Direct imaging of anthrax intoxication in animals reveals shared and individual functions of CMG-2 and TEM-8 in cellular toxin entry

The virulence of Bacillus anthracis is linked to the secretion of anthrax lethal toxin and anthrax edema toxin. These binary toxins consist of a common cell-binding moiety, protective antigen (PA), and the enzymatic moieties, lethal factor (LF) and edema factor (EF). PA binds either of two specific cell surface receptors, capillary morphogenesis protein-2 (CMG-2) or tumor endothelial marker-8 (TEM-8), which triggers the binding, endocytosis, and cytoplasmic translocation of LF and EF. The cellular distribution of functional TEM-8 and CMG-2 receptors during anthrax toxin intoxication in animals is not fully elucidated. Herein, we describe a novel assay to image anthrax toxin intoxication in live animals, and we use the assay to visualize TEM-8- and CMG-2-dependent intoxication. Specifically, we generated a chimeric protein consisting of the N-terminal domain of LF fused to a nuclear localization signal-tagged Cre recombinase (LFn-NLS-Cre). When PA and LFn-NLS-Cre were co-administered to transgenic mice that ubiquitously express a red fluorescent protein in the absence of Cre activity and a green fluorescent protein in the presence of Cre activity, anthrax toxin intoxication could be visualized at single-cell resolution by confocal microscopy. By using this assay, we show that CMG-2 is critical for intoxication in the liver and heart, whereas TEM-8 is required for full intoxication in the kidney and spleen. Other tissues examined were largely unaffected by single deficiences in either receptor, suggesting extensive overlap in TEM-8 and CMG-2 expression. The novel assay will be useful for basic and clinical/translational studies of Bacillus anthracis infection and for identifying on- and off-targets for reengineered toxin variants in the clinical development of cancer treatments. Background Assays for imaging of anthrax toxin intoxication in animals are not available. Results Anthrax toxin-Cre fusions combined with fluorescent Cre reporter mice enabled imaging of anthrax toxin intoxication in animals. Conclusion Shared and distinct functions of toxin receptors in cellular entry were uncovered. Significance. A simple and versatile assay for anthrax toxin intoxication is described.

