Glycosaminoglycans are specific endosomal receptors for Yersinia pseudotuberculosis Cytotoxic Necrotizing Factor

The Cytotoxic Necrotizing Factor Y (CNFY) is produced by the gram-negative, enteric pathogen Yersinia pseudotuberculosis. The bacterial toxin belongs to a family of deamidases, which constitutively activate Rho GTPases, thereby balancing inflammatory processes. We identified heparan sulfate proteoglycans as essential host cell factors for intoxication with CNFY. Using flow cytometry, microscopy, knockout cell lines, pulsed electron–electron double resonance and bio-layer interferometry, we studied the role of glucosaminoglycans in the intoxication process of CNFY. To analyze toxin-glucosaminoglycan interaction we utilized a truncated CNFY (CNFY709-1014). Especially this C-terminal part of CNFY, which encompasses the catalytic activity, binds with high affinity to heparan sulfates. CNFY binding with the N-terminal domain to its protein receptor seems to induce a first conformational change supporting the interaction between the C-terminal domain and heparan sulfates, which seems sterically hindered in the full toxin. A second conformational change occurs by acidification of the endosome, probably allowing insertion of the hydrophobic regions of the toxin into the endosomal membrane. Our findings suggest that heparan sulfates play a major role for intoxication within the endosome, rather than being relevant for an interaction at the cell surface. Lastly, cleavage of heparin sulfate chains by heparanase is likely required for efficient uptake of the toxic enzyme into the cytosol of mammalian cells. Author Summary The RhoA deamidating Cytotoxic Necrotizing Factor Y (CNFY) from Yersinia pseudotuberculosis is a crucial virulence factor that is important for successful infection of mammalian cells by the pathogen. The mode of action by which CNFY is able to intoxicate cells can be divided into the following steps: Binding to the cell surface, internalization, translocation from the endosome to the cytosol and deamidation of RhoA. We show, that CNFY uses heparan sulfates to maximize the amount of molecules entering the cytosol. While not being necessary for toxin binding and uptake, the sugars hold a key role in the intoxication process. We show that CNFY undergoes a conformational change at a low endosomal pH, allowing the C-terminal domain to be released from the endosomal membrane by the action of heparanase. This study reveals new insights into the CNFY-host interaction and promotes understanding of the complex intoxication process of bacterial toxins.


