Mycobacterial infection induces the secretion of high-mobility group box 1 protein

Mycobacteria-induced HMGB1 release in vitro

To evaluate the release of HMGB1 in mycobacterial infection, BMDM cultures and cultures of macrophage-like J774A.1 cells were stimulated with M. tuberculosis, Mycobacterium bovis BCG, Mycobacterium avium and environmental mycobacteria such as Mycobacterium scrofulaceum. The levels of HMGB1 in the culture medium were subsequently measured by immunoblotting analysis. HMGB1 was detected in the culture medium of BMDMs (Fig. 1A) and J774A.1 cell lines (Fig. 1B) after stimulation with both M. tuberculosis and M. bovis BCG at multiplicity of infection (MOI) 10:1. HMGB1 was not detected in culture medium in the absence of stimuli. The release of HMGB1 was also observed with other less virulent species of mycobacteria, such as M. avium and M. scrofulaceum (Fig. 1C). To determine the effect of MOI of mycobacterium on HMGB1 release, J774A.1 cells were stimulated with M. bovis BCG. HMGB1 was not observed in culture medium at MOI 1:1, but was released maximally at MOI 100:1 (Fig. 1D). Thus, infection of mycobacterium effectively induced HMGB1 release both in macrophage and monocytic cell cultures.

We examined the induction of HMGB1 release in J774A.1 cells by different mycobacterial cell components. Whole-cell lysate, cell wall fraction, cell membrane fraction, cytoplasmic fraction and culture filtrate proteins (CFP) induced HMGB1 translocation and extracellular release, as shown by Western blotting (Fig. 1E). CFP induced the maximum amount of HMGB1 release among all cellular fractions, and other cellular components induced similar levels of HMGB1 release (Fig. 1F). To determine whether the mycobacterial cell and its components had a cytotoxic effect on cultured cells, the levels of cytotoxicity in J774A.1 cells was measured using lactate dehydrogenase (LDH) release assay. M. bovis BCG, M. avium and M. scrofulaceum induced a cytotoxicity of 10–15% at MOI of 10:1 whereas M. tuberculosis induced a cytotoxicity of more than 20% at 24 h of incubation (data not shown). All cellular fractions of mycobacterial cell induced a cytotoxicity of less than 5% (data not shown). These observations suggest that these components of the mycobacteria contain TLR ligands that are inducing active secretion of HMGB1 into the extracellular milieu.

In order to understand the role of HMGB1 in mycobacterial infection further, we cultured murine BMDMs with anti-HMGB1 antibody, infected them with M. tuberculosis H₃₇R_v and estimated the levels of pro-inflammatory cytokines such as TNF-α (Fig. 1G) and IL-1β (Fig. 1H). The results of ELISA demonstrated that HMGB1 antibody reduced the TNF-α and IL-1β production induced by M. tuberculosis during time-course of infection, indicating the significance of HMGB1 in inflammatory response during progressive TB. No change was observed in TNF-α and IL-1β biosynthesis when macrophages were incubated with isotype control antibody during M. tuberculosis infection.

To determine whether stimulation with mycobacterium affects HMGB1 cellular localization, J774A.1 infected with M. bovis BCG cultures were immunostained with monoclonal anti-HMGB1 Abs. Quantification of HMGB1 was done by measuring the area either FITC-positive (HMGB1) or -negative in the nucleus and cytoplasm of uninfected or BCG-infected J774 cells. Each compartment (nuclear and cytoplasmic) was outlined using bright field and DAPI fluorescence. Un-stimulated macrophages constitutively expressed HMGB1 and maintained an intracellular pool of HMGB1 in the cytoplasm and nuclear regions (Fig. 2A, panel a). The pattern and localization of HMGB1 staining were noticeably altered as early as 16 h after stimulation with M. bovis BCG (Fig. 2A, panel b). The HMGB1 appeared to be diffusely distributed in both the cytoplasm and nucleus regions of un-stimulated macrophage cultures, but was observed predominantly and uniformly in the cytoplasm of BCG-infected macrophages as numerous aggregated granules with more intense staining (Fig. 2A, panel b). Similar changes of HMGB1 localization have been observed by others for monocytes in response to LPS stimulation (Wang et al., 1999; Chen et al., 2004). Actively multiplying mycobacterial bacilli were observed in the cytoplasm of infected macrophages (Fig. 2A, panel c). A total of 10 cells were counted in each sample for quantitative analysis of HMGB1. Uninfected cells had an average of 45% of the cytoplasm HMGB1-positive, whereas BCG-infected cells reached 65% HMGB1-positive (Fig. 2B). This suggests that macrophages infected with mycobacterium translocate nuclear HMGB1 to the cytoplasm before releasing it into the extracellular milieu. The cellular HMGB1 mRNA levels in cells were determined using real-time polymerase chain reaction (RT-PCR). Consistent with earlier observations (Wang et al., 1999), cultured cells constitutively expressed HMGB1 mRNA and stimulation with LPS, M. bovis BCG and CFP did not significantly change its cellular mRNA levels in macrophage cultures (Fig. 2C).