Anthrax is contracted through inhalation, ingestion, or cutaneous inoculation of endospores of the Gram-positive bacterium Bacillus anthracis. Spores germinate following their introduction to the body and disseminate to cause a systemic infection, which, if left untreated, is associated with high mortality rates. Upon the death of the host, Bacillus anthracis forms spores that are resistant to chemical insults, heat exposure, and dehydration and remain infectious for long periods (1,2).
The virulence of Bacillus anthracis results from the release of three proteins into the circulation: protective antigen (PA), lethal factor (LF), and edema factor (EF). These three proteins are individually nontoxic, and PA combines with either LF to form anthrax lethal toxin or with EF to form anthrax edema toxin. The systemic administration of anthrax toxin to animals closely mimics experimental infection with Bacillus anthracis, and vaccination against the toxin components is protective, indicating that anthrax is largely a toxin-mediated disease (1,2). Anthrax toxins exert their cytotoxic actions in a three-step activation process that involves: a) the binding of PA to the surface of target cells, b) the translocation of LF and EF to the cytoplasmic compartment of the target cells, and c) the enzymatic action of LF and EF on cytoplasmic substrates (1,2). Anthrax toxin intoxication is initiated by the binding of PA to either of two receptors, capillary morphogenesis protein-2 (CMG-2) or tumor endothelial marker-8 (TEM-8) (3,4).
Subsequently, PA is cleaved at the sequence, 164 RKKR 167 , by cell surface-localized furin or furin-like proprotein convertases (5). This endoproteolytic cleavage is absolutely required for toxin activation and triggers all subsequent steps of the activation process. The C-terminal 63 kDa fragment of PA (PA63) remains bound to the cell surface after endoproteolytic cleavage and undergoes a conformational change that leads to the formation of a PA63 heptamer or octamer that subsequently binds up to four molecules of LF or EF with high affinity (6)(7)(8). The complex of PA63 and LF or EF is then endocytosed, and PA63 undergoes pH-induced conformational changes in the endosomal/lysosomal compartment to form a channel that facilitates the unfolding and translocation of LF and EF to the cytoplasm. EF is an adenylate cyclase proposed to lead to the formation of supraphysiological intracellular levels of cyclic AMP (9). LF is a zinc-dependent metalloproteinase that can cleave and inactivate several mitogen-activated protein kinase kinases (10). Although essential for intoxication, the cellular distribution of CMG-2 and TEM-8 and the function of each receptor in the intoxication in specific organs remain to be fully elucidated.
Notably, assays for direct visualization of anthrax toxin intoxication in vivo are not available, and the tissue and cellular targets for anthrax toxin during in vivo infection have been inferred only indirectly from analysis of tissues from intoxicated animals or from biochemical and genetic analysis of anthrax toxin targets (11,12).
LF is stable in circulation when administered alone and only becomes cell surface-associated after the binding of PA to CMG-2 or TEM-8 and its subsequent proteolytic cleavage to PA63. It has long been noted that LF residues 1-254 suffice to achieve translocation of a variety of "passenger" polypeptides and other molecules into the cytoplasm of the cells in a PA63-dependent manner (13,14).
These include other bacterial toxins and bacterial proteins (13,(15)(16)(17)(18)(19), fluorescent proteins (20), viral proteins (21)(22)(23)(24), eukaryotic proteins (25)(26)(27)(28), and radioisotopes (29-32). Thus, the fusion or conjugation of LF to imageable moieties could provide ideal agents for studying the cellular intoxication by anthrax toxin in vivo. A significant caveat to this approach, however, is the low number of LF molecules successfully translocated to the cytoplasm through the PA pore, which makes most imaging modalities poorly suited to study anthrax toxin intoxication in whole-animal systems (18,27). A second challenge to whole-animal imaging approaches is that most imaging modalities, such as radionuclides, enzymes such as horseradish peroxidase and b-galactosidase, and fluorescent proteins, likely would not discriminate between productive intoxication (i.e. PA-dependent entry into the cytoplasm) and non-productive interactions of the labeled toxin with cells, such as cell surface retention, fluid phase pinocytosis, and endosomal/lysosomal accumulation of intact or partially degraded toxin conjugates.
Spleen extracts from reporter mice carrying a Cre-activated b-galactosidase transgene have been shown to express increased b-galactosidase activity when infected with Salmonella enterica serovar Typhimurium carrying the type III secreted protein, SopE, fused to bacteriophage P1 Cre recombinase (33). Although single-cell resolution was not achieved, presumably due to a low number of cells being infected (33), the study provided evidence that bacterial protein-Cre fusion proteins may display sufficient enzymatic activity in animals to induce LoxP-dependent recombination.
In this study, we used a combined biochemical and genetic approach to image anthrax toxin intoxication in animals. Specifically, we generated a tripartite fusion protein that consists of the Nterminal domain of LF fused to a nuclear localization signal-tagged bacteriophage P1 Cre recombinase (LFn-NLS-Cre). When PA and LFn-NLS-Cre were co-administered to transgenic mice that ubiquitously express a red fluorescent protein (tdTomato) in the absence of Cre activity and a green fluorescent protein (eGFP) in the presence of Cre activity (hereafter mTmG mice), anthrax toxin intoxication could readily be visualized at single-cell resolution by using confocal microscopy of unfixed and unprocessed organs. By superimposing individual genetic deficiencies of either TEM-8 or CMG-2 in mTmG mice, we were able to directly establish the importance of each receptor in anthrax toxin intoxication in individual tissues.
When used in conjunction with modified PA variants that are activated by specific cell surface proteases, the assay may also be suitable for in vivo imaging of specific cell surface proteolytic activity in a variety of physiological and pathological processes.

Recombinant proteins
Plasmids for expressing proteins having LFn (LF aa 1-254) fused to Cre recombinase were constructed using the Champion pET SUMO vector (Invitrogen, Carlsbad, CA), which expresses proteins fused at the C-terminus of a His6-SUMO tag. DNA-encoding residues 1-254 of anthrax lethal factor originated from Hideo Iwai (51). The His6-SUMO polypeptide and the His6-tagged SUMO protease were subtractively removed by passage through Ni-NTA resin. The LFn-Cre proteins were further purified by chromatography on hydroxyapatite to achieve purities of >95%. The LFn-NLS-Cre protein selected for the animal imaging studies was obtained in yields of >20 mg/L of culture.