Introduction
Yersinia pseudotuberculosis is a gram-negative, enteric pathogen, which causes selflimiting gastrointestinal infections. The bacterium produces several virulence factors like adhesins and a type-III secretion system, which injects diverse effector proteins (Yersinia outer proteins, Yops) into mammalian cells. Moreover, Y. pseudotuberculosis secretes the protein toxin Cytotoxic Necrotizing Factor (CNFY) via outer membrane vesicles into the environment (1). The toxin belongs to a family of deamidases modifying small GTPases of the Rho family (2,3). Rho GTPases are molecular switches, which govern a wide variety of signaling pathways, including the rearrangement of the actin cytoskeleton, gene synthesis and survival (4). Small GTPases are regulated by binding and hydrolyzing GTP. The CNFY-catalyzed modification blocks intrinsic as well as the GAP-stimulated GTP hydrolysis, because it converts a specific glutamine (Gln63 in RhoA) required for GTP hydrolysis to glutamate, leading to constitutive activation of the targeted small GTPases (5,6). CNFY improves translocation of Yops into host cells thereby supporting inflammatory responses of the host (7). Following binding to mammalian cells, the toxin is taken up by receptormediated endocytosis, however, a specific cellular toxin receptor is still not known.
Subsequently, the catalytic part of the toxin is released from acidified endosomes into the cytosol where it deamidates RhoA (8).
In previous work we showed that the Escherichia coli toxin CNF1 (61% identity to CNFY) binds with high affinity to the Lutheran/Basal Cell Adhesion Molecule (Lu/BCAM). This interaction is required for intoxication of mammalian cells (9). In contrast, CNFY does not bind to Lu/BCAM. Previous results of cell culture experiments with pharmacological inhibition of heparin sulfation led to the assumption that proteoglycans may be involved in CNFY action (10). Here, we show that glycosaminoglycans (GAG) are involved in recruiting CNFY to mammalian cells. GAG are linear, unbranched negatively charged poly-sugars, for example heparin, heparan sulfate, chondroitin sulfate A or dermatan sulfate, covalently attached to the core protein of a proteoglycan (PG). The GAG synthesis is very complex and involves many different proteins.
The first enzyme transferring Xylose to a specific serine within the recognition sequences of the core protein is the Xylose transferase (XYLT) (11). Addition of Xylose to the core protein is required for the biosynthesis of the growing GAG chain. Heparan sulfates are synthesized by five different glucosyltransferases (EXT1, EXT2, EXTL1, EXTL2 and EXTL3) and may contain more than 100 sugar units (12). Some of the sugar units are frequently modified further for example by sulfation or epimerization, leading to generation of a wide variety of different GAG. PG are components of the extracellular matrix, the plasma membrane and are part of secretory vesicles. Main plasma membrane PG are either syndecans, which are integral membrane proteins or glypicans, attached to the surface by a glycosyl-phosphatidylinositol (GPI) anchor. Cell membrane anchored PG are well known receptors or co-receptors for a variety of macromolecules and viruses (13). Recently, sulfated glucosaminoglycans have been identified to interact with the bacterial toxin Clostridium difficile toxin A and to mediate its uptake into target cells (14). The aim of this study was to understand the role of PG and their GAG side chains in the CNFY intoxication process. Using flow cytometry and microscopy, mutant cell lines and bio-layer interferometry we show that heparan sulfates are crucial for CNFY-mediated intoxication of mammalian cells. They do not only serve as surface receptors, but are functionally essential components for proper toxin translocation within the endosomal compartment.

Results
Yersinia pseudotuberculosis CNFY is an important virulence factor, which is crucial for successful infection of mammals. An infection with a bacterial pathogen results in strong induction of the host immune response. CNFY has been shown to influence inflammatory processes in mice (7). Therefore, we intended to analyze the toxins direct activity on immune cells.

Immune cells are not the preferred targets of CNFY
In a first set of experiments, we studied toxin binding to different human and mouse cells. Therefore, we labeled GST-CNFY with the fluorescent marker DyLight488. Cells were incubated with the labeled toxin and washed. Bound fluorescence was then analyzed by flow cytometry. Surprisingly, tested mouse immune cells: monocytes/macrophages (RAW 264.7) and B lymphocytes (J558L) and most of the human immune cells: T lymphocytes (Molt-4) as well as B lymphocytes (Ramos) did not bind or only marginally bound the toxin (Fig. 1A). In contrast, mouse fibroblasts (MEF cells), human epithelial cells (HeLa cells) as well as human monocytes/macrophages (THP-1 cells) bound the labeled toxin. We further analyzed uptake of CNFY by directly visualizing the toxin-catalyzed modification of RhoA.
Deamidated RhoA shifts to apparent higher molecular weight in urea SDS-PAGE and therefore can be distinguished from the unmodified GTPase. As expected, and in line with the results for toxin binding, CNFY was taken up into mouse MEF, human THP-1 and HeLa cells, whereas there was no shift of RhoA detectable in the other tested immune cells (Fig. S1). Our data are summarized in table 1.

