Although epidemic cholera had been absent from the Americas for more than a century, it suddenly reappeared in Peru in 1991 and then quickly spread to neighboring countries (40). Soon thereafter epidemic cholera caused by the new O139 serogroup appeared in South Asia. This had particularly devastating consequences, since the new O139 strains afflicted adults with partial immunity to the previously dominant O1 strains, as well as young and immunologicaly naive children who depended on them. Given these new threats to public health, the National Institute for Diagnoses and Reference in Mexico (INDRE) began a surveillance program for O139 strains in 1993, in the event that they might also arrive in Mexico (17). Although antibody-based tests for serogroup O1 strains were well established, no such tests for O139 strains were available in Mexico at that time. Our group in Mexico therefore elected to screen cultures of non-O1 V. cholerae from patients with diarrhea for toxigenicity in mammalian cell culture, as a surrogate marker for possible epidemic strains. Instead of CT-induced cytotoxicity, we found that many of these strains produced a striking vacuolating activity, initially reminiscent of the vacuolating cytotoxin of Helicobacter pylori. Further analysis, described here, showed that this V. cholerae activity was entirely different and was due to the hemolysin (hlyA gene product) of V. cholerae.

Bacterial strains and media. A total of 335 independent non-O1 V. cholerae isolates (268 isolated in 1993 and 67 isolated in 1995) were studied. They were derived from fecal samples of patients with diarrheal disease and from their contacts. The patients and contacts originated from various regions of Mexico (13 of 33 states; six coastal and seven inland). Their isolation was carried out either at State Health Institutes or at INDRE. All specimens were transported in Cary-Blair modified medium (Difco). The V. cholerae strains used most intensively in the present study are described in Table 1. Strain 52201 in particular, was isolated from a 30-year-old female patient with gastroenteritis. Escherichia coli strains S-17 (13) and DH5α (20) were used as hosts for cosmid and plasmid cloning, respectively.

TABLE 1 Bacterial V. cholerae strains specifically referred to in this paper

Standard methods (9) were used to isolate and characterize V. cholerae strains. The methods used included: (i) enrichment growth in alkaline-peptone water (pH 8.0) and formation of yellow colonies on thiosulfate citrate-bile salt-sucrose selective agar (Difco); (ii) oxidase and indole positivity and fermentation of glucose, sucrose, and lactose, with no gas or sulfhydric acid production; (iii) lysine and ornithine decarboxylase production but not arginine dihydrolase or hydrogen sulfide production; and (iv) motility in soft (7.5%) agar. Isolates were also tested for agglutination with anti-serogroup O1-specific antiserum that had been produced and validated at INDRE according to international standards (21).

Alkaline-peptone water (pH 8.0) contained 10 g of peptone and 10 g of sodium chloride, made up to 1 liter with distilled water and adjusted to pH 8.0. Thiosulfate citrate-bile salt-sucrose agar was prepared from a Difco mix. Luria-Bertani broth and agar contained 10 g Difco Bacto tryptone, 5 g of Difco Bacto yeast extract, and 10 g NaCl per liter (pH 7.0) and 15 g of Difco Bacto agar per liter (37), plus 50 μg of ampicillin per ml or 30 μg/ of kanamycin per ml when needed. Craig's medium contained 30 g of Casamino Acids (Difco), 4 g of Bacto yeast extract (Difco), 2 g of glucose, and 0.5 g of potassium dibasic phosphate (K₂HPO₄) per liter of distilled water (pH 7.0) (9). Luria broth with 30% glycerol was used to preserve stock cultures at −70°C. Ringer's solution contained 155 mM NaCl, 3 mM KCl, 2 mM CaCl₂, 1 mM MgCl₂, 3 mM NaH₂PO₄, 5 mM HEPES, and 10 nM glucose with a final pH of 7.4 (38). Dulbecco's modified Eagle medium (DMEM) (Gibco-BRL, Bethesda, Md.) was supplemented with 10% fetal bovine serum.

