To investigate the cell entry and intracellular trafficking of anthrax oedema factor (EF) and lethal factor (LF), they were C-terminally fused to the enhanced green fluorescent protein (EGFP) and monomeric Cherry (mCherry) fluorescent proteins. Both chimeras bound to the surface of BHK cells treated with protective antigen (PA) in a patchy mode. Binding was followed by rapid internalization, and the two anthrax factors were found to traffic along the same endocytic route and with identical kinetics, indicating that their intracellular path is essentially dictated by PA. Colocalization studies indicated that anthrax toxins enter caveolin-1 containing compartments and then endosomes marked by phoshatidylinositol 3-phoshate and Rab5, but not by early endosome antigen 1 and transferrin. After 40 min, both EF and LF chimeras were observed to localize within late compartments. Eventually, LF and EF appeared in the cytosol with a time-course consistent with translocation from late endosomes. Only the EGFP derivatives reached the cytosol because they are translocated by the PA channel, while the mCherry derivatives are not. This difference is attributed to a higher resistance of mCherry to unfolding. After translocation, LF disperses in the cytosol, while EF localizes on the cytosolic face of late endosomes.
Anthrax is a zoonosis caused by infection with toxigenic strains of Bacillus anthracis, a spore-forming Gram-positive bacterium widespread in the environment (Beyer and Turnbull, 2009; Koehler, 2009). Depending on the site of entry of spores or bacteria, humans may develop three forms of anthrax: cutaneous, gastrointestinal or inhalational anthrax, the latter form being the most dangerous (Dixon et al., 1999). Virulent B. anthracis strains are characterized by the expression of three major virulence factors: an anti-phagocytic polyglutamic capsule, which largely prevents phagocytosis by neutrophils and macrophages (Fouet, 2009), and two A-B type toxins. The toxins are capable of binding and entering almost any cell type, where they derange two major cell signalling pathways (Ascenzi et al., 2002; Moayeri and Leppla, 2009). The two toxins were termed oedema toxin (EdTx) because, when injected sub-cutaneously, it caused oedema, and lethal toxin (LeTx) because it led to rapid death in guinea pigs and in Fischer 344 rats (Smith and Keppie, 1954; Smith et al., 1955; 1956; Beall et al., 1962). These two toxins share the same B binding component, a protein of 83 kDa, named protective antigen (PA) from the fact that it is an immunogen that provides effective protection against B. anthracis infection in several animal species, including humans (Cybulski et al., 2009). The A part of LeTx is lethal factor (LF, 90 kDa) and that of EdTx is oedema factor (EF, 89 kDa). PA binds to two cell surface receptors: Tumor Endothelial Marker 8 (TEM8) and Capillary Morphogenesis Protein 2 (CMG2), both widely expressed in different tissues and cell types. The removal of a 20 kDa N-terminal domain converts PA into the PA63 form, which self-associates into heptamers (PA63)7 capable of binding LF and EF (Young and Collier, 2007). The (PA63)7 + LF complex (LeTx) promotes its own endocytosis by entering surface rafts and inducing specific signalling events on the cytosolic face of the membrane. Similar data are not available for the (PA63)7 + EF (EdTx) complex, and it remains to be determined if this toxin follows the same binding and intracellular trafficking route followed by LeTx. After entering early endosomal compartments, LeTx reaches late endosomes, wherefrom LF is delivered into the cytosol (Abrami et al., 2005 and Puhar and Montecucco, 2007).
The process of cell entry of EF and LF has been studied using cell fractionation and electron microscopic methods, and in several cases a construct between the N-terminus of LF and diphtheria toxin (for recent review see Collier, 2009; Moayeri and Leppla, 2009; Van Der Goot and Young, 2009). We have attempted to improve the knowledge of this essential process by producing chimeras consisting of the full-length LF and EF molecules fused at their C-termini to different fluorescent proteins: enhanced green fluorescent protein (EGFP) or monomeric Cherry protein (mCherry) (Shaner et al., 2004; 2005). Fusion at the N-terminus was avoided because this part of the molecule is essential for membrane translocation (Collier, 2009). We have found that the two toxins follow exactly the same pathway of entry with the same time-course. Both EF and LF enter the cytosol from late endosomal compartments only when coupled to EGFP, but not to the mCherry protein. The two toxins reach different cytosolic locations: LF diffuses in the cytosol while EF appears to remain bound to the cytosolic face of the limiting membrane of late endosomes.
