INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Haemophilus ducreyi is a gram-negative organism responsible for chancroid (soft chancre, ulcus molle), a sexually transmitted disease characterized by painful, slowly healing genital ulcers and swollen regional lymph nodes (40). Renewed attention has been focused on H. ducreyi, since human immunodeficiency virus transmission in developing countries has been associated with genital ulceration (40). The general pathogenic mechanisms and immunity in chancroid, in particular, the virulence factors responsible for ulceration, remain unclear. The chancroid bacterium produces a toxin (HdCDT) which belongs to the recently discovered family of cytolethal distending toxins (CDTs) elaborated by various species of gram-negative bacteria (24). Three linked genes, cdtABC, encode these toxins, which are composed of the A, B, and C proteins, with molecular masses of 26, 30, and 20 kDa, respectively. Expression of all three genes is needed to create an active toxin, but it is not clear whether the mature (active) holotoxin contains one, two, or all three components. The cdtABC genes from different species are related to each other. The highest sequence homology (>95%) was noted between HdCDT and the corresponding CdtABC proteins from Actinobacillus actinomycetemcomitans (33), whereas homology to Escherichia coli CDTs is much lower (5). Among the three components, CdtB is the most conserved, with 49% identity even between the most distantly related CDTs (24).

All the CDTs known to date induce the same cytotoxic effect in sensitive epithelial target cells: an irreversible arrest in the G₂ phase of the cell cycle (4, 6, 45). This effect is accompanied by a conspicuous enlargement of most cells (24), but T lymphocytes are an exception (33). The earliest toxin-induced biochemical alteration identified so far is tyrosine phosphorylation of the cyclin-dependent kinase cdc2 (4), which was detectable within 6 h in cells intoxicated with HdCDT (6). As a result of tyrosine phosphorylation, this kinase remains inactive and thus is unable to promote the transition of cells from the G₂ phase into mitosis. The molecular events responsible for this effect have not yet been clarified for any of the CDTs. This effect resembles the so-called G₂ checkpoint response (44), known to occur after DNA damage induced by certain chemicals or exposure to ionizing radiation. However, no DNA damage was detectable in CDT-treated cells prior to tyrosine phosphorylation of cdc2, when measured as DNA synthesis (4, 6) or the presence of DNA strand breaks (32). The CDTs, particularly HdCDT, are extremely potent and might enzymatically affect one of the many components involved in the regulation of cdc2 activity. However, neither the putative enzymatic activity nor the molecular substrate has yet been defined.

Despite these uncertainties, it is reasonable to assume that the primary cellular target of HdCDT and the other CDTs is located intracellularly and that these toxins therefore have to enter cells before being able to exert their effect. Most bacterial protein toxins known to act on intracellular targets have been shown to enter cells by endocytosis (17). However, for instance, the heat-stable enterotoxin STa of E. coli induces its intracellular effect by triggering a transmembrane signal from outside the cells (42), and a similar signaling action is a possibility for the CDTs as well. Thus, the aim of the present work was to determine if cellular internalization of HdCDT is required for its cytotoxic action and, if so, to clarify the major events in the internalization pathway. The general experimental approach was to test whether the toxin would be able to intoxicate cells in which the endocytic pathway was blocked at defined stages.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Materials. HdCDT was purified from H. ducreyi by immunoaffinity chromatography using the neutralizing monoclonal antibody (MAb) M4D4 (26; A. Frisk, M. Lebens, H. Ahmed, C. Johansson, L. Svensson, K. Ahlman, and T. Lagergård, unpublished data). Immunoblotting with specific MAbs (Frisk et al., unpublished) showed that the toxin preparation contained both the B and the C components, while the A component was not detectable. The protein concentration in the stock solution of this partially purified preparation was 25 µg/ml. A rabbit antiserum reacting with the B and C components was prepared against this toxin preparation. L-[³H]leucine (specific activity, 40 to 60 Ci/mmol) was obtained from NEN Life Science Products (Boston, Mass.). Other reagents were obtained from Sigma Aldrich Sweden AB unless otherwise stated.

