TNF Influences Chemokine Expression of Macrophages In Vitro and That of CD11b⁺ Cells In Vivo during Mycobacterium tuberculosis Infection¹

The hallmark of Mycobacterium tuberculosis infection is granuloma formation. The granuloma is a focal accumulation of immune cells, including macrophages, lymphocytes, and, at times, neutrophils (1). Within the granuloma, effector and bystander cells need to communicate in a coordinated manner. Signals required from CD4 and CD8 lymphocytes include IFN-γ and TNF, which activate macrophages harboring M. tuberculosis (reviewed in Ref.2). Activated macrophages have enhanced reactive nitrogen intermediate production and phagolysosome fusion, both of which can lead to destruction of M. tuberculosis bacilli. In addition, CD8 cytotoxic lymphocytes can kill the bacillus directly through granulysin-dependent mechanisms or kill the macrophages harboring the bacilli (reviewed in Ref.3).

In recent years, the use of anti-inflammatory drugs targeting TNF, such as Infliximab and Enbrel, for treatment of chronic inflammatory diseases, has highlighted the importance of TNF in controlling M. tuberculosis, because there is an increased risk of reactivation tuberculosis in patients treated with these drugs (4, 5). TNF has many potential effects within the lungs of an infected individual. In vitro, TNF has been reported to affect apoptosis of infected macrophages (6, 7), influence secondary signaling to macrophages for NO production (reviewed in Ref.8), and regulate the activity of other inflammatory cytokines (9). In vivo, pathology, control of bacterial numbers, and chemokine expression (in the liver) have been shown to be affected by TNF (10, 11). The role of TNF within the lungs in chemokine expression after M. tuberculosis infection has not been explored.

Neutralization of TNF in chronically infected mice led to severe pathology with the loss of granuloma structure (12, 13). Bacterial loads in the TNF-neutralized, chronically infected mice increased initially, but stabilized at a level below that generally considered fatal in C57BL/6 mice. Nonetheless, TNF-neutralized mice succumbed quickly. Granuloma structural deficiencies and uncontrolled infection in mice lacking TNF have been reported by our group and others, in both acute and chronic infection (10, 12, 13, 14, 15); however, the mechanisms by which TNF controls granuloma development and maintenance have yet to be elucidated. Studies in the liver of M. smegmatis and M. tuberculosis i.v.-infected mice showed that gene expression of a subset of chemokines was reduced in TNF-deficient mice (11), but these studies did not reveal the localized chemokine expression patterns in the lungs, the primary site of M. tuberculosis infection and control after aerosol infection.

We hypothesized that TNF influences macrophage chemokine expression and that these localized differences in expression lead to changes in granuloma formation. Our results described in this study demonstrate for the first time that TNF affects chemokine gene and protein expression by M. tuberculosis-infected macrophages both in vivo and in vitro. We also demonstrate that during the early phases of acute infection, TNF-deficient mice have similar numbers and types of immune cells migrating to the lungs, indicating that the lack of granuloma formation is a result of a migration defect within the lung tissue, rather than a defect in the migration of cells to the lungs. These data support the idea that TNF affects the expression of many chemokines at the level of macrophage gene expression, suggesting that differences in localized chemokine gradients may result in disorganized cellular infiltrate within the lungs.

C57BL/6 female mice (Charles River, Rockland, MA) and TNF receptor (TNFR)³p55^−/− (breeding pairs from Dr. T. Mak, Ontario Cancer Institute, Toronto, Canada) (16), 8–14 wk old, were used in all experiments. TNFRp55^−/− mice were bred in a specific pathogen-free facility at University of Pittsburgh. All infected mice were maintained in the BSL3 animal laboratories and routinely monitored for murine pathogens by means of serological and histological examinations. The university institutional animal care and use committee approved all animal protocols used in this study.

All chemicals were purchased from Sigma-Aldrich (St. Louis, MO) unless otherwise noted. Middlebrook 7H9 liquid medium and 7H10 agar were obtained from Difco (Detroit, MI). Abs used in flow cytometric analyses were obtained from BD PharMingen (San Diego, CA). The cell line MP6XT-22 (anti-TNF) was obtained from DNAX (Palo Alto, CA) and used to generate ascites (Harlan Bioproducts, Indianapolis, IN). Abs were purified endotoxin-free from ascites fluid as previously described (12), and mice were treated with Ab i.v. every 3 days (0.5 mg/dose) starting 4 mo postinfection. Control mice received normal rat IgG (Jackson ImmunoResearch Laboratories, West Grove, PA) with the same dosing regimen. In cell culture MP6XT-22 was used at a concentration of 0.1 mg/ml.