Characterization of a possible CNFY-GAG interaction
We previously showed, that sodium chlorate, which inhibits sulfation of GAG chains, without affecting the size of the sugar chains, reduces intoxication of HeLa cells by CNFY, indicating that proteoglycans may be relevant for toxin binding (10,15).
Following data from the human protein atlas, proteoglycans are widely expressed by most mammalian cells, however, circulating blood cells only marginally express them ( Fig. S1A). Therefore, we studied expression of GAG on the same cell lines analyzed above. Cells were incubated with a FITC coupled antibody detecting HSPG and washed. Bound fluorescence was then analyzed by FACS measurements. In line with toxin binding, MEF and HeLa cells bound the antibody (Fig. 1B). Surprisingly, there was no bound fluorescence detectable on THP-1 cells, suggesting that this cell line may not express HSPG and also other PG (for example chondroitin sulfate or dermatan sulfate) may bind the toxin. In Molt-4 cells (human T-lymphocytes) most of membrane GAG like syndecans and glypicans were not expressed (Fig. S1A). Only glypican-2 RNA was detected more prominently in Molt-4 and THP-1 compared to HeLa cells. If poly-sugars like GAG are involved in toxin interaction, binding kinetics should show no saturation. Therefore, we incubated HeLa cells and THP-1 cells, respectively with increasing concentrations of DL488-GST-CNFY. As shown in Fig. 2A, increasing concentrations of the protein led to increased binding without obvious saturation, indicating a highly abundant structure (like sugars) present on mammalian cells as interaction partner of the toxin. However, data fit best to regression with two curves, one saturated, most likely a protein receptor, the other one not saturated, most likely a sugar. This suggests a combined function of at least two binding surfaces on the cell ( Fig. S2). To study which part of the toxin may interact with GAG, we repeated the experiments with the labeled C-terminal domain (GST-CNFY 709-1014 ), because it was shown for the homologue toxin family member CNF1 that this domain mediates binding to its receptor Lu/BCAM (16). As for the full toxin, we could not observe saturated binding of GST-CNFY 709-1014 to HeLa cells, indicating that the C-terminal part of CNFY is sufficient for GAG binding ( To directly study the involvement of GAG on the cell surface in CNFY binding, we specifically removed HS-chains from GAG by treating cells with a recombinant active human heparanase 1 (HPSE). Therefore, HeLa cells were incubated with active HPSE, washed and analyzed for toxin binding. As control for HPSE activity, we studied also binding of an anti HSPG antibody to the cells. As shown in Fig. 2C, treatment of HeLa cells with HPSE led to reduced binding of labeled GST-CNFY. Binding of GST-CNFY 709-1014 was even slightly stronger inhibited (by about 40%) equal to the reduced binding of a FITC-labeled anti HSPG antibody (Fig. 2C). The data indicate that HSPGs are involved in CNFY (more specifically CNFY 709-1014 ) binding to cells.

Identification of the relevant GAG
To study which GAG might be involved in CNFY binding, we carried out competition studies with several soluble GAG. To this end, HeLa cells were incubated with DL488-GST-CNF proteins in the presence and absence of, heparin (H), chondroitin sulfate A (CS), dermatan sulfate (DS), hyaluronic acid (HA), dextran sulfate (DxS) and fondaparinux (F) (Fig. 3A). As a control, also binding of labeled Clostridium difficile toxin A (TcdA) was studied. It was recently shown that TcdA binds to sulfated GAG (14). In the presence of heparin, hyaluronic acid and dextran sulfate, respectively, there was reduced binding of TcdA to Hela cells (Fig. 3B). Also binding of CNFY full length toxin was diminished by addition of dextran sulfate. Importantly, there was nearly complete inhibition of binding of CNFY 709-1014 in the presence of heparin, indicating that CNFY Cterminal domain has high affinity to heparin and presumably binds to heparan sulfate HSPG on the cell surface, which is in line with the results shown above (using heparanase 1). Heparin, dermatan sulfate and dextran sulfate showed an inhibitory effect on CNFY 709-1014 . However, reduced binding of the full toxin was only marginally inhibited (significant only with dextran sulfate) and was not sufficient, to block intoxication of cells. RhoA was modified in all cells treated with CNFY or CNF1, respectively, independent of the competitors added. As expected, the catalytically inactive mutants of both toxins (C866S) did not induce a shift of RhoA (Fig. S3).
Importantly, the reduced amount of toxin bound to the cell surface seems still sufficient to modify the complete pool of RhoA within the cells. Taken together, the isolated Cterminal domain of CNFY interacts strongly with heparin and binds to dermatan and dextran sulfate. Binding of the full toxin was not influenced dramatically in the presence of polysugars.

