Porphyromonas gingivalis induce apoptosis in human gingival epithelial cells through a gingipain-dependent mechanism

Background The oral pathogen Porphyromonas gingivalis has been shown to modulate apoptosis in different cell types, but its effect on epithelial cells remains unclear. Results We demonstrate that primary human gingival epithelial cells (HGECs) challenged with live P. gingivalis for 24 hours exhibit apoptosis, and we characterize this by M30 epitope detection, caspase-3 activity, DNA fragmentation and Annexin-V staining. Live bacteria strongly upregulated intrinsic and extrinsic apoptotic pathways. Pro-apoptotic molecules such as caspase-3, -8, -9, Bid and Bax were upregulated after 24 hours. The anti-apoptotic Bcl-2 was also upregulated, but this was not sufficient to ensure cell survival. The main P. gingivalis proteases arginine and lysine gingipains are necessary and sufficient to induce host cell apoptosis. Thus, live P. gingivalis can invoke gingival epithelial cell apoptosis in a time and dose dependent manner with significant apoptosis occurring between 12 and 24 hours of challenge via a gingipain-dependent mechanism. Conclusion The present study provides evidence that live, but not heat-killed, P. gingivalis can induce apoptosis after 24 hours of challenge in primary human gingival epithelial cells. Either arginine or lysine gingipains are necessary and sufficient factors in P. gingivalis elicited apoptosis.


Background
Chronic inflammatory periodontal disease is initiated by a bacterial biofilm called dental plaque that causes inflammation affecting the supporting structures of teeth, leading eventually to bone and tooth loss. Porphyromonas gingivalis is a Gram-negative anaerobe of dental plaque and a putative pathogen in chronic periodontitis [1]. The plaque bacteria possess numerous virulence factors including factors that aid intracellular invasion, intracellular persistence and host cell apoptosis [2].
Apoptosis or programmed cell death is triggered by two distinct signaling pathways; the intrinsic or stress-activated and the extrinsic or receptor-activated apoptotic pathway [3]. Both pathways activate their respective initiator caspases and converge to trigger executioner caspases 3, 6 and 7. The caspase cascade cleaves key cellular components responsible for the hallmarks of apoptosis such as chromatin condensation, pyknosis DNA fragmentation, cytoskeleton collapse, blebbing and formation of apoptotic bodies. Apoptosis is prevalent in the gingiva at sites of chronic bacteria-induced inflammation [4,5], particularly in the superficial cells of the junctional epithelium [5] and the fibroblasts and leucocytes of the connective tissue [4,5]. In vitro studies show that P. gingivalis can modulate apoptosis in the following cell types: fibroblasts [6,7], endothelial cells [8][9][10][11] and lymphocytes [12] and apoptosis has been proposed as a mechanism to explain the extensive tissue destruction in chronic periodontitis lesions.
It is not clear how P. gingivalis influences apoptosis in epithelial cells. In agreement with studies in fibroblasts, endothelial cells, cardiac myoblasts and lymphocytes, several authors [13,14] have shown induction of apoptosis in epithelial cells. In contrast, other laboratories [15][16][17] have shown inhibition of apoptosis by P.gingivalis. The reason for the discrepancies between these studies remains unknown, although variable challenge conditions were used. In this regard, the dose of bacteria and the duration of P. gingivalis challenge may be a critical parameter in determining whether induction or inhibition of apoptosis will occur. Thus, the aim of the current study was to characterize P. gingivalis-induced apoptosis of epithelial cells under various conditions, utilizing a wide array of apoptosis assays and gene expression profiling.

HGECs challenged with live P. gingivalis show early signs of apoptosis in a time-and dose-dependent manner
HGECs were challenged with live or heat-killed P. gingivalis 33277 at an MOI:10, MOI:100 and MOI:1000 for 4 and 24 hours and M30 epitope detection was performed with immunohistochemistry. M30 is an antibody that recognizes a specific caspase cleavage site within cytokeratin 18 that is not detectable in native cytokeratin 18 of vital cells. This occurs early in the apoptosis cascade, before Annexin-V reactivity or positive DNA nick labeling. Untreated cells were used as a negative control and cells treated with camptothecin 4 μg/ml for 4 hours, an apoptosis-inducing agent, were the positive control. Cells challenged with live or heat-killed bacteria at an MOI:10 showed no positive staining at any time point (data not shown). Cells challenged with live or heat-killed bacteria at an MOI:100 and MOI:1000 did not show any positive staining at 4 hours (data not shown). The epithelial cells appeared morphologically normal under all of the above conditions. However, challenge with live P. gingivalis at an MOI:100 for 24 hours increased the detachment of cells, while the remaining attached cells showed signs of blebbing, had pyknotic nuclei, and stained positive for M30 epitope, an early sign of apoptosis (Fig. 1C). In contrast, cells challenged with heat-killed P. gingivalis at an MOI:100 for 24 hours did not show any signs of apoptosis (Fig. 1D). Cells challenged with live P. gingivalis at an MOI:1000 for 24 hours completely detached from the plate, thus MOI:1000 was not used for subsequent experiments.