M. tuberculosis strain Erdman (Trudeau Institute, Saranac Lake, NY) was used to infect mice via aerosol (50–100 CFU) as previously described (17). In acute experiments in which a higher inoculum was used, the concentration of M. tuberculosis in the nebulizer was increased, and day 1 CFU were determined in the lungs. The tissue bacillary load was quantified by plating serial dilutions of the tissue homogenates onto 7H10 agar as described previously (18). In all chronic M. tuberculosis experiments, infections were performed using the low dose (50–100 CFU) aerosol infection, and mice were used at 4 mo postinfection.

Macrophages were derived from C57BL/6 bone marrow based on adherence, as previously described (19). On day 6 of culture macrophages were infected with M. tuberculosis (multiplicity of infection, 4). Four hours postinfection the supernatant was removed, cells were washed, and fresh medium was added.

To determine cellular infiltrate in the lung, at 10-day intervals lungs were removed for flow cytometric analysis. Lungs were subjected to a short period (20 min) of digestion with 1 mg/ml collagenase A and 25 U/ml DNase (Roche, Mannheim, Germany) at 37°C. The suspension was pushed through a cell strainer as previously described (20). RBC were lysed with RBC lysis buffer (NH₄Cl/Tris solution), and the single-cell suspension was counted. The samples were stained with anti-CD4, anti-CD8, anti-CD69, or anti-Gr1 and anti-CD11b in FACS buffer as previously described (20). The cells were fixed in 4% paraformaldehyde and collected on a FACSCaliber (BD Biosciences, Mountain View, CA). Analysis was performed using CellQuest software (BD PharMingen, San Diego, CA).

Tissue samples for histological studies were fixed in 10% normal buffered formalin, followed by paraffin embedment. For histopathological studies, 5- to 6-μm sections were stained with Harris’ H&E.

To isolate CD11b⁺ cells from the lungs of C57BL/6 mice, a MACS column isolation protocol (Miltenyi Biotech, Auburn, CA) was followed. The lungs were removed, rinsed in T cell medium, and pushed through a 70-μm pore size cell strainer (Fisher, Pittsburgh, PA) using the end of a 5-ml syringe. After RBC lysis and washing, the cells were labeled with CD11b microbeads in MACS buffer (2 mM EDTA and 0.5% SDS in PBS) for 15 min at 8°C. The cells were washed with MACS buffer and then resuspended in 0.5 ml of buffer. The cells were poured over a 40-μm pore size cell strainer and then applied to a prewashed MS⁺ column. After initial flow-through, the column was washed three times with 0.5 ml of buffer. The column was removed from the magnet, and CD11b⁺ cells were eluted from the column with 1 ml of buffer. The cells were spun down and resuspended in 1 ml of TRIzol, and the RNA was isolated.

RNA was isolated from the lung or cells using the TRIzol isolation protocol with modifications. The lung was homogenized in 3 ml (cells were lysed in 1 ml of TRIzol/2 × 10⁶ cells) of TRIzol reagent, and then two chloroform extractions were performed. After an isopropanol precipitation, the RNA was washed with 70% ethanol and treated with RNase inhibitor (Applied Biosystems, Foster City, CA) for 45 min. After treatment at 65°C for 15 min (to fully resuspend the RNA), the RNA was cleaned, and DNase was digested using the Qiagen RNA isolation kit, as directed by the manufacturer (Qiagen, Valencia, CA).

The RNA was reverse transcribed using Superscript II enzyme, as directed by the manufacturer (Invitrogen, Carlsbad, CA). For real-time RT-PCR we used the relative gene expression method (21). hypoxanthine phosphoribosyltransferase served as the normalizer, and uninfected lung or macrophages served as the calibrator. Each primer and probe set was tested for efficiency (results efficient >97% for all primer/probe sets). All samples were run in triplicate and with no-reverse transcriptase controls on an ABI PRISM Sequence Detector 7700. Relative gene expression was calculated as 2^{(−ΔΔ cycle threshold (Ct))}, where ΔCt = Ct (gene of interest) − Ct (normalizer) and the ΔΔCt = ΔCt (sample) − ΔCt (calibrator). Results are expressed as relative gene expression to uninfected samples. The primer and probe concentrations were used as suggested by Applied Biosystems, with the final concentration of each primer at 400 nM and that of probe at 250 nM. The primers (Life Technologies, Grand Island, NY) and probes (Applied Biosystems) used in this assay were designed using Applied Biosystems Primer Express software and the sequences are in Table I.

The quantities of CXCL9 and CXCL10 in the macrophage culture supernatants were determined with the Quantikine mouse CXCL9 and CXCL10 immunoassays (R&D Systems, Minneapolis, MN). These kits were used exactly as suggested by the manufacturer. The readout was performed using the Emax precision microplate reader (Molecular Devices, Sunnyvale, CA).