Direct interaction of CNFY 709-1014 with heparin
To further characterize the binding affinities of CNFY 709-1014 with heparin, we utilized the bio-layer interferometry (BLI) assay. Binding of a protein to immobilized heparin results in a shift within the light interference pattern that can be monitored in real-time. The catalytic inactive full toxin CNFY C866S was used as binding control, whereas biotinylated hyaluronic acid and biotinylated cellulose served as sugar control. As shown in Fig. 4A, the full toxin only marginally interacted with heparin and did not bind to hyaluronic acid or cellulose, as expected from the competition studies (Fig. 3B). In contrast, there was strong binding of CNFY 709-1014 to heparin but no interaction to the sugar controls ( has high affinity to heparin with a K D in the low µmolar range (about 10 µM). In contrast, the affinity of the full CNFY to heparin was too low to estimate a K D . CNFY 709-1014 contains a typical heparin binding motif rich in lysine ( 772 VKKTKF 779 ). To analyze its potential involvement in GAG binding, we mutated each lysine in this motif to alanine (structure in S4A). We then labeled the respective recombinant proteins (CNFY 709-1014 ) with DL488 and studied binding to HeLa cells in the presence and absence of heparin.

GAG are cellular attachment factors for CNFY but not sufficient for intoxication
To study the possible involvement of GAG in toxin binding and uptake, we made use  17)). These cells express GAG, which carry shortened heparan sulfate chains, because EXT2 is specifically required for the elongation of the HS chain and no other types of GAG (12). FACS experiments with these cells showed 72% reduced binding of full length CNFY, as well as for CNFY 709-1014 , about 87% reduction (Fig. 5D), respectively, compared to wildtype HeLa cells. As a control for the functional knockout, we used TcdA, as well as a FITC-HSPG antibody. Also binding of TcdA and the HSPG-antibody to the EXT2 -/cells was 70% respectively 87% reduced (Fig. 5D). The data show that heparan sulfate chains are involved in CNFY attachment to the cell surface, but they may not be the unique binding structures. We again studied intoxication of HeLa and EXT-2 -/cells by analyzing the shift of deamidated RhoA in the lysates of toxin-treated cells. However, although we measured binding of CNFY to EXT2 -/cells, no modification of RhoA was detectable (Fig. 5E). Interestingly, these data suggest that the toxin did bind to EXT2 -/cells but was not able to enter the cytosol.

CNFY as well as CNFY 709-1014 are enriched in early endosomes
Importantly, the C-terminal part of the toxin (CNFY 709-1014 ) comprises the catalytic domain. Therefore, it is surprising that this domain binds with high affinity to heparin and heparan sulfate side chains whereas the full toxin did not. We asked what function sugar binding to the catalytic domain of a toxin could hold and where this interaction might occur. We studied, whether the isolated catalytic domain is sufficient to be taken up by endocytosis. Therefore, we analyzed localization of labeled GST-CNFY full and GST-CNFY 709-1014 , respectively following incubation of HeLa cells by co-staining with an antibody against Rab5, a marker for early endosomes. As shown in Fig. 6, both, the full toxin and the C-terminal domain co-localize with Rab5 and therefore appear enriched in early endosomes. This indicates that the interaction of CNFY 709-1014 with HSPG is sufficient for endocytosis.
Together our data suggest that CNFY binding to cells with shortened sugar chains on the surface GAG is reduced but not completely missing. However, intoxication of the cells, which requires endocytosis and release from the acidified endosomes is eventually blocked, suggesting a specific function of GAG within the endosome.