Vacuolation assays. Bacterial cultures were grown for 16 to 18 h at 37°C in 5 ml of Craig's medium with shaking and pelleted by centrifugation (in an Eppendorf microfuge) at 5,000 rpm at 4°C. Supernatants were collected and filter sterilized through 0.45-mm-diameter filters (Uniflo). For cell physiology experiments, filter-sterilized supernatants were further dialyzed for 1 h at 4°C against Ringer's solution.

Vero cells from the American Type Culture Collection (CCL-81) were grown to confluence at 37°C in a 5% CO₂ atmosphere in 100-mm-diameter petri plates (Falcon). They were harvested by treatment with trypsin and diluted with DMEM to obtain approximately 5 × 10⁴ cells/100 μl. Aliquots of 180 μl were loaded into each well of 96- or 24-well flat-bottom microtiter plates (Falcon). The cells were allowed to settle, attach, and grow for 24 or 48 h prior to use (80% confluence). Twenty microliters of filter-sterilized V. cholerae culture supernatants or 10-fold serial dilutions thereof were gently mixed into the medium overlying these Vero cells, and incubation was continued for between 30 min and 24 h as appropriate. The vacuolation titer was defined as the inverse of the highest dilution that gave at least 50% vacuolated cells in 24 h under these conditions (cytotoxic dosage). For neutral red vital dye staining, Vero cell monolayers pretreated with appropriately diluted V. cholerae culture supernatants were rinsed twice with Ringer's solution and treated with 100 μl of a 3.3-mg/ml solution of the dye. Results were recorded by time-lapse video as described below.

Hemolysis assays. Hemolytic activity was assessed by streaking bacterial cultures on Difco blood agar base medium containing 5% sheep red blood cells and scoring formation of transparent halos around single colonies after overnight incubation. Green halos, which are formed by some V. cholerae strains, were not considered to be indicative of hemolysis, since they generally result from nonspecific lysis caused by other bacterial metabolites released into the medium (26). More precise estimates of hemolytic activity were obtained by adding aliquots of diluted supernatants of V. cholerae cultures to a suspension of rabbit red blood cells (2%) in phosphate-buffered saline in 96-well (round-bottom) microtiter plates (Corning) and incubating overnight at 4°C, as suggested previously (26).

Protease sensitivity assays. Proteinase K-agarose beads (Sigma) were suspended at 1 mg/ml in distilled water and washed three times with ice-cold Tris HCl at pH 7.2, and the pellet from 1 ml of this washed suspension (1 U) was mixed with 500 μl of V. cholerae culture supernatant. The mixture was incubated for 10 min at 4 or 25°C with 1 mM CaCl₂, or without CaCl₂ as a negative control (proteinase K activity depends on Ca²⁺ ions). The proteinase beads were removed by centrifugation for 1 min at 13,000 rpm (in an Eppendorf microfuge), and bead-free supernatants were then tested for vacuolating cytotoxic activity on Vero cells as described above.

Time-lapse video microscopy. Vero cells were plated on 22- by 40-mm no. 1 1/2 glass coverslips at a density of 10⁵ cells/ml and cultured overnight in DMEM with 10% fetal calf serum without antibiotics or growth factors. Thereafter, they were exposed for 0.5 to 3 h to a 1/500 dilution of the supernatants of V. cholerae cultures grown for 18 h. For “still” light microscopy the cells were fixed with 2% glutaraldehyde, then with 1% tannic acid, and finally with 1% osmium tetroxide and then embedded in epoxy resin exactly as would be done for traditional electron microscopy, but they were then mounted flat for viewing by phase-contrast light microscopy. In certain experiments, cells were exposed for 1 to 15 min at 37°C to a 10-mg/ml concentration of exogenous horseradish peroxidase (HRP) (Sigma type VI) in Ringer's solution containing 1% bovine serum albumin (BSA) (RB-BSA) and then washed in RB-BSA for various periods to chase HRP into various parts of the endosomal system.

