 100-fold excess (50 g/ml) of, or without, unlabeled Stx1 or Stx2. After 24 h at 37 C, counts per minute in the basolateral chamber were determined with a gamma counter and 50 l from the basolateral chamber (45,000 cpm) was run on a 1.5-mm-thick sodium dodecyl sulfate (SDS) polyacrylamide gel. The gel was dried with a model 583 Bio-Rad gel dryer and exposed to film for 3 days. Inverted CaCo2A monolayers were used as opposed to monolayers with cells seeded right side up because adding a set concentration of 125 I-Stx1 to the bottom apical chamber (600 l) and sampling from the top basolateral chamber (100 l) serves to concentrate the toxin, thus facilitating toxin visualization on a gel. Statistics. Statistics were performed with the Instat statistics program for Macintosh computers. Unpaired Student s t tests were used to compare differences in sets of data; P 0.05 was considered significant. RESULTS Stx1 and Stx2 translocation. Intestinal cell monolayers demonstrated stable electrical resistance after having grown for 8 to 10 days on collagen-coated filters. To insure that each monolayer remained intact and represented an epithelial barrier during the course of the experiment, the widely used paracellular marker [ 3 H]inulin was added with toxin in each experiment (21). Only a small amount of [ 3 H]inulin (3 to 10% over 24 h) is able to move across polarized monolayers unless cell injury occurs or tight junctions are altered (4). Experiments using collagen-coated filters containing no cells and thus representing no barrier are expected to allow 85.7% of the total amount of [ 3 H]inulin or toxins added to cross from the top to the bottom chamber of the Transwell membrane. This presumption of equilibration is based on the volumes of the top (100 l) and bottom (600 l) chambers as well as on the random-equilibrium principle. [ 3 H]inulin, Stx1, and Stx2 were found to freely equilibrate across these filters in the absence of cells, demonstrating that the collagen-coated filter itself offers no significant barrier to these molecules (data not shown). Stx1 and Stx2 at doses of 1 g/ml added to the apical compartment of intestinal epithelial cell monolayers did not compromise the integrity of monolayers in a 24-h period. Resistance and [ 3 H]inulin movement were not significantly altered in the presence of either toxin (data not shown). CaCo2A cells

3 6672 HURLEY ET AL. INFECT. IMMUN. TABLE 1. Comparison of Stx1 and Stx2 movement across two different intestinal epithelial lines grown on polycarbonate filters as measured by biological activity Cell line n Mean % movement (SD) of: Stx1 [ 3 H]inulin c Stx2 [ 3 H]inulin d CaCo2A (3.1) a 9.0 (3.2) 0.6 (0.4) a 8.4 (2.9) HCT (0.5) b 4.5 (0.3) 0.3 (0.3) b 4.5 (0.4) a P b P c In cells treated with Stx1. d In cells treated with Stx2. TABLE 3. Comparison of Stx1 and Stx2 translocation across CaCo2A cells c Toxin n Mean % movement (SD) Mean % movement b (SD) of: [ 3 H]inulin HRP Stx (1.0) a 6.1 (0.2) 0.05 (0.02) Stx (0.1) a 6.2 (0.5) 0.08 (0.09) a P b Movement across cells treated with toxin indicated in the left column. c Molecular masses: toxin, 72 kda; inulin, 5 kda; HRP, 44 kda. were also shown to be highly insensitive to both Stx1 and Stx2. Concentrations as high as 10 g of toxin/ml did not affect protein synthesis of CaCo2A cells (unpublished observations). A comparison of the movement of Stx1 with that of Stx2 across CaCo2A monolayers showed that there was approximately 10- fold-less biologically active Stx2 translocating across CaCo2A cells in a 24-h period than Stx1 (P 0.008) (Table 1). The same trend was observed in HCT-8 cells, although it was less pronounced and failed to reach statistical significance (P 0.051) (Table 1). There was no difference in [ 3 H]inulin movement between cells treated with Stx1 and cells treated with Stx2 (Table 1). To address the question of whether this observed difference in toxin movement resulted from a loss of Stx2 biological activity or an increase in Stx1 biological activity during the translocation process, the amounts of translocated toxin in several samples were measured by both biological activity determination and ELISA. Table 