To determine the protein concentrations of TNF, C-C chemokine ligand 2 (CCL2), CCL3, CCL5, CCL12, and CXCL2, a multiplex sample testing service was used. Supernatants were collected from uninfected and infected macrophages using a pipette. These supernatants were then were filter-sterilized using an Acrodisc 13-mm syringe filter with a 0.2-μm HT Tuffryn membrane (Pall Corp., Ann Arbor, MI) attached to a 5-ml syringe. The supernatants were stored in 0.5-ml aliquots at −20°C. An aliquot was sent to Pierce (Rockford, IL) on dry ice (Boston, MA) for quantitative measurement using multiplexed sandwich ELISAs.

The pGEM-T Easy Vector System (Promega, Madison, WI) was used to clone CCL5, CXCL9, and CXCL10 PCR products. Primers were designed to amplify CCL5, CXCL9, and CXCL10 with restriction enzyme sites specific for the pGEM-T vector cloning sites. After amplification, the product was purified using the QIAquick Nucleotide Removal Kit (Qiagen, Valencia, CA). The ends were digested, and the product was ligated to cut pGEM-T vectors. Competent Escherichia coli (JM109) was transformed with the ligation reactions, using the suggested protocol (Promega, Madison, WI). Transformed bacteria were selected based on resistance to ampicillin and interruption of the lacZ promoter, and the selected colonies were grown up in LB⁺AMP liquid medium. The plasmid DNA was purified from this culture by the Qiagen Midiprep protocol (Qiagen). After verification of the insert, the plasmids were linearized and used to generate probes.

Labeled RNA probes were synthesized by in vitro transcription of DNA cloned downstream of T7 or SP6 RNA polymerase in the pGEM-T vector using [³⁵S]UTP as substrate. The MAXIscript In Vitro Transcription Kit (Ambion, Austin, TX) was used for this protocol, with slight modification. Probes were stored at −80°C.

Lungs were fixed in 4% paraformaldehyde at 40°C for 5 h, washed twice overnight with PBS, followed by a sucrose gradient, and then snap-frozen and stored at −80°C. Cryosections (14 μm) were cut, and slides were fixed in 4% paraformaldehyde for 20 min, dehydrated in an ethanol gradient (70% ethanol for 20 min, 80% ethanol for 5 min, then 95% ethanol for 5 min). After air-drying the slides, Ag retrieval was performed in 0.01 M sodium citrate, with warming in the microwave. After the slides were cooled, the tissues were acetylated twice in 0.25% acetic anhydride-0.1 M triethanolamine solution. After treatment in another ethanol dehydration gradient, the slides were air-dried.

Tissue was hybridized to the ³⁵S-labeled probes as previously described (22). In short, 5 × 10⁴ cpm/μl sense or antisense probe was hybridized to the tissue under siliconized coverslips (in 1× hybridization buffer containing 10% dextran sulfate formamide, 0.6 M NaCl, yeast tRNA (100 μg/ml), and 0.4 M DTT) overnight at 50°C. Coverslips were removed, and slides were washed in a series of SSC buffers: 5× SSC and 0.1 mM DDT at 42°C for 30 min, 2× SSC/50% formamide and 0.1 mM DTT at 60°C for 20 min, ribowash solution (0.1 M Tris, 0.05 M EDTA, and 0.4 M NaCl) at 37° C for 10 min, ribowash solution at 37°C for 10 min, ribowash solution and 6.25 U of an RNase A⁺ T1 mix at 37°C for 30 min, ribowash solution at 37°C for 10 min, 2× SSC at 37°C for 10 min, and 0.2× SSC at 37°C for 10 min. Slides were then dehydrated in 0.3 M ammonium acetate and an ethanol gradient. After the slides were completely dried, they were dipped in prewarmed radiography emulsion (Eastman Kodak, Rochester, NY) and dried. Slides were wrapped in a dark box for 1 wk. Development was performed in D19 developer, followed by fixation in Rapid Fix. Slides were counterstained in hematoxylin.

Three or four mice per group per time point were used for all studies. Statistical analysis was performed on the data using PRISM software for an unpaired t test. For bacterial numbers and cell numbers, log transformation was performed before statistical analysis to normalize the data. A value of p < 0.05 was considered significant.

Mice deficient in TNF have increased susceptibility to M. tuberculosis infection (reviewed in Ref.23). After low dose aerosol infection, bacterial burdens in TNF-neutralized mice (MP6-XT22 treated) and TNFRp55^−/− mice began to diverge from wild-type (WT) levels by 14 days postinfection, and the mice had significantly higher bacterial burdens than control mice by 18–21 days postinfection (Fig. 1 A and data not shown). The TNF-deficient mice succumbed by 28 days postinfection.