For viewing, coverslips were assembled into Zigmond chambers (42) and flushed with RB-BSA at 37°C. They were then placed on a 37°C temperature-controlled stage of an inverted microscope (IM-35, Carl Zeiss, Inc. Thornwood, N.Y.) and observed using a 63× NA1.25 Zeiss Antiflex lens designed for phase-contrast optics, interference reflection optics, or fluorescence optics. Time-lapse images were collected at 1 frame per s using an optical memory disk recorder (Panasonic OMDR TQ3038) and transferred to a computer for final generation of still frames or Quicktime movies.

To test various agents for their effects on vacuolation, test solutions were flowed over Vero cell monolayer cultures that had been prevacuolated by treatment with a 500-fold dilution of the supernatant of a highly vacuolating strain of V. cholerae (strain 52201) for 1 to 2 h. When appropriate, cultures were then stained by subsequently flowing a solution of acridine orange or neutral red at 10 ng/ml in RB-BSA into the Zigmond chamber.

DNA extraction. Cosmid or small plasmid DNAs were extracted by alkaline lysis and use of Qiagen columns. For cloning and subcloning, cosmid and plasmid DNAs were cut with appropriate restriction enzymes and fragments were separated by gel electrophoresis in 1% agarose–1× Tris-acetate-EDTA (TAE) at 80 V for 20 min. Cloned DNA fragments were recovered by cutting out the corresponding gel slice and purifying the desired DNA fragment with a Gene Clean kit (Bio 101, Inc., La Jolla, Calif.).

Genomic DNA samples for Southern blot analysis were prepared using a standard sodium dodecyl sulfate (SDS)-lysozyme lysis method (5), including cetyltrimethylammonium bromide to eliminate carbohydrates in some cases.

Five micrograms of genomic DNA was digested with BglI and electrophoresed in 0.8% agarose gels in 1× TAE buffer at 1.4 V/cm for 18 h. The gels were then blotted overnight onto nylon membranes (Hybond; positively charged) prepared in 20× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate), and the transferred DNA was fixed with UV light (Stratagene). Standard prehybridization and hybridization solutions contained 6× SSC, 5× Denhardt's solution, 0.5 (wt/vol) SDS, and 20 ng of herring sperm DNA per ml. Hybridization using ³²P-labeled probe DNA was carried out overnight at 65°C. The filters were then washed at low stringency (20 min in 6× SSC or 4× SSC at room temperature) or higher stringency (2× SSC at 40°C for 10 min) stringency as needed to remove nonspecifically bound label.

For colony hybridization, two sets of 50 colonies were grown for 4 h on Luria agar at 37°C and then directly transferred to nylon filters (18). Hybridization was carried out as described above, with a high-stringency wash (2× SSC at 40°C for 10 min, twice) to avoid nonspecific background hybridization. The H. pylori vacA DNA probe consisted of a 3.2-kb EcoRI fragment from plasmid pCTB6 (12). The ctxA (564-bp) and ctxAB (1,019-bp) CT probes consisted of PCR products from genomic DNA of V. cholerae O1 strain C-6706 (20). The ctxA probe was generated using the primers ctx 2 (5′-CGG GCA GAT TCT AGA CCT CCT G) and ctx 3 (5′-CGA TGA TCT TGG AGC ATT CCC AC). The ctxAB gene probe was generated using the ctx 2 primer and ctxB (5′-GCC ATA CTA ATT GCG GCA ATT GC). PCR was carried out using standard amplification protocols (8). Products were electrophoresed in agarose gels, purified using a Gene Clean Kit (Bio 101), and labeled by the random hexamer primer procedure (Stratagene) with [α-³²P]dCTP (Amersham).

Random amplified polymorphic DNA (RAPD) typing was carried out as described for H. pylori (2), using primers 1247 (5′-AAGAGCCCGT 3′) and 1281 (5′-AACGCGCAAC 3′).

Cosmid cloning and subcloning. Genomic DNA from V. cholerae strain 52201 was partially digested with Sau3A, and a 30- to 50-kb fraction was selected after electrophoresis in 0.6% agarose (8). The DNA population was cloned into pLAFR-5 vector DNA (9) that had been digested with BamHI and then treated with calf intestine alkaline phosphatase to prevent self-ligation. DNA samples in the ligation mix were packaged into lambda phage heads (Amersham kit), and the phage particles were used to transduce E. coli S-17 to kanamycin resistance. In subsequent subcloning experiments, cosmid DNA was partially digested with Sau3A and size fractionated as described above. DNA in the 2.5- to 3.5-kb size range was subcloned into BamHI- and phosphatase-treated pBluescript plasmid DNA, and ligated DNAs were used to transform E. coli DH5α.