2 demonstrates that the amount of translocated toxin measured as bioactivity equals the amount measured by ELISA for both Stx1 and Stx2. Thus toxin bioactivity is proportional to the quantity of toxin protein. In this set of experiments, there was again a significant (P ) reduction in Stx2 movement, as measured by ELISA, compared with Stx1 movement (Table 3), with no difference in [ 3 H]inulin movement (Table 3). In addition, we measured the movement of HRP (44 kda), which more closely approximates the two toxins (72 kda) in size than does the much smaller [ 3 H]inulin (5 kda). HRP thus served as a control to demonstrate that these intestinal cell barriers maintain their ability to be selective in retaining large molecules. The lack of difference in HRP movement in Stx1- versus Stx2-treated cells confirms that the differences in movement between the two toxins were not simply experimental variation. Our observations on differences in toxin movement across CaCo2A cells were highly reproducible. In a compilation of a larger set of data including five separate experiments (n 10), the translocation of Stx1 and Stx2 in paired sets of CaCo2A monolayers was measured. The amount of Stx1 crossing was 2.9% / 1.5% (mean / standard deviation [SD]), while the amount of Stx2 was only 0.07% / 0.07% (P ), representing a 40-fold difference. Movement of Stx1 across CaCo2A cells has previously been shown to be dependent on an active metabolic process (4). Cooling the cell monolayers to 4 C results in a dramatic reduction in Stx1 translocation with only a minor reduction in [ 3 H]inulin translocation (4). These observations were confirmed as shown in Table 4: there was a 2-fold drop in [ 3 H]inulin movement and a statistically significant 200-fold drop in Stx1 movement over 4 h at 4 C compared with movement at 37 C for 4h(P 0.001). Stx2 showed a similar trend, but the drop in translocation at 4 C was only 20-fold (P 0.019) (Table 4). The inhibition in movement of both toxins by reduced temperature was reversible by rewarming to 37 C. There were no significant differences in movement of [ 3 H]inulin, Stx1, and Stx2 between wells that had been at 4 C for 4 h followed by 37 C for 24 h and wells that had been at 37 C throughout the 28-h period (data not shown). Directional movement. In this set of experiments, CaCo2A cells were grown either on upright or inverted filters with [ 3 H]inulin and toxin added to either the top or the bottom chamber in an effort to investigate the effects on translocation of reversing the intestinal epithelial cell polarity (Table 5). [ 3 H]inulin translocation was not affected by intestinal cell polarity (Table 5). Stx1 showed a preference for apical-to-basolateral movement irrespective of whether Stx1 was moving from the top to the bottom chamber (P 0.028) or from the bottom to the top chamber (P ) (Table 5). Stx2 translocation exhibited a preference for apical-to-basolateral movement when moving from the top to the bottom chamber (P 0.029); however, there was no significant polarity preference for Stx2 movement from the bottom to the top chamber (P 0.311) (Table 5). Binding experiments. Using 125 I-labeled Stx1 and Stx2, we performed a Scatchard analysis on the binding of Stx1 and Stx2 to CaCo2A cells seeded on 96-well plates in order to determine whether there were any differences in binding that might explain the observed differences in translocation. Experiments were performed at 4 C to prevent toxin internalization. The TABLE 2. Comparison of biological activity and ELISA measurements of translocated Stx1 and Stx2 across CaCo2A cells Toxin n Mean % movement (SD) by: Bioactivity ELISA Mean % movement of [ 3 H]inulin c (SD) Stx (1.0) a 1.9 (1.0) b 4.4 (0.4) Stx (0.2) a 0.2 (0.01) b 6.3 (0.4) a P b P c Movement across cells treated with toxin indicated in the left column. Temp ( C) TABLE 4. Effect of temperature on the translocation of Stx1, Stx2, and [ 3 H]inulin across CaCo2A monolayers n Mean % movement (SD) of: Stx1 [ 3 H]inulin c Stx2 [ 3 H]inulin d (0.002) a 1.0 (0.1) (0.0002) b 1.0 (0.1) (0.03) a 2.0 (0.1) (0.003) b 2.1 (0.4) a P b P c In cells treated with Stx1. d In cells treated with Stx2.

