HeLa cells were regularly cultured in DMEM low glucose (1 g/L) with L-glutamine (Sigma Aldrich, St. Louis, MO, USA), supplemented with 10% fetal bovine serum, 1% non-essential amino acids and penicillin/streptomycin. Cells were cultured at 37°C in a 5% CO₂ atmosphere. Media was replaced every 2–3 days. For experiments where phosphorylation levels of p65 were determined, cells were serum-starved overnight. Experiments were performed at passage number 24–32. Regular tests demonstrated the cells to be free of Mycoplasma contamination.

The following antibodies were used in the study: NF-κB p65 D14E12 (Cell Signaling Technology, Danvers, MA, USA), phospho-NF-κB p65 (Cell Signaling Technology), FLAG M2 (Sigma-Aldrich, St. Louis, MO, USA), Rab5 (Cell Signaling Technology), Rab6 (Santa Cruz Biotechnology, Dallas, TX, USA), Rab7 (Cell Signaling Technology), CD63 H5C6 (Developmental Studies Hybridoma Bank, Iowa City, IA, USA), beta COP1 (Abcam, Cambridge, United Kingdom), GM130 (Cell Signaling Technology), alpha tubulin DM1A (Sigma-Aldrich), GAPDH FL335 (Santa Cruz), HRP-labeled goat anti rabbit (Dianova, Hamburg, Germany), HRP-labeled goat anti mouse (Dianova), Cy3-labeled goat anti rabbit (Dianova) Cy3-labeled goat anti mouse (Dianova).

The following chemicals were used in this study: Amiloride (Sigma-Aldrich), Bafilomycin A1 (invivoGen, San Diego, CA, USA), BAY 11-7085 (Calbiochem, San Diego, CA, USA), 2′7-bis(2-carboxyethyl)-5(6)-carboxy-fluoresceinacetoxy-methylester (Sigma-Aldrich), 6,P-toluidinyl-naphtalene-2-sulphonate (TNS) (SigmaAldrich), 5-(and 6-) carboxynaphthofluorescein succinimidyl ester, mixed isomers (Thermo Fisher Scientific, Waltham, MA, USA), Cytochalasin D (Sigma-Aldrich), DAPI (Sigma-Aldrich), DAKO Mounting Medium (DAKO, Glostrup, Denmark), Dynasore (Enzo Life Sciences, Farmingdale, NY, USA), Draq5 (BioStatus, Shepshed, United Kingdom), Filipin (Sigma-Aldrich), IL1β (Sigma-Aldrich), Lysotracker red DND99 (Thermo-Fisher Scientific), Methyl-β-cyclodextrin (Sigma-Aldrich), N-Lauroylsarcosine sodium salt (Sigma-Aldrich), Nocodazole (Sigma-Aldrich), Propidium iodide (Sigma-Aldrich), Trypan blue (Sigma-Aldrich).

The prediction of the potential (PTDs) was performed by Ü. Langel (CePeP, Sweden) using the prediction tool MPEx (Snider et al., 2009), which is based on the Wimley-White hydrophobicity scale (White and Wimley, 1999).

Plasmids for the overexpression of proteins were constructed using the pET24b(+) 3xFLAG vector (Lubos et al., 2014) in order to generate rNleC or rTat NleC with C-terminal 3xFLAG tags for detection of the protein and 6xHis for purification. To insert the nlec gene into the vector, a restriction-free cloning approach was chosen (Chen et al., 2000; van den Ent and Löwe, 2006) using the primer pairs F-NleC 3xFLAG/R-NleC 3xFLAG or F-Tat NleC 3xFLAG/R-NleC 3xFLAG (Table 1) and genomic EPEC E2348/69 DNA. Restriction-free cloning was performed as described in Lubos et al. (2014). Deletion mutants were generated using an inverse PCR approach with the insertion of restriction sites for EcoRI using deletion primer pairs (F-NleC Δ183–187/R-NleC Δ183–187 or F-NleC Δ208–257/R-NleC Δ208–257) employing pET24b(+): NleC 3xFLAG as a template. The resulting PCR product was digested with EcoRI and DpnI and ligated. In order to generate rPTD1+2 GFP the sequence for ptd1+2 was inserted into pET24b(+) GFP 3xFLAG (Lubos et al., 2014) via inverse PCR with the primer pairs F-PTD1+2GFP and R-PTD1+2GFP generating pET24b(+) PTD1+2 GFP 3xFLAG. The corresponding primer sequences are listed in Table 1. The resulting PCR product was digested with DpnI and ligated. The plasmid pET24b(+) GFP 3xFLAG had been constructed previously (Lubos et al., 2014). Plasmids were transformed into ClearColi® (Lucigen), expressed at 37°C (OD₆₀₀ = 0.7) and induced with 1 mM IPTG for 4 h. Bacterial cultures were harvested by centrifugation. Bacterial pellets were resuspended in 50 mM Tris HCl, pH 8.0, 500 mM NaCl, 10 mM Imidazole, 10% Glycerol, and 0.1% Triton X-100. Bacterial lysis was performed by sonication. The crude extract was incubated with 2% N-Lauroylsarcosine for 1 h prior to centrifugation at 7,200 × g for 15 min. The soluble fraction was added to Protino® Ni-NTA Agarose (Macherey-Nagel) for 1 h at 4°C with rotation. The resin was washed twice with 50 mM Tris HCl, pH 8.0, 500 mM NaCl, 20 mM Imidazole, 10% Glycerol, and 0.5% Triton X-100 and once in 50 mM Tris HCl, pH 8.0, 500 mM NaCl, 20 mM Imidazole, 10% Glycerol, and 0.1% Triton X-100. Proteins were eluted from the Ni-NTA agarose matrix with 50 mM Tris HCl, pH 8.0, 500 mM NaCl, 60 mM Imidazole, and 10% Glycerol, dialyzed twice against phosphate-buffered saline (PBS) and concentrated using Centricon centrifugal filters (Milipore). The protein concentration was determined by a BCA assay (Thermo Scientific). Proteins were stored at 4°C in PBS until use.