DNA sequencing. DNA cloned in pBluescript plasmids was sequenced using standard primers specific for sequences flanking the cloning site (T7 [5′-GTA AAA CGA CGG CCA GT 3′] and M13-reverse [5′-GGA AAC AGC TAT GAC CAT G 3′] [Stratagene]), using the Sequenase version 2.0 kit (Amersham) with ³⁵S labeling (37).

PCR-based cloning of the hlyA (hemolysin) gene. The hlyA gene was PCR amplified from genomic DNAs from various V. cholerae strains. The primers used were Hly Fwd (5′-CTG TCT AGA [XbaI] AGT GAG GTT TAT ATG CCA AAA CTC AAT CGT) and Hly Rev (5′-CTG CTC GAG [XhoI] TTA GTT CAA ATC AAA TTG AAC CCC TTT CAC CAA). PCR was carried out in a volume of 15 μl containing 0.5 U of TaqPlus DNA polymerase (Stratagene), 1.5 μl of buffer high, 2.5 mM each deoxynucleoside triphosphate, and 10 pmol of each primer for 30 cycles of 94°C for 1 min, 60°C for 1 min, and 72°C for 1 min. DNA amplified with these primers was cloned into XbaI- and XhoI-digested pBluescript DNA for directional cloning or into pBS prepared by T addition (28) when needed for cloning undigested PCR fragments.

Vacuolating activity of non-O1 V. cholerae strains. Dilutions of culture supernatants from 111 of the 335 non-O1 V. cholerae strains isolated from Mexican patients with diarrhea in 1993 and 1995 (88 of 268 and 23 of 67 tested, respectively) were found to cause vacuolation of cultured Vero cells (Fig. 1). Arbitrarily primed PCR DNA fingerprinting (RAPD) of 37 such isolates yielded a reproducibly different pattern in each case, as illustrated in Fig. 2. Thus, vacuolation was a feature of many different V. cholerae strains, not just of one or a few widespread epidemic clones. Vacuolating activities had been found independently in nonepidemic V. cholerae strains from India (29) and Brazil (10), indicating that this phenomenon is widespread.

FIG. 1 Temporal progression of cell vacuolation caused by V. cholerae hemolysin (phase-contrast light microscopy; magnification, ×200). Top row, control cells, not exposed to toxin, showing no vacuolation. Second row, control for the exacerbation of (more ...)

FIG. 2 RAPD fingerprinting of representative V. cholerae strains using two different informative primers, 1281 and 1247. The five-digit number above each pair of lanes indicates the strain tested. Both 20 and 5 ng of DNA were used in duplicate PCR to determine (more ...)

Initially this vacuolation seemed reminiscent of that caused by the vacuolating cytotoxin (VacA) of H. pylori, although far stronger and faster acting. No hybridization was detected, however, between genomic DNAs of any of 30 vacuolating V. cholerae strains and a 3.2-kb H. pylori vacA gene fragment, even under lowest possible stringency conditions (6× SSC, room temperature), either in dot blot or in more sensitive Southern blot hybridizations. Similarly, no hybridization was detected with CT gene probes (ctxA and ctxAB), indicating that vacuolation was not due to an unusual form of CT. Further tests (see below) showed that the V. cholerae and H. pylori vacuolating activities are unrelated mechanistically.

The level of vacuolating activity varied among V. cholerae strains. One of the most active was that of strain 52201: a 2,000- to 5,000-fold-diluted aliquot of its supernatant was sufficient to cause vacuolation of all Vero cells in a monolayer within 2 h. Striking Vero cell rounding, detachment, and lysis were evident with less dilute aliquots of culture supernatants, but vacuolation was still evident when Vero cells were observed soon after addition of V. cholerae culture supernatant. Equivalent cell rounding was seen with supernatants of each of the other vacuolating strains. The cell rounding effect seen here was reminiscent of that attributed to the RTX toxin of epidemic strains (27), but gene cloning (see below) indicated that these two cell rounding phenomena are distinct. Most further analysis of the vacuolating activity detailed below was carried out using the hyperactive 52201 strain.