4 VOL. 67, 1999 TRANSLOCATION OF Stx1 AND Stx2 ACROSS EPITHELIUM 6673 TABLE 5. Directional movement of Stx1, Stx2, and [ 3 H]inulin Mean % movement (SD) of indicated molecule added from d : Direction of movement Top chamber b Bottom chamber c Stx1-treated cells e Stx2-treated cells f Stx1-treated cells g Stx2-treated cells g Stx1 [ 3 H]inulin Stx2 [ 3 H]inulin Stx1 [ 3 H]inulin Stx2 [ 3 H]inulin Apical to basolateral 4.3 (5.2) 4.4 (0.4) 0.03 (0.009) 4.2 (0.5) 0.4 (0.4) 1.0 (0.6) 0.03 (0.04) 0.7 (0.1) Basolateral to apical 0.7 (1.1)/ (0.4)/NS (0.005)/ (0.3)/NS 0.01 (0.02)/ (0.3)/NS 0.06 (0.05)/NS 0.7 (0.1)/NS a [ 3 H]inulin and toxin concentrations were kept constant when added to either the upper (100 l) or lower chamber (600 l), and values are percentages of the amount of starting material. b For measurement of apical-to-basolateral movement, cells were grown on the upright sides of permeable filters, with toxin and [ 3 H]inulin added to the top chamber (apical) and recovered from the bottom chamber (basolateral). For measurement of basolateral-to-apical movement, cells were grown on the undersides of permeable filters, with toxin and [ 3 H]inulin added to the top chamber (basolateral) and recovered from the bottom chamber (apical). c For measurement of apical-to-basolateral movement, cells were grown on the undersides of permeable filters, with toxin and [ 3 H]inulin added to the bottom chamber (apical) and recovered from the top chamber (basolateral). For measurement of basolateral-to-apical movement, cells were grown on the upright sides of permeable filters, with toxin and [ 3 H]inulin added to the bottom chamber (basolateral) and recovered from the top chamber (apical). d Statistical significance is indicated by P values or NS (not significant) following the slashes in the basolateral-to-apical row of values. e n 12. f n 7. g n 6. association constant (K a ) of Stx1 binding was / M 1, whereas the K a for Stx2 binding was / M 1. This represents a significant fourfold difference (P 0.008). The numbers of toxin binding sites per cell were similar for the two toxins ( / for Stx1 and / for Stx2) (P 0.548). Given that these cells are grown on a solid surface rather than a Transwell filter, we cannot confirm that the monolayer represents a polarized barrier by measuring electrical resistance. Therefore, we wanted to confirm that the fourfold difference in the binding of Stx1 versus that of Stx2 observed in cells grown on 96-well plates was consistent with cells grown on Transwell filters, where resistance can be tested. We chose a single concentration of 125 I-Stx1 and 125 I-Stx2 and corrected for specific activity. The counts associated with the monolayers were 3,905 / 274 cpm for 125 I-Stx1 and 980 / 238 cpm for 125 I-Stx2, representing about a fourfold difference (P 0.008). Effect of a high concentration of Stx2 on Stx1 movement. Based on the above data, we speculated that Stx1 and Stx2 use separate pathways across intestinal epithelial monolayers. Previous data showed that the translocation of Stx1 across the CaCo2A monolayer was a saturable process; as the amount of Stx1 added to the apical surface increased, the overall percentage of translocated Stx1 decreased (4). If Stx1 and Stx2 were following different pathways, we hypothesized that excess Stx2 would not reduce the amount of Stx1 translocation. As shown in Table 6, increasing amounts of Stx2 added to the apical surface (up to 1,000-fold in excess of Stx1) had no effect on Stx1 or [ 3 H]inulin movement. Stx1 was detected by an ELISA using an Stx1-specific monoclonal capture antibody (4D3), which does not cross-react with Stx2 (9). An unrelated protein (HRP) in 1,000-fold excess also had no effect on either [ 3 H]inulin or Stx1 movement (Table 6). To further demonstrate that Stx1 and Stx2 use separate routes, we measured the movement of 125 I-Stx1 across CaCo2A monolayers in the presence of a 100-fold excess of unlabeled Stx1 or Stx2 or in the absence of unlabeled Stx1 and Stx2. The translocated 125 I-Stx1 was detected by running an aliquot of the translocated material on an SDS-polyacrylamide gel electrophoresis followed by autoradiography. Figure 1 clearly shows that a 100-fold excess of unlabeled Stx1 (lane 2) dramatically reduces the amount of 125 I-Stx1 translocating to the basolateral chamber compared with the control, which contains an equal amount of 125 I-Stx1 but no unlabeled toxin (lane 1). A 100-fold excess of unlabeled Stx2 (lane 3) had no effect on the amount of 125 I-Stx1 translocating across the monolayer (lanes 1 and 3). Effect of colchicine, monensin, and brefeldin A on Shiga toxin movement. To further investigate the movement of Stx1 and Stx2 across polarized CaCo2A monolayers, we employed various drugs that interfere with different cellular processes. As before, all CaCo2A monolayers were grown for 8 to 9 days TABLE 6. Effect of increasing doses of Stx2 on Stx1 translocation across CaCo2A monolayers a Molecule (amt [ng]) added to top (apical) chamber Stx1 Mean % movement (SD) of: [ 3 H]inulin Control 2.1 (1.4) 3.3 (0.5) Stx2 (100) 2.5 (2.0) 3.3 (0.6) Stx2 (1,000) 2.6 (1.7) 3.3 (0.3) Stx2 (10,000) 2.7 (0.9) 3.2 (0.4) HRP (10,000) 2.3 (0.9) 3.4 (0.2) a Stx1 (10 ng/well) and [ 3 H]inulin were added to the upper chambers with increasing doses of Stx2 or HRP. Samples were collected from the bottom, and Stx1 was measured by ELISA using the Stx1-specific monoclonal antibody (4D3). n 4 for each condition. FIG. 1. Effects of excess unlabeled Stx1 and Stx2 on the translocation of 125 I-Stx1 across CaCo2A monolayers. 125 I-Stx1 (0.5 g/ml) was added to the apical chamber of inverted CaCo2A monolayers. After 24 h at 37 C, material from the basolateral chamber was removed and run on an SDS 15% polyacrylamide gel and exposed to film for 3 days. Lane 1, 125 I-Stx1 alone; lane 2, 125 I-Stx1 with a 100-fold excess of unlabeled Stx1; lane 3, 125 I-Stx1 with a 100-fold excess of unlabeled Stx2.