In vivo release of HMGB1 in mycobacterial infection

As HMGB1 acts as a marker of tissue damage and inflammatory response, we hypothesized that vaccination would limit the extent of tissue damage and hence the levels of HMGB1 secretion in a guinea pig model of TB. Guinea pigs were immunized with BCG and infected with low-dose aerosol infection of M. tuberculosis. The levels of HMGB1 in bronchoalveolar lavage (BAL) cells were determined by Western blotting of nuclear and cytoplasmic extracts of cells (Fig. 2A) at day 60 after infection. As shown in Fig. 3A, HMGB1 was found in nuclear extracts of BAL cells of both different groups of guinea pigs. It was detected in very small amount in the cytoplasm of uninfected guinea pigs, and the levels of HMGB1 detected in cytoplasmic extracts of BCG-vaccinated animals were lower than those found to be released in cytoplasm of cells of guinea pigs not vaccinated with BCG. The HMGB1 protein was also detected in BAL fluid of the infected guinea pigs (Fig. 3B). The amount of HMGB1 in BAL fluid of BCG-vaccinated guinea pigs was observed to be much lesser than those found in BAL fluid of unvaccinated guinea pigs.

In order to extend these findings, we performed immunofluorescence staining of cytospin preparations of BAL cells using monoclonal anti-HMGB1 antibody. Each compartment (nuclear and cytoplasmic) was outlined using bright field and DAPI fluorescence. HMGB1 staining appeared to be diffusely distributed in the nucleus and cytoplasm of cells of uninfected guinea pigs (Fig. 3C, panel A), but intense staining of HMGB1 was observed in cytoplasm of many cells of infected guinea pigs (Fig. 3C, panel B). Some BAL cells (marked with arrows) in infected guinea pigs were not highly positive for HMGB1 staining, and we assume that it could be either because of absence of infection in those cells or the total release of HMGB1 out of the cells. The predominant staining of HMGB1 in cytoplasm was not observed in cells of animals which were first immunized with BCG and were then infected with M. tuberculosis. (Fig. 3C, panel C). The log₁₀ colony-forming units (cfu) recovered from lungs and spleens of guinea pigs were significantly lower (P ≤ 0.001) in BCG-vaccinated group compared with unvaccinated controls at 60 days post challenge (Fig. 4A). As release of HMGB1 is an indication of tissue injury in lungs, we also estimated expression of cytokines in lungs of infected guinea pigs using RT-PCR in order to detect the level of inflammation. The levels of IFN-γ in lungs of guinea pigs were significantly higher (P ≤ 0.01) in BCG-vaccinated group compared with unvaccinated controls (Fig. 4B). The mRNA levels of TNF-α were found to be significantly higher (P ≤ 0.001) in unvaccinated controls when compared with the BCG-vaccinated group (Fig. 4B). The expression of IL-1β mRNA was found to be significantly higher (P ≤ 0.001) in unvaccinated controls when compared with that of BCG-immunized animals. The levels of TGF-β in lungs of guinea pigs were significantly higher (P ≤ 0.001) in BCG-vaccinated group compared with unvaccinated controls (Fig. 4C). However, no difference was found in gene expression of inducible NO synthase (iNOS) when two experimental groups were compared (Fig. 4C).