After TNF neutralization of chronically infected mice, the bacterial burden increased by 8 days postinfection (Fig. 1,B), and differences were significant by 18 days postinfection. We previously reported that the mean survival time for chronically infected TNF-neutralized mice was 44 days when infection was initiated by i.v. infection with moderate doses of M. tuberculosis (12). However, using low dose inoculum delivered via aerosol to set up the chronic infection, neutralization of TNF resulted in faster progression of disease, with a mean survival time of 21 days after initiation of Ab treatment (Fig. 1 C). In the following experiments, we analyzed the effects of TNF deficiency in acute and chronic models of infection, using aerosol delivery for all studies.

During acute mycobacterial infections, TNF-deficient mice do not form proper granulomas (14, 15, 24). The cell populations migrating into the lungs were assessed by flow cytometry. No defects in the appearance of T lymphocytes and macrophages in the lungs early after infection were observed in TNF-deficient mice compared with controls (Fig. 2). By 21 days postinfection, when bacterial burdens were already significantly higher in anti-TNF Ab-treated mice, the number of neutrophils was significantly higher. The increased neutrophilic infiltrate was probably a result of the increased mycobacterial burden in these mice. Histological analysis of the lung infiltrate indicated that although the cells were migrating into the lungs, they were not forming organized granuloma structures, as previously reported (Refs.14 and 15 and data not shown). These data indicate that TNF is not required for early migration of cells into the lung after infection, but, rather, may influence migration within the lung.

Similar results were obtained when analyzing TNF-neutralized, chronically infected mice. The cell populations in the lungs of anti-TNF Ab-treated and IgG control mice were not significantly different in the anti-TNF Ab-treated mice (Fig. 3). Although there was a trend toward increasing cell numbers in the anti-TNF-treated mice, there was variability among the mice during the chronic stage of infection (see error bars for day 0 time points, Fig. 3), and mice in the TNF-neutralized group succumbed to reactivation between days 12 and 25 after Ab treatment (Fig. 1 C). There was no statistical change in the numbers of cells in the lungs compared with the number of cells on day 0 of Ab treatment.

Histological changes in granuloma structure were observed as early as 9 days after anti-TNF Ab treatment (representative sections are shown in Fig. 4). By 20 days of Ab treatment all surviving anti-TNF treated mice had diffuse cellular infiltrate in their lungs accompanied by edema within alveolar spaces (Fig. 4).

These data indicate that TNF plays a direct or indirect role in controlling migration of immune cells within the lungs and maintaining granuloma structure. A possible mechanism for control of migration of cells is through effects on chemokine expression.

The effects of TNF on M. tuberculosis-induced chemokine expression by macrophages were first examined in vitro. After infection of bone marrow-derived macrophages with M. tuberculosis, the expression of many inducible C-C and C-X-C chemokines increased relative to that of uninfected macrophages, and mRNA levels peaked between 4 and 12 h postinfection depending on the gene (Fig. 5 and data not shown). RNA expression of CXCL10, CCL12, CCL2, and CCL4 peaked at 4 h, whereas expression of CXCL9, CXCL11, CCL3, and CCL5 did not peak until 12 h postinfection. Although RNA expression decreased rapidly after 12 h of infection, the levels remained elevated compared with those of uninfected macrophages.

As early as 4 h postinfection, TNF expression was induced (Fig. 5). Anti-TNF Ab treatment at the time of infection reduced the expression of many chemokines (Fig. 5), indicating that optimal expression of these chemokines was dependent on TNF. The expression of some chemokines, such as CCL5, was sustained at a higher level as infection progressed in vitro, and in the presence of anti-TNF Ab, the expression of CCL5 was reduced 30–70% (depending on the time point analyzed). CCL2 expression was reduced by TNF neutralization only at the 4 h point (Fig. 5). TNF neutralization modestly reduced CCL3 expression early in infection, but high levels of CCL3 mRNA were sustained in the anti-TNF Ab-treated, but not control, macrophage cultures (Fig. 5). CCL12 also had reduced expression in TNF-neutralized cultures early after infection, but this was up-regulated in the presence of anti-TNF Ab by 24 h postinfection and decreased again by 48 h (data not shown).

In a separate set of experiments, anti-TNF Ab was added 4 h after infection (when unincorporated bacteria were removed from the culture); the expression of chemokines was examined 24 and 48 h after infection (data not shown). The mRNA levels of CCL5, CXCL10, and CXCL11 were still significantly reduced at 24 and 48 h postinfection (60, 58, and 84%, respectively). The expression of CCL12 was significantly increased (data not shown). Treatment of infected macrophages with isotype control Ab did not alter chemokine expression (data not shown). Similar results were obtained using TNFRp55^−/− macrophages (data not shown). Addition of recombinant TNF to the infected macrophage cultures did not lead to additional chemokine expression (data not shown), suggesting that the TNF induced by M. tuberculosis infection was sufficient to affect the expression of the chemokine genes. These data indicated that TNF induced the expression of CCL5, CXCL9, CXCL10, and CXCL11, but down-regulated the expression of CCL12.

There are little data about post-transcriptional regulation of chemokine expression (25, 26). To confirm the gene expression results, we quantified protein production. Supernatants collected from uninfected and infected macrophages at 0, 4, 8, 12, 24, and 48 h postinfection (from the cultures used to make RNA in Fig. 5) were used in the immunoassay for CXCL9 and CXCL10 and were used for multiplex sample testing to quantify CCL2, CCL3, CCL5, CCL12, CXCL8, and TNF (Fig. 6).

Protein expression of CXCL9 was reduced 30–70% in the presence of anti-TNF Ab compared with control macrophages, depending on the time point. The expression of CXCL10 was reduced 20–30%, that of CCL5 was reduced 50–90%, that of CCL3 was reduced 10–90%, and that of CXCL8 was reduced 25%–70% (data not shown). The expression of CCL2 and CCL12 (data not shown) followed a similar pattern; in the presence of anti-TNF Ab, expression was reduced initially, then increased, but by 48 h postinfection there was clearly a reduction in the amount of protein present.

Protein analysis of supernatants from cultures to which anti-TNF Ab was added at 4 h postinfection confirmed the mRNA macrophage data, in that CCL2, CCL5, CCL12, CXCL9, and CXCL10 were induced by M. tuberculosis infection, and CXCL9, CXCL10, and CCL5 were down-regulated in the presence of anti-TNF Ab (data not shown). In contrast, although there was no difference in the expression of CCL2 mRNA in anti-TNF-treated macrophages, there was a significant reduction in the amount of protein.

Based on the results from macrophages and the findings in the spleen and liver of infected mice from a previous study (11), we examined the expression of chemokines in the whole infected lung by both RNase protection assay and real-time RT-PCR. We expected to observe differences similar to our findings in the in vitro macrophage experiments, but after low dose aerosol infection, the induction of chemokines over uninfected lung was not detectable until 10–12 days postinfection. In the aerosol infection model used, only 50–100 CFU were deposited in the lungs, and only a small percentage of macrophages were infected, leading to very low level chemokine expression. There is very little time between the detectable induction of expression in the whole lung and increasing bacterial growth in TNF-deficient mice, complicating analysis of the role of TNF in control of chemokine expression.

Gene expression measured in whole lung tissue might not accurately reflect the relevant gene expression within the granuloma (27) due to the local nature of the immune response within the granuloma. For this reason, chemokine expression was measured from RNA obtained specifically from granulomatous tissue procured by laser capture microscopy (LCM). Lung sections from early time points after MP6 or IgG treatment were chosen to assess the effects of TNF neutralization during a time where bacterial numbers remained similar in both groups. Differences between anti-TNF Ab-treated mice and control mice were not evident using LCM (data not shown). This may be due to collecting many cell types from the granulomas including T lymphocytes, B lymphocytes, macrophages, and neutrophils.

Macrophages are a major reservoir for M. tuberculosis and probably play a key role in orchestrating granuloma formation. As a more sensitive technique for assessing the effect of TNF on macrophage chemokine expression, we enriched for macrophages by isolating CD11b⁺ cells from the lungs of acutely infected mice at various time points after low dose aerosol infection. RNA was isolated from the CD11b⁺ population, and real-time RT-PCR was performed. On day 10 postinfection, the CD11b⁺ population of cells was 50% F4/80⁺ and 46% Gr1⁺ after MACS column isolation, indicating that this technique purified both macrophages and neutrophils from the lungs. Using this method, induction of chemokines in CD11b⁺ cells was still difficult to detect before 9 days postinfection; at this time point there was reduced relative expression of chemokines in the TNF-deficient mice compared with WT, but the expression was very low (data not shown). Just 3 days later (day 12) chemokine expression rose dramatically, and the TNF-deficient mice had higher expression of many chemokines (data not shown).