To assess the catalytic activity of rNleC and of the different fusion/deletion versions of NleC, 2.2 × 10⁶ cells were seeded into a 10 cm dish (one 10 cm dish per sample). The following day, HeLa cells were lysed with 350 μl cold radioimmunoprecipitation assay buffer (RIPA buffer: 25 mM Tris-HCl pH 8.0, 137 mM NaCl, 0.1% /w/v) SDS, 0.5% (w/v) Na-deoxycholate, 10% (v/v) Glycerol, 1% (v/v) Nonident P40) in the absence of protease inhibitors and gentle agitation for 30 min at 4°C. Lysates were pooled and cell debris was removed by centrifugation at 10,000 × g for 10 min at 4°C. 350 μl of the supernatants were incubated with 250 μg protein for 30 min at 4°C. This equals the maximum amount of protein that was added to cell supernatants to test whether the applied amount of protein was sufficient to cleave p65. 25 μg/ml equals 576 nM rNleC, 560 nM rTatNleC, 581 nM rNleC Δ183–187, 659 nM rNleC Δ208–257, 827 nM rGFP, or 662 nM rPTD1+2 GFP. The reaction was stopped by the addition of 4x Laemmli buffer and subsequently the cellular proteins were denatured by incubation for 10 min at 95°C. The enzymatic activity was confirmed by Western blotting using an α-NF-κB p65 antibody (1:1,000 dilution in 0.5% skim milk, 0.1% Tween 20 in PBS) for detection of p65 and its cleavage product. To examine the ability of rNleC to cleave p65 in intact cells, cells were seeded in 10 cm dishes, grown to 80% confluence and serum starved overnight. Cells were incubated for up to 4 h in a volume of 5 ml with recombinant protein with the indicated concentration and stimulated with 10 ng/ml IL1β for 20 min unless indicated otherwise. In order to block activation of the NF-κB signaling cascade, cells were pre-incubated with the inhibitor BAY 11-7085 (10 μM/1 h). For inhibiting endosome acidification 10 nM Bafilomycin A1 was added to the cells 2 h prior to protein incubation and subsequent IL1β stimulation. Cells were lysed in RIPA buffer for 30 min at 4°C and—for removal of cell debris—the lysates were centrifuged with 16,000 × g at 4°C for 10 min. 4 × Laemmli buffer was added and the samples were denatured further for 10 min at 95°C. For Western blotting, proteins were transferred onto nitrocellulose membranes using a semi-dry transfer chamber (Transblot SD Semi-dry transfer cell, BioRad Laboratories) for 1 h at 15 V. Unspecific binding sites were blocked with 5% skim milk, 0.1% Tween 20 in PBS for 1 h at RT. Primary antibodies were incubated at RT for 1 h or overnight at 4°C with α-NF-κB p65 (1:1,000 dilution in 0.5% skim milk, 0.1% Tween 20 in PBS), α-phospho-NF-κB p65 (1:1,000 dilution in 0.5% skim milk, 0.1% Tween 20 in PBS) antibodies. The recombinant proteins were detected in lysates with α-FLAG (1:1,000 dilution in 0.5% skim milk, 0.1% Tween 20 in PBS) antibody. As loading control, we used the reaction with α-tubulin (1:1,000 dilution in 0.5% skim milk, 0.1% Tween 20 in PBS) or α-GAPDH (1:1,000 dilution in 0.5% skim milk, 0.1% Tween 20 in PBS) antibodies. Enzymatic activity of rNleC and its derivatives was monitored by the formation of a p65 double band. For detection of primary antibodies, peroxidase labeled goat α-mouse, or goat α-rabbit secondary antibodies were incubated with the nitrocellulose membranes for 1 h at RT in a 1:10,000 dilution in 0.5% skim milk, 0.1% Tween 20 in PBS. Blots were washed in between incubation times three times 0.1% Tween in PBS. For detection, Pierce™ (ECL) Western Blotting Substrate (Thermo Fisher Scientific) was used as a substrate. The detection of emitted light was performed using a Lumi-Imager F1 (Boehringer Mannheim) with the software LumiAnalyst 3.1 (Boehringer Mannheim) (Figure 2) or Odessey Fc (Li-Cor Biosciences) with the software ImageStudio™ Lite (Li-Cor Biosciences) (Figures 3B, 5B, 8B and Figure S6A). Densitometric analysis was performed with the corresponding software programs. For evaluation, band intensities of cleaved p65 (lower p65 band) was divided by band intensities of total p65 (the sum of uncleaved (upper) and cleaved (lower) p65 band).