5 6674 HURLEY ET AL. INFECT. IMMUN. on collagen-coated filters at which point they developed electrical resistance levels of 750 to 900 /cm 2. Colchicine, which interferes with microtubule formation, was used at doses ranging from 10 9 to 10 2 M. The CaCo2A monolayers were pretreated with various doses of colchicine for 1 to 2hat37 C. Stx1 or Stx2 and [ 3 H]inulin were then added, and samples were taken from the basolateral chamber after 24 h as in previous experiments. There was no significant change in the movement of either [ 3 H]inulin or Stx1 at colchicine doses less than 10 7 M. At 10 7 M colchicine, the amount of Stx1 crossing the monolayer was 67.5% ( / 14.1%) of that crossing untreated control monolayers (Fig. 2A). There was no effect on [ 3 H]inulin movement at this colchicine dose. As the concentration of colchicine was further increased, the amount of Stx1 crossing the monolayer decreased. At 10 5 to 10 3 M colchicine there was an increase in [ 3 H]inulin movement, indicating that the monolayers became leaky at these higher doses, which suggests the loosening of tight junctions or damage to cells (20). Despite this increase in [ 3 H]inulin (5 kda) movement, the amount of Stx1 (72 kda) translocating was still lower at these doses than the amount translocating in the control. The effect of colchicine on Stx2 movement was completely different. As the dose of colchicine was increased, the amount of Stx2 that crossed the CaCo2A monolayer increased and was greater than that in the untreated control at doses of colchicine that did not affect [ 3 H]inulin movement (Fig. 2B). As both toxins were quantified by the ELISA method, we ran control ELISAs in the presence of colchicine to determine whether colchicine had any effect on the assay and found none (data not shown). Translocation of Stx1, Stx2, and [ 3 H]inulin in CaCo2A cells treated with monensin was also measured. This drug, which interferes with receptor-mediated endocytosis, was used in concentrations from 10 8 to 10 4 M. As the dose was increased, the amount of Stx1 translocating across decreased in a dose-dependent fashion with a maximum decrease at 10 6 M monensin of 55.3% ( / 8.9%) of the untreated control. This dose had no effect on [ 3 H]inulin movement (Fig. 2C). At higher doses of monensin, both toxin and [ 3 H]inulin began to leak across the monolayer (Fig. 2C). Monensin had similar effects on Stx2, although less pronounced (Fig. 2D). Stx2 movement decreased with increasing concentrations of monensin, reaching the maximal decrease at 10 7 M monensin of 65.2% ( / 12.8%) of control. As monensin was further increased, the amounts of [ 3 H]inulin and Stx2 translocation increased (Fig. 2D). Brefeldin A, which disrupts the Golgi apparatus, had no apparent effect on Stx1 or Stx2 translocation until doses were high enough to cause [ 3 H]inulin leaks (Fig. 2E and F). At this point the amounts of both Stx1 and Stx2 movement increased over that of the untreated controls (Fig. 2E and F). When Stx1 and Stx2 ELISAs were performed in the presence and absence of monensin and brefeldin A, no effect on the ELISA results was found. DISCUSSION Shiga toxins are generally thought to be responsible for the development of both HC and the life-threatening condition HUS, which occur in patients infected with STEC (28). Strong evidence suggests that these complications arise as a result of Shiga toxins directly targeting microvascular endothelial cells in the kidney and other sites (25). Human microvascular endothelial cells may be exquisitely sensitive to Shiga toxin (25). Since STEC infection is localized to the luminal side of the epithelial cell barrier in the intestine, it is important to understand how Shiga toxins gain access to the endothelial cell beds of the colon, kidney, and elsewhere. An in vitro model developed to address this question demonstrated that high doses of Stx1 are not cytotoxic to the intestinal epithelial cell lines CaCo2A and T84 (4). Stx1 has no effect on barrier function; however significant amounts of biologically active Stx1 are capable of crossing the intact epithelial monolayer (4). Stx1 translocation is drastically reduced