Heparin interaction with CNFY does not induce a conformational shift
To answer the question whether the shift to acidic pH or the interaction of HSPG with CNFY induces a conformational change, we measured the distance between the two single cysteine residues (C134 and C866) in CNFY using nitroxide spin labeling in combination with PELDOR spectroscopy (18). The PELDOR method exploits the distance dependence of the interaction between the magnetic dipole moments of two spin labels. Nitroxide spin-labeled CNFY samples were frozen and subjected to a fourpulse PELDOR sequence. The resulting echo decays were first baseline corrected ( Fig.   7, blue curves), then analyzed for dipolar oscillations, and their frequencies correlated to distance distributions (Fig. 7, black curves) between the points of highest electron spin-density in the two nitroxide radicals covalently attached to C134 and C866.
Prominent peaks in the distance distributions can be observed between 4 and 5 nm.
Specifically, CNFY at physiological pH 7.4 exhibits a peak distance of 4.1 nm (Fig. 7A), which shifts to 4.7 nm upon lowering the pH to 5 (Fig. 7C). Addition of heparin did not alter these values within the experimental error (Fig. 7, panels B and D). Hence, PELDOR data indicate that the acidic pH rather than heparin binding induces a conformational change of the toxin.

Interaction of CNFY with heparin is unaffected by the pH
Because the acidic pH of endosomes is required for toxin release, which was previously shown by bafilomycin A1, an inhibitor of the endosomal proton pump (1, 16), we studied whether heparin interaction with CNFY is pH dependent. Therefore, we acidified the cell culture medium to mimic the pH present in late endosomes on the surface of HeLa cells. Under these conditions competition assays with heparin were performed. Therefore, HeLa cells were incubated with the labeled CNFY or labeled CNFY 709-1014 at pH 7.4 or pH 5.0 in the presence or absence of heparin, respectively as indicated in Fig. 8. In the absence of heparin, binding of labeled CNFY to the cells was slightly diminished at acidic pH. A further inhibition by heparin was visible, but not significant (Fig. 8A). In contrast, binding of labeled CNFY 709-1014 was even increased at acidic pH but significantly inhibited in the presence of heparin at acidic and neutral pH, respectively, indicating that binding of the C-terminal part of CNFY to heparin is not diminished at acidic pH (Fig. 8B). Acidification of the endosomes seems not sufficient to separate the toxins C-terminal domain from membrane bound HSPG. Membrane interaction is required for the passage through the membrane. However, coming off the membrane should be a prerequisite for the toxins release into the cytosol.

Cleavage of heparin is required for the release of CNFY from the endosome
High affinity binding of the C-terminal domain to heparan sulfate within the endosome would interfere with its release into the cytosol. It is well known that an acidic pH not only induces a conformational change of proteins and activates proteases but also leads to higher activity of heparanases (19). Cleavage of heparin would also relieve the toxin from its attachment site at the inner endosomal membrane. Therefore, we intended to inhibit the activity of heparanase by reducing its expression with siRNA and to study its effect on cell intoxication. As shown by Western Blotting, siRNA significantly reduced the amount of heparanase in HeLa cells (Fig. 9A, B). In line with the assumption of a crucial role of heparinase in the intoxication process, deamidation of RhoA in siRNA-treated cells occurs slower compared to cells treated with control siRNA (Fig. 9C).

Model for binding and uptake of CNFY
Our data lead to the following model for the cellular uptake of CNFY (Fig. 10).
CNFY binds with its N-terminal domain to an unknown protein receptor, which may already lead to a conformational change of the toxin allowing an interaction of the Cterminal domain with HSPGs thus enhancing avidity. Following receptor-mediated endocytosis and acidification, a conformational change of CNFY leads to cleavage retaining high affinity contact of the C-terminal domain to HS side chains at the membrane. Acidification allows incorporation of the translocation domain into the endosomal membrane and leads to activation of heparanase. The C-terminal part of CNFY would remain attached to the inner endosomal membrane by its high affinity interaction to HS. Release into the cytosol therefore requires cleavage of HS by the activated heparanase.