Cell culture and toxin treatment. Human epidermoid larynx carcinoma cells (HEp-2; ATCC no. CCL-23) were cultivated in Eagle's minimum essential medium (MEM) supplemented with 10% fetal bovine serum, 5 mM L-glutamine, penicillin (100 U/ml), and streptomycin (100 µg/ml) in a humid atmosphere containing 5% CO₂. Human cervix carcinoma cells transfected with dominant-negative dynamin (HeLa^dynK44A) (kindly provided by K. Sandvig, Institute for Cancer Research, Montebello, Oslo, Norway) were cultivated in Dulbecco's modified Eagle's medium with the same supplements as HEp-2 cells plus 400 µg of Geneticin per ml, 200 ng of puromycin per ml, and no tetracycline or 1 µg of tetracycline per ml. Exponentially growing cells (in 96- or 24-well plates) were cooled on ice for 15 min before the addition of ice-cold toxin (25 ng/ml unless otherwise stated). After toxin exposure for 15 min at 0°C, the cells were washed three times with cold Hanks' balanced salt solution (HBSS). Eagle's MEM was added, and the cells were further incubated at 37°C. Intoxication was scored as the accumulation of cells in the G₂/M phase by flow cytometry or as tyrosine phosphorylation of cdc2 as described below. These two parameters of intoxication were always in agreement.

Cell cycle analysis by flow cytometry. Cells growing in 25-cm² culture flasks (Costar, Cambridge, Mass.) were treated with 25 ng of HdCDT per ml, incubated for various times, rinsed with HBSS, scraped off (Cell Scraper; Nunc, Naperville, Ill.), and processed for flow cytometric analysis of the DNA content as previously described (6). The data from 10⁴ cells were collected with a FACSort flow cytometer (Becton Dickinson) and analyzed using CellQuest software (Becton Dickinson).

Western blotting. Cells were lysed in 300 µl of sodium dodecyl sulfate (SDS) electrophoresis sample buffer, and samples were boiled for 10 min. The proteins were separated on SDS-10% polyacrylamide gels, transferred to polyvinylidene diflouride membranes (Millipore), and probed with antiphosphotyrosine (Upstate Biotechnology) or anti-cdc2 or anti dynamin (Transduction Laboratories, Lexington, Ky.) antibodies. Blots were developed with an ECL Western blotting analysis kit (Amersham Pharmacia) using a peroxidase-labeled secondary antibody. The amount of protein in the lysates was determined by the Bio-Rad protein assay using bovine serum albumin as a standard; 25 to 30 µg was loaded per lane.

Quantification of tyrosine phosphorylation. Western blots were scanned, and densitometric analysis of tyrosine-phosphorylated cdc2 bands was performed using Image Quant software (Molecular Dynamics). Each experiment comprised four lanes: (i) a control without toxin or drug (C), (ii) a positive control with toxin only (T), (iii) the actual test with toxin and drug (TD), and (iv) a control with drug only (D). After densitometer scanning of these four bands (using a Personal Densitometer from Molecular Dynamics), the value for C was subtracted from that for T, and the value for D was subtracted from that for TD. The value for TD  D was divided by that for T  C, and the resulting ratio was multiplied by 100. This final value indicates the cdc2 tyrosine phosphorylation in the test, given as a percentage of the cdc2 tyrosine phosphorylation in the positive control.

Treatment with proteases. HEp-2 cells were exposed to HdCDT (15 min, 0°C), washed three times with HBSS, and treated with trypsin (2.5 mg/ml) or proteinase K (100 µg/ml) for 9 min at 37°C. The detached cells were seeded in a new well with normal medium and incubated for 24 h. Control cultures were treated (15 min, 0°C) with toxin (2.5 µg/ml) that had been preincubated for 9 min at 37°C with trypsin or proteinase K and further diluted 1/100 in normal medium to give a final concentration of 25 ng/ml. Samples for Western blotting were taken after 24 h.

Treatment with antibodies. (i) Neutralization before addition of toxin to HEp-2 cells. HdCDT was incubated for 30 min at 37°C with antibody diluted 1/100. Cells were treated with the toxin-antibody mixture (15 min, 0°C) and washed three times with HBSS. Cells were then incubated for 24 h at 37°C in fresh medium.