both when cells are chilled to 4 C and when cells are treated with the metabolic inhibitor 2,4-dinitrophenol, indicating that Stx1 translocation is an energy-requiring process (4). Since the epithelial monolayers maintain high resistance in the presence of Stx1 and since Stx1 movement across the monolayer is an active process, we hypothesized that Stx1 translocates across the CaCo2A monolayer via a transcellular rather than a paracellular route (4). This idea is supported by the study of Philpott et al. (29). Using T84 cells as a model, they proposed that Stx1 translocation was transcellular and demonstrated immunogold localization of Stx1 within endosomes and exclusion from the paracellular space (29). The most striking observation in the present work is that the translocation of Stx2 across polarized intestinal epithelia in vitro differs significantly from that of Stx1. While neither toxin affects the barrier function of CaCo2A cells, we consistently observed in numerous experiments that less Stx2 than Stx1 translocates across CaCo2A cells in a 24-h period. Data from experiments in which levels of translocation of both Stx1 and Stx2 in the same set of CaCo2A monolayers were compared showed that there was, on average, a 40-fold difference in movement. Both toxins retain biological activity after translocation. A comparison of biological activity with antigen amount showed that both Stx1 and Stx2 were neither activated nor inactivated. Temperature had less of an inhibiting effect on Stx2 movement than it had on Stx1 movement. Stx1 movement was clearly directional, with a preference for apical-to-basolateral translocation, whereas the directional preference of Stx2 movement was not clear. Stx2 appeared to also prefer apicalto-basolateral translocation, but this preference was not maintained when Stx2 moved across the monolayer against gravity. The reasons for this difference are unknown. Using iodinated toxins we found that Stx1 had a binding affinity to the apical surface of CaCo2A higher than that of Stx2; however the difference was only fourfold, insufficient to explain the difference in toxin movement. We hypothesize that while both toxins are capable of crossing the intestinal epithelial monolayer, they do so via different pathways. This model was supported by our competition experiments in which a 100-fold excess of Stx1 blocked the translocation of 125 I-Stx1, whereas a 100-fold excess of Stx2 had no blocking effect. As much as a 1,000-fold excess of Stx2 had no effect on Stx1 translocation. Several intracellular vesicle pathways rely upon microtubule assembly and disassembly for the shuttling of vesicles throughout the cell (5, 7, 36). Transcytosis of proteins within vesicles from the apical to the basolateral surface or from the basolateral to the apical surface is thought to be microtubule dependent. Agents that disrupt microtubule function can cause a significant decrease in the amount of protein that is transcytosed (5, 7). Colchicine binds to tubulin dimers and inhibits polymerization, preventing microtubule assembly (20). Our study demonstrates that high doses of colchicine reduce the amount of Stx1 translocation across CaCo2A cells in a 24-h period by about 60 to 70%. This reduction in Stx1 movement was maintained even at doses of colchicine that caused leakage of [ 3 H]inulin. [ 3 H]inulin (5 kda) is a much smaller molecule than Stx1 (72 kda); thus small breaches in tight-junction in-

6 VOL. 67, 1999 TRANSLOCATION OF Stx1 AND Stx2 ACROSS EPITHELIUM 6675 FIG. 2. Effects of increasing doses of colchicine, monensin, and brefeldin A on the translocation of [ 3 H]inulin (solid circles) and toxin (open circles) across CaCo2A monolayers. The x axis represents the dose of drug added to the cell monolayer, with control representing the absence of drug. The percent translocation of [ 3 H]inulin is shown on the left y axis. The control value for the translocation of toxin was set at 100%, and values within the same experiment are reported as percentages of the control value (shown on the right y axis). Each data point represents the mean ( SD) of pooled data (n 2 to 9) from multiple experiments.