Immunohistochemcal studies of the lungs from non-vaccinated guinea pigs or those vaccinated with BCG, at day 60 post infection showed that endothelial cells and macrophages stained positively for HMGB1 (Fig. 4D). No staining of HMGB1 was observed in isotype antibody control-treated sections (Fig. 4D, panel A). The histologic appearance and HMGB1 staining were clearly different between the two groups of animals. Lungs from the saline-treated group had extensive areas of pulmonary parenchyma effaced by granulomatous inflammation and coalescing foci of necrosis. The cells in the outer and middle layers of granuloma stained strongly positive for HMGB1 in lungs of saline-treated control animals (Fig. 4D, panel B). The distribution of HMGB1 was found to be less in normal areas of tissue of the lungs in both saline-treated and BCG-vaccinated animals (Fig. 4D, panels C and E). The regions of the granulomas that contained a necrotic core with many dead cells were also areas of strong staining for HMGB1 in lungs of the saline-treated control group (Fig. 4D, panel D). The BCG-treated animals had fewer and much smaller lesions in which mononuclear cells were found to be positively stained for HMGB1 in cytoplasm (Fig. 4D, panel F). However, the inflammatory regions in the lungs of saline-treated control animals were more intensely stained than those of BCG-vaccinated animals.

Effects of HMGB1 on the cytokine release against mycobacterial infection and mycobacterial growth

High-mobility group box protein 1 released by necrotic cells acts as a potent adjuvant in vivo. HMGB1 stimulates further release of TNF-α and other pro-inflammatory cytokines (Kalinina et al., 2004). We hypothesized that HMGB1 may be involved in the modulation of mycobacterium-specific inflammatory responses. J774A.1 cells were stimulated with recombinant HMGB1 (Sigma) and then were infected with M. bovis BCG at MOI of 10:1. The levels of cytokines TNF-α (Fig. 5A) and IL-1β (Fig. 5B) were estimated in culture supernatants by ELISA at various time points after adding M. bovis BCG. Different concentrations of rHMGB1 were tested and optimum concentration of 200 ng ml⁻¹ was used (data not shown). HMGB1 alone did not significantly affect TNF-α or IL-1β production, while BCG alone induced both TNF-α and IL-1β, maximally at 24 h for the former and 72 h for the latter. In combination, the concentrations of TNF-α and IL-1β were significantly increased (P ≤ 0.001), although the kinetics of secretion was not altered. As HMGB1 has been found to have antibacterial activity against the airway pathogen Moraxella catarrhalis and also Escherichia coli and Bacillus megaterium (Zetterstrom et al. 2002; 2006), it was hypothesized that HMGB1 could also kill mycobacteria. J774A.1 cells were cultured in the presence of different concentrations (0.25–1 µM) of rHMGB1 (Sigma) and M. bovis BCG at MOI of 10:1. The cells were lysed at different time points and lysates were plated on Middlebrook 7H11 agar. No difference was found in the growth of BCG in cultured cells in the presence of HMGB1 as observed by log₁₀ cfu of bacteria (Fig. 5C). These observations indicate that HMGB1 does not have anti-mycobacterial activity.

Toll-like receptor ligands of mycobacteria induce active secretion of HMGB1

The release of HMGB1 by cells has been observed using TLR ligands, such as LPS (TLR4), lipoteichoic acid (TLR2) and polyinosinic-polycytidylic acid (TLR3) (Jiang and Pisetsky, 2006). To investigate the release of HMGB1 by mycobacterial ligands, we incubated RAW 264.7 cells with known TLR ligands from mycobacteria such as lipomannan, mannosylated lipoarabinomannan (ManLAM) arabinosylated lipoarabinonannan, HSP65, HSP70 and 38 kDa lipoprotein. It was found that treatment of RAW264.7 cells with these ligands could cause the release of HMGB1 (Fig. 6A). Among these ligands, HSP65 was found to induce the maximum release of HMGB1 (Fig. 6B) after 24 h. This release of HMGB1 was not due to LPS contamination of these components, as levels of endotoxin were found to be lower than 0.005–0.01 ng mg⁻¹ as observed using Limulus amebocyte lysate test (data not shown). This value is much below the toxicity level for the cells cultured in vitro. Purified TLR ligands were not toxic to the cells as less than 2% cytotoxicity was observed using LDH release test during culture of TLR ligands with RAW264.7 cells (Fig. 6C). To further confirm our results, we first treated RAW264.7 cells with TLR antibodies and then cultured them with TLR ligands. As observed in Fig. 6D, the release of HMGB1 by ManLAM and HSP65 was reduced when cells were treated with TLR2 and TLR4 antibodies respectively. This suggests that mycobacterial TLR ligands were also capable of inducing active release of HMGB1 like LPS and viral DNA.