Potassium depletion. HEp-2 cells were cultivated in six-well plates, washed two times with HEPES buffer (pH 7.4) with 100 mM NaCl, and hypotonically shocked for 5 min at 37°C with K⁺ depletion buffer (140 mM NaCl, 2 mM CaCl₂, 1 mg of glucose per ml, 25 mM HEPES) diluted 1:1 in water. The cells were incubated in isotonic K⁺-free medium (50 mM HEPES, 100 mM NaCl, 2 mM CaCl₂, 1 mg of glucose per ml) for 1 h and then treated with HdCDT in the same medium. After the toxin treatment, the cells were incubated without K⁺ for 1 h and then in normal medium for 24 h.

18°C treatment. Cells were exposed to HdCDT (15 min, 0°C) and incubated for 15 min at 37°C to permit endocytic uptake into early endosomes. Thereafter, the cells were incubated at 18°C for 12 h or at 18°C for 12 h followed by 37°C for 12 h.

Treatment with drugs. Cells were preincubated in medium with a drug, exposed to HdCDT (15 min, 0°C), and postincubated in the presence of the drug. To exclude the possibility that the drug inhibited the binding of the toxin, the cells in each situation were treated with the toxin plus the drug and postincubated in drug-free medium. Cells were preincubated with various concentrations of filipin, chlorpromazine, imipramine, and ilimaquinone for 1 h at 37°C; with brefeldin A (BFA) for 45 min at 37°C; and with nocodazole for 30 min at 4°C. Preincubation with all other drugs was done for 30 min at 37°C. Treatments with ilimaquinone and filipin were performed with serum-free medium.

DT treatment. Cells were incubated with 5 µg of diphtheria toxin (DT) per ml for 30 min at 37°C, washed three times with HBSS, and incubated in normal medium for 1.5 h. Cells were then incubated with [H³]leucine (1 µCi/ml) in leucine-free medium for 1 h at 37°C. After trypsinization, the cells were treated for 30 min at 4°C with trichloroacetic acid (final concentration, 10%). The macromolecular fraction was precipitated by centrifugation (13,000 rpm in a Biofuge 13 [Heraeus] for 10 min), washed in 10% trichloroacetic acid, and dissolved in 2% Na₂CO₃-1% SDS-0.1 M NaOH, and 3 ml of scintillation liquid (OptiPhase HiSafe; Fisons Chemicals) was added. Sample counts were determined with a liquid scintillation counter (LKB Wallac).

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Cell surface-bound HdCDT is not accessible to activation with proteases or antibodies. Preincubation of HdCDT with trypsin or proteinase K before addition to cells inactivated the toxin. However, when the cell surface-bound toxin was treated under the same conditions (9 min at 37°C) with the same concentrations of trypsin or proteinase K, the cells were still strongly intoxicated, as measured by cdc2 tyrosine phosphorylation. A similar result was found with the polyclonal antiserum against the toxin. Incubation of HdCDT with the antiserum for 30 min at 37°C before the binding step neutralized the toxin, whereas antiserum added directly after the binding step was no longer able to neutralize the toxic effect. These findings suggested that after binding to the cell surface and upon elevation of the temperature, the toxin either is conformationally changed to become resistant to these treatments or is rapidly internalized into a compartment where it is no longer accessible to proteases or antibodies.

HdCDT is internalized by endocytosis. Cells treated with ammonium chloride or methylamine were partially protected against intoxication by HdCDT (Fig. 1), cdc2 tyrosine phosphorylation being reduced to 33.9% ± 13% and 74.0% ± 7% by ammonium chloride and methylamine, respectively. Upon treatment with monensin, the toxic effect was inhibited, as shown by the lack of tyrosine phosphorylation of cdc2 (Fig. 1) and flow cytometric analysis (Fig. 2). There was no evident accumulation of cells in the G₂ phase 12 h after toxin exposure, in contrast to the results for toxin-treated control cells without monensin. None of these drugs blocked the cell surface binding of the toxin per se, as tested both by preincubating the cells with a drug and by allowing its presence during, but not after, the toxin-binding step (data not shown). In conclusion, HdCDT appears to require passage via an intracellular low-pH compartment in order to intoxicate cells with full efficiency, suggesting that the toxin is internalized by endocytosis.