Discussion
In a recent publication by Heine et al., a relevant influence of CNFY on proinflammatory IL-6 production in mice was reported. Surprisingly, binding and intoxication studies indicated that CNFY does not influence immune cells directly. IL-6 is expressed by immune cells but also by fibroblasts and endothelial cells. Following our results presented here, the later are more likely targets of CNFY. Based on these findings, we intended to study the reason for the missing immune cell response by analyzing binding and uptake of CNFY in more detail. Our group showed before, that treatment of cells with sodium chlorate, which induces under-sulfation of cellular proteoglycans, inhibits intoxication of HeLa cells (10), indicating that sulfated sugars, which are rarely present on immune cells, could be involved in a crucial step of the toxins entry path. Interestingly, it has been shown that heparan sulfate proteoglycan (HSPG) binding of viruses allows movement on the cell surface until the protein receptor is bound to mediate endocytosis (20). To verify our former results, we directly removed HSPG form the cell surface by heparanase treatment. A specific antibody recognizing an epitope of heparan sulfate (21)  HSPGs are extracellular matrix components (27), they are found in secreted vesicles of mast cells (serglycin, (28)), described as sequestering co-receptors for growth factors and cytokines (fibroblast growth factor (FGF), (29,30)) or binding partners for molecules undergoing transcellular transport (31). In principle, HSPG binding could be sufficient for endocytosis of the toxin. Therefore, we studied whether or not CNFY would be enriched within endosomes. Using microscopy, we found that endocytosis is not inhibited by knockout of Exostosin 2. However, no deamidation of RhoA occurred, indicating that the release into the cytosol is not possible. We suggest an additional function of HS binding within the endosome. Different from Diphtheria toxin, CNFs do not form a disulfide bond between receptor binding and catalytic domain. Cleavage would therefore release the catalytic domain into the lumen of the endosome and subsequent degradation. It has been shown, that acidification of endosomes is crucial for the release of many bacterial toxins into the cytosol. Bafilomycin A1, which inhibits acidification by blocking the endosomal proton pump, completely interferes with intoxication of HeLa cells (10). Acidification in the late endosomes leads to unfolding of the protein toxin, which allows insertion of two hydrophobic helices (located in the translocation domain) into the endosomal membrane. In PELDOR studies we showed that a conformational change of CNFY occurs following acidification, whereas there is no detectable change of conformation upon heparin binding. Moreover, heparin binding takes place at neutral and acidic pH. For CNF1 we showed that the protein chain is cleaved by a protease. This may also depend on the pH. The high affinity interaction to a membrane component should inhibit translocation from the endosome to the cytosol. It has not yet been shown that CNFY is cleaved by proteases. However, we assume that this is the case because of the high homology of the toxins (8). However, it is the C-terminal, catalytically active part, which shows affinity for heparin. Concerning our results presented here, a third function of acidification could be the activation of heparanase. This enzyme cleaves HS chains from HSPG within the endosome in a pH dependent manner. Recently, it was shown that acidification-dependent activation of heparanase is required for the release of a cytosol-penetrating antibody (19).

Knockdown of heparanase clearly diminished intoxication of HeLa cells with CNFY,
indicating a crucial function of this enzyme for intoxication. Therefore, GAG binding of the catalytic domain of CNFY and cleavage of heparin seems to be crucial for intoxication. However, the deficiency of sugar chains should also lead to missing exposure of glycoproteins on the cell surface, among which the protein receptor may exist. We cannot study nor exclude this possibility, since the potential protein receptor is not yet known.

Materials
Used antibody, application and supplier: Mutation sites and primer sequence:

Cloning, Mutagenesis and Purification of Recombinant TcdA protein
The used recombinant TcdA was expressed and purified as described previously (32).
Then culture medium was replaced by 2.

SDS Page
Cell lysate was loaded on a 12.5% SDS-Gel. In the case of RhoA deamidation, a 12.5% urea SDS-Gels was used. After protein separation by SDS-PAGE, the gels were either stained with Coomassie or electro transferred onto a PVDF membrane.