Physical characteristics of vacuolating factor. Vacuolating activity was essentially abolished by heating for 5 min at 55°C but not by heating for 5 min at 50°C. It was also 75% inactivated by 10 min of treatment with proteinase K-agarose beads at 25°C, provided that 1 mM Ca²⁺, an ion essential for proteinase K activity, was present. Collectively, these results suggested that vacuolation was caused by a secreted protein.

Time-lapse light microscopic characterization of vacuolation. Time-lapse viewing of acute vacuolating effects of V. cholerae supernatants provided a very different image from that obtained with H. pylori VacA protein. First, the vacuolating effect appeared as early as 30 min after addition of relatively undiluted supernatant and appeared proportionately later when more dilute aliquots of supernatant were used (60 to 90 min with a 1/500 dilution and overnight with a 1/5,000 dilution). Second, the vacuolating effect initially involved membrane compartments distributed throughout the cell (Fig. 1, row 3). Addition of 300 mM sucrose promptly reversed this diffuse vacuolation by making the medium of Vero cells hypertonic (Fig. 3). This is analogous to the vacuolating effect of the Aeromonas aerolysin toxin (even including its osmotic sensitivity [G. Van Der Goot, personal communication]). It is unlike the action of VacA, which we have found does not cause acute vacuolation of cells, even in massive cytotoxic doses (T. L. Cover and J. E. Heuser, unpublished data).

FIG. 3 Osmotic reversal of the vacuolating effect of V. cholerae hemolysin. Left column, after 3 h of exposure to a 1/500 dilution of V. cholerae supernatant, all cells are clearly vacuolated (the top two more diffusely than the lower one). Center column, addition (more ...)

The characteristically wide distribution of vacuoles initially induced by V. cholerae supernatants evolved with time or increasing dose to a more limited, perinuclear distribution of large multivesicular bodies (MVBs) indistinguishable from that produced by H. pylori VacA toxin (Fig. 1, row 5). To determine whether these late multivesicular vacuoles were endosomal in origin, as they are in the case of VacA (11), we prevacuolated cells with a sufficiently high dose of V. cholerae supernatant to create the large late vacuoles and then allowed them to take up the extracellular tracer HRP (Fig. 4). The location of HRP was then determined by standard diamine berecidine (DAB) histochemistry, which gives a dark reaction product visible by bright-field or phase-contrast microscopy. Normally, exogenous HRP promptly enters the endosomal system of all cultured cells (Fig. 4, row 1). Likewise, in prevacuolated cells it appeared to enter in normal amounts (Fig. 4, row 2) and at normal rates (Fig. 4, row 3). Nevertheless, this endocytosed HRP did not gain access to the large phase-lucent vacuoles generated by V. cholerae supernatant. This indicated that these vacuoles were not part of the Vero cells' normal endocytic circuit.

FIG. 4 The endocytotic tracer protein HRP is denied access to V. cholerae hemolysin-induced vacuoles, and thus they either are not derived from endosomes or are at least “off the circuit” of normal endosomal trafficking. Top row, control cells (more ...)

To further test the inference that the V. cholerae vacuoles are endosomal in origin, other cultures were treated in the opposite sequence, namely, by loading with HRP first before treatment with V. cholerae supernatants (Fig. 5). A mild swelling of the HRP-containing endosomes was produced by very high doses of V. cholerae supernatant, a mild effect on this organelle. An additional, more important feature revealed was centripetal movement of HRP-loaded endosomes caused by exposure to V. cholerae supernatant (Fig. 5, rows 2 to 4), culminating in tight clustering of HRP endosomes at the microtubule-organizing site (MTOC) just beside the nucleus. This was most striking when viewed by time-lapse light microscopy, but was also readily apparent in still images (Fig. 5). Earlier experiments had shown that such movement is pathognomic of cell alkalinization (23), and thus we conclude that the V. cholerae agent causes severe cell alkalinization.