Fluorescent chimeras of EF and LF
The four chimeras EF-EGFP, EF-mCherry, LF-EGFP and LF-mCherry were produced as recombinant molecules with an N-terminal 6xHis tag, which allows easier purification (Fig. S1) and was shown to confer to LF an increased capacity to enter and translocate across the (PA63)7 transmembrane channel (Neumeyer et al., 2006). The two LF derivatives and the two EF derivatives were shown to retain the ability to perform their respective biochemical activities in cell culture. BHK cells were used because of their flattened shape, which favours fluorescence microscopy, and because they express both PA receptors, as the majority of sensitive animal cells do. Figure 1 shows that the EGFP derivatives of both LF and EF are active in the cell cytoplasm, though with lower levels of activity with respect to their nonconjugated counterparts, as if a smaller number of molecules is capable of reaching the cytosol when EGFP is present. However, these chimeric toxins do reach the cytosol and therefore can report on each step of the cell entry process. On the contrary, the two mCherry derivatives displayed no intracellular activity. The most likely explanation was that the two mCherry chimeras were unable to unfold and translocate into the cytosol across the (PA63)7 channel (Collier, 2009). To test this possibility, we measured translocation of the chimeric proteins through pores formed by PA63 in planar phospholipid bilayers (Fig. 2). The EGFP derivatives of both LF and EF translocated with slightly lower efficiency and time-course than the wild-type proteins, whereas the mCherry derivatives did not translocate at all. We have not investigated in detail the reason for the difference between mCherry and EGFP, but this is likely to be due to a different energy requirement for their unfolding, as shown by their different resistance to guanidinium-induced unfolding (Fig. S2). Protein unfolding has been documented before to be necessary for the translocation across the PA63 channel and this result provides a further example (Collier, 2009). Accordingly, only EF-EGFP and LF-EGFP were used in the study of the final stage of cell intoxication.
Cell binding of the EGFP derivatives of EF and LF
In the present work, we have attempted to use conditions that mimic the in vivo situation. In particular, we have not used cells transfected with TEM8 or CMG2 in order to avoid the possibility that a high receptor density may distort the picture of toxin cell surface binding by creating arrays or clusters of PA receptors. In addition, the toxins were added at 37°C, to avoid the cell shape change that takes place when the cells are shifted from the cold (to avoid endocytosis) to 37°C. PA was added and allowed to be processed to PA63 for 12 min; LF or EF were then added for 2 min and cells were fixed and processed for fluorescence microscopy. Figure 3 shows that both the LF and EF fluorescent derivatives bind to the BHK plasma membrane in a spotty manner, consistent with the biochemical findings of Abrami et al. (2003) in CHO cells showing that the binding of LF to (PA63)7 oligomers induces the partition of LeTx in plasma membrane rafts. Clearly, the presence of the EGFP did not prevent binding of EF or LF to (PA63)7. However, the strength of the fluorescence signal was not optimal and was improved by using anti-GFP antibodies, which were found not to alter the patterns of fluorescence distribution.
No difference in the spotty binding of LF-EGFP and EF-EGFP was ever noticed in hundreds of cells observed, and the two fluorescent toxins always overlapped, indicating that LF and EF distribute similarly on the cell surface after binding. Partial colocalization of the fluorescence of the two toxins was found with caveolin-1 (∼50%, n ≥ 5) and lysenin (∼35%, n ≥ 3), a protein that binds specifically to sphingomyelin enriched in raft membrane domains (Yamaji et al., 1998); there was little colocalization (Fig. 4) with the clathrin light chain (∼3%, n ≥ 4). This finding is in agreement with the previous report that (PA63)7-LF enters into membrane rafts of the plasma membrane that contain caveolin-1 (Abrami et al., 2003; 2008a), and also with the finding that dynamin is involved in the cell entry of LF (Boll et al., 2004), as dynamin is essential for the endocytosis of caveolin-1 rafts. The similar distribution of EGFP fluorescence always seen with EF and LF derivatives indicates also that (PA63)7-EF segregates into plasma membrane rafts, as previously found for (PA63)7-LF.