View larger version (42K): [in this window] [in a new window]   FIG. 1.   Effects on HdCDT-induced intoxication of drugs which inhibit endosomal acidification. Cells were preincubated with methylamine (10 mM), NH4Cl (20 mM), or monensin (10 µM), treated with the toxin for 15 min, and postincubated for 12 h in the presence of the respective drug. Tyrosine phosphorylation of cdc2 in the treated cells was determined by Western blotting as described in Materials and Methods. Blots are representative of two different experiments with each drug.

View larger version (30K): [in this window] [in a new window]   FIG. 2.   Effect of monensin on HdCDT-induced intoxication. Flow cytometry was used to analyze control cells [HdCDT()] and toxin-treated cells [HdCDT (+)] in the presence (+) or absence () of monensin (10 µM). Samples were prepared 12 h after toxin treatment, and DNA was stained with propidium iodide. The G1 and G2/M regions were marked, and the percentages of cells in these phases are shown. One representative experiment of two is shown. FL3-H, relative fluorescence.

HdCDT is endocytosed via clathrin-coated pits. We tested the hypothesis that endocytosis of HdCDT takes place via clathrin-coated pits. Three different experimental approaches were used: (i) removal of clathrin coats by K⁺ depletion, (ii) treatment with drugs known to inhibit receptor clustering in coated pits, and (iii) use of a cell line genetically manipulated to fail in endocytosis via clathrin-coated pits.

View larger version (18K): [in this window] [in a new window]   FIG. 3.   Effect of K+ depletion and chlorpromazine treatment on tyrosine phosphorylation of cdc2 in HdCDT-treated cells. Cells were depleted of K+ as described in Materials and Methods and treated with HdCDT; samples were prepared 24 h after toxin exposure. For chlorpromazine treatment, cells were preincubated with chlorpromazine (25 µg/ml, 1 h), treated with HdCDT, and postincubated for 8 h with 10 µg of the drug per ml. Intoxication with DT and measurement of DT activity (inset) were performed as described in Materials and Methods. Error bars indicate standard deviations.

View larger version (27K): [in this window] [in a new window]   FIG. 4.   Effect of HdCDT on HeLadynK44A cells. Tyrosine phosphorylation of cdc2 after HdCDT treatment was measured in cells expressing only endogenous dynamin (tetracycline +) and in cells overexpressing dominant-negative dynamin (tetracycline ). To obtain overexpression, cells were grown in the absence of tetracycline for 2 days before the toxin treatment. Samples were taken 24 h after toxin exposure. Blots are representative of four different experiments. P-Tyr, phosphotyrosine.

HdCDT requires transport from early endosomes to downstream vesicular compartments. The next few experiments were designed to clarify whether the toxin can be translocated to the cytosol directly from early endosomes or needs to be transported to some downstream compartment(s) before being able to act. (i) Bafilomycin A1 (BafA1) is a specific inhibitor of vacuolar proton ATPases and is known to raise the vesicular pH and block protein transport from early to late endosomes (1). Pretreatment of HEp-2 cells with BafA1 completely inhibited the intoxication induced by HdCDT (Fig. 5A). (ii) The microtubule-disrupting agent nocodazole is known to block the fusion of endosomal carrier vesicles with downstream compartments, such as late endosomes, lysosomes, or the Golgi complex (1). We observed that nocodazole treatment of HEp-2 cells almost completely inhibited HdCDT-induced intoxication, whereas its addition 1 h after toxin exposure could not prevent the intoxication (Fig. 5B). (iii) Incubation of cells at 18°C is another treatment known to block the transport of proteins from endosomes to lysosomes and/or the Golgi complex (28). After the toxin-binding step (15 min, 0°C), followed by a 15-min incubation at 37°C to allow initial endocytic uptake, the cells were incubated for 12 h at 18°C. This treatment strongly reduced the intoxication scored as tyrosine phosphorylation of cdc2 (Fig. 5C). After incubation at 18°C, a parallel set of similarly treated cells was incubated for another 12 h at 37°C. This shift in temperature resulted in complete intoxication compared to the results obtained for control cells incubated at 37°C for the full 24 h (Fig. 5D). These results demonstrate that HdCDT was unable to act when trapped in early endosomes.