FIG. 5 The dynamics of HRP-bearing endosomes indicate that V. cholerae hemolysin causes extreme cell alkalinization. Vero cells were prepared as described for Fig. 4, but were photographed with bright-field rather than phase-contrast microscopy to better see (more ...)

Further time-lapse viewing demonstrated that maximum centripetal movement could be achieved by placing Vero cells in hypertonic sucrose to block vacuolation during treatment with V. cholerae supernatant (Fig. 3) and releasing the sucrose block minutes after alkalinization by washing the Vero cell monolayer with fresh isotonic medium (Fig. 5, row 4). This treatment caused an immediate vacuolation, indicating that V. cholerae toxin had indeed acted on these cells during the osmotic inhibition of vacuolation. These results showed that the ionic imbalance leading to vacuolation and cytoplasmic alkalization proceeded even in the presence of sucrose and that centripetal endosome movement was retarded by the growing vacuoles themselves.

Studies of accumulation of weak basic dyes further differentiated the vacuoles observed here from those produced by H. pylori VacA. All H. pylori VacA vacuoles rapidly accumulated neutral red or acridine orange (Fig. 6, bottom row), swelled in weak bases like ammonia, and promptly collapsed when vacuolated cells were exposed to inhibitors of the vacuolar-type proton pump such as bafilomycin and concanamycin (32; J. Heuser, unpublished data), which indicated that they are strongly acidic. In contrast, about 25% of the early V. cholerae vacuoles were acidic enough to accumulate visible amounts of neutral red (Fig. 6, top row), and only 25% of them—even the large late V. cholerae vacuoles—became further enlarged when vacuolated cells were exposed to various proton pump inhibitors (Fig. 1, row 4 versus row 3; row 2 shows the relevant control).

FIG. 6 Neutral red uptake in Vero cells vacuolated with V. cholerae hemolysin (top row) versus H. pylori VacA (bottom row). Top row, Vero cells were exposed to a 1/500 dilution of a V. cholerae supernatant for 3 h and then allowed to take up neutral red for (more ...)

These conclusions should be qualified in one respect. Most large, late V. cholerae vacuoles displayed many small internal vesicles that appeared to move in Brownian motion when viewed by time-lapse microscopy. These promptly precipitated when Vero cells were exposed to neutral red (as do vesicles found in MVBs of H. pylori VacA-treated cells [J. Heuser, unpublished data]). This indicates that the V. cholerae vacuoles had accumulated at least traces of this weak base and thus were at least mildly acidic, even though they did not stain overtly with neutral red. Other studies have indicated that the internal vesicles in multivesicular bodies are highly negatively charged on their surfaces (16), which is why they precipitate when agents like neutral red enter the MVBs.

V. cholerae vacuolation is caused by hemolysin (hlyA gene product). To search for the V. cholerae gene whose product caused vacuolation, we constructed a cosmid library from partially Sau3A-digested genomic DNA of strain 52201. Sets of 20 cosmid-containing clones were pooled, and 100 such pools were tested for Vero cell vacuolating activity. Ten pools with vacuolating effects were found, and the responsible cosmids were identified and fingerprinted by HindIII digestion. Different but related arrays were obtained from each of the six cosmids tested (Fig. 7), as expected of cosmids generated from partial Sau3A digest products. The gene responsible for vacuolating activity was defined more closely by subcloning and screening for vacuolating activity. The sequence was 93% matched to nucleotides 648 to 949 upstream of the V. cholerae hlyA gene under GenBank accession no. Y00557 (3) and 90% matched to nucleotides 2560 to 2348 downstream of the hlyA gene under GenBank accession no. D58374 (24). Independent studies of Indian and Brazilian V. cholerae strains had similarly attributed their vacuolating activities to hemolysin (10, 29).

FIG. 7 HindIII restriction fragment length polymorphism analysis of representative cosmid clones with vacuolating activity from the V. cholerae strain 52201 clone library. (A) Agarose gel electrophoresis of HindIII-digested DNAs. (B) Hybridization of HindIII (more ...)