To further investigate the role of pro-inflammatory mediators in HMGB1 release by mycobacterial TLR ligands, we chose HSP65 which induced maximum secretion of HMGB1. Mycobacterial HSP65 protein is known to induce TNF-α and reactive nitrogen intermediates (Bulut et al., 2005); therefore, we investigated the role of TNF-α and nitric oxide in HSP65-induced release of HMGB1. The cells were treated with neutralizing antibody to TNF-α and stimulated with mycobacterial HSP65. As shown in Fig. 6E, the neutralizing antibody to TNF-α inhibited HMGB1 release by RAW 264.7 cells stimulated by HSP65. To evaluate the role of NO in HMGB1 release, the specific iNOS inhibitor 1400W was used. As shown in Fig. 6F, the addition of 1400W in RAW 264.7 cells reduced HMGB1 release by HSP65. 1400W alone did not result in HMGB1 release in RAW 264.7 cells (data not shown).

Inhibition of c-Jun NH₂-terminal kinase activation attenuated HMGB1 release induced by HSP65

We further studied the role of signal transduction pathways involved in HSP65-induced HMGB1 release by cells. Because activation of MAPK is required for the production of mediators such as TNF-α, IL-6, IFN-γ and NO, the role of MAPK in the release of HMGB1 was investigated. To demonstrate the inhibition of MAPK activation, RAW 264.7 cells were pre-treated with various specific MAPK inhibitors, followed by stimulation with HSP65 for 24 h. Inhibition of p38 using SB203850 did not inhibit HMGB1 release induced by HSP65 (Fig. 7A). Similarly, other p38-specific inhibitor SB202190 did not reduce HMGB1 release, even at a concentration of 200 nM (Fig. 7B).

To examine the dependence of HMGB1 release on the extracellular signal-regulated kinase (ERK) signalling pathway, specific inhibitors (U0126 and PD98059) of MEK, an upstream activator of ERK1 and ERK2, were used in parallel experiments. U0126, at concentrations up to 150 nM did not affect HSP65-induced HMGB1 release in cultures of RAW 264.7 cells (Fig. 7C). Similarly, PD98059, even at concentrations up to 25 µM, did not affect HSP65-induced HMGB1 release (Fig. 7D), indicating that mycobacterial HSP65 induces HMGB1 release through an ERK1/2-independent mechanism. Inhibition of c-Jun NH₂-terminal kinase (JNK) activation using specific inhibitor JNK inhibitor II blocked HMGB1 release induced by HSP65, even at 1 µM concentration (Fig. 7E). To assess the status of TNF-α in relation to HMGB1 secretion pathway, we pre-treated RAW 264.7 with JNK-specific inhibitor and iNOS inhibitor 1400W, stimulated the cells with HSP65 and determined the presence of TNF-α in cell culture supernatants using ELISA (Fig. 7F). JNK-specific inhibitor inhibited the production of TNF-α whereas 1400W did not inhibit the HSP65-induced TNF-α secretion. Inhibition of TNF-α secretion using anti-TNF-α antibody served as positive control. Taken together, our observations suggest that JNK controls the secretion of both HMGB1 and TNF-α induced by mycobacterial HSP65.

Cells and animals

J774A.1 and RAW 264.7 cells were purchased from American Type Culture Collection and cultured in DMEM medium supplemented with 10% FBS. M. tuberculosis H₃₇Rv, M. bovis BCG Pasteur, M. avium and M. scrofulaceum were grown from low-passage seed lots in Proskauer–Beck liquid media containing 0.05% Tween 80 to early mid-log phase. Cultures were aliquoted into 1 ml tubes and frozen at −70°C until used. Mycobacterial cell components were provided by TB Vaccine Testing and Research Materials Contract at Colorado State University. To obtain bone marrow-derived macrophages (BMDM), femurs from C57BL/6 mice (The Charles River Laboratory) were dissected free of connective tissue and flushed with DMEM medium. Bone marrow cells were seeded at 1 × 10⁶ cells ml⁻¹ in the presence of 30% conditioned medium from L929 cells. On day 7 of culture, non-adherent cells were removed by vigorous washing with DMEM medium. The adherent BMDMs were cultured in Opti-MEM (Invitrogen Life Technologies) medium without FBS for experiments. Specific pathogen-free, outbred, Hartley strain guinea pigs (Charles River Breeding Laboratories, Wilmington, MA) were individually housed, fed commercial guinea pig chow and maintained in a temperature and humidity-conditioned environment on a 12 h light/dark cycle. Each animal was randomly assigned to a vaccination treatment group. All protocols for animal use were approved by the Animal Care and Use Committee of the Colorado State University.

Cell culture

J774A.1 and RAW 264.7 cells (3 × 10⁶ cells) were plated in 6-well culture plates in Opti-MEM medium without FBS for 4 h and washed twice with Opti-MEM medium. Cells were then stimulated with M. tuberculosis, M. bovis BCG, M. tuberculosis cell components, M. avium, M. scrofulaceum and mycobacterial components for 24 h, and supernatants were collected for cytokine and HMGB1 assay. Similar experimental conditions were used in experiments with BMDMs.

BCG vaccination and AEROSOL challenge

Guinea pigs were vaccinated with 10³ cfu of M. bovis BCG (BCG Pasteur) intradermally. Non-vaccinated guinea pigs received an equal volume of pyrogen-free saline. The guinea pigs were rested for 8 weeks before being infected. A Madison chamber aerosol generation device was used to expose the animals to an aerosol of M. tuberculosis which was calibrated to deliver approximately 20–30 bacilli into the lungs. Guinea pigs from non-vaccinated and BCG-vaccinated groups were euthanized at day 60 after aerosol infection and the lungs were removed aseptically. To assess bacterial loads, individual organ homogenates were plated on nutrient Middlebrook 7H11 Bacto agar (Becton Dickinson Microbiology Systems, Cockeysville, Md.). Bacterial colonies were counted after 2–3 weeks of incubation at 37°C.

Preparation of lung lavage cells

To obtain the BAL cell population, BAL was performed by instilling ice-cold 2% heparin in phosphate-buffered saline (pH 7.4) into the lungs of euthanized guinea pigs. Lavage cells were washed twice in DMEM medium with 10% FBS by centrifugation at 320 g for 10 min at 4°C. After the second wash, the cell pellet was resuspended in DMEM medium supplemented with 2 µM l-glutamine, 100 U of penicillin ml⁻¹ and 10% FBS. Viable BAL cells were enumerated by trypan blue exclusion. An aliquot (70 µl) of the BAL cell population (10⁵ cells) was spun for 5 min at 140 g in a cytospin centrifuge, the slides were air-dried, fixed in 4% paraformaldehyde and stained for immunoflourescence microscopy. Nuclear and cytoplasmic extracts of lung lavage cells were prepared using CeLytic NuCLEAR Extraction Kit (Sigma) according to manufacturer's instructions.

Immunohistochemical analysis

Lungs of guinea pigs were infused with 30% OCT (Tissue-Tek) in PBS through the trachea. Once removed from the pulmonary cavity, lungs were embedded in OCT, frozen in a bath of liquid nitrogen for a few seconds, and then stored at −80°C. Serial sections, 9 µm thick, from each lung, were cut on a cryostat (Leica; CM 1850) using the Instrumedics tape transfer system, fixed in cold acetone, and air-dried. The sections were washed, and unspecific Ab binding was blocked with 3% BSA-PBS solution. Thereafter, the sections were incubated overnight at 4°C with HMGB1 or Isotyoe control antibody. All sections were washed three times in PBS and incubated with the secondary detection antibody conjugated to HRP (Jackson Immunoresearch). Finally, the reaction was developed using aminoethylcarbazole (BioGenex) as substrate. The sections were counterstained with Meyer's hematoxylin and thereafter mounted with crystal/mount (Biomedia).

Biochemical and immunochemical assays

The levels of TNF-α and IL-1β in supernatants were measured from the cultured J774A.1 cells using OptEIA ELISA kits (BD Biosciences), and assays were performed in accordance with manufacturer's instructions. For Western blotting of HMGB1, supernatants of cultured cells collected after 20–24 h of culture, lavage fluids, nuclear and cytoplasmic extracts were concentrated using Amicon ultra-4 centrifugal filter (Millipore) as described previously (Rendon-Mitchell et al., 2003). Protein concentration of concentrated supernatants, lavage fluids, nuclear and cytoplasmic extracts was determined by the BCA method (Pierce BCA kit) to ensure equal loading, and 25 µg of each sample was resolved on 12% Tris-SDS polyacrylamide gel. Protein was transferred to nitrocellular membranes, blocked with 5% bovine serum albumin in PBS-Tween, and blotted with mouse anti-HMGB1 monoclonal antibody (Sigma). The membrane was then incubated with HRP-conjugated anti-mouse IgG (Jacksonimmuno Research), and developed using ECL chemiluminiscence kit (Pierce). The relative levels of HMGB1 in culture supernatants were quantified using Quantity One software (Bio-rad) after capturing Western blot images by a charge-coupled device camera.

RNA isolation and RT-PCR

Total RNA was isolated from J774A.1 cells using the Trizol Method (Invitrogen) as per manufacturer's instructions. To the RNA, 4 µl of a 10 mM dNTP mix (Invitrogen), 2 µl of 10× first-strand buffer (500 mM Tris–HCl, pH 8.3, 750 mM KCl, 30 mM MgCl₂, 50 mM dithiothreitol), 2 µl of Random Decamer primers (50 µM), 100 U Moloney Murine Leukemia Virus–Reverse Transcriptase (100 U ml⁻¹, Invitrogen) and nuclease-free water were added to yield a final reaction volume of 25 µl. Reaction mixtures were incubated at 42°C for 1 h and at 95°C for 10 min to inactivate the reverse transcriptase. cDNA (5 ng) was amplified with a standard reaction mix containing 1× PCR buffer (10 mM Tris–HCl, pH 8.3, 50 mM KCl), 1.5 mM MgCl₂, 0.125 mM each dNTP, 0.4 µM HMGB1 primers and 1.25 U AmpliTaq Gold DNA polymerase (5 U ml⁻¹) (Applied Biosystems). Nuclease-free water was added to yield the final reaction volume of 25 µl. A 648 bp open reading frame of HMGB1 was amplified using HMGB1 primers described earlier (Wang et al., 1999).

Measurement of LDH release from J774A. 1 cell lines

Estimation of necrosis in cultured cells was determined by measurement of LDH release using the CytoTox96 non-radioactive cytotoxicity assay (Promega). Background release of LDH was obtained with untreated control cells, and maximum release of LDH with lysed uninfected cells. Percentage cytotoxicity was calculated by the formula: [release of LDH from infected cells (OD₄₉₀)/maximum LDH release (OD₄₉₀)] × 100.

Immunofluorescence microscopy

J774A.1 cells (1 × 10³ cells) were cultured in 8-well chamber slides (Nunc) and stimulated with M. bovis BCG for 18–20 h. The cells were fixed with paraformaldehyde and washed three times with PBS. J774A.1 cells and guinea pig lung lavage cytospin cells were permeabilized with 0.05% Triton X-1000 solution for 60 min at room temperature. The cells were then incubated with mouse anti-HMGB1 monoclonal antibody (Sigma) for 60 min at room temperature and again washed three times with PBS. After washing, the cells were incubated with FITC-conjugated anti-mouse IgG1 antibody/Alexa-fluor 488 conjugated anti-mouse IgG1 (Invitrogen) for 60 min at room temperature in dark. Finally, the cells were washed with PBS and mounted with 3:1 Vectashield : DAPI (Vector Laboratories) and photographed with a cooled Princeton change-coupled device camera mounted on an Olympus BX-60 fluorescence microscope.

Real-time polymerase chain reaction

Total RNA was isolated from guinea pig lung homogenates using the Trizol Method (Invitrogen) as per manufacturer's instructions. The cDNA was prepared by reverse transcription and RT-PCR, performed by using SYBR Green I double-stranded DNA binding dye (Applied Biosystems). The real-time primers used for guinea pig IFN-γ, TNF-α, TGF-β1, IL-1β and iNOS were published earlier (Allen and McMurray, 2003). The level of induction of mRNA was determined from the threshold cycle values normalized to the β-actin gene and then normalized to the value derived from healthy TB-free animals.

Statistical analysis

Two-way comparison between test and control group was performed using Student's t-test. The data are given as means ± standard error of the mean.