Endocytosis of EF-EGFP and LF-EGFP
Binding of the chimeric fluorescent LF and EF is followed by rapid endocytosis, as a significant amount of EF/LF is found inside cells only 5 min after EF or LF addition at 37°C. Cells were fixed and observed at different time points (3, 5, 7 and 10 min) after addition of PA and LF/EF. Figure 5 shows the images obtained after an incubation of 10 min at 37°C; the two toxins are within early endosomal compartments that contain phosphatidylinositol 3-phosphate (PI3P) (∼30%, n ≥ 3) and Rab5 (∼40%, n ≥ 3), but apparently not transferrin (Tfn) or early endosomal antigen 1 (EEA1) (13% and 12%, respectively, n ≥ 4). This is not unprecedented, as it has been reported that cholera toxin B- and Simian virus 40-containing organelles are distinct from classical EEA1- and Tfn-positive endosomes, but communicate with early endosomes via a pathway regulated by Rab5 (Pelkmans et al., 2001; Nichols, 2002; Parton and Simons, 2007). These data show that the endocytosis of LF/EF mediated by the binding of PA to its receptors in BHK cells is rapid and efficient, with undetectable LF/EF remaining on the cell surface after 10 min. This kinetics reflects the fact that the (PA63)7-LF triggers its own endocytosis via modifications of the PA receptor, which take place on the cytosolic side of the protein and includes ubiquitination, palmitoylation and induction of specific phosphorylation of the receptors associated proteins (Abrami et al., 2004, 2006, 2008b).
After 40 min, both EF and LF are within late endosomal compartments, as indicated by the extensive colocalization (∼94%, n ≥ 3) with the lipid molecule lysobisphosphatidic acid (LBPA) (Fig. 6), a marker of these compartments (Kobayashi et al., 1998). The fluorescence patterns of EF-mCherry and LF-EGFP intracellular distribution are identical up to this time point (Fig. 7), as indicated by a high degree of colocalization (∼80%, n ≥ 3). Controls with equimolar amount of PA and the EGFP or the mCherry proteins were performed and revealed no fluorescence staining (data not shown), indicating that the observed pattern is due to LF and EF. However, it should be recalled that preliminary experiments had indicated that the conjugation with the mCherry protein prevents cell intoxication, by blocking the translocation of EF/LF across the trans-membrane PA channel. To exclude that the colocalization data are altered by the difference in the translocation properties of mCherry chimera, the fluorescence pattern of the inverted chimeras LF-mCherry and EF-EGFP was characterized and gave identical results (not shown). However, to study the intracellular distribution of LF and EF at later time points, only the EGFP derivatives were used.
Different intracellular localization of EF and LF
Figure 8 shows that, after 90 min from their addition, EF and LF have different intracellular localization, with LF being clearly cytosolic, though its fluorescence signal is weak, a fact that is expected because the signal is dispersed in a large volume (Fig. 8B). Notwithstanding the low fluorescence signal of LF, its distribution is not exactly homogeneous throughout the cytoplasm, as there appears to be some concentration on cellular organelles; this interpretation is in agreement with the fact that LF is active on MEK isoforms that are known to be localized on organelles (e.g. MEK1 and 2 on Golgi and endosomes, and MEK6 on mitochondria) (Wunderlich et al., 2001; Poderoso et al., 2008). At variance, EF-EGFP gives a spotty and perinuclear fluorescence distribution and shows an extensive colocalization with the LBPA-specific marker of late endosomes (Fig. 8A). Together with the fact that, after 90 min, EF-EGFP has translocated across the late endosomal membrane and has already caused a large increase in cAMP level (Fig. 1), this picture clearly indicates that in vivo EF translocates from the lumen of late endosomes to the cytosol in such a way as to remain associated to the cytosolic surface of these intracellular compartments. We are currently investigating the molecular basis of this specific association, and have already excluded that the N-terminal domain of EF is involved.
It has been recently shown that diphtheria toxin and Clostridium botulinum C2 toxin cross the endosomal membrane and refold into the cytosol with the contribution of chaperone proteins that are inhibited by radicicol and cyclosporin A (Ratts et al., 2003; Kaiser et al., 2009). The presence of these two drugs together did not cause any reduction in the entry into the cytosol of the native LF and EF proteins or of their conjugates with EGFP, monitored by enzymatic activity, Western blotting and fluorescence (not shown).