To test the generality of our results, the hlyA gene was PCR amplified from six non-O1 strains and cloned directionally into a pBluescript plasmid vector, and vacuolating titers of supernatants of E. coli carrying these cloned DNAs were determined (Table 2). Considerable vacuolation was obtained in each case, but at a level that was strain specific and not particularly correlated with the activity detected in the ancestral V. cholerae strain. Most striking was the potent activity observed with hlyA cloned from strain 52453 (titer, >5,000), a strain that had not been seen to produce a vacuolating activity itself. The cloned hlyA genes from the other putatively nonvacuolating strain tested, 64401, and from the epidemic El Tor O1 strain C-6706 also produced modest but significant vacuolating activity in E. coli. No such activity was detected with control clones containing hlyA from strain O395, which contains a frameshift null mutant allele (4). Vacuolating activities from E. coli supernatants were not much increased by sonication (Table 2), indicating that low activities were not due to a specific secretion defect. These differences in relative vacuolating activities before and after cloning suggested that V. cholerae strains differ in the regulation of synthesis or secretion of their hemolytic and vacuolating toxin.

TABLE 2 Vacuolating titer following hlyA gene cloning in E. coli

Relatively undiluted (10- to 500-fold-diluted) supernatants of E. coli carrying hlyA from the hyper producer V. cholerae strain 52201 caused Vero cell rounding, detachment, and lysis equivalent to that seen with supernatants of the original V. cholerae strains themselves (see above). This implied that the HlyA protein has a cytotoxic effect superficially resembling that of the unrelated RTX toxin of epidemic strains.

Divergence between vacuolating and hemolytic activities. We compared vacuolation and hemolytic titers from several representative V. cholerae strains to learn whether these two activities depended on exactly the same features of HlyA protein. The vacuolation titer of strain 52201 was far higher than its hemolytic titer (5,120 versus <10), an imbalance also reflected in a cosmid clone and its 3-kb subclone derivative, whereas each activity from strain 69750 was high. Different ratios of vacuolation to hemolytic titers were also observed with other strains tested (Table 3). Thus, vacuolation and hemolysis may depend on somewhat different features of the HlyA protein, features that are polymorphic in the nonepidemic V. cholerae population.

TABLE 3 Strain specificity of ratio of vacuolating versus hemolytic activity

We found that many strains of V. cholerae associated with diarrheal disease in Mexico produce a potent, fast-acting eukaryotic cell vacuolating activity, and we showed that this activity is due to the HlyA hemolysin. Equivalent hlyA-encoded vacuolating activity has been found in many diarrheal but nonepidemic V. cholerae strains in India and Brazil (10, 29). Vacuolation is also caused by aerolysin from Aeromonas hydrophila and hemolysin from Serratia marcescens (1, 22), which are not closely related to HlyA of V. cholerae. Thus, a capacity for target cell vacuolation may be quite widespread among pathogens.

The V. cholerae HlyA-induced cell vacuolation was distinct from the much-studied vacuolation induced by the VacA protein of virulent H. pylori strains. The V. cholerae hlyA and H. pylori vacA genes do not share significant homology. More important, the vacuoles induced by these two toxins behaved differently in response to perturbants of the endosomal pathway, such as spermidine and NH₄Cl, which stimulated H. pylori-induced but not V. cholerae-induced vacuolation. In addition, all H. pylori-induced vacuoles took up neutral red uniformly (33, 35), whereas only half of the V. cholerae-induced vacuoles were stained. That V. cholerae hemolysin-induced vacuolation is independent of the endosomal system is emphasized by the enhancement of V. cholerae-induced vacuolation by concanamycin, a bafilomycin-like agent that reverses H. pylori-induced vacuolation by interfering with vacuolar-type H⁺-ATPases (14). A previous study of cell vacuolation by V. cholerae hemolysin (10) reported that bafilomycin did not block vacuolation, as noted here, but rather exacerbated vacuolation. This may reflect their having used Vero cell cultures 24 h after inducing cell vacuolation (10), by which time most vacuoles have progressed to a late, relatively acidic, autophagic-vacuole morphology. In contrast, we found that proton pump inhibitors strongly potentiate vacuolation, particularly at the early stages of V. cholerae hemolysin action, when most vacuoles are relatively nonacidic and are composed primarily of what appears to be endoplasmic reticulum (ER) membrane.