Live but not heat-killed P. gingivalis induce caspase-3 activation in HGECs in a time-dependent manner
HGECs were challenged with live or heat-killed P. gingivalis 33277 at an MOI:100 for 4 and 24 hours and caspase-3 activity was measured fluorometrically. Caspase-3 is an executioner caspase involved in both the extrinsic and intrinsic pathway of apoptosis. Caspase-3 activation plays a key role in the initiation of cellular events during the early apoptotic process. Untreated cells were used as a negative control and cells treated with camptothecin were the positive control. There was no significant increase in caspase-3 activity after 4 hours challenge with live or heatkilled bacteria (Fig. 2). However, after 24 hours challenge with live P. gingivalis, caspase-3 activity increased more than 2-fold compared to the negative control. In contrast, 24 hours challenge with heat-killed P. gingivalis resulted in a caspase-3 activity level similar to the negative untreated M30 epitope immunohistochemistry was used to detect cas-pase-cleaved cytokeratin-18 which is detectable in early stages of apoptosis Figure 1 M30 epitope immunohistochemistry was used to detect caspase-cleaved cytokeratin-18 which is detectable in early stages of apoptosis. Images are fluorescent confocal staining at ×600 magnification. The negative control was unchallenged HGECs with only media added (A). The positive control was HGECs treated with camptothecin 4 μg/ml for 4 hours (B). HGECs challenged with live P. gingivalis 33277 at MOI:100 for 24 hours show marked staining (C), while HGECs challenged with heat-killed bacteria under the same conditions show no detectable apoptosis (D). Challenging HGECs with an MOI:100 for 4 hours or MOI:10 for 4 and 24 hours showed no positive staining (no apoptosis) (data not shown). l control. These results are in accordance with our previous results, confirming that challenge with live, but not heatkilled, P. gingivalis at an MOI:100 for 24 hours can induce apoptosis in human gingival epithelial cells.

HGECs challenged with live P. gingivalis undergo DNA fragmentation in a time-and dose-dependent manner
HGECs were challenged with live or heat-killed P. gingivalis 33277 at an MOI:10 and MOI:100 for 4, 24 and 48 hours and DNA fragmentation was detected by ELISA, as well as by TUNEL. Untreated cells were used as a negative control and cells treated with camptothecin or DNase 1000 U/ml were used as a positive control. Once the caspase cascade has been activated, the inhibitor of caspaseactivated DNase (ICAD) is cleaved liberating this DNase and resulting in fragmentation of the chromosomal DNA. The Cell Death Detection ELISA can detect internucleosomal degradation of genomic DNA during apoptosis and provide relative quantification of histone-complexed DNA fragments (mono-and oligo-nucleosomes). There was no significant increase in DNA fragmentation after 4 hours challenge with live or heat-killed bacteria (Fig. 3). However, 24 hours challenge with live P. gingivalis, resulted in DNA fragmentation 3-fold higher than the negative control. On the other hand, 24 hours challenge with heat-killed P. gingivalis resulted in negligible increase in DNA fragmentation, suggesting that, although some apoptosis is evident after challenge with heat-killed bacteria, the effect is not statistically significant ( Fig. 3). At 48 hours, DNA fragmentation was at similar levels as at 24 hours. These results were also confirmed by TUNEL. The TUNEL assay measures and quantifies apoptosis by labeling and detection of DNA strand breaks in individual cells by fluorescence microscopy. The assay uses an optimized terminal transferase (TdT) to label free 3'OH ends in genomic DNA. Cells challenged with live or heat-killed bacteria at an MOI:10 did not show any positive staining at any time point (data not shown). Cells challenged with live or heat-killed bacteria at an MOI:100 did not show any positive staining at 4 hours (data not shown). The epithelial cells appeared morphologically normal under all of the above conditions. However, the cells challenged with live P. gingivalis at an MOI:100 for 24 hours showed signs of blebbing and pyknotic nuclei and stained positive for TUNEL (Fig. 4C), confirming our previous observations using M30 epitope detection (Fig. 1C). In contrast cells challenged with heat-killed P. gingivalis at an FIENA was used to detect caspase-3 activation, a key mole-cule in initiation of apoptosis Figure 2 FIENA was used to detect caspase-3 activation, a key molecule in initiation of apoptosis. HGECs were challenged with live or heat-killed P. gingivalis 33277 at MOI:10 and MOI:100 for 4 and 24 hours. Negative control was unchallenged HGECs. Positive control was HGECs challenged with camptothecin 4 μg/ml. Values represent the means ± SD of at least two experiments. Statistical comparisons are to the unchallenged negative control cells (* P < 0.05, ** P < 0.01).
Cell Death Detection ELISA was used to detect DNA frag-mentation, a hallmark of apoptosis Figure 3 Cell Death Detection ELISA was used to detect DNA fragmentation, a hallmark of apoptosis. HGECs were challenged with live and heat-killed P. gingivalis 33277 at MOI:10 and MOI:100 for 4, 24, and 48 hours. Negative control was unchallenged HGECs in media. Positive control was HGECs challenged with camptothecin 4 μg/ml. Values represent the means ± SD of at least two experiments. Statistical comparisons are to the unchallenged negative control cells (* P < 0.05, ** P < 0.01).
MOI:100 for 24 hours did not show any signs of DNA fragmentation (Fig. 4D).

P. gingivalis-induced apoptosis in HGECs is dependent on either Arg-or Lys-gingipains
P. gingivalis-induced apoptosis has been shown previously to depend on gingipain activity in fibroblasts and endothelial cells [7,8,10,11]. Gingipains are cysteine proteases produced by P.gingivalis that cleave after an arginine (Arg) or a lysine (Lys) residue. To elucidate the role of gingipains in our P. gingivalis-induced apoptosis model, HGECs were challenged with whole live bacteria (Fig. 4) as well as filtered bacterial supernatant (Fig. 5) of the following strains: wild-type P. gingivalis 33277; wild-type W50; the Arg-gingipain (RgpA/ RgpB) double mutant E8; the Lys-gingipain (Kgp) mutant K1A; or the Arg-Lys-gingipain (RgpA/RgpB/Kgp) triple mutant KDP128. All strains were utilized live at an MOI:100 and the filtered supernatants at a 10× dilution. DNA fragmentation was assessed by TUNEL after 24 hours. HGECs were also challenged with live wild-type P. gingivalis 33277 or its filtered supernatant previously incubated with leupeptin, a specific Rgp inhibitor, zFKck, a specific Kgp inhibitor, TUNEL assay to detect DNA fragmentation by confocal microscopy  or a cocktail of both gingipain inhibitors. Untreated cells were used as a negative control and cells treated with DNase 1000 U/ml were used as a positive control. Increased detachment of cells was observed upon challenge with 33277 and W50 detached from the plate, while the remaining cells showed signs of blebbing and pyknotic nuclei and stained positive for TUNEL (Fig. 4C, H), confirming our previous observations on live wild-type P. gingivalis. When any of the gingipain deficient mutants was used for the live challenge, DNA fragmentation was not evident (Fig. 4I, J, K, Fig. 5G, H, I), suggesting that the presence of either Arg-and Lys-gingipains is necessary for apoptosis and that depletion of any one of them completely abolishes P. gingivalis-induced apoptosis in HGECs (Fig. 6). Furthermore, cell detachment was still observed to a lesser extent with both E8 and K1A, suggesting that apoptosis is independent of cell detachment (Fig. 4I, J, K). The difference between the strains is unlikely to be due to differences in bacterial viability, since the viability over time in culture was similar for all strains examined (Fig.  7). The role of gingipains in HGEC apoptosis was also confirmed by using specific gingipain inhibitors (Fig. 4E, F, G). Furthermore, apoptosis was still observed when HGECs were challenged with filtered supernatant of P. gingivalis 33277 culture (Fig. 5C), but not when the challenge was per-TUNEL assay to detect DNA fragmentation by confocal microscopy Figure 5 TUNEL assay to detect DNA fragmentation by confocal microscopy. Images are fluorescent confocal staining at ×600 magnification. Negative control was unchallenged HGECs at 24 h (A formed with supernatant pre-incubated with gingipain inhibitors (Fig. 5D, E, F) or supernatant derived from the gingipain-deficient mutants (Fig. 5G, H, I). These results suggest that apoptosis is not dependent on bacterial invasion and although invasion might influence the apoptotic process our data reaffirm that gingipains are sufficient to invoke this process.

Purified gingipains can induce detachment and apoptosis in HGECs
Our previous experiments with live bacteria and bacterial culture supernatant suggest that either Arg-or Lys-gingipains are necessary for apoptosis in HGECs. In order to determine if specific purified gingipains are also sufficient to induce apoptosis, HGECs were challenged with purified HRgpA, RgpB and Kgp for 2, 4, 8, 15 and 24 hours and DNA fragmentation was assessed by TUNEL (Fig. 6). All three gingipains were able to induce cell detachment and apoptosis, although at different time points. For HRgpA, signs of apoptosis were already evident at 2 hours post-challenge, while for RgpB and Kgp, TUNEL positive cells appeared at 4 and 8 hours respectively. For all three gingipains, the percentage of apoptotic and detached HGECs increased progressively over time. By 24 hours, HGECs challenged with HRgpA and Kgp had completely detached from the plates, while some clumped cells still remained on the plates challenged with RgpB (Fig. 6).

Different WT P. gingivalis strains induce apoptosis with similar kinetics
HGECs were challenged with live P. gingivalis 33277 or W50 at an MOI:100 for 4, 8, 12 and 24 hours and phos-TUNEL assay to detect DNA fragmentation by confocal microscopy Figure 6 TUNEL assay to detect DNA fragmentation by confocal microscopy. Images are fluorescent confocal staining at ×600 magnification.

phatidylserine (PS) externalization was measured by
Annexin-V staining. Untreated cells were used as a negative control. A slow gradual increase in both Annexin-V single and Annexin-V/7-AAD double positive cells was noted for HGECs challenged with both strains compared to the unchallenged control over 12 hours (Fig. 8). The percentage of apoptotic cells was 4-5 fold higher than the unchallenged control 24 hours after challenge with either WT strain. The results of this kinetic study confirm our previous observations that apoptosis occurs late upon P. gingivalis challenge. Furthermore, the similarity in the kinetics of the response between the two strains suggests that the observed apoptosis is a characteristic of P. gingivalis and not an attribute of a single strain.

P. gingivalis challenge of HGECs results in upregulation of genes related to apoptosis
HGECs were challenged with live or heat-killed P. gingivalis 33277 at an MOI:100 for 4 and 24 hours and qPCR was performed on a focused panel of 86 apoptosis-related genes (Fig. 9). Live bacteria strongly upregulated apoptosis related pathways, as indicated by caspase cascade activation and apoptotic signaling activation in response to DNA damage. More specifically, the pro-apoptotic molecules caspase-3, -8, -9, Bid and Bax were upregulated at 4 and strongly upregulated at 24 hours, while the antiapoptotic Bcl-2 was also upregulated at 24 hours. Both the intrinsic and extrinsic pathways appear to be involved, as indicated by the activation of mitochondrial apoptosis signaling, as well as the Fas signaling pathway, TNFR and IL-1R signaling pathways (TNF, TRADD, FADD, IL-1b, IL-1R1, IRAK-2). The effect of heat-killed bacteria was less pronounced, indicating that higher doses or longer challenge times would be necessary to induce apoptosis.

Discussion
We demonstrate that primary HGECs challenged with live P. gingivalis for 24 hours exhibit apoptosis, evidenced by M30 epitope detection, caspase-3 activity, DNA fragmentation and Annexin-V staining. Apoptosis was dose and time dependent and live bacteria strongly upregulated Bacterial viability was determined following epithelial cell challenges Figure 7 Bacterial viability was determined following epithelial cell challenges. From each challenge assay reported in Fig. 4 Flow cytometry for Annexin-V staining to detect PS externalization, an early apoptotic event In the present study, live P. gingivalis induced considerable apoptosis in human gingival epithelial cells between 12 and 24 hours at MOI:100, as evidenced by M30 epitope detection (Fig. 1), increased caspase-3 activity (Fig. 2), DNA fragmentation (Fig. 3, Fig. 4) and Annexin-V staining (Fig. 8). These results agree with previous reports on fibroblasts [7,18], endothelial cells [9] and lymphocytes [12]. In contrast, heat-killed Porphyromonas gingivalis did not induce apoptosis.
Apoptosis is a complex process regulated by multiple pathways such that no single molecule gives sufficient information on the dynamics of apoptosis. After an apoptotic stimulus, a subset of pro-apoptotic molecules is upregulated and others such as Bcl-2, an anti-apoptotic molecule, downregulated, with cellular fate depending on the fine tuning of all pathways involved. We used a focused array of 86 apoptosis-related genes to elucidate the apoptotic process (Fig. 9). Live P. gingivalis strongly Focused qPCR-Array consisting of 86 genes relevant to inflammation and apoptosis Figure 9 Focused qPCR-Array consisting of 86 genes relevant to inflammation and apoptosis. HGECs were challenged with live or heat-killed P. gingivalis 33277 at MOI:100 for 4 and 24 hours. Negative control was unchallenged HGECs in media. The mRNA fold change between each sample and the negative control was calculated based on the ΔΔCt method and Log10 foldincrease was used to generate the heatmap using MeV v4.1 release software and hierarchical clustering with Pearson correlation. (A) represents a heatmap of the 86 genes and (B) represents specific apoptotic markers with color coding: Magenta (upregulated genes) to Green (down-regulated genes). The apoptotic markers in (B) and the fold differences are shown in Table 1.
upregulated apoptosis pathways: evidenced by caspase cascade activation and apoptotic signaling in response to DNA damage. Both the intrinsic and extrinsic pathways appear to be involved in this process: evidenced by activation of mitochondrial apoptosis signaling, as well as Fas signaling, TNFR signaling and IL-1R signaling pathway (Table 1). In terms of individual molecules, the pro-apoptotic caspase-3, -8, -9, Bid, Bax, TNF, TRADD, FADD, IL-1b, IL-1R1, IRAK-2 were upregulated after 24 hours. On the other hand, the anti-apoptotic Bcl-2 was also upregulated, but this did not appear to be sufficient to ensure cell survival, as indicated by the apoptosis assays (Fig. 1, Fig.  2, Fig. 3, Fig. 4, Fig. 8). The upregulation of Bcl-2 is in agreement with Nakhjiri et al [16], underlining the fact that single molecule and single time point assessments alone can be misleading.
It has been suggested that apoptosis due to P. gingivalis challenge of human cells involves the gingipains [7,8,10,11,14]. Gingipains are cysteine proteases produced by P. gingivalis that are either secreted or membrane bound and arginine or lysine specific. In the present study, the mechanism used by P. gingivalis to induce apoptosis in gingival epithelial cells was shown to be dependent upon both Arg-and Lys-gingipains (Fig. 4). Gingipain deficient P. gingivalis mutants did not cause apoptosis as evidenced by a lack of DNA fragmentation indicating that gingipains are necessary for apoptosis to occur and that their depletion abolishes P. gingivalis' ability to induce apoptosis in HGECs. This suggests a step-wise enzymatic action of these gingipains on substrates such that action of one alone is not sufficient. Similarly, inhibition of apoptosis was also observed when the wild-type P. gingivalis was pre-treated with specific gingipain inhibitors, providing evidence that the observed lack of apoptosis is due to the lack of gingipains and not other potential differences between the wild-type strains and the mutants.
Furthermore, filtered cell-free supernatant derived from wild-type P. gingivalis culture, as well as purified gingipains, retained the ability to induce apoptosis in HGECs (Fig. 5, Fig. 6), providing evidence that the gingipains are sufficient for the induction of apoptosis and that the presence of whole cells is not necessary for this process. This suggests that apoptosis is not dependent on bacterial invasion and although invasion might influence the apoptotic process our data reaffirm that gingipains are sufficient to invoke this process. The ability of the bacterial culture supernatant to induce apoptosis was lost when it was preincubated with specific gingipain inhibitors, while bacterial culture supernatant derived from gingipain-deficient mutants did not result in apoptosis (Fig. 5). These results are in agreement with previous studies in endothelial cells [10,11]. The mechanism of action of gingipains has been shown to be both caspase-dependent and caspase-independent [11] and in vitro evidence suggests that gingipains may activate caspase-3 by cleaving procaspase-3 [7].
In addition to variable bacterial strain virulence and variable host resistance, local factors, such as MOI or length of exposure, could vary across different areas of the lesion and inter-laboratory differences in apoptosis studies may reflect these variables. Thus, results from different laboratories and studies may supplement rather than conflict each other in elucidating the actions of P. gingivalis on host epithelial cells. In areas where the bacteria to epithe- HGECs were challenged with live or heat-killed P. gingivalis 33277 at MOI:100 for 4 and 24 hours. Negative control was unchallenged HGECs in media. The data shown represent log-fold differences in gene expression (means ± SD) between the respective sample and the negative control. A value of 1 indicates no change, less than one indicates down-regulation and greater than one, up-regulation (*P < 0.05 ** P < 0.01, *** P < 0.001) lial cells ratio is low or the exposure time is short, bacterial invasion [19,20] may result in cell survival [15][16][17], contributing to the chronicity of the periodontal lesion. On the other hand, in areas with high bacteria to epithelial cell ratio or longer exposure time, the bacterial insult may result in apoptosis [7,9], contributing to extensive tissue destruction. Further translational studies are needed to determine which scenarios predominate in the pathogenesis of periodontitis.