This is the first study of the trafficking of the two anthrax toxins, LF and EF, in the same cell at the same time, a situation that mimics the in vivo one in which both toxins are released by the infecting B. anthracis. We found that EF and LF clearly follow the same pathway of entry from cell surface beginning from the concentration into plasma membrane microdomains, which leads to a rapid endocytosis involving multiple routes to early endosomes. The two toxins end up in the same late endosomes, wherefrom LF and EF translocate from the lumen into the cytosol, but reach two different intracellular localizations. EF remains associated with late endosomes, while LF diffuses into the cytosol. This is in agreement with the recently acquired knowledge that LF cleaves different MEK proteins that are known to havedifferent intracellular distributions (Fanger et al., 1997; Vitale et al., 2000; Wunderlich et al., 2001; Poderoso et al., 2008), while EF generates a intracellular gradient of cAMP from the perinuclear area to the cortical sub-plasma membrane region (Dal Molin et al., 2006; Puhar et al., 2008).
The binding of LF and EF to cell surface bound (PA63)7 gives a patched distribution on the plasma membrane,which is fully consistent with the two toxins entering cholesterol-enriched microdomains. This was established previously by Abrami et al. (2003) for LF and it is extended here to EF. Here, we also found that several of these microdomains are enriched in sphingomyelin, as there is colocalization with lysenin, a sphingomyelin binding protein (Yamaji et al., 1998). Cell surface binding was found to be rapidly followed by endocytosis. From the present study, it appears that the (PA63)7-LF and (PA63)7-EF complexes may enter various types of endocytic vesicles to reach early endosomal compartments, most of them marked by the presence of Rab5 and PI3P. It should be noted that, here, we have deliberately chosen not to use any method that may alter the physiological process of cell entry of these toxins, such as cholesterol depletion, cross-linking, inhibitors, siRNA, overexpression of different proteins, etc. The only ‘cell manipulation’ procedure used was that of expressing the light chain of clathrin coupled to mRFP, but this protein was not found to colocalize substantially with the two toxins. This result is at variance with the previous report of a clathrin-mediated endocytosis of LF in CHO cells (Abrami et al., 2003), and we have no satisfactory explanation for such a difference. However, different cell lines were used and fibroblasts are known to have multiple pathways of endocytosis (Doherty and McMahon, 2009). On the other hand, the present result is not surprising in light of the fact that other toxins that bind via an oligomeric binding protomer enter cells via nonclathrin-dependent trafficking (Sandvig et al., 2008) and that the majority of ligands that bind to raft microdomains are preferentially taken up by clathrin-independent endocytosis (Nichols, 2003).
No matter which initial traffic route is taken by LF and EF to early endosomes, eventually they reach late endosomal compartments, wherefrom they translocate into the cytosol. This translocation is not assisted by chaperones inhibited by radicicol or cyclosporin A, as was found to be the case for diphtheria toxin and Clostridium botulinum C2 toxin (Kaiser et al., 2009).
A remarkable finding presented here is that both LF and EF are capable of pulling the conjugated EGFP through the PA channel with high efficiency, while this is not the case of the mCherry fluorescent protein, which is more resistant to unfolding than the EGFP protein (Fig. S2). Taken together, these data provide further evidence in favour of the model of translocation of LF that envisages a low pH driven unfolding of the polypeptide chain to enter the PA channel and translocation of the unfolded chain that refolds in the neutral pH of the cytosol (Krantz et al., 2005; Collier, 2009). At the same time, the present report shows that mCherry may not be an appropriate choice as reporter of intracellularly acting toxins.
Cells, antibodies and reagents
BHK cells were maintained in DMEM (Gibco) containing 10% heat-inactivated foetal calf serum (FCS, Euroclone), penicillin (100 U ml−1) and streptomycin (100 mg ml−1). Antibodies were obtained from the following sources: anti-His tag monoclonal antibody from Novagen, monoclonal and polyclonal anti-GFP, -RFP, -EF and -LF polyclonal antibodies from Abcam, anti-caveolin and anti-EEA1 antibodies from BD Transduction Laboratories; anti-PI3P from Echelon, anti-Rab5 from Synaptic System, anti-LBPA (6C4) was a kind gift of J. Gruenberg (University of Geneva, CH), Tfn-Alexa555 and fluorescently labelled secondary antibodies from Molecular Probes. Lysenin and lysenin antiserum were from Peptide Institute; FuGENE HD from Roche Diagnostics Corporation. The plasmid encoding mRFP-Clathrin Light Chain was from Addgene. Reagents were Sigma-Aldrich and Calbiochem.