Our studies have not yet revealed (i) the exact subcellular origin of V. cholerae HlyA-induced vacuoles, (ii) how exactly they are formed, (iii) why concanamycin enhances their formation, and (iv) how V. cholerae hemolysin causes cell alkalinization. It is certain, however, that the vacuoles found many hours after intoxication by V. cholerae HlyA stain uniformly with neutral red and hence are acidic inside, while at early times (1 to 4 h after exposure) they generally do not stain with this dye and hence are relatively nonacidic. In a separate study of V. cholerae vacuolation, mixed populations of stained and nonstained vacuoles were found in HeLa cells many hours after initial exposure (29). Perhaps the peristence of such mixed vacuole populations reflects the toxin's apparent lower potency in this cell type. In any case, the diversity in early vacuole phenotypes observed here indicates a mechanism of vacuolation very different from that produced by H pylori VacA and more akin to that operating in aerolysin intoxication (1).

The much-studied mechanism of cell lysis (hemolysis) by HlyA seems to involve intercalation of single monomeric HlyA proteins into cell membranes, their coalescence to form highly ordered multimeric pores that allow selective leakage of certain ions (e.g., K⁺ but not Ca²⁺), and eventual cell disruption (43). HlyA protein exhibits remarkable differences in potency for erythrocytes from different animal species and cell types (44) and a dual specificity for cholesterol and ceramides (45). We propose that HlyA-induced vacuolation is distinct mechanistically from pore formation, based on the finding that strains differ in relative vacuolating and hemolytic activities. These differences are in accord with a formal model in which somewhat different structural components or domains of HlyA mediate cytolytic and vacuolating activities (25). One possibility is that vacuolation involves HlyA monomers, by analogy with vacuolation by H. pylori VacA (which must change from an inactive oligomer to an active monomer) (35). In this scenario, certain amino acid sequence differences between HlyA proteins from diverse V. cholerae strains could affect the relative stabilities or rates of interconversion of monomers and oligomers.

A functional hlyA gene is present almost universally in nonepidemic but pathogenic V. cholerae strains, but not in putative avirulent environmental strains (19, 26). Defective hlyA genes have also accumulated in epidemic (CT-producing) strains. This might illustrate how other, more potent virulence determinants can override the need for the milder toxic effects of HlyA protein (34). It had been parsimonious to imagine that HlyA-induced cytolysis was important to the virulence of human infections, but our results illustrate the need to also consider possible roles of vacuolation. As with vacuole formation caused by VacA of H. pylori or reagents like monensin (39), perhaps HlyA-induced vacuolation reflects interference with normal intracellular trafficking of other molecules, such as lysosomal hydrolases, and an impairment of lysosomal function in target cells (36). HlyA vacuolation might equally interfere with trafficking needed for antigen presentation and protective immune responses to infection (30). Alternatively, it might contribute rather nonspecifically to virulence, for example by wasting cellular energy through synthesis of excess internal membranes and thereby disrupting cellular homeostasis and intestinal cell function and integrity. Might vacuolation and cytolysis each contribute to virulence, but differently? Given the phenotypic diversity among HlyA activities from different clinical isolates and the ease of gene replacement and mutational analyses of V. cholerae (15), it should now be feasible to learn just which domains of HlyA protein are needed for which of its activities and the roles of each in human infection and disease.

Research at Washington University was supported by NIH grants AI38166 and DK53727 to D.E.B., GM29647 to J.E.H., and P30 DK52574 to Washington University. P.F.-A. was the recipient of a scholarship from CONACyT, Mexico.

We thank Timothy L. Cover for many helpful comments and suggestions; Lucina Gutiérrez and the Cholera Laboratory authorities for isolation, characterization, and typing of V. cholerae strains; and for allowing us access to the INDRE collection of V. cholerae strains.