View larger version (25K): [in this window] [in a new window]   FIG. 5.   Effects of BafA1 (50 µM), nocodazole (30 µM), and 18°C incubation on tyrosine phosphorylation of cdc2 in HdCDT-treated cells. (A) Cells were exposed to BafA1 before and after toxin treatment. The postincubation time was 24 h. (B) Cells were exposed to nocodazole before and directly after toxin treatment (panel 1) or received nocodazole 60 min after toxin treatment (panel 2). The postincubation time was 8 h. (C and D) Cells were cultivated in eight individual petri dishes; four were kept as controls, and four were exposed to the toxin at the same time. (C) One pair of plates (HdCDT  and HdCDT +) was incubated (Inc.) at 37°C and the other was incubated at 18°C. Samples were prepared 12 h after toxin treatment. (D) One pair of plates was incubated at 37°C for 24 h. The other was incubated at 18°C for the first 12 h (Inc. 1) and then transferred to 37°C for another 12 h (Inc. 2). Blots are representative of three different experiments with each treatment.

An intact Golgi complex is required for intoxication. BFA is a drug well known for its ability to disrupt the Golgi complex (2). When HEp-2 cells were treated with BFA before and after toxin exposure, intoxication was completely inhibited, as scored by flow cytometric analysis after 6, 8, and 12 h (Fig. 6A to C). Since BFA has been reported to interfere with the cell cycle as well (19), an additional experiment was performed to ensure that the inhibition by this drug was actually due to disruption of the Golgi complex. BFA was added only 1 h after toxin exposure; in this situation, it did not interfere with intoxication (Fig. 6D), implying that internalization was the step prevented by BFA, as shown in Fig. 6A to C. The protective effect of BFA was also determined as a lack of tyrosine phosphorylation of cdc2 (Fig. 7). Similar results were obtained after treatment of the cells with ilimaquinone (Fig. 8), another Golgi complex-disrupting drug (39). Interestingly, after the removal of BFA from toxin-treated cells, intoxication did develop within 12 h (data not shown). Thus, the endocytosed toxin remained active in some compartment that could fuse with Golgi vesicles only upon restoration of the Golgi complex. Taken together with the finding that HdCDT could not act when trapped in endosomes, these observations strongly suggest that the toxin is transported to the Golgi complex and delivered to the cytosol either from there or from the endoplasmic reticulum (ER).

View larger version (60K): [in this window] [in a new window]   FIG. 6.   Effect of BFA on HdCDT-induced intoxication. Flow cytometry was used to analyze control [HdCDT ()] and toxin-treated [HdCDT (+)] cells in the presence (+) or absence () of BFA. Cells were pretreated with BFA (2.5 µg/ml) for 45 min, exposed to the toxin, and postincubated in normal medium with BFA. Samples were prepared 6 h (A), 8 h (B), and 12 h (C) after toxin treatment. (D) Sample from cells treated with toxin, postincubated for 45 min at 37°C in fresh medium to allow internalization of the toxin, and exposed for 12 h to BFA. Percentages of cells in G1 and G2/M are shown. One representative experiment of three is shown. FL3-H, relative fluorescence.

View larger version (42K): [in this window] [in a new window]   FIG. 7.   Effect of BFA on the tyrosine phosphorylation of cdc2 in HdCDT-treated cells 8 h (A) and 12 h (B) after toxin treatment. Blots are representative of three different experiments. P-tyr, phosphotyrosine.

View larger version (47K): [in this window] [in a new window]   FIG. 8.   Effect of ilimaquinone (Ilimaq.) on HdCDT-induced intoxication. Shown are Flow cytometric analysis (A) and tyrosine phosphorylation of cdc2 (B) in control [HdCDT ()] and toxin-treated [HdCDT (+)] cells with (+) or without () ilimaquinone. One representative experiment of two is shown.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Like all macromolecules having intracellular sites of action, protein toxins acting intracellularly have to cross the plasma membrane at some stage before being able to attack their targets. The known pathways for receptor-mediated uptake of physiologically important macromolecules are endocytosis via clathrin-coated pits or clathrin-independent endocytosis, for example, via caveolae (30). Regardless of the mode of primary uptake, endocytic vesicles fuse and form endosomes that mature by additional fusion processes. After this stage, endocytosed macromolecules may be delivered either to the lysosomes for degradation or to the Golgi complex for other kinds of processing. In the Golgi compartment, the endocytic and exocytic pathways meet each other. Mechanisms exist for retrograde transport from the Golgi compartment to the ER of macromolecules containing appropriate signaling sequences (3, 12, 13).