7 6676 HURLEY ET AL. INFECT. IMMUN. tegrity may result in differential molecular movement across paracellular space depending on the size of the molecule. We interpret the large decrease in Stx1 movement when cells are exposed to colchicine as further evidence for the transcellular movement of Stx1. In contrast, colchicine did not inhibit Stx2 movement. In fact, as colchicine concentrations increased, Stx2 movement also increased at doses that did not compromise the monolayer. This further implies that the toxins are moving across by different pathways, suggesting a microtubule-dependent vesicle pathway for Stx1 translocation and a microtubuleindependent pathway for Stx2. Monensin is an ionophore that interferes with vesicular Na /H exchange. At low doses (10 7 to 10 5 M), it is reported to prevent the acidification of intracellular compartments, including ER, Golgi apparatus, and endosomes (33). Monensin is capable of disrupting transcytosis in epithelial cells (7, 15); more specifically, it is thought to inhibit receptormediated endocytosis because the rise in endosomal ph prevents receptor-ligand disassociation (7, 33). Monensin inhibits movement of Stx1 by 40 to 50% at 10 6 M, further suggesting a transcellular route for Stx1 translocation and also indicating that receptor-mediated endocytosis may be involved. Monensin resulted in a decrease in Stx2 movement, but one not as large as the decrease in Stx1 movement seen. Brefeldin A disrupts Golgi stacks by interfering with vesicle movement from the ER to the Golgi apparatus (12). It has been shown to both increase (38) and decrease the transcytosis of proteins in epithelial cells (7). An interesting property of brefeldin A is that it protects Shiga toxin-sensitive cells from intoxication by disrupting the Golgi apparatus and interfering with the retrograde transport of Shiga toxin into the cytoplasm (12). CaCo2A cells are not toxin sensitive, although they do express low levels of the Gb3 toxin receptor (17). Since brefeldin A interferes with the vesicle trafficking leading to intoxication, we looked for an effect on Stx1 or Stx2 translocation across the epithelium but found none. From this work it remains unclear whether transcytosing Stx1 and Stx2 bind to an apical receptor and enter the cell via receptor-mediated endocytosis or via nonspecific fluid phase endocytosis. However, the monensin data suggests that receptor-mediated endocytosis may be more involved in Stx1 translocation than in Stx2 translocation. Monensin does not inhibit the movement of HRP, which has been reported to transcytose by a fluid phase nonreceptor mechanism (15). The only wellestablished receptor for Stx1 and Stx2 is Gb3, which is involved in the delivery of Shiga toxins to the cytoplasm of cells for intoxication. Gb3 is expressed at significantly lower levels on CaCo2A cells than on cells that are sensitive to Shiga toxins such as Vero cells (4, 17). Sodium butyrate increases Gb3 receptor expression and sensitivity to Stx1 in CaCo2A cells. When numbers of receptors are increased by this mechanism, Stx1 translocation is significantly decreased (4). Although the evidence suggests that Stx1 appears to be internalized via receptor-mediated endocytosis and translocates via transcytotic vesicles, it is not clear what the role of Gb3 is in this pathway, if any. Transcytosis of numerous proteins across various epithelia have been described previously (5, 7, 15, 36, 38); these include cholera toxin, a bacterial toxin produced in the intestines of individuals infected with Vibrio cholerae (19). While Stx1 and, to a lesser extent, Stx2 can move across epithelial layers in vitro, the significance of Shiga toxin translocation in vivo is unknown. Given the apparent greater pathogenicity of Stx2- producing STEC in clinical situations, it was surprising to discover that Stx2 crossed human intestinal epithelial cells significantly less efficiently than did Stx1. However, it has been reported that intestinal endothelial cells are 10-fold more sensitive to Stx2 than to Stx1 and that glomerular endothelial cells are 1,000-fold more sensitive