Western Blot
Proteins were transferred onto a PVDF membrane (Carl Roth) using the Wet tank plotting system (Mini Trans-Blot ® , BioRad). The membrane was blocked with 5% skimmed milk for 1 h at RT. After washing the membrane 3 times with TBS-T for 10 min, primary antibody was added and incubated overnight at 4°C. Unbound antibody was removed by washing 3 times with TBS-T for 10 min. The secondary HRP-linked antibody was subsequently incubated with the membrane for 1 h at RT. Following washing, membranes were developed using SignalFire™ ECL Reagent (Cell signalling) and finally exposed with the imaging system Amersham Imager 600 (GE Healthcare Life Science).

FACS analysis
Toxins were labelled with DyLight488 Maleimide (

Competition Assay with GAGs or unlabelled CNF
Cells were co-incubated with DL488-GST-CNFY and different concentrations of glycosaminoglycan or unlabelled GST-CNF, respectively as indicated for 20 min at 4°C.
Unbound toxin was removed by washing with ice-cold PBS. Cell bound toxin was measured using a FACS Melody (BD Bioscience).

Confocal images for co-localization
Cells were seeded on coverslips and fixed in 4% para-formaldehyde in PBS, permeabilized with 0,3% Triton X-100, 5% FCS in PBS at RT. Samples were incubated with the Rab5 antibody overnight at 4°C. After washing and incubation with the secondary antibody for 1 h at RT, coverslips were mounted with Prolong diamond + DAPI. Confocal images for co-localization of toxin and the early endosomal marker Rab5 were acquired with a LSM 800 Confocal Laser Scanning Microscope (Zeiss), equipped with a 63x /1.4 NA oil objective and Airyscan detector (Zeiss). Airyscan images were processed by using Zen software (Zeiss)
Concentrated sample was loaded into 3.8 mm (outer diameter) synthetic-quartz EPR tubes (Qsil GmbH) and frozen in liquid nitrogen, either directly or after adding heparin.

PELDOR
Dead-time-free four-pulse electron-electron double resonance (PELDOR) experiments (33) were carried out at Q-band frequency (33.8 GHz) with Bruker Elexsys E580 instrumentation equipped with a PELDOR unit. Microwave pulses were amplified using a 3-W solid-state amplifier. A Bruker dielectric-ring resonator (EN 5107-D2) was used that was immersed in a continuous-flow helium cryostat (CF-935, Oxford Instruments).
The temperature during the measurements was 60 K and stabilized to +/-0.1 K by a temperature control unit Oxford Instruments ITC-503. In all measurements, the pump frequency was set to the center of the resonator dip as well as to the maximum of the nitroxide EPR spectrum. The observer frequency was shifted up by approximately 45 MHz. All measurements were performed with observer pulse lengths of 20 and 40 ns for π/2-and π-pulses, respectively, and a pump-pulse length of 40 ns.
PELDOR analyses have been carried out using the analysis tool GloPel (34), which is available free of charge at https://www.radicals.uni-freiburg.de/de/software/glopel-1.
Standard Tikhonov regularization was used for the analysis of all PELDOR time traces.
All resulting distance distributions have been validated using a method described earlier (35).

Regression analysis of FACS-derived toxin binding data.
Regression analysis was performed with Sigma Plot (Systat Sofware). Toxin binding data from FACS experiments were depicted as scatter plots. Utilizing the Sigma Plot curve fitting tool, "best fit" regression curves were modeled for different binding kinetics.
The following regression functions were used for modelling: One binding site, f(x)=Bmax1*x/(KD1+x).

Figures
Statistical analysis was performed using student's t-test. * p < 0.05.
(C) Time dependent shift assay: SiRNA-treated HeLa cells were incubated with GST-CNFY for 1-4 h. RhoA was detected by Western blotting following urea SDS PAGE.
GAPDH was used as a loading control. The experiment was repeated three times independently with similar results.