Toxins are masters of exploiting normal cellular physiologic processes for their own benefit and can enter cells after binding to almost any surface structure and using either a clathrin-dependent or a clathrin-independent pathway. Some toxins, for instance, ricin, are able to use both clathrin-dependent and clathrin-independent pathways for initial entry into cells (30). With respect to the subsequent internalization steps, two major groups of toxins have been defined: (i) toxins translocated directly from endosomes to the cytosol (DT [28] and anthrax toxin [37]) and (ii) toxins which are transported all the way to the ER via the Golgi complex before being translocated to the cytosol. Recent data indicate that this latter pathway is used by the majority of toxins studied to date. Thus, Shiga and Shiga-like toxins (13, 29), pertussis toxin (8, 10), cholera toxin (21), E. coli heat-labile enterotoxin (14), Pseudomonas exotoxin A (46), and the plant toxin ricin (27, 36) are transported in a retrograde manner to the ER via the Golgi complex. To date, no toxin has been reported to translocate at the level of the Golgi compartment.

We have previously reported that HdCDT, like other CDTs, causes irreversible arrest in the G₂ phase of the cell cycle, although the molecular target of the toxin is not yet known (6). In the present study, we provide biochemical data strongly supporting the notion that HdCDT requires internalization by endocytosis in order to intoxicate HEp-2 and other cells. We demonstrate that HdCDT uses a clathrin-dependent pathway. This finding was indirectly suggested by rapid disappearance of the toxin from the cell surface, consistent with rapid macromolecular entry via the clathrin-dependent pathway. Apparently, the time needed for warming up the cells enough to permit proteases and antibodies to act sufficed for entry of the toxin or at least for its clustering into coated pits in a form inaccessible to antibodies or proteases. More direct evidence for the clathrin-dependent pathway was established using cells depleted of potassium or treated with chlorpromazine or imipramine, conditions under which clathrin-dependent uptake of ligands into cells is known to cease (38, 43) and which protect cells from intoxication by DT (20).

The clathrin-dependent uptake of HdCDT in HEp-2 cells was corroborated by the finding that HeLa^dynK44A cells (overexpressing the dominant-negative dynamin) were resistant to the toxin. The importance of dynamin in clathrin-mediated endocytosis is well established (7, 31). It is needed for the constriction of coated pits and the subsequent budding of the coated vesicles. Overexpression of dominant-negative dynamin was shown to specifically block endocytic clathrin-coated vesicle formation, although fluid-phase uptake continued (7). The possibility that active toxin could enter nonspecifically by fluid-phase uptake was thus excluded. The roles of dynamin and its partners in other intracellular trafficking events are less clear. However, the toxic effect of ricin was recently shown also to be inhibited by the overexpression of dominant-negative dynamin, despite the fact that this toxin can enter cells via clathrin-independent endocytosis (9, 35). In this situation, the delivery of ricin from endosomes to the Golgi complex was inhibited (16). The complete inhibition of the effect of HdCDT in HeLa^dynK44A cells might therefore be due to both the inhibition of clathrin-dependent endocytosis and an effect on a later step of the endocytic pathway.

The partial inhibition of intoxication by methylamine and ammonium chloride, in contrast to the complete inhibition by monensin or BafA1, suggests that a raised intraendosomal pH per se can only delay HdCDT-induced intoxication. Monensin at the concentration used in this study has been reported to affect the Golgi complex as well (18), and this effect might explain its complete inhibition of intoxication. In addition, monensin was previously shown to inhibit the proliferation of cells exposed for more than 24 h (15), implying that it might interfere with the cell cycle. However, we used a shorter incubation time (12 h), and the flow cytometric analysis did not indicate any major change in cell cycle distribution (Fig. 2).

In HEp-2 cells, BafA1 appears not to affect the formation and maturation of multivesicular bodies, which are the endosomes of this particular cell line. However, delivery from such mature endosomes to lysosomes was reduced by BafA1 (41), and this drug was also reported to block transport from early to late endosomes in HeLa cells (1). With this background, the complete inhibition of intoxication by BafA1 in HEp-2 cells, in contrast to the partial inhibition by agents that only neutralize the endosomal pH, suggested that HdCDT might need some further transport along the endocytic pathway before being able to act. This conclusion was also supported by the inhibition obtained with the microtubule-disrupting agent nocodazole added before toxin internalization and by the fact that incubation of toxin-treated cells at 18°C was protective against the toxic effect. The complete restoration of intoxicating ability upon transfer of the cells back to 37°C (Fig. 5C) confirmed that the toxin was entrapped in vesicles that were not able to fuse with compartments downstream of the endocytic pathway at 18°C. Furthermore, the toxin was obviously protected from degradation during the delay in these vesicles.