to Stx2 (16, 25). If these various in vitro observations represent the in vivo situation, the increase in sensitivity of the endothelium to Stx2 over Stx1 may compensate for the diminished Stx2 translocation. We acknowledge that the situation in the human intestine is far more complex, involving many different cell types, including both intestinal and inflammatory cells. Thus a better understanding of how toxins interact with the epithelium in vivo will require further study. Developing a further understanding of how these very similar toxins differ from one another will help to explain the epidemiological association of Stx2 in human STEC disease. It will also facilitate the development of therapies that are geared toward preventing toxins from crossing the epithelium at the early stages of the disease, which may aid in preventing systemic sequelae. ACKNOWLEDGMENTS Research support for this study included the following grants from the National Institutes of Health, Bethesda, Md.: HL-55660, A , and A , as well as P 30 DK for the Center for Gastroenterology Research on Absorptive and Secretory Processes. C.M.T. was supported by NIH Training Grant T32-A We thank Anne V. Kane for critical comments and Thao N. Ngo for technical assistance. REFERENCES 1. Acheson, D. W. K., S. De Breucker, A. Donohue-Rolfe, K. Kozak, A. Yi, and G. T. Keusch Enzyme immunoassay for direct detection of Shiga-like toxins in stool samples, abstr. B-62, p. 39. In Abstracts of the 94th General Meeting of the American Society for Microbiology American Society for Microbiology, Washington, D.C. 2. Acheson, D. W. K., A. Donohue-Rolfe, and G. T. Keusch The family of Shiga and Shiga-like toxins, p In J. E. Alouf and J. H. Freer (ed.), Sourcebook of bacterial protein toxins. Academic Press Ltd., London, United Kingdom. 3. Acheson, D. W. K., M. Jacewicz, A. V. Kane, A. Donohue-Rolfe, and G. T. Keusch One step high yield affinity purification of shiga-like toxin II variants and quantitation using enzyme linked immunosorbent assays. Microb. Pathog. 14: Acheson, D. W. K., R. Moore, S. De Breucker, L. Lincicome, M. Jacewicz, E. Skutelsky, and G. T. Keusch Translocation of Shiga toxin across polarized intestinal cells in tissue culture. Infect. Immun. 64: Apodaca, G., M. Bomsel, J. Arden, P. P. Breitfield, K. Tang, and K. E. Mostov The polymeric immunoglobulin receptor: a model protein to study transcytosis. J. Clin. Investig. 87: Boerlin, P., S. A. McEwen, F. Boerlin-Petzold, J. B. Wilson, V. Johnson, and C. L. Gyles Associations between virulence factors of Shiga toxinproducing Escherichia coli and disease in humans. J. Clin. Microbiol. 37: Deffback, M. E., C. J. Bryan, and C. M. Roy Protein movement across cultured guinea pig trachea: specificity and effect of transcytosis inhibitors. Am. J. Physiol. 271:L744 L Donohue-Rolfe, A., D. W. K. Acheson, A. V. Kane, and G. T. Keusch Purification of Shiga and Shiga-like toxins I and II by receptor analog affinity chromatography with immobilized P1 glycoprotein and the production of cross-reactive monoclonal antibodies. Infect. Immun. 57: Donohue-Rolfe, A., M. A. Kelley, M. Bennish, and G. T. Keusch Enzyme-linked immunosorbent assay for Shigella toxin. J. Clin. Microbiol. 24: Endo, Y., K. Tsurugi, T. Yutsudo, Y. Takeda, T. Ogasawara, and K. Igarashi Site of action of vero toxin (VT2) from Escherichia coli O157:H7 and of Shiga toxin on eukaryotic ribosomes. RNA N-glycosidase activity of the toxins. Eur. J. Biochem. 171: Francis, D. H., R. A. Moxley, and C. Y. Andraos Edema disease-like brain lesions in gnotobiotic piglets infected with Escherichia coli serotype O157:H7. Infect. Immun. 57: Garred, O., E. Dubinina, P. K. Holm, S. Olsnes, B. Van Deurs, J. V. Kozlov, and K. Sandvig Role of processing and intracellular transport for optimal toxicity of Shiga toxin and toxin mutants. Exp. Cell Res. 218: Griffin, P. M Escherichia coli O157:H7 and other enterohemorrhagic Escherichia coli, p In M. J. Blaser, P. D. Smith, J. L. Ravdin, H. B. Greenberg, and R. L. Guerrant (ed.), Infections of the gastrointestinal tract. Raven Press, Ltd., New York, N.Y.