The aerosol inoculum was increased 20-fold to increase the number of macrophages initially infected in the lungs, and induction of chemokines in the macrophage population was detected as early as 4 days postinfection. There were differences between TNFRp55^−/− and WT mice in the expression of CXCL9, CXCL10, CCL3, CCL4, CCL5, and CCL12 (Fig. 7,A). By 7 days postinfection the bacterial burdens in the TNFRp55^−/− mice were significantly higher at this dose (data not shown). There were still reduced levels of expression of CXCL9, CXCL10, and CCL4, but now there was increased expression of CCL5 and CCL12 in TNFRp55^−/− mice compared with that in WT mice (Fig. 7 A). By day 10 the expression was 10-fold higher than on day 7, and the differences between the mouse strains were no longer evident (data not shown).

CD11b⁺ cells were also isolated ex vivo from the lungs of M. tuberculosis chronically infected mice (treated with IgG or anti-TNF Ab), and real-time RT-PCR was performed on isolated RNA. The expressions of CCL3, CCL4, CCL5, CXCL9, and CXCL10 were transiently reduced in CD11b⁺ cells isolated from chronically infected MP6-treated mice (Fig. 7 B). The expression of these genes increased as bacterial burdens increased, and by 12 days postinfection, when the bacterial burdens in MP6-treated mice were substantially higher than those in IgG-treated mice, chemokine mRNA levels were higher than those in control mice.

To examine the effects of TNF on the localized chemokine expression in the lungs early in infection, in situ hybridization was performed to visualize the expression of CXCL9 and CXCL10. An algorithm was designed to accurately assess the location and quantity of gene expression with respect to pathology within the lung (Table II). Single cells or nonclustered cells (<20 cells together) that expressed the gene of interested were classified as category 1. Infiltrating clusters of cells were classified as category 2 or 3 on the basis of histology and regardless of gene expression. Category 2 infiltrates were clusters or infiltrates of 20–100 cells, whereas category 3 infiltrates were defined as clusters of >100 cells. At the early time points after low dose aerosol infection (before day 14), many of the infiltrates were near blood vessels, with little apparent organization in both control and anti-TNF Ab-treated mice. After classifying the infiltrates, the expression of genes was analyzed in each infiltrate. If no signal was detected for the gene of interest the infiltrate was designated negative. If signal was detected above background, but <50% of the cells within a cellular aggregation were positive for expression, the infiltrate was designated positive, but if >50% of the cells within a cellular cluster were signaling for expression, the infiltrate was considered double positive. Therefore each category 2 or 3 infiltrate could be either negative, positive, or double positive.

The number of category 1 areas expressing CXCL9 or CXCL10 was counted in 20 ×20 fields. Category 1 (normal tissue) expression of CXCL10 was detected as early as 3 days postinfection, and the expression increased gradually as infection progressed. By 12 days postinfection the number of category 1 areas expressing CXCL10 (in the 20 fields counted) increased to an average of 8 ± 2 in the IgG-treated mice and 4 ± 1 in the anti-TNF Ab-treated mice. Expression of CXCL9 was detected on day 6 in some, but not all, mice. By 12 days postinfection the number of category 1 areas counted was 4 ± 1 in the IgG-treated mice and 1 ± 0.8 in the anti-TNF Ab-treated mice in the 20 fields counted. These data suggest that TNF could affect early local CXCL9 and CXCL10 expression before granuloma formation.

The overall number of category 2 and 3 infiltrates was not different between MP6-treated and control mice (data not shown) up to 12 days postinfection. This was not surprising, because at these early time points the overall number of cells migrating into the lungs was not significantly affected by anti-TNF Ab treatment (Fig. 2), and this is before true granuloma formation. The number of CXCL9 category 3 double-positive clusters was significantly lower in the anti-TNF Ab-treated mice by 12 days postinfection, but there were no differences in the number of CXCL9 category 2 positive clusters (Fig. 8 and data not shown). CXCL10 gene expression followed a different pattern; there were significantly fewer CXCL10 category 2 positive clusters in MP6-treated mice, but no differences in the number of CXCL10 category 3 positive clusters (Fig. 8 and data not shown). These data confirm that TNF neutralization affects the expression of a subset of chemokines in the lung, and that effects on the timing of chemokine expression may be important for granuloma formation.

TNF is required for granuloma formation, and the absence of TNF affects the expression of many inducible chemokines in the acute and chronic models of tuberculosis. In the absence of TNF, CCR5 and CXCR3 ligand expression was reduced both in vivo and in vitro, and the reduced expression was specific and localized to macrophages and CD11b⁺ cells. In the absence of TNF, the cell-mediated immune response fails to control infection, and the bacterial burden increases. Interestingly, in the chronic model of infection, the mice succumb to infection before the bacterial burdens rise to moribund levels, indicating that lung pathology is leading to a decrease in lung function and death (12). Deciphering the difference between the influences of TNF and bacterial burden on chemokine expression in a model in which TNF affects bacterial burden was technically challenging. Focusing attention at early time points in the experimental models allowed us to diminish the impact of higher bacterial burdens in TNF-deficient mice. In addition, we focused on macrophage chemokine expression because this cell is a major M. tuberculosis reservoir and is responsive to TNF. This enabled detection of the effects of anti-TNF Ab on chemokine expression before changes in pathology, granuloma formation, and increased bacterial burden. After increased bacterial burdens, the chemokine gradients were even more dramatically altered. The initial changes in chemokine expression may lead to a cascade of inappropriate cellular communication, impaired cell migration within the lung, loss of cell-mediated immune response, and substantial lung pathology.