Cell culture assays
Efficacy of LFn-Cre protein translocation into cells was tested in CV1-5B cells (52,53). Cells were plated in 96-well plates in DMEM high glucose medium with 10% fetal bovine serum, cultured at 37˚C at 10% CO2, and used when they were at low and high confluency. PA was added at 1 µg/mL in all wells, and LFn-Cre proteins were diluted serially at 3.14-fold. After 40 h, cells were washed in phosphate buffered saline (PBS) containing 2 mM MgCl2 and fixed in PBS, 5 mM EGTA, 2 mM MgCl2, and 0.2% glutaraldehyde for 30 min. After again washing in PBS with 2 mM MgCl2, the cells were stained for bgalactosidase activity with PBS, 2 mM MgCl2, 0.1% Triton X-100, 0.1% NaN3, and 1 mg/mL chlorophenolred-b-D-galactopyranoside. Absorbance was measured at 570 nm after 20 min to quantify conversion of substrate by b-galactosidase (54).

Imaging anthrax toxin intoxication in mice
LFn-NLS-Cre and PA proteins alone or in combination in PBS were delivered intraperitonially or via tail vein injection. The mice were tail vein-injected with 100 µL of 6 mg/mL Hoechst dye (Thermo Fisher Scientific, Waltham, MA) 4-6 h prior to termination of an experiment to visualize nuclei (57). Mice were euthanized by CO2 inhalation and perfused with ice-cold PBS using cardiac puncture. Organs were immediately removed and cut into ~1-2 mm thick slices using a scalpel. The organ slices were placed on a MatTek glass bottom microwell dish (MatTek Corporation, Ashland, MA) and imaged using a 20x 0.75 NA Air or 60x 1.27 NA Water objective (Nikon, Tokyo, Japan) on an A1R + MP confocal microscope system (Nikon, Tokyo, Japan). Large images were composed of stitched images with a 10% overlap using NIS-Elements software (Nikon, Tokyo, Japan).

Generation of LFn-Cre recombinase fusion proteins capable of PA-dependent cytoplasmic translocation
We have previously shown that PA-dependent translocation of a LF-b-lactamase fusion protein can be imaged in cultured cells by using a cell-penetrating b-lactamase quenched fluorescence resonance energy transfer substrate (18). The adaptation of this assay for imaging anthrax toxin intoxication in whole animals, while hypothetically feasible, is prohibited by the high cost of the b-lactamase substrate and, likely, by logistic problems associated with systemic delivery of the substrate to animals. We therefore explored the possibility of combining biochemical and genetic approaches to imaging anthrax toxin intoxication in whole animals. Specifically, we generated a series of proteins that consist of the PA- Cre was administered at a 40-fold higher concentration ( Figure 1D).

Imaging anthrax toxin intoxication in mice
The above studies showed that LFn-Cre fusion proteins could translocate to the cytoplasm in a PA-dependent manner, that the Cre moiety (alone or as an intact fusion protein) thereafter was imported to the nucleus, and that it retained its recombinase activity after nuclear translocation. This indicated that To test if PA and LFn-NLS-Cre could mediate LoxP-dependent recombination in a whole-animal context, we first designed PCR primer sets that would selectively amplify, respectively, the nonrecombined and the recombined mTmG transgene (Figure 2A). Interestingly, a PCR product derived from the recombined transgene was readily detected in the heart, lungs, liver, kidney, spleen, lymph nodes, thymus, uterus, esophagus, trachea, tongue, and bone marrow of mTmG +/0 mice injected with LFn-NLS-Cre and PA but not in these organs from non-injected mTmG +/0 mice ( Figure 2B). This PCR product was not observed in the intestine, skin, and brain of mice injected with LFn-NLS-Cre and PA ( Figure 2B).
We next determined the ability to detect fluorescence in mTmG +/0 mice by confocal microscopy of unfixed whole organ slices, which could serve as a convenient readout for Cre activity. Wildtype mice were analyzed in parallel as a control for autofluorescence. To obtain semi-quantitative estimates of fluorescence intensities of the examined organs, in this and the following experiments, we generated composite images of the entire organ slice from confocal images acquired at low magnification. As expected, red fluorescence of variable intensity was observed in multiple organs of mTmG +/0 mice but not in the corresponding organs from wildtype mice imaged using the identical conditions (Supplemental Compatible with the PCR analysis, green fluorescence was weak or absent in intestine (Supplemental To determine when an eGFP signal is first detectable after the administration fo PA and LFn-NLS-Cre, we injected mTmG +/0 mice with 75 µg of each protein and examined the heart, kidney, liver, lungs, and spleen by confocal microscopy at 6, 8, 10, and 12 h (Supplemental Figure 9). Whereas no signal was observed at any of these time points in the heart (Supplemental Figure 9 A'