8 VOL. 67, 1999 TRANSLOCATION OF Stx1 AND Stx2 ACROSS EPITHELIUM 6677 Editor: J. T. Barbieri 14. Hecht, G., A. Koutsouris, C. Pothoulakis, J. T. LaMont, and J. L. Madara Clostridium difficile toxin B disrupts the barrier function of T84 monolayers. Gastroenterology 102: Heyman, M., A. Crain-Denoyelle, S. K. Nath, and J. F. Desjeux Quantification of protein transcytosis in the human colon carcinoma cell line CaCo-2. J. Cell. Physiol. 143: Jacewicz, M. S., D. W. K. Acheson, D. G. Binion, G. A. West, L. L. Lincicome, C. Fiocchi, and G. T. Keusch Responses of human intestinal microvascular endothelial cells to Shiga toxins 1 and 2 and pathogenesis of hemorrhagic colitis. Infect. Immun. 67: Jacewicz, M. S., D. W. K. Acheson, M. Mobassaleh, A. Donohue-Rolfe, K. A. Balasubramanian, and G. T. Keusch Maturational regulation of globotriaosylceramide, the Shiga-like toxin 1 receptor, in cultured human gut epithelial cells. J. Clin. Investig. 96: Keusch, G. T., and D. W. K. Acheson Thrombotic thrombocytopenia purpura with Shiga toxins. Semin. Hematol. 34: Lencer, W. L., S. Moe, P. A. Rufo, and J. L. Madara Transcytosis of cholera toxin subunits across model human intestinal epithelia. Proc. Natl. Acad. Sci. USA 92: Ma, T. Y., D. Hollander, L. T. Tran, D. Nguyen, N. Hoa, and D. Bhalla Cytoskeletal regulation of Caco-2 intestinal monolayer paracellular permeability. J. Cell. Physiol. 164: Madara, J. L Regulation of the movement of solutes across tight junctions. Annu. Rev. Physiol. 60: Melton-Celsa, A. R., and A. D. O Brien Structure, biology, and relative toxicity of Shiga toxin family members for cells and animals, p In J. B. Kaper and A. D. O Brien (ed.), Escherichia coli O157:H7 and other Shiga toxin-producing E. coli strains. American Society for Microbiology, Washington, D.C. 23. Milford, D. V., M. C. Taylor, P. E. Rose, T. C. F. Roy, and B. Rowe Immunologic therapy for hemolytic-uremic syndrome. J. Pediatr (Letter to the editor.) 24. Nataro, J. P., and J. B. Kaper Diarrheagenic Escherichia coli. Clin. Microbiol. Rev. 11: Obrig, T. G Interaction of Shiga toxins with endothelial cells, p In J. B. Kaper and A. D. O Brien (ed.), Escherichia coli O157:H7 and other Shiga toxin-producing E. coli strains. American Society for Microbiology, Washington, D.C. 26. Ostroff, S. M., P. L. Tarr, M. A. Neill, J. H. Lewis, N. Hargrett-Bean, and J. M. Kobayashi Toxin genotypes and plasmid profiles as determinants of systemic sequelae in Escherichia coli O157:H7 infections. J. Infect. Dis. 160: Parkos, C. A., C. Delp, M. A. Arnaout, and J. L. Madara Neutrophil migration across a cultured intestinal epithelium. Dependence on a CD11b/ CD18-mediated event and enhanced efficiency in a physiological direction. J. Clin. Investig. 88: Paton, J. C., and A. W. Paton Pathogenesis and diagnosis of Shiga toxin-producing Escherichia coli infections. Clin. Microbiol. Rev. 11: Philpott, D. J., C. A. Ackerley, A. J. Kiliaan, M. A. Karmali, M. H. Perdue, and P. M. Sherman Translocation of verotoxin-1 across T84 monolayers: mechanism of bacterial toxin penetration of epithelium. Am. J. Physiol. 273:G1349 G Pickering, L. K., T. G. Obrig, and F. B. Stapleton Hemolytic-uremic syndrome and enterohemorrhagic Escherichia coli. Pediatr. Infect. Dis. J. 13: Sandvig, K., and B. Van Deurs Endocytosis, intracellular transport, and cytotoxic action of Shiga toxin and ricin. Physiol. Rev. 76: Scotland, S. M., G. A. Willshaw, H. R. Smith, and B. Rowe Properties of strains of Escherichia coli belonging to serogroup O157 with special reference to production of vero cytotoxins VT1 and VT2. Epidemiol. Infect. 99: Tartakoff, A. M Perturbation of vesicular traffic with the carboxylic ionophore monensin. Cell 32: Tesh, V. L., J. A. Burris, J. W. Owens, V. M. Gordon, E. A. Wadolkowski, A. D. O Brien, and J. E. Samuel Comparison of the relative toxicities of Shiga-like toxin type 1 and type 2 for mice. Infect. Immun. 61: Thomas, A., H. Chart, T. Cheasty, H. R. Smith, J. A. Frost, and B. Rowe Vero cytotoxin-producing Escherichia coli, particularly serogroup O157, associated with human infections in the United Kingdom: Epidemiol. Infect. 110: Tome, C. D Transport of beta-lactoglobulin and alpha-lactalbumin in enterocyte-like Caco-2 cells. Reprod. Nutr. Dev. 35: Wadolkowski, E. A., L. M. Sung, J. A. Burris, J. E. Samuel, and A. D. O Brien Acute renal tubular necrosis and death of mice orally infected with Escherichia coli strains that produce Shiga-like toxin type 2. Infect. Immun. 58: Wan, J., M. E. Taub, D. Shah, and W. C. Shen Brefeldin A enhances receptor-mediated transcytosis of transferrin in filter-grown Madin-Darby canine kidney cells. J. Biol. Chem. 267: