Necrosis of Lung Epithelial Cells during Infection with Mycobacterium tuberculosis Is Preceded by Cell Permeation

Bacterial cultures and subcellular fractions.

M. tuberculosis strain Erdman, M. tuberculosis strain CDC1551 (46), and Mycobacterium smegmatis strain LR222 were obtained from the Centers for Disease Control and Prevention culture collection (Atlanta, Ga.). M. bovis BCG strain Pasteur was a kind gift from Barry Bloom (Harvard School of Public Health, Boston, Mass.). Stock 1-ml cultures of the mycobacterial strains were grown by serial 10-fold passage to 100 ml in Middlebrook 7H9 oleic acid, albumin, dextrose, and catalase (OADC) broth until the optical density (600-nm wavelength) was 0.5 to 0.7, corresponding to 1 × 10⁷ to 10 × 10⁷ CFU/ml. This 100-ml stock solution was then dispersed to make a single-cell suspension by applying a 1-min pulse at a setting of 10 to cultures in closed polypropylene tubes at 4°C using a Branson 450 cell sonicator fitted with a cup horn attachment (Branson Ultrasonics, Danbury, Conn.). Aliquots of 0.5 and 1.0 ml were made from this suspension, and these aliquots were immediately quick frozen in an ethanol–dry-ice bath and stored at −70°C until needed.

Before the beginning of each experiment, the aliquots of mycobacteria were thawed, and any bacterial clumps caused by freezing were disrupted by pulse sonication (25, 35). Bacteria were then diluted by serial dilution into tissue culture medium prior to addition to the cell monolayers. All dilutions were verified for uniform dispersal, and accurate multiplicities of infection (MOIs) were confirmed directly for each experiment by performing viable counts of CFU on Middlebrook 7H10 OADC agar.

For experiments using heat-killed or γ-irradiated M. tuberculosis, the aliquots of M. tuberculosis CDC1551 prepared as described above were inactivated either by incubation at 80°C for 60 min or by exposure to 1.2 megarads of γ-irradiation over 12 h at −70°C. CFU counts for these aliquots were determined prior to inactivation, and inactivation of cultures following these treatments was confirmed by plating onto 7H10 plates and observing these plates over a 5-week period. Subcellular fractions from γ-irradiated M. tuberculosis CDC1551 were provided as aseptic lyophilisates by John Belisle (Colorado State University, NIAID contract NOI-AI-75320).

Cell cultures.

Human lung epithelial cells (A549) were obtained from the American Type Culture Collection (Manassas, Va.) at passage number 76. Cells were maintained in modified Eagle's medium (MEM) with Earle's salts and l-glutamine (Gibco/BRL, Grand Island, N.Y.), supplemented with 5% heat-inactivated fetal bovine serum (FBS) (HyClone Laboratories Inc., Logan, Utah), and incubated at 37°C with 5% CO₂. Prior to use in experimental assays, the cells were released from the culture flask with 0.25% trypsin (Gibco), washed twice with fresh medium, and seeded onto 6- or 24-well microculture plates as needed. Viable cell counts were confirmed from seeded test wells prior to each experiment using trypan exclusion as previously described (33). All experiments were performed between cell passage numbers 78 and 95.

Necrosis and apoptosis determination in lung epithelial cells.

Analysis of cytotoxicity in lung epithelial cells was performed using an enzyme-linked immunosorbent assay (ELISA) that differentiates between necrosis and apoptosis by measuring the release or sequestering of histone-DNA complexes in affected eukaryotic cells (Roche, Indianapolis, Ind.). Both phenotypes have been measured in A549 and other epithelial cells using this ELISA (8, 28, 34).

Initially, the pore-forming detergent digitonin (1%) and the apoptosis inducer camptothecin (0.005 mM) were used in our model to determine the time of incubation required to detect necrosis and apoptosis with this ELISA. Digitonin or camptothecin was added to uninfected monolayers that were 75% confluent in 24-well plates (∼1 × 10⁵ cells/ml) and incubated at 37°C with 5% CO₂. Tissue culture plates were then centrifuged for 5 min at 200 × g, and culture supernatants from the monolayers were collected and filtered through a 0.22-μm-pore-size syringe filter (Millex/GV filters; Millipore Corp., Bedford, Mass.) for the measurement of necrosis. The remaining cells were then treated with a lysis buffer (supplied by Boehringer Mannheim) to lyse the cells and release intracellular histone-DNA complexes. The cell lysates were then collected and filtered through a 0.22-μm-pore-size syringe filter (Millipore). Culture supernatants and cell lysates from each individual well were immediately tested for histone-DNA complexes using the cell death ELISA according to the manufacturer's directions. The levels of necrosis and apoptosis from treated and untreated cells were measured spectrophotometrically at 405 nm (with the reference wavelength of 490 nm subtracted per the manufacturer's directions) and were compared at 4, 24, and 72 h. Untreated monolayers were incubated and measured by ELISA at identical time points to determine the background levels of necrosis and apoptosis in this model. All experiments were tested in triplicate and performed in duplicate.

For the infections, lung epithelial cells were seeded into 24-well tissue culture plates as described above and monolayers were infected with each mycobacterial strain at an equal MOI of 10 bacilli per cell, except for M. smegmatis strain LR222. This strain was used at a viable count equal to an MOI of 50 bacilli per cell to control for the effects of mycobacterial extracellular growth on necrosis or apoptosis in this tissue culture model. All infected cell cultures were incubated at 37°C with 5% CO₂. Tissue culture plates were centrifuged, and culture supernatants from infected and uninfected monolayers were collected at 1, 3, and 5 days postinfection as described above. The remaining monolayers were then lysed, and cellular lysates collected as described above. Both the culture supernatants and cell lysates from each individual well were immediately tested as described above, and all data were normalized by subtraction of background levels at each time point. All infections were tested in triplicate, and experiments were repeated three times.

For measurement of necrosis and apoptosis caused by heat-killed or γ-irradiated M. tuberculosis, an inoculum of these inactivated aliquots equal to an MOI of 10 (from CFU per milliliter determined prior to inactivation) was added to A549 cell monolayers. Tissue culture plates were incubated at 37°C with 5% CO₂, and mock-infected monolayers were treated for measurement of necrosis and apoptosis as described above.

For measurement of necrosis and apoptosis caused by incubation of subcellular fractions of M. tuberculosis, the protein yields of culture filtrate (CF) and of subcellular fractions of cell wall (CW) and whole-cell lysate (WCL) from 10⁷ CFU of M. tuberculosis/ml were first determined. Briefly, a 10-ml culture of M. tuberculosis CDC1551 was grown in glycerol alanine salts medium (44) for 10 days. A 1-ml aliquot from this preparation was removed and then added to 9 ml of 7H9–OADC, and this preparation was sonicated and plated with serial dilutions onto 7H10 plates as described above. The CF was harvested from the remaining 9 ml of GAS culture, and crude CW and WCL preparations were made from the bacterial pellet of this culture by standard methods (16). These analytical preparations of CW, WCL, and CF were quantitated for their protein content by bicinchoninic acid (BCA) analysis (Pierce Chemical Co., Rockford, Ill.), and the results were compared with the CFU determination. A total of 8 μg of CF protein, 20 μg of WCL protein, and 2.5 μg of CW protein were found to be present in 1 ml of a 10⁷-CFU culture. Tissue culture plates were seeded with A549 cells as above, and viable A549 cells were determined prior to each experiment as above. CF, CW, and WCL fractions were then added at microgram-of-protein-per-CFU equivalents of MOIs of 100 and 500. These 10- and 50-fold excesses were used to address bacterial replication or loss of activity in these γ-irradiated subcellular fractions. Plates were incubated, and cell lysates and supernatants were collected and measured by ELISA as above. All experiments were performed in triplicate.

Effects of inhibitors on the necrosis of A549 cells.

The requirement of bacterial and eukaryotic protein synthesis for the necrosis of A549 cells was determined by incubation with streptomycin sulfate (10) or cycloheximide (43) during the infection of A549 cells with M. tuberculosis strain CDC1551. The MIC of streptomycin sulfate (Sigma, St. Louis, Mo.) for M. tuberculosis strain CDC1551 was determined to be 2 μg/ml (data not shown), and we used a concentration of 10 μg/ml that completely inhibited bacterial growth but was not bactericidal (data not shown). A549 cells were counted and seeded onto 24-well plates as described above. Cells were infected at an MOI of 10 bacilli per cell with M. tuberculosis strain CDC1551 that had been preincubated for 24 h at 37°C either with 10 μg of streptomycin sulfate/ml in MEM–5% FBS or with MEM–5% FBS alone. These preparations of bacilli in either MEM–5% FBS–streptomycin or MEM–5% FBS were directly overlaid onto cell monolayers and were allowed to remain on the cell monolayers throughout the course of infection. Uninfected cells with or without streptomycin sulfate were also included as controls for subtraction of background necrosis, and all data were normalized to these controls. The infected cell cultures were incubated for 5 days, and necrosis was measured as described above at days 1, 3, and 5 postinfection. All infections were conducted in triplicate, and experiments were performed in duplicate.

For experiments to test the effects of cycloheximide on necrosis, we chose a concentration of cycloheximide (50 μg/ml) that inhibited protein synthesis and the exogenous release of chemokines in A549 cells (5, 23, 30, 45), but did not interfere with M. tuberculosis protein synthesis or intracellular growth (27). A549 cells were counted and seeded onto 24-well plates as described above and were treated with cycloheximide for 3 h prior to infection. Cell monolayers with or without cycloheximide were then infected at an MOI of 10 bacilli per cell with M. tuberculosis strain CDC1551. Uninfected cells with or without cycloheximide were included as controls for background necrosis, and all data were normalized to these controls. The infected cell cultures were incubated for 5 days, and necrosis was measured as described above at days 1, 3, and 5 postinfection. All infections were conducted in triplicate wells, and experiments were performed in duplicate.

Measurement of intracellular and extracellular growth of mycobacterial strains.

Six-well plates of A549 cell monolayers (∼1 × 10⁵ cells/ml) were infected with M. tuberculosis strain Erdman or CDC1551 or with M. bovis BCG at an MOI of 10 and then incubated at 37°C with 5% CO₂. Entry of M. bovis BCG into A549 cells was verified by electron microscopy of infections at 24 h postinoculation using previously described methods (33). CFU assays were performed at 6, 24, 48, 72, 96, and 120 h postinfection by standard methods (25, 35). Briefly, for determination of extracellular growth, extracellular medium was removed at the above time points from tissue culture wells. Wells were washed twice with phosphate-buffered saline (PBS) to remove any additional extracellular or adherent bacilli. Media and washes were combined, and this suspension was sonicated to achieve single-cell suspensions as described above and was then serially diluted and plated onto 7H10 plates. For determination of intracellular growth, 1 ml of lysis buffer (PBS with 0.25% Tween 80 and 0.016% digitonin) was added to each washed monolayer and tissue culture plates were incubated for 10 min at 37°C according to the standard protocol (25, 35). The lysate was collected, sonicated, serially diluted, and plated as described above. Colonies were counted from extracellular and intracellular growth plates following 14 to 21 days of incubation. All infections were conducted in duplicate and repeated twice.

Measurement of cell permeability and cell membrane distortion in infected lung epithelial cells.

Propidium iodide (PI), a cell membrane-impermeant nuclear stain that stains the nuclei of necrotic porous cells only (53), and annexin V (AV), a dye that binds to negatively charged phospholipids on the outer surface of the plasma membrane of all apoptotic cells and necrotic cells undergoing membrane translocation (48), were used to stain infected A549 cells to determine when M. tuberculosis-infected cells became permeable to PI or susceptible to AV staining.

Lung epithelial cells were seeded into chamber slides (∼1 × 10⁵ cells/ml), and monolayers were infected in duplicate at an MOI of 10 bacilli per cell with mycobacteria as above. At 24 h postinfection, the extracellular bacilli were removed and cell monolayers were washed twice with PBS to remove residual adherent bacilli. Monolayers were overlaid with HEPES buffer supplied by Roche. Individual wells of infected and control cultures were stained with PI and AV (Roche) according to the manufacturer's methods. Slides were washed and fixed with 10% buffered formulin, and cells were viewed at a magnification of ×400 and were counted and photographed using a Zeiss fluorescent microscope (488-nm excitation and 515-nm long pass filter). Uninfected monolayers were also treated with 1% digitonin or 0.005 mM camptothecin and analyzed at 4 h as positive controls for detection of necrotic and apoptotic cells, respectively. Ten fields were counted and recorded for each infection and control assay. All data reported represent the mean and standard deviations of 10 different fields for each infection and control.

Measurement of LDH release from infected A549 cells.

A549 cells were seeded into 24-well plates in a manner similar to that described above, except that the cells were laid down using MEM with 1% FBS, and monolayers were infected with mycobacteria at an MOI of 10 as described above. The release of the cytoplasmic enzyme lactate dehydrogenase (LDH) from infected, permeabilized A549 cells was measured at 6, 24, 48, and 72 h postinfection by using the colorimetric kit from Roche and following the manufacturer's instructions. Background release of LDH was obtained by measurement of untreated cells, and maximum release of LDH was obtained by lysis of uninfected cells per the manufacturer's protocol. The percent LDH release was then determined using the following calculation: [(release of LDH from infected cells background release)/(maximum release of LDH − background release)] × 100. Similarly, heat-killed and γ-irradiated preparations of M. tuberculosis were added to A549 cell monolayers at inoculations equivalent to an MOI of 10 (determined prior to inactivation); subcellular fractions of CF, CW, and WCL were added to monolayers at protein concentrations equivalent to MOIs of 100 and 500; and LDH release was measured from A549 cells following incubation with these preparations as described above. No significant LDH was detected in the MEM–1% FBS tissue culture medium alone, in this medium with each of the above mycobacterial strains, or in the inactivated and subcellular mycobacterial preparations in this medium. All infections, mock infections, and protein overlays were conducted in triplicate and performed in duplicate.

Statistical analysis.

All tests for significance were performed using the Student two-tailed t test, with the statistical program provided with Excel software 97 (Microsoft Corp., Redmond, Wash.). Unless otherwise stated, data were considered significant at a P value of <0.05.

Necrosis of infected lung epithelial cells correlates with cytotoxicity.

In previous studies, M. tuberculosis induced clearing of confluent monolayers upon infection, but M. bovis BCG did not (32). Similar findings were observed in our laboratory when monolayers were infected with M. tuberculosis strain Erdman or CDC1551 versus M. bovis BCG (data not shown). To determine the role of necrosis and apoptosis in the cytotoxicity of lung epithelial cells by infection with M. tuberculosis, a quantitative ELISA was used to measure the release of histone-DNA complexes from the supernatants (necrosis) and lysates (apoptosis) of infected and uninfected lung epithelial cells.

A549 epithelial cells were first tested for susceptibility to necrosis and apoptosis by incubation with either 1% digitonin or 0.005 mM camptothecin. Treatment of A549 cells with digitonin resulted in cellular necrosis within 3 days postoverlay (Fig. 1A), whereas camptothecin treatment resulted in cellular apoptosis within 1 day postoverlay (Fig. 2A), demonstrating the susceptibilities of this cell line and the ability to distinguish between these two cell death phenotypes using this assay.

The mode of cytotoxicity after infection of A549 cells by four different strains of mycobacteria was then determined. M. tuberculosis strains Erdman and CDC1551 induced significantly greater necrosis (Fig. 1B) than apoptosis in the infected monolayers (Fig. 2B) by 3 (P = 0.035 and P = 0.021) and 5 (P = 0.021 and P = 0.001) days postinfection, indicating that necrosis was the primary cause of cytotoxicity to the A549 epithelial cell monolayers, and that the time course for induction of necrosis was similar to that shown with digitonin (Fig. 1A). M. tuberculosis strains Erdman and CDC1551 were significantly more necrotic to the monolayers than M. bovis BCG after 5 days postinfection (Fig. 1B) (P = 0.026), correlating to the visual observations of cytotoxicity by these strains (data not shown). Interestingly, M. smegmatis strain LR2222 caused very little cytotoxicity (Fig. 1B), despite the fact that a higher MOI was used (50:1) and a large number of extracellular bacilli had grown in the medium over the 5 days of incubation (data not shown). Although M. tuberculosis strain CDC1551 induced greater necrosis than strain Erdman by 3 days postinfection (Fig. 1B), levels of necrosis by CDC1551 and Erdman were not significantly different (P = 0.776) by day 5 postinfection.

Necrosis of M. tuberculosis-infected lung epithelial cells is reduced by incubation with streptomycin, but not by incubation with cycloheximide.

A549 cells infected with M. tuberculosis strain CDC1551 treated with streptomycin demonstrated a significant reduction in necrosis by day 3 postinfection compared to untreated M. tuberculosis CDC1551-infected monolayers (Fig. 3A) (P = 0.027), and this reduction remained significant through day 5 postinfection (P = 0.003), suggesting that bacterial replication or protein synthesis was important for induction of necrosis.

However, when lung epithelial cells were infected with M. tuberculosis strain CDC1551 after pretreatment of these cells with, and in the presence of, cycloheximide, no reduction in necrosis was observed over the course of the infection (P = 0.372) (Fig. 3B). Eukaryotic protein synthesis, therefore, did not appear to be required for necrosis of A549 cells upon infection with M. tuberculosis.

Growth rates of mycobacteria in A549 cells.

To determine if differences in mycobacterial growth were related to the cytotoxic phenotype, we measured the intracellular and extracellular growth of both strains of M. tuberculosis and that of M. bovis BCG during infection of A549 cells. Verification that M. bovis BCG entered A549 cells was obtained by electron microscopy (data not shown), as has been shown previously for M. tuberculosis (6, 32, 33). Although the numbers of M. tuberculosis CFU obtained from the A549 lysates after 6 h of incubation were greater than those obtained for M. bovis BCG (Fig. 4A), the numbers of intracellular and extracellular CFU obtained for all mycobacterial strains measured at day 1 were similar (Fig. 4). At 3 days postinfection, only M. tuberculosis strain CDC1551 demonstrated more growth than the other two mycobacteria, and this increase was seen only for extracellular growth of this mycobacterium (Fig. 4B). However, this increase was only transient, as all mycobacteria tested demonstrated similar CFU numbers for extracellular growth by day 5 postinfection (Fig. 4B). In addition, while M. tuberculosis CDC1551 continued to grow intracellularly in the A549 cells at day 5 postinfection, the growth of M. tuberculosis Erdman began to wane, and CFU numbers of M. tuberculosis Erdman and M. bovis BCG were similar (Fig. 4A).

High or differential growth rates were infrequently observed and did not explain the necrosis induced by infection of A549 cells with mycobacteria. Little extracellular growth occurred for any mycobacteria throughout the course of the experiments (Fig. 4B; also observations by others [7, 29, 35]). Only M. tuberculosis strain CDC1551 demonstrated a greater number of CFU from extracellular growth, and only on day 3 postinfection (Fig. 4B). M. tuberculosis CDC1551 also demonstrated greater CFU numbers for intracellular growth than both M. tuberculosis Erdman and M. bovis BCG at day 5 postinfection (Fig. 4A), though necrosis by both M. tuberculosis strains at day 5 was similar (Fig. 1B). Lastly, M. bovis BCG did not demonstrate any appreciable level of necrosis of A549 cells throughout the course of the experiments (Fig. 1B), despite the fact that similar numbers of intracellular and extracellular CFU were obtained for M. tuberculosis strain Erdman and M. bovis BCG (Fig. 4).

Necrosis of M. tuberculosis-infected lung epithelial cells is preceded by cell permeation.

We next examined cell membrane permeability and distortion to determine if loss of integrity of A549 cellular membranes correlated with necrosis. Initially, A549 cell monolayers were infected with mycobacteria and stained with both PI and AV after 24 h postinfection to identify changes in permeability and/or ionic charge of A549 cell membranes, respectively.

Enumeration of PI-stained cells demonstrated that M. tuberculosis strain Erdman- and M. tuberculosis CDC1551-infected monolayers contained significantly greater numbers of PI-stained cells (7.2 ± 2.5 and 7.8 ± 2.7, respectively) than either BCG-infected (4.2 ± 2.7; P = 0.003) or uninfected (4.7 ± 2.8; P = 0.01) cell monolayers. No difference was observed between monolayers infected with the two strains of M. tuberculosis (P = 0.614) or between BCG-infected and uninfected monolayers (P = 0.689). No significant staining of the cells with AV was evident in any of the monolayers infected with either M. tuberculosis or M. bovis BCG, or in the uninfected monolayers (data not shown). These initial findings suggested that changes in cell permeability, not membrane distortion, correlated with necrosis in this model. In order to better quantitate the permeation of M. tuberculosis-infected A549 cells, we measured the release of the eukaryotic cytoplasmic enzyme LDH from infected A549 cells over time. LDH release was seen only upon infection with M. tuberculosis throughout the course of infection of A549 cells (P = 0.007 for M. tuberculosis strain Erdman versus M. bovis BCG at 48 h postinfection, and P < 0.0001 for M. tuberculosis strain CDC1551 versus M. bovis BCG at 48 h postinfection (Fig. 5). In fact, only a low level of LDH release from M. bovis BCG-infected cells was seen at 48 h, and it was never higher than the background release of uninfected cells at all the other time points (Fig. 5). These data correlated with our results for necrosis (Fig. 1B) and those reported for cell monolayer clearing (32).

Interestingly, significant levels of LDH were detected within 24 h postinfection from M. tuberculosis strain CDC1551-infected cells, compared with those from M. tuberculosis strain Erdman-infected cells (P < 0.0001), before any significant growth of the mycobacteria occurred (Fig. 4), suggesting that this rapid release of LDH may be due to inherent properties of this strain. No significant difference in LDH release was seen between M. tuberculosis strains by 48 h postinfection (P = 0.408) (Fig. 5). Consistent with our earlier observations with PI staining and necrosis of A549 cells, the release of LDH from infected cells was specific to infection with M. tuberculosis and could not be explained by differential growth of the mycobacteria.

Necrosis and cell permeation are not induced by incubation of A549 cells with nonviable M. tuberculosis and mycobacterial secreted and subcellular fractions.

We next measured the necrosis of A549 cells following incubation with either nonviable M. tuberculosis or M. tuberculosis subcellular fractions in an attempt to determine what bacterial factors were responsible for this phenotype.

Incubation of A549 cells with either heat-killed or γ-irradiated M. tuberculosis resulted in little to no significant necrosis over the 5-day incubation period (Fig. 6). Further, the addition of a higher inoculum of either preparation did not increase necrosis (data not shown). Necrosis was then measured in A549 cells after addition of high doses of subcellular fractions from γ-irradiated M. tuberculosis to determine whether this phenotype could be induced by these partially purified components. Again, little to no necrosis was observed over the 5-day incubation period with either the mock 100-MOI inoculum (Fig. 7A) or the 500-MOI inoculum (data not shown). M. tuberculosis and its lipoglycan constituents have been shown to induce apoptosis of affected cells (12, 20, 39, 42). Although we did not observe apoptosis of A549 cells infected with viable M. tuberculosis (Fig. 2B), we measured apoptosis induction by these subcellular fractions to determine if similar observations could be made in our laboratory. Induction of A549 cell apoptosis was observed following exposure to these fractions, with the high-lipoglycan-containing CW fraction demonstrating the highest level of apoptosis (Fig. 7B). These results indicate that these preparations contained properties similar to those already reported (39, 42).

We next measured LDH release after incubation of A549 cells with either inactivated or subcellular fractions of M. tuberculosis to determine if the early events of cellular necrosis could be attributed to these preparations. Incubation of A549 cells with subcellular mycobacterial fractions and inactivated mycobacteria did not yield any significant release of LDH from A549 cells over the 3-day period (data not shown). In fact, LDH release following exposure to most of these preparations was lower than the background LDH release (data not shown). These results suggest that the cell permeation and subsequent necrosis we observed in A549 cells required infection with viable M. tuberculosis and/or the expression and secretion of factors unique to M. tuberculosis upon exposure to host cells.