HdCDT failed to intoxicate cells pretreated with BFA, suggesting that the relevant downstream compartment is the Golgi complex. BFA disrupts the Golgi complex, with redistribution of its proteins to the ER as well as inhibition of vesicular transport from the ER to the Golgi complex (2). It is well established that toxins internalized via the Golgi complex can be inhibited by BFA. This finding has been demonstrated for all the above-mentioned toxins, which undergo retrograde transport to the ER via the Golgi complex. For CDTs, however, the interpretation is complicated by the fact that BFA per se can interfere with the cell cycle, arresting cells to a certain extent in the G₁ phase (19). For this reason, incubations with BFA were kept as short as possible (6 to 12 h), thereby avoiding pronounced G₁ arrest due to BFA. In addition, intoxication did occur when BFA was added 1 h after the toxin (Fig. 6D). This result indicates that BFA interfered with a step taking place during the first hour of intoxication, i.e., the internalization. Our observations with BFA were corroborated with ilimaquinone, another drug causing the fragmentation of Golgi membranes and their dispersion throughout the cytoplasm (22, 39). The actions of BFA and ilimaquinone differ in that the latter does not induce retrograde transport of Golgi enzymes to the ER (39). Thus, two drugs that disrupt the Golgi complex in different ways inhibited HdCDT-induced intoxication.

In conclusion, all our results support the notion that HdCDT undergoes clathrin-dependent endocytosis and vesicular transport at least to the Golgi complex before it can induce arrest in the G₂ phase of the cell cycle. It remains to be determined experimentally whether HdCDT is transported in a retrograde manner from the Golgi complex to the ER and translocated to the cytosol (or possibly to the nucleus) from there. We have not yet succeeded in detecting HdCDT in any specific intracellular compartment, probably because the toxin is very potent and only a few molecules may enter at the final destination. This is a common problem with all toxins transported in a retrograde manner; for ricin, this problem was recently solved by introducing N-glycosylation sites to demonstrate passage via the ER (27).

All the toxins previously found to enter via the Golgi complex have been either implied to or actually demonstrated to continue to the ER before being translocated to the cytosol. The mechanism for retrograde transport is not yet clear. The KDEL retrieval system is exploited by Pseudomonas exotoxin A but not by Shiga-like toxin I (12). Toxins translocated from the ER to the cytosol have been suggested to disguise themselves as misfolded proteins, thereby succeeding in being transported across the ER membrane (11, 17). The fate of proteins being extruded from the ER in this manner is usually ubiquitination and proteasomal degradation (25). How toxins avoid this fate is not known, but an interesting hypothesis is that they resist ubiquitination because of a relative lack of lysines (11). Indeed, toxins transported in a retrograde manner have very few lysines, which are located only near the N or C termini of the active toxin component. In contrast, both DT and anthrax toxin, which are translocated from early endosomes (28, 37), have several lysines scattered all along their active fragments. Interestingly, the entire CdtB component of HdCDT contains only 3 lysines, which all are near the N terminus, whereas the CdtC component has 13 lysines scattered along the entire molecule. A similar pattern of lysine distribution is found in the CdtB and CdtC components of all the other CDTs. Recent work suggested that the A. actinomycetemcomitans CdtB component alone is capable of inducing G₂ arrest (33, 34). Considering the lysine pattern and the internalization data presented here, this suggestion is consistent with the possibilities that the CdtB protein of the CDTs is the active component and that it is transported to the Golgi complex and from there probably to the ER.

In conclusion, this work indicates that HdCDT exerts its action intracellularly and not by transmembrane signaling. This is the first member of the family of CDTs for which cellular internalization and some details of the pathway have been demonstrated. Based on the similarities in sequence and action among all the CDTs, it is likely that they all need to be internalized via the Golgi complex before being able to intoxicate cells.