Conclusion
The present study provides evidence that live, but not heat-killed, P. gingivalis can induce apoptosis after 24 hours of challenge in primary human gingival epithelial cells. Either arginine or lysine gingipains are necessary and sufficient factors in P. gingivalis elicited apoptosis.

Cell isolation and culture
Gingival tissue biopsies were obtained with informed consent from periodontally healthy patients undergoing crown lengthening procedures at the University of Louisville School of Dentistry Graduate Periodontics Clinic, according to an IRB approval. The gingiva was treated with 0.025% trypsin and 0.01% EDTA overnight at 4°C and human gingival epithelial cells (HGECs) were isolated as previously described [21]. The HGECs were seeded in 60mm plastic tissue culture plates coated with type-I collagen (BD Biocoat, Franklin Lakes, NJ, USA) and incubated in 5% CO 2 at 37°C using K-SFM medium (Invitrogen, Carlsbad, CA, USA) containing 10 μg/ml of insulin, 5 μg/ ml of transferrin, 10 μM of 2-mercaptoethanol, 10 μM of 2-aminoethanol, 10 mM of sodium selenite, 50 μg/ml of bovine pituitary extract, 100 units/ml of penicillin/streptomycin and 50 ng/ml of fungizone (complete medium). When the cells reached sub-confluence, they were harvested and sub-cultured as previously described [22].

Purified gingipains and gingipain inhibitors
Purified HRgpA, RgpB and Kgp were isolated as previously described [25][26][27]. The purified gingipains were used at a final concentration of 8 μg/ml for HRgpA, 5.2 μg/ml for RgpB and 3 μg/ml for Kgp (all equivalent to 113 units of Rgp activity/ml or 12.4 units of Kgp activity/ml) in the presence of 5 mM L-cysteine [10]. For the gingipain inhibition assays, live P. gingivalis 33277 or its culture supernatant was incubated with gingipain inhibitors for 15 min at 37°C, just prior to the HGEC challenge. zFKck, a specific Kgp inhibitor [28], was used at a final concentration of 10 μM. Leupeptin (Sigma), a specific Rgp inhibitor, was used at a final concentration of 100 μM. The final concentrations used were the maximum inhibitory doses that retained specificity, as determined by a dose-response using the enzymatic substrate hydrolysis of N-α-benzoyl-DL-arginine-p-nitroanilide (BAPNA) (Sigma), for Rgp activity, or acetyl-lysine-p-nitroanilide (ALNA) (Bachem), for Kgp activity.

Bacterial challenge
HGEC cultures at the fourth passage were harvested and seeded at a density of 0.5 × 10 5 cells/well in a 6-well culture plate coated with type-I collagen or in a 35-mm collagen-coated glass bottom culture dishes (Mat-tek Corp., Ashland, MA, USA), and maintained in 2 ml of complete medium. When they reached confluence (approximately 10 6 cells/well), the cells were washed twice with fresh media and were challenged with live or heat-inactivated bacteria in antibiotic-free medium at MOI:10 (10 7 bacteria/well) and MOI:100 (10 8 bacteria/well) at 37°C in 5% CO 2 for 4 or 24 hours. For each experiment the final concentration of the suspension was determined by measurement of A 600 and appropriate dilutions were made to achieve the desired MOI. The bacterial number was confirmed by viable counting of colony forming units (cfu) on blood agar plates incubated at anaerobically at 37°C.

M30 epitope detection
The M30 epitope released by caspase-cleaved cytokeratin-18 was detected using a commercially available kit (Cyto-DEATH Fluorescein kit, Roche Applied Science, Indianapolis, IN, USA), according to the manufacturer's instructions. Briefly, the cells were washed three times with PBS, fixed with ice-cold pure methanol for 30 minutes at -20°C and then incubated with the M30 antibody