Single-cell resolution imaging of anthrax toxin intoxication
Using the knowledge gained from the above experiments, we next tested the ability of the assay to image intoxication in individual cells in unprocessed organs. For this purpose, mice received three intravenous injections of 25 µg LFn-NLS-Cre and 25 µg PA at 0 h, 24 h, and 48 h. 72 h after the first injection, confocal images of red, green, and blue (nuclei) fluorescence of slices of the heart, kidney, liver, lungs, and spleen were acquired at high magnification ( Figure 3). This analysis showed that in tissues of these five organs, non-intoxicated and intoxicated individual cells were readily distinguishable by their red and their green or yellow membrane-confined fluorescence, respectively.

Effect of genetic elimination of CMG-2 and TEM-8 on anthrax toxin intoxication
We next interbred previously generated CMG-2-deficient (Cmg2 -/-) and TEM-8-deficient (Tem8 -/-) mice to mTmG +/0 mice to generate, respectively, Cmg2 -/-;mTmG +/0 and Tem8 -/-;mTmG +/0 mice. These respectively, anthrax lethal toxin and anthrax edema toxin (11,56). Assuming that both receptors were close to ubiquitously expressed in tissues, this was tentatively suggested to be a consequence of a more than 10-fold lower affinity of PA for TEM-8 than for CMG-2. These findings are compatible with the current imaging study, showing that TEM-8 was essentially unable to support the intoxication in the heart, liver, and lungs, despite repeated toxin exposure through multiple injections. It should be noted, however, that full intoxication in other organs, including the spleen and kidney, was dependent upon TEM-8 but not CMG-2. This uequivocally demonstrates that TEM-8 is a functional receptor for anthrax toxin in vivo despite its reported lower affinity for PA.
A curious discrepancy between the aforementioned genetic studies and our current imaging study pertains to the intestine: Genetic studies have definitively established intestinal epithelial cells as direct targets for anthrax edema toxin (11). Nevertheless, we were consistently unable to observe the intoxication in intestinal epithelium. Intestinal epithelial cells of the mTmG +/0 reporter mice have previously been shown to undergo recombination in vivo and express eGFP in response to constitutive or inducible Cre expression, showing that the mTmG transgene is not inherently refractory to Cre-mediated recombination in intestinal epithelium (55). An attractive explanation for this discrepancy pertains to the lack of toxicity of LFn-NLS-Cre/PA as compared to edema toxin. We speculate that edema toxin may initially be unable to access this barrier tissue, but because damage to other visceral organs progresses as a consequence of EF intoxication, endothelial and/or epithelial barrier breakdown may allow entrance to the intestinal epithelial cells.
The imaging assay described here is simple and robust, and, importantly, it does not require the handling of toxic proteins. Therefore, it should be amenable for and adaptable to diverse research settings.
The assay should be useful for answering a number of basic research questions regarding the pathogenicity of anthrax toxins, as well as assisting clinical/translational efforts aimed at optimizing the treatment of individuals accidentally or deliberately exposed to Bacillus anthracis.
Considerable effort is currently being expended on the development of modified anthrax toxins as novel agents for the treatment of human malignancies. Strategies employed include the reengineering of PA to bind tumor cell surface-enriched proteins (34,35) and the reengineering of PA to be proteolytically activated by proteases enriched in the tumor microenvironment, including matrix metalloproteinases (36)(37)(38)(39)(40)(41)(42)(43), urokinase plasminogen activator (38,40,42,(44)(45)(46)(47)(48)(49), and testisin (50). The assay described here is imminently suited for assessing the efficiency of LF delivery to tumor-relevant cell populations by these modified PAs, as well as to systematically delineate off-targets, which may be invaluable for dose and route of delivery optimization. Last, but not least, by using PA variants selectively cleaved by specific cell surface proteases (43,46,50), the assay may be used for in vivo imaging of specific cell surface proteolytic activity in diverse physiological and pathological settings.