For studying uptake kinetics of the different recombinant proteins continuous time-lapse quenched uptake assays were performed as described previously (Lubos et al., 2014). Briefly, trypsinized HeLa cells were incubated with 25 μg/ml fluorescein isothiocyanate (FITC)-labeled proteins or 576 nM FITC-labeled peptides. Proteins were labeled using Fluorescein Isothiocyanate (FITC) the FluoReporter® FITC Protein labeling Kit (Thermo Fisher Scientific, Waltham, MA, USA) and labeling was conducted following the manufacturer's instructions. FITC-labeled peptides were synthetized and purchased from GenScript (Piscataway, NJ, USA). At the indicated time points samples were taken, quenched with 0.2% Trypan Blue (final concentration) and subjected to FACS analysis using a FACScan flow cytometer (BD BioScience). To measure a potential influence of the recombinant proteins on the integrity of cellular membranes, cells were incubated with 1 μg/ml propidium iodide (PI) in parallel. In order to block different endocytic uptake routes, cells were pre-incubated for 1 h with the appropriate inhibitor (30 mM dynasore to block dynamin—and clathrin-mediated endocytosis, 200 μM cytochalasin D to block F-actin elongation, 16.5 mM nocodazole to block microtubule polymerization, 19 mM amiloride to inhibit macropinocytosis, and 3.8 mM filipin or 50 mM methyl-β-cyclodextrin to block lipid raft mediated endocytosis). Cells were incubated with 25 μg/ml FITC-labeled recombinant proteins for 4 h at 37°C, washed with PBS, trypsinized, quenched with a final concentration of 0.2% Trypan Blue and immediately measured using the FACScan flow cytometer (BD BioScience). To measure the impact of protein on the acidification of endosomes, 50 μg/ml protein were incubated with HeLa cells. 3 h prior to the end of an experiment, 400 nM lysotracker DND99 was added to the cells to stain acidified endosomes. Cells were washed, trypsinized and subjected to FACS analysis using a FACScan flow cytometer (BD BioScience). To only measure protein residing in a neutral environment within the cell, endpoint assays with 5-(and 6-)carboxynaphthofluorescein (NF)-labeled proteins were conducted. Proteins were labeled using the dye 5-(and-6)-carboxynaphthofluorescein succinimidyl ester mixed isomers (Thermo Fisher Scientific, Waltham, MA, USA) and labeling was performed according to the manufacturer's instructions. For the uptake assays with NF-labeled proteins, HeLa cells were grown to 80% confluence in a 6-well plate format and incubated with 25 μg/ml NF-labeled protein for the indicated times. To remove putative extracellular signals the cells were washed with PBS and quenched with 0.2% glycine in PBS for 5 min. Cells were trypsinized and fluorescence intensities were measured using FACScan flow cytometer (BD BioScience). To exclude cross-signals arising from proteins residing in neutral early endosomes, pulse-chase assays were performed with NF-labeled proteins. Cells at 80% confluence were pre-chilled on ice for 30 min, followed by incubation with 50 μg/ml NF-labeled protein for additional 30 min on ice and then the cells were transferred to 37°C for 1 h. Cells were washed with PBS and fresh media was added for the indicated times. To quench signals from the outside of the cell, cells were incubated with 0.2% glycine in PBS for 5 min and trypsinized. Cells were directly measured using FACScan flow cytometer (BD BioScience). In each FACS experiment the fluorescence intensity (GeoMean) of 10,000 cells was measured in triplicates. Each experiment was conducted at least three times.

HeLa cells were seeded on cover slips and grown over night. Labeled proteins or peptides were added to cells in the indicated concentrations and at the given time points. To stain acidified lysosomes, 400 nM Lysotracker DND99 was incubated 3 h prior to the end of an experiment. Cells were washed twice for 20 min with 0.05% Tween, fixed with 4% paraformaldehyde (PFA) for 20 min at RT, quenched with 0.2% glycine in PBS and permeabilized with 0.2% Triton X 100. For co-localization studies, the specimen were blocked in 5% goat serum in PBS and incubated with the appropriate antibodies for 1 h at RT. In order to demonstrate co-localization with different markers of the endosomal-lysosomal pathway, specimen were visualized with primary antibodies raised against Rab5 (1:50) as a marker for early endosomes, Rab7 (1:25) as a marker for late endosomes and CD63 (1:600) as a marker for late endosomes/early lysosomes. In order to investigate co-localization with markers of the retrograde trafficking route, specimen were incubated with primary antibodies raised against β-COP1 (1:100) as a marker for trafficking endosomes, Rab6 (1:100) to mark the trans-Golgi/ER network and exocytic vesicles or GM130 (1:25) to stain for the Golgi apparatus. For detection the appropriate α-mouse or α-rabbit Cy3-labeled antibodies were employed. Where indicated, filamentous actin was stained with Phalloidin-TRITC (Sigma) for 30 min at RT in a 1:500 dilution. The nucleus was stained with DAPI (1:500) (Sigma) or Draq5 (1:500) (Biostatus) for 30 min at RT. Cells were washed 3 times with PBS after each step. Before mounting in DAKO mounting medium, cover slips were washed once in ddH₂O. Cells were visualized using a Zeiss LSM 510, Zeiss LSM 780, or Zeiss LSM 800 confocal laser-scanning microscope. Quantification of co-localization was performed with the free available BioImageXD software (Kankaanpää et al., 2012). Briefly, LSM files were uploaded to BioImage XD to analyze a pixel-based co-localization between two channels of interest. Thresholds were adjusted automatically and statistics were calculated to obtain reliable quantitative results.

To measure the release of LDH in the culture medium, 1 × 10⁴ HeLa cells per well were seeded in a 96 well plate. The next day, 50 μg/ml recombinant protein was incubated with HeLa cells for 1, 4, and 24 h in a volume of 100 μl. The quantification of LDH was performed using the CytoTox 96® Non-Radioactive Cytotoxicity Assay kit (Promega GmbH, Mannheim, Germany) which is based on a colorimetric two-step enzymatic reaction. The assay was performed following the manufacturer's instructions.

Conformational changes of recombinant proteins in vitro were measured using 6,P-toluidinyl-naphthalene-2-sulfonate (TNS) (Sigma-Aldrich) (Albani, 2009). 1.5 μM recombinant protein was incubated in buffers ranging from pH 7.5 to pH 4.0 (150 mM NaCl, 100 mM ammonium acetate adjusted to pH 4.0, 4.5, 5.0, and 5.5, 150 mM NaCl, 100 mM MOPS adjusted to pH 6.0 and 6.5 or 150 mM NaCl, 100 mM HEPES adjusted to pH 7.0 and 7.5) together with 150 μM TNS for 15 min at RT in non-autofluorescent 96 well plates (black, 655079, Greiner) TNS fluorescence was measured using the photospectrometer SpectraMax MS (Molecular Devices GmbH) (excitation wavelength: 270 nm, emission spectra: 400–500 nm). The background fluorescence (buffer + TNS) was subtracted from the signal. For easier comparison of the capacity to alter their conformations between different recombinant proteins, the fold change between pH 7.5 and pH 4.0 was calculated.

We used blastp (including or excluding E. coli) to reconstruct the affiliations of NleC to AIP56 and APSE-2 ORF D. A subsequent blastp using all three queries revealed the complex pattern of related sequences shown in Figure 9A.

The multiple sequence alignment tool Clustal Omega (http://www.ebi.ac.uk/Tools/msa/clustalo/) was used for pre-aligning protein sequences under default settings. This alignment was used for a manual curation and removal of non-overlapping sequence regions in Se-Al Sequence Alignment Editor v2.0a11 (http://tree.bio.ed.ac.uk/software/seal/). Figure 9B shows a schematic representation of the aligned sequence regions. The final alignment is presented in fasta format as Figure S11. For AA identity or similarity calculations, we used the NCBI blastp suite for aligning two sequences.

The SplitsTree4 software (version 4.10) was used for computing an unrooted phylogenetic network. Settings were Uncorrected_P distance, NeighborNet, and bootstrap with 10,000 replicates.

In our previous work, we found the effector proteins YopM from Y. enterocolitica and SspH1 from S. enterica serovar Typhimurium to have cell-penetrating abilities (Rüter et al., 2010; Lubos et al., 2014). Hence, we investigated whether also other effector proteins might share this ability to overcome membranes of targeted host cells. Therefore, we first analyzed known effector proteins of EPEC in silico for the presence of putative protein transduction domains (PTDs). Here, we identified NleC, a zinc-dependent metalloprotease, which cleaves p65 of the NF-κB pathway and thus contributes to the abrogation of the host's innate immune response. The in silico analysis indicated that NleC might potentially harbor two PTDs near the C-terminus. These regions form α-helical structures and exhibit a high arginine content, which would favor attachment to and uptake across eukaryotic membranes (Figure 1A and Figure S1A; Tang et al., 2013). For further analysis, we generated additional versions of NleC (Figure 1B). For rNleC Δ183–187 the HEXXH catalytic motif as part of the catalytic domain has been deleted. This version should be taken up by host cells, but should be catalytically inactive and unable to cleave p65. rNleC Δ208–257 lacks the in silico-predicted PTDs and should thus be unable to get across eukaryotic membranes, however, should be catalytically active. Our initial strategy to control catalytic activity included a point mutation of histidine 187 to tyrosine, which was shown in a previous publication to abolish the catalytic activity of NleC (Yen et al., 2010). The validity of the point mutation had been confirmed by sequencing. However, when we investigated the activity of this mutant in HeLa cell lysates, we observed quite some residual catalytic activity (Figure S1B). Therefore, we decided to delete the whole HEXXH signature. rTat NleC is a fusion protein of NleC with the well-described CPP Tat (“transactivator of transcription”; Frankel and Pabo, 1988) at the N-terminus. This protein served as a control to compare its ability to enter cells with that of the wt rNleC and of the two deletion mutants (Figure 1B).

First, we tested the catalytic functions of rNleC and rNleC deletion mutants in vitro. In Figure 2A rNleC and different NleC variants were incubated with HeLa cell lysates. Only rNleC and rTat NleC proteolytically cleaved the NF-κB subunit p65, whereas the deletion mutants were unable to cleave p65. NleC Δ208–257 harbors the HEXXH motif that had been identified as part of the active Zn-binding site and hence serves as a signature for Zn-metalloproteases (Silva et al., 2010) and should in theory be able to cleave p65. However, a comparison of the predictions of the 3D structures for NleC and NleC Δ208–257 suggested extended conformational changes close to the catalytic center in the deletion mutant, which might explain its inability to cleave p65 (Figure 2B). Furthermore, the amino acid Y227, which is deleted in NleC Δ208–257, coordinates the zinc ion and its deletion might result in loss of protease activity (Li et al., 2014; Turco and Sousa, 2014). The functional rNleC protein was tested for catalytic activity in intact HeLa cells. However, rNleC was unable to cleave p65 even after 4 h of incubation (Figure 2C), although the protein was shown to be functional in cell lysates (Figures 2A,C). In non-stimulated cells, NF-κB is retained in the cytosol by the inhibitor κB (IκB), which is phosphorylated and degraded upon stimulation. During EPEC infection the NF-κB signaling cascade is activated due to the recognition of pathogen-associated molecular patterns (PAMPS) and IκB is degraded (Johannessen et al., 2013). In this scenario, T3SS-injected NleC cleaves the p65 subunit of NF-κB, potentially facilitated by the activation of the NF-κB pathway during infection and the ensuing lack of competition with IκB for the p65 substrate. To test this assumption, cells were activated by incubation with IL1β (10 ng/ml), which induces NF-κB signaling (Figure 2D). Indeed, after 20 and 45 min of activation, we could observe cleavage of p65 as indicated by the appearance of the p65 double band in Western blotting. This clearly showed that the NF-κB pathway needed to be activated for efficient cleavage of p65 by rNleC and implies a preference of NleC to free p65 rather than to IκB bound p65. To exclude an impact of IL1β on the integrity of the cells, cells were pretreated with 50 μg/ml rNleC for 30 min together with 1 μg/ml propidium iodide (PI) followed by stimulation with 10 ng/ml IL1β for 3.5 h. The uptake of PI was measured using flow cytometry (Figure S2). No significant differences in PI uptake between IL1β treated and IL1β untreated cells could be observed.

Intact cells need to be activated with IL1β for rNleC to cleave p65 of NF-κB. To demonstrate that this effect was due to the stimulation of the NF-κB pathway, we compared the effects of rNleC and rTat NleC in IL1β-stimulated cells to the effect of rNleC and rTat NleC in cells that had been pre-treated for 1 h with the NF-κB inhibitor BAY 11-7085 which shuts down NF-κB signaling (Figure 2E). Pre-incubation with the inhibitor BAY 11-7085 nearly completely abrogated p65 cleavage by rNleC and rTat NleC in IL1β activated cells compared to BAY 11-7085 untreated cells. In summary, we demonstrated that rNleC only cleaves p65 in stimulated cells with an activated NF-κB pathway since inhibition of NF-κB signaling with the inhibitor BAY 11-7085 also blocks p65 cleavage by rNleC.

Previously, Rüter et al. had shown that the T3SS-effector YopM of Yersinia enterocolitica as a recombinant expressed and purified protein autonomously entered eukaryotic cells independent of the T3SS thereby revealing properties of a cell-penetrating effector protein (CPE) (Rüter et al., 2010). Using an algorithm developed by Snider et al. we identified several other putative effector proteins of Gram-negative bacteria including SspH1 from Salmonella enterica serovar Typhimurium that potentially carry a PTD (Snider et al., 2009; Lubos et al., 2014). In this study, we found that rNleC reached and cleaved its intracellular target p65 when rNleC was added to intact, IL1β-stimulated cells. To further describe the kinetics of rNleC-uptake, FITC-labeled rNleC was added to trypsinized HeLa cells and the uptake of the protein was monitored using a FACS-based time-lapse quenched uptake assay as described by Lubos et al. (2014) (Figure 3A). rNleC showed a linear uptake for up to 8 h which was even more pronounced than internalization of rTat NleC, a fusion protein of rNleC with an established CPP (Hauber et al., 1987). rGFP was not internalized and in this context served as negative control (Figure 3A, left). To exclude that rNleC uptake might be due to cytotoxic effects impairing membrane integrity, the uptake of PI was measured in parallel. Although uptake of PI increased over time, there was no significant difference between cells treated with the various (rNleC) proteins or the medium control indicating that the uptake of PI was not mediated by cytotoxic effects (Figure 3A, right). Importantly, the fluorescent-labeled rNleC retained its catalytic activity as FITC-NleC could proteolytically cleave the NF-κB subunit p65 in HeLa cell lysates (Figure 3B). Uptake of these proteins was visualized additionally by fluorescence microscopy (Figure 3C). FITC-labeled proteins localized in dotted structures within cells after 1 and 4 h of incubation, which suggests their localization in vesicles. rGFP could not be detected inside cells and was employed as a negative control. Interestingly, cellular uptake of rNleC was more pronounced than uptake of rTat NleC. To elucidate whether the rNleC Δ208–257 and rNleC Δ183–187 deletion mutants were able to overcome eukaryotic membranes, the uptake of FITC-labeled rNleC deletion mutants was compared to FITC-labeled rNleC (Figure 3D) in a FACS-based continuous-quenched time-lapse uptake assay. Interestingly, not only the deletion construct lacking the predicted PTDs (rNleC Δ208–257), but also the inactive mutant (rNleC Δ183–187) showed only residual uptake when compared to wild type rNleC. However, based on structure predictions using Swiss-PdbViewer4.1.0., deletions of these amino acid sequence segments might have induced conformational alterations preventing efficient uptake. Hence, the uptake data suggest that already small changes in the three-dimensional structure of NleC reduce uptake efficiency to a basal level (Figure 3D, left). To exclude toxic effects induced by the rNleC Δ208–257 and rNleC Δ183–187 deletion constructs, we measured PI uptake in parallel. In this case, we could observe increased PI uptake following incubation with the recombinant proteins than for the medium control (Figure 3D, right). To further confirm that these deletion mutants did not exert a cytotoxic effect we additionally employed LDH release assays to monitor possible membrane damage. No cytotoxic effects could be observed (Figure S3). To exclude that the observed cleavage (Figure 2D) occurred due to increased uptake of rNleC or rTatNleC after stimulation with IL1β, we incubated 25 μg/ml rNleC FITC for 30 min followed by stimulation with 10 ng/ml IL1β for 3.5 h. Cells were washed trypsinized and quenched using TB and uptake of rNleC-FITC was measured using flow cytometry (Figure S4).

The in silico prediction suggested the presence of two potential PTDs in the C-terminal half of NleC (Figure 1). In accordance with these predictions uptake of the deletion variant rNleC Δ208–257, which lacks both of the predicted domains, was reduced. However, uptake of rNleC Δ183–187 lacking only the HEXXH Zn-binding metalloprotease signature region was also reduced. As CPPs (PTDs) are defined as small peptides, which are able to facilitate also cellular entry of cargo, we investigated whether PTD1 and/or PTD2 could autonomously enter cells and whether they might be able to transport cargo. For this, a FACS-based time-lapse quenched uptake assay was performed with a fusion protein of PTD1+2 and GFP (Figure 4A, left). The uptake of this recombinant protein was compared to FITC-labeled rNleC and to rGFP alone. We did not observe enhanced uptake of rPTD1+2 GFP compared to rGFP, whereas FITC-labeled rNleC was clearly detected inside cells. Cytotoxic effects of rNleC could be excluded since the parallel uptake observed for PI did not differ from the medium control (Figure 4A, right panel). As the uptake efficacy of CPPs might also depend on the nature of the cargos that are fused to the CPP (El-Andaloussi et al., 2007) we investigated the uptake of the FITC-labeled peptides PTD1 and PTD2 without any fused cargo and compared this to the well-described CPP Tat and a non-penetrating peptide (NPP) (Choi et al., 2012) in a FACS-based quenched time-lapse uptake assay as well as by fluorescence microscopy (Figure 4B and Figure S5). We could only observe a low uptake of PTD1 compared to NPP whereas PTD2 uptake did not exceed that of the negative control NPP, suggesting that neither the small increase in uptake for PTD1 nor for PTD2 would explain the remarkable uptake we observed for rNleC. In addition, we could observe a reduction of uptake for the deletion mutant of rNleC Δ183–187 to the same level as rNleC Δ208–257 (Figure 3D). Therefore, in summary, we can exclude that the predicted PTDs mediate the uptake of rNleC and its derivatives. Thus, rNleC cannot be designated as a CPE, as it does not fulfill the two main criteria for CPPs/CPEs: The predicted PTDs cannot enter eukaryotic cells and they are neither necessary nor sufficient to transport cargo into cells. Thus, far, our overall data suggest, that uptake of rNleC depends mostly on the native conformation of NleC and appears not to be mediated by distinct linear segments of amino acids as it is the case for CPEs.

Thus far, we have demonstrated efficient uptake of rNleC by eukaryotic cells (Figures 3A,C). However, the mechanism underlying the uptake process has not been addressed. Therefore, we asked whether uptake of the protein might proceed by an endocytic pathway which is followed by many viruses, CPE and bacterial toxins (Sandvig and van Deurs, 1996; Lakadamyali et al., 2004; Rüter et al., 2010).

To identify which endocytic pathway might be important for efficient rNleC uptake, HeLa cells were pre-incubated with different endocytosis inhibitors for 1 h followed by incubation of FITC-labeled rNleC for 4 h. After trypsinization, Trypan Blue-quenched fluorescence was measured using a cytometric approach (Figure 5A). Additionally, it was excluded that the presence of endocytic inhibitors reduces catalytic activity of rNleC (Figure S6A). A reduction in the uptake of more than 50% could be observed only for those cells pre-treated with methyl-β-cyclodextrin (MβCD), an inhibitor of lipid-raft mediated endocytosis. Inhibitors, such as cytochalasin D, which blocks actin-mediated processes, dynasore, which inhibits dynamin—and clathrin-mediated uptake, or nocodazole preventing microtubule—and clathrin-mediated uptake, did not reduce the uptake of rNleC significantly. Also, amiloride, an inhibitor blocking Na⁺-channels and thus micropinocytosis, or filipin, which binds to cholesterol, did not affect uptake of rNleC (Figure 5A). A putative cytotoxicity of the different inhibitors was monitored by measuring uptake of PI (1 μg/ml added together with rNleC) in parallel (Figure S7). Moreover, pre-incubation with MβCD strongly abrogated p65 cleavage by rNleC in IL1β activated cells compared to untreated cells (Figure 5B), indicating that the major uptake route for rNleC is lipid raft-mediated endocytosis.

As rNleC very likely enters cells via endocytosis we were interested in the subsequent trafficking of the endocytosed protein. Endosomal contents can be sorted into different endocytic routes, e.g., the endosomal-lysosomal route, which ultimately leads to degradation of the endocytosed material (Hu et al., 2015), retrograde trafficking along the trans-Golgi network and ER (Johannes and Popoff, 2008), or exocytosis which leads to recycling to the outside of the cell (Ward et al., 2005). To investigate the fate of intracellular rNleC, we analyzed a potential co-localization of FITC-labeled rNleC with markers of the endosomal-lysosomal pathway by fluorescence microscopy (Figure 5C). Markers for the endosomal-lysosomal pathway included Rab5 for early endosomes, Rab7 for late endosomes and CD63 for late endosomes/early lysosomes (Metzelaar et al., 1991; Christoforidis et al., 1999; Vanlandingham and Ceresa, 2009). A low co-localization of rNleC with Rab5 was detected after 30 min and 1 h. However, already after 1 h Rab5 co-localization could no longer be observed which is indicative of the well-documented fast shuttling through early endosomes. Between 30 min and 3 h rNleC strongly co-localized with Rab7 and subsequently also with CD63. These results strongly point toward uptake of rNleC via endocytosis with subsequent routing to lysosomes. To address whether rNleC could escape from endosomes or would undergo retrograde trafficking as it is known for example for Shiga toxin or other “long-trip” A-B toxins (Sandvig et al., 2010), we performed co-localization studies with markers for retrograde trafficking, such as Rab6, COP1, and GM130 (Figure S6B) (Nakamura et al., 1995; Grigoriev et al., 2007; Beck et al., 2009). As none of these markers co-localized with rNleC, we excluded retrograde transport as a possible mechanism for endosomal escape of rNleC.

To investigate whether and to what extend rNleC is degraded by lysosomes we quantified the fluorescence signal of cells treated with Bafilomycin A1, a potent inhibitor of H⁺-ATPases (V-ATPases) which blocks the acidification of lysosomes and thus inhibits lysosomal degradation, and compared it to cells treated with rNleC-FITC without inhibitor (Figure 6A). Our results indicated that partial degradation of the protein took place at 24 h of incubation. As an internal control, we confirmed the functionality of Bafilomycin A1 by quantifying the percentage of acidified lysosomes stained with Lysotracker red DND99 (Figure S8). However, we could still measure an increase in protein uptake after 24 h in Bafilomycin untreated cells, indicating that more protein was taken up after 24 h than degraded. To visualize degradation with lysotracker, we performed co-localization with the pH-sensitive marker Lysotracker red DND99. We confirmed, that rNleC co-localized with acidified lysosomes after 6 h, which could lead to a clearance of the protein from the eukaryotic cell (Figure 6B). Yet, we found rNleC residing in lysosomes for up to 24 h, indicating that, although lysosomes are functional, rNleC is not cleared completely from the cells after 24 h of incubation (Figure 6B). Although relative values compared to Bafilomycin-treated cells seemed to be relatively low, the absolute fluorescence in the cell was indeed significant. Furthermore, a more diffuse signal for rNleC after 16 h and especially after 24 h was observed, indicating that rNleC might have escaped from endosomes and has entered the cytosol. When quantifying the co-localization of rNleC and Lysotracker red DND99, we observed an increase in co-localization up to 16 h, whereas the co-localization decreased after 24 h of rNleC incubation, further indicating a separation of rNleC from lysosomes (Figure 6C). Incubation of cells with rNleC and rNleC variants suggested that these recombinant proteins had no effect on endosomal acidification (Figure 6D). With these experiments, we could show that rNleC, to some extent, is indeed degraded by lysosomes. However, cleavage of p65 after incubation with stimulated whole cells (Figure 2D) and separation of rNleC from lysosomes after 24 h indicate that at least some rNleC molecules can escape from the endo-lysosomal compartment.

To elucidate whether rNleC can escape from the endosomal-lysosomal pathway to reach its target in the cytosol as it had been shown for rYopM (Rüter et al., 2010), experiments with naphthofluorescein (NF)-labeled proteins were conducted. NF is a pH-sensitive dye, which emits fluorescence signals in neutral environments, whereas its fluorescence is quenched in acidic environments. Thus, NF-labeling of proteins can be used to measure only those proteins that have reached the neutral environment of the cytosol or the nucleus, whereas the fluorescence signal is quenched in an acidic environment as for example in late endosomes or lysosomes (Qian et al., 2015). To verify whether this fluorescent dye was a suitable marker for delivery of proteins into the cytosol, we incubated cells with NF-labeled BSA, which is endocytosed but subsequently degraded in lysosomes (Humphries et al., 2012). We could only measure background fluorescence when cells were treated with NF-BSA, whereas we could observe a significant increase in fluorescence after artificially disrupting endosomes with chloroquine (Figure S9). Thus, NF-BSA is a suitable control for a protein that is endocytosed but is not delivered into the cytosol. To follow uptake into the neutral environment (e.g., the cytosol) over time, NF-labeled proteins were added to cells for 1, 3, 6, and 12 h and uptake was measured by monitoring increasing fluorescence using flow cytometry (Figure 7A). During 12 h of incubation fluorescence of NF-BSA increased only very little in comparison with the NF-labeled rYopM, rNleC, and rTat NleC proteins. The strongest uptake could be observed for rNleC alone. This is in line with previous results, where also a reduced uptake for rTat NleC compared to rNleC was found (Figure 3A). In this experimental set-up, NF-labeled proteins were incubated with the cells during the complete incubation time, which means that proteins could also be continuously introduced into the cells. Since early endosomes are only slightly acidic and thus NF-labeled proteins residing in early endosomes might also emit fluorescence, we performed pulse-chase experiments to account for a putative continuous uptake of proteins. In this way, we were able to exclude a contribution of fluorescence signals derived from proteins residing in early endosomes. In the pulse-chase assays we confirmed for all time points higher fluorescence values for the NF-labeled rNleC, rYopM, and rTat NleC proteins compared to NF-labeled BSA. This indicated that rNleC, rYopM, and rTat NleC were able to escape from endosomes (Figure 7B). After 1 h, a reduction in fluorescence signals was observed which might be either due to degradation of proteins or might be caused by the maturation of early and non-acidified endosomes to late and acidified endosomes. Yet, the fluorescence signals remained at the same level between 3 h and 12 h suggesting that these proteins, once reaching the neutral environment, were stable within the cell and were not further degraded. To clarify the intracellular behavior of the deletion mutants, we performed the NF-uptake assay (Figure 7C) and the pulse-chase experiment (Figure 7D) with NF-labeled NleC deletion mutants and observed similar fluorescence intensities as for NF-labeled BSA, indicating that these proteins were apparently unable to be delivered into the cytosol.

NleC is structurally and functionally homologous to the A subunit of the zinc-metalloprotease A-B toxin AIP56 of (Phdp). AIP56 is the major virulence factor for colonization and dissemination in sea bass, the natural host for Phdp. Just like NleC, AIP56 also harbors a HEXXH motif, which is essential for cleaving its target, NF-κB p65. However, in contrast to NleC, AIP56 does not down-regulate pro-inflammatory cytokines, such as TNFα, IL1β, or IL8, but leads to apoptosis (Do Vale et al., 2005; Silva et al., 2013; Pereira et al., 2014). Besides the sequence homology of NleC and the A subunit of AIP56 including the catalytic center and their identical cellular target protein (Figure S10A), NleC and the A subunit of AIP56 also share a remarkable similarity in their predicted 3D-structure (Figure S10B). AIP56 induces its endosomal escape due to structural changes upon acidification within the endosomes (Pereira et al., 2014). Since NleC and the A subunit of AIP56 are also quite similar regarding the distribution of polar and hydrophobic amino acids (Figure S10C), we asked whether alterations in pH could lead to conformational changes in rNleC, which would promote its escape from acidified endosomes/lysosomes. As an indicator for conformational changes induced by acidic pH we used the hydrophobic molecular probe 6,P-toluidinyl-naphtalene-2-sulphonate (TNS) (Golczak et al., 2001; Albani, 2009) and monitored its alterations in fluorescence between pH 7.4 and pH 4.0. TNS is a molecular probe that is quenched in hydrophilic environments, whereas it emits fluorescence in hydrophobic environments. Hence, conformational changes resulting in increasingly exposed hydrophobic surfaces would increase binding of TNS and also fluorescence intensities. For easier comparison, alterations in TNS fluorescence have been depicted as fold changes at 440 nm (Figure 8A and Figure S11). Interestingly, rNleC as well as rTat NleC showed remarkably high fold changes in TNS fluorescence, whereas the rNleC deletion constructs exhibited fold changes in TNS fluorescence comparable to rGFP. This suggests that acidification of lysosomes induces substantial alterations in the conformation of rNleC and thus might be instrumental for rNleC to escape from endosomes, which is reminiscent of AIP56. Furthermore, these results might explain why deletions in NleC result in an abrogated escape from endosomes compared to wt rNleC. To assess whether acidification is the underlying mechanism facilitating endosomal escape of rNleC, we blocked endosome acidification with Bafilomycin A1 in HeLa cells for 2 h. Next, the cells were incubated with rNleC for 30 min and subsequently stimulated with IL1β to activate NF-κB signaling. As readout, the cleavage of p65 was determined by Western blotting (Figure 8B). Under these conditions the cleavage of p65 was reduced by 45%, which suggests that the acidification of endosomes is an important step for rNleC to reach its intracellular target p65.

The uptake of rNleC into endosomes and the subsequent escape from endosomes into the cytosol displays features known from short-trip A-B toxins. Because NleC and the A subunit of the short-trip A-B toxin AIP56 of Photobacterium damselae subsp. Piscicida (Phdp) show a high sequence similarity, we investigated their evolutionary relationship in more detail. A blastp search indicated that the E. coli NleC metalloprotease exhibits 38% identity (58% similarity) to the 294-amino acid N-terminal part of the plasmid-encoded apoptosis gene product, AIP56 of Phdp (Figure 9A and Data Sheet 1). This strongly suggests a potential relationship of these two bacterial virulence factors that both target, albeit by different mechanisms, the NF-κB signaling system. The 189 amino acid C-terminal region of AIP56 has 53% identity (67% similarity) to the γ-proteobacterium Candidatus Hamiltonella defensa (Entero-bacteriaceae) genome-integrated prophage product APSE-2 ORF D (A. pisum secondary endosymbiont; CP001277.1; position 1520562–1521326; Figure 9A). The annotated hypothetical protein D is 100% identical to the corresponding region in the isolated bacteriophage genome product APSE-2 (NC_011551.1; position 7577–8341; A. pisum secondary endosymbiont phage; Moran et al., 2005). A possible association of the N-terminal domain of AIP56 to NleC and the C-terminal domain to a hypothetical protein of the bacteriophage APSE-2 had been proposed previously by Silva et al. (2013).

In an expanded blastp screening of the NCBI non-redundant protein sequences using NleC, AIP56, and APSE-2 ORF D AA query sequences, we retrieved significant hits for diverse Enterobacteriaceae and Vibrionaceae. Remarkably, along with the bacterial and phage hits, some Eukaryonta were also represented with annotated hypothetical proteins similar to the C-terminal AIP56/APSE ORF D query sequences, with the highest similarity to the sequences of APSE-2 ORF D (46–60%). To visualize the relationships of these hits, we derived an evolutionary network (Figure 9B) that shows a clear, significantly separated cluster of the NleC-coding Enterobacteriaceae (green; bootstrap value 100%). AIP56-like sequences are dispersed over many Vibrionaceae including diverse species from the genus Vibrio as well as from P. damselae and the Enterobacteriaceae Arsenophonus nasoniae (blue species names in Figure 9B). Eukaryotic representatives as well as the bacteriophage APSE-2 sequence and Arsenophonus sequences are connected in the red section of the net.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.