Control of cell migration is a complex process involving not only chemokines and receptors, but also adhesion molecules. The complexity of this process has been illustrated in this study. We also examined the expression of adhesion molecules using gene expression filter arrays (GEArray Q Series; SuperArray Bioscience, Frederick, MD), but no striking differences were observed after anti-TNF Ab treatment (data not shown). Clearly, changes in pathology are compounded by the other roles that TNF plays in control of M. tuberculosis infection, and the effects on chemokine expression are probably not solely responsible for the severe pathology observed in TNF-deficient mice. Studies on low dose aerosol M. tuberculosis infection in CCR2 and CCR5 chemokine receptor transgenic knockout mice (Ref.20 and data not shown) and in vivo models of CXCL9 and CXCL10 neutralization (Ref.28 and data not shown) have revealed some differences in cell migration, but even in these models the pathology does not mirror the destructive pathology observed in the absence of TNF, where signaling by some of the same chemokines that were affected by TNF neutralization is diminished. CCR2^−/− mice have fewer macrophages throughout infection, but have sufficient numbers to control low dose infection, although the mice are quite susceptible to high dose infection (20, 29). CCR5^−/− mice had a greater lymphocytic infiltrate in their lungs, but controlled infection.⁴ CXCL10^−/− and mice treated with both anti-CXCL9 and anti-CXCL10 Abs controlled infection with no apparent lymphocyte migration deficiencies (H. M. Scott Algood, J. L. Flynn, and A. D. Luster, unpublished observations) to the lung, but a slight delay in lymphocyte migration to the lymph nodes.

A recently published study provided data that anti-CXCL9 Ab treatment in M. tuberculosis-infected mice did not affect the control of infection, but the mice had smaller, although still organized, granulomas (28). These authors suggested that CXCL9 was produced by neutrophils and played a key role in granuloma formation. However, the differences we observed in pathology and granuloma formation in TNF-deficient mice were not found in the anti-CXCL9 Ab-treated mice. Therefore, TNF-mediated control of CXCL9 expression may be only one factor in the influence this cytokine has on granuloma formation.

Our studies on whole lung chemokine RNA expression consistently (in three experiments) revealed differences at early time points postinfection and after Ab treatment when comparing control and TNF-deficient mice, although the differences were not statistically significant (data not shown). The LCM technique was used to examine more localized gene expression patterns in granulomas. To isolate enough RNA from fixed, embedded tissue to perform real-time RT-PCR on the LCM samples, multiple granulomas had to be isolated from each mouse. Although the granulomas isolated were from the same lobe of the lung, there was still high variability among mice and probably among granulomas, because granuloma size and the number of bacteria in each granuloma can be variable (data not shown). Although the LCM data revealed that chemokine expression was higher within the granuloma compared with whole lung, the pattern of expression was similar to whole lung chemokine expression.

We next focused our in vivo studies on the CD11b⁺ population of cells, enriching for macrophages. With this technique we were able to specifically localize reduced expression of CXCR3 and CCR5 ligands to the CD11b⁺ cells, confirming our in vitro studies. The technique used removed epithelial cells and lymphocytes from the lung homogenates, but the neutrophils (which are CD11b⁺) were not removed. We have focused on the macrophages, because direct effects of TNF on chemokine production by this population was observed in vitro. However, TNF may also affect chemokine production by neutrophils. The number of neutrophils at early time points postinfection and after Ab treatment in chronically infected mice was similar to the number of macrophages present in the lungs(20) (Figs. 3 and 4). There were 3–5 × 10⁴ macrophages and 3–5 × 10⁴ neutrophils 6 days postinfection, and 4 × 10⁵ of each cell type at 4 days after initiation of Ab treatment in the chronic model.

Recently, there has been published evidence of TNF affecting chemokine levels in the liver during i.v. M. smegmatis and M. tuberculosis infection (10, 11). Our study has gone beyond these by focusing on localized chemokine expression in lungs after aerosol infection and by addressing chemokines that were not studied. We also carefully controlled for the confounding effects of bacterial number on chemokine expression by studying very early time points in the infection or neutralization. We demonstrated that chemokine deficiency in the absence of TNF was at the macrophage level, using in vivo and in vitro studies. The macrophage probably plays a key role in granuloma formation as an initial site of infection, a reservoir for M. tuberculosis, and a potent producer of both TNF and chemokines.

The chemokines affected by TNF are known to influence the migration of cell types found within the lung during M. tuberculosis infection. CCL5, produced in the lung and by macrophages at high levels, induces migration of T lymphocytes, monocytes, and immature dendritic cells through CCR1 and CCR5. Disruption of the CCL5 gradient, as in TNF-deficient mice, probably affects the migration of these cells within the lung. T lymphocyte migration in the lungs could also be affected by the reduction in CXCL9, CXCL10, CCL3, and CCL4. Monocyte and macrophage migration may be affected by similar chemokines, including CCL3, CCL4 through CCR5, CCL2, and CCL12 through CCR2. The increased expression of these chemokines after increased bacterial load also probably leads to increased cell migration to the lungs. Therefore, we propose that initial reduction in chemokines in the absence of TNF leads to localized changes in the gradients, decreased cell responsiveness, and disorganized cellular infiltrate. As a result the cells do not interact with each other in a coordinated manner, thus reducing the cell-mediated response (11). The reduced cell-mediated response leads to increased bacterial burden, increased chemokine expression, and an increase in cells migrating into the lungs.

The chemokine expression studies performed in vitro advance our understanding of the role of TNF in chemokine expression and in post-transcriptional regulation of chemokines. Gene expression of most chemokines tested diminished by 48 h postinfection (except CCL5); this may indicate that there was a negative feedback loop for gene expression. CCL3 mRNA expression peaked at a similar point in anti-TNF Ab-treated and control macrophages, but mRNA levels remained higher in TNF-neutralized cultures. However, the protein level peaked at a lower point and was reduced in TNF-neutralized cultures. Anti-TNF Ab-treated macrophages may continue to express the gene in these cultures, because the protein has not accumulated to a level to turn down transcription. In addition, the expression of CCL2 was only reduced very early after infection in the presence of anti-TNF Ab, but protein expression was reduced as late as 48 h postinfection. This suggests that CCL2 and CCL3 post-transcriptional regulation may be affected by TNF.

The in situ hybridization analysis of CXCL9 and CXCL10 expression highlighted the extensive variability between different stages of cellular infiltrates, granuloma formation, and their chemokine expression. First, analyzing the number and size of the infiltrates confirmed that at early time points cellular infiltration in the lungs was not different in anti-TNF Ab-treated mice compared with control mice. As infection progressed, control mice developed organized granulomas, whereas TNF-deficient mice did not. Early patterns of chemokine expression presumably affect this organization of the cellular infiltrate. The size of the collection of cellular infiltrate in the lungs can reasonably be attributed to the time of formation; as the infiltrate grows and collects more cells, the granuloma forms. The expression of CXCL10 in smaller (early) infiltrates (category 2) was affected by TNF, but CXCL9 expression was only influenced by TNF in larger infiltrates (category 3). This may suggest a differential timing of chemokine expression in granuloma formation, or that the expression of these chemokines is sensitive to different levels of TNF. Alternatively, in the larger infiltrates there are more cellular interactions and possibly more cytokines to signal CXCL10 expression, overcoming a deficiency in TNF. In macrophages, CXCL9 was also more dramatically affected by anti-TNF Ab than was CXCL10. The kinetics of chemokine expression are clearly complex in the lungs, and the dynamics of granuloma formation may depend on TNF signaling the expression of different chemokines at specific times.

In summary, this in-depth study of localized chemokine expression patterns during acute and chronic M. tuberculosis infections presents evidence for TNF playing a direct role in chemokine expression and granuloma formation. We demonstrated that macrophages in vitro and CD11b⁺ cells in vivo are partially dependent on TNF for the expression of many inducible chemokines after M. tuberculosis infection. Our findings are relevant to understanding the immune response and granuloma formation during aerosol M. tuberculosis infection. Anti-TNF therapies have clearly illustrated the requirement for TNF in controlling latent tuberculosis in humans, and our studies have defined one possible mechanism of TNF in the lungs during infection. These studies will contribute to an appreciation of the potential effects of anti-inflammatory or antichemokine therapies on infections.

We are grateful to Amy Myers for excellent technical assistance, and to the members of the Flynn, Chan, and Nau laboratories for helpful discussions. We are grateful for the Reinhart laboratory’s technical assistance with the ISH protocol, particularly the guidance of Craig Fuller. We appreciate Sergio Onate providing assistance and access to the laser capture microscopy equipment.