Cloning, expression and purification of chimeric proteins
The EGFP gene was PCR-amplified from pEGFP-N1 (Clontech) using the following primers: forward 5′-AAAGAGCTCATGGTGAGCAAGGGCG-3′ and reverse 5′-AAAGAATTCCTTGTACAGCTCGTCCAT-3′. The mCherry gene was PCR-amplified from pREST B, a generous gift from RY Tsien, using the following primers: forward 5′-AAAGAGCTCATGGCCACTGGTGGACAG-3′ and reverse 5′-AAAGAATTCTAGGCGCCGGYGGAGT-3′. Both fragments were digested with SacI and EcoRI and inserted in pRSET A (Invitrogen) containing LF or EF, respectively, as previously described (Dal Molin et al., 2006), downstream from an N-terminal His-tag coding region. The sequences were confirmed by DNA sequencing. LF chimeric derivatives were expressed in Escherichia coli BL21(DE3) and EF chimeric proteins in E. coli BL21 (DE3)-Codon Plus-RIL (Stratagene) grown at 37°C in LB broth containing 100 mg ml−1 ampicillin or 100 mg ml−1 ampicillin and 34 mg ml−1 chloramphenicol. After 4 h of induction with 1 mM isopropyl-1-thio-a-D-galactopyranoside at 30°C, the pellet was resuspended in buffer A (50 mM Na2HPO4, 500 mM NaCl, pH 8) and lysozyme (0.1 mg ml−1). Bacterial cells were disrupted by ultrasonic dispersion and centrifuged, and the supernatant was loaded onto a Hi-trap column charged with Cu2SO4 and equilibrated with buffer A. The column was washed with buffer A, the protein was eluted with a 0–100 mM imidazole gradient, and the fractions containing chimeric proteins were pooled and dialysed against binding buffer (50 mM Tris, 20 mM NaCl and 1 mM EDTA, pH 7.5) to remove imidazole and NaCl. The identities of chimeric proteins were assessed by immunoblotting with anti-His tag, anti-LF, anti-EF and anti-GFP or anti-mRFP antibodies. We used this antibody, which recognizes a conserved epitope on mCherry, because no other was available for this red variant.
FRET imaging of cAMP intracellular dynamics
BHK cells (2.0 × 105) were co-transfected with 1 µg of two pcDNA3.1 plasmids, one carrying the catalytic (C) subunit of PKA fused to YFP (C-YFP) and one carrying the regulatory (R) subunit of PKA fused to CFP (RII-CFP) (Lissandron et al., 2005) using FuGENE HD reagent, following manufacturer's instructions. Forty-eight hours after transfection, cells were incubated in a balanced salt solution (NaCl 135 mM, KCl 5 mM, KH2PO4 0.4 mM, MgSO4 1 mM, HEPES 20 mM, CaCl2 1.8 mM, glucose 5.4 mM, pH 7.4) in a microscope-adapted micro-incubator equipped with a temperature controller (HTC, Italy) at 37°C and constant 5% CO2 pressure. Toxins were added after about 15 min of imaging, and images were taken every 20 s for the indicated time periods. Integration time was 200 ms. At each time point, the intracellular cAMP level was estimated by measuring the ratio between the background-subtracted CFP emission image and the YFP emission image upon excitation of CFP (R CFP/YFP) (Mongillo et al., 2005). Images were acquired using an oil immersion 40× PlanApo 1.4 NA objective on a Leica DMI6000 microscope. FRET images were collected through a BP 436/20-nm excitation filter and a custom-made optical beam splitter built with a 515 nm dichroic mirror and ET 480/40-nm and ET 535/30-nm emission filters. All optical filters were obtained from Chroma Technologies. A cooled camera from OES (Padova, Italy) with a 1,4 Megapixel CCD and a sensor resolution of 1360 × 1024 Pixel was used. The acquisition software was from OES (Padova, Italy). Recorded images were processed with WCIF ImageJ v1.40 (http://rsb.info.nih.gov/ij).
MEK3 cleavage by chimeric lethal factor
BHK cells (1.5 × 104) were incubated with PA 400 nM and 200 nM LF or LF-EGFP in DMEM plus 1% BSA at 37°C for different time periods in a 96-well plate. After removal of the culture medium, the cells were lysed, subjected to SDS-PAGE, and immunoblotted for the isoform 3 of MEK with a specific polyclonal antibody from Santa Cruz Biotechnology (USA). Samples were developed with ECL plus detection system (Amersham Biosciences), and chemiluminescence emission was detected with ChemiDoc™ XRS (Biorad). The antibody detects two splicing isoforms of MEK3, MEK3b and MEK3a, a variant lacking of the first 29 residues. Only the MEK3b is cleaved by LF and the percentage of cleavage was quantified considering the ratio between the MEK3b band and the band intensity at lower molecular weight, normalizing the result against the control. Band intensities were quantified with the Quantity One software from Biorad.
Sub-confluent BHK cells grown on glass coverslips were rinsed two times with DMEM plus 2% w/v BSA and treated for different periods of time with EF-mCherry and/or LF-EGFP (200 nM) and PA (600 nM). Then cells were washed with PBS, fixed with ice-cold acetone for 5 min at room temperature to localize the toxins along the endocytic pathway or with PFA 4% (10 min at room temperature) to mark endocytic lipids and incubated sequentially with a mixture of primary antibodies and a mixture of fluorescent secondary antibodies. To monitor cell surface binding, cells were first treated with PA83 for 12 min, washed, incubated with LF or EF derivatives for 2 min at 37°C and immediately washed and fixed. All antibody incubations were performed for 1 h at room temperature. Images were acquired sequentially with a FITC and Texas Red® filter set (Chroma Technology corp., USA) with 250 ms or longer integration times by using an oil immersion 63× PlanApo 1.40 NA objective on a Leica DMIRE3 widefield inverted microscope equipped with a DC 500 digital camera with 1300 × 1030 pixels resolution from Leica. The acquisition software was FW4000 (Leica). Images were processed with ImageJ v1.40 (http://rsb.info.nih.gov/ij) and colocalization analysis on raw images was performed using the JaCoP plugin (Bolte and Cordelières, 2006) and OBCOL plugin (Woodcroft et al., 2009) under ImageJ.
Colocalization studies were performed with an object based analysis with plugins which autonomously detect objects within an image stack (z-planes ≥ 4) and analyse them as individual objects. Before segmentation, each channel was preprocessed to remove background fluorescence below a given intensity cut-off. Cut-offs between 40 and 50 and 50–60 intensity units were applied to the red and the green channels respectively. An overlap of three pixels was required to join objects and the objects of dimensions < 10 pixel were excluded. Images were segmented to automatically define discrete objects, at least 120 objects were defined for each examined cell. Colocalization was expressed as the percentage of objects containing green and red pixels on the objects that contain only red pixels.
Translocation of chimeric proteins across artificial membranes
Planar phospholipid bilayers were formed by standard methods (Janowiak et al., 2009). Once a membrane was formed, WT PA63 prepore (25 pM) was added to the cis compartment, held at a Δψ = +20 mV with respect to the trans compartment. Free PA63 not inserted into the membrane was removed by perfusion. Binding cargo (LF, EF, or chimeric proteins) was added to the cis compartment (1 µg ml−1), and the progress of binding to PA channels was monitored by the decrease in conductance. Free was removed by buffer exchange. Translocation was initiated by raising the pH of the trans compartment to pH 7.2 with 2 M KOH, while maintaining the cis compartment pH at 5.5. At the same time, we increased the membrane potential from Δψ = +20 mV to Δψ = +50 mV. Experiments were normalized to controls lacking cargo protein (n ≥ 3). All planar phospholipid experiments were performed in a Warner Instruments Planar Lipid Bilayer Workstation (BC 525D, Hamden, CT).
This work has been supported by the MIUR programme FIRB Internazionalizzazione and the Istituto Superiore di Sanità and NIH grants 1F32 AI077280 to BEJ and 5R01 AI022021 to RJC, who holds equity in PharmAthene.