A chemoattractant cytokine associated with granulomas in tuberculosis and silicosis

Mycobacterium bovis bacillus Calmette–Guérin (American Type Culture Collection no. 35734) was grown from a frozen stock for 3 days (to OD₆₀₀ ≈1.2) in Middlebrook 7H9 broth with 0.5% glycerol, 0.05% Tween 80, and ADC enrichment (Difco). BCG diluted 1:10 in RPMI medium 1640 with 1% fetal calf serum was added to J774 cells (bacteria:macrophage ratio approximately 10:1). The streptomycin-dependent strain of Escherichia coli, sd-4 (American Type Culture Collection no. 11143), was cultured in Luria–Bertani medium with streptomycin at 624 μg/ml. Frozen stocks in 20% glycerol were made from stationary-phase cultures after 2 days growth in Luria–Bertani medium containing 25 μg/ml streptomycin. On the day of infection these bacteria were thawed, were resuspended in 7H9 medium, and were diluted as for BCG (bacteria:macrophage approximately 1:1). These two ratios were required to achieve comparable percentages of infected macrophages; limited growth of the E. coli in the absence of streptomycin necessitated a lower ratio. Latex beads (0.8 μm, carboxylate-modified, Sigma no. L-1398) were treated with 70% ethanol before use.

J774 cells (American Type Culture Collection no. TIB 67) were grown without antibiotics in RPMI medium 1640 with 10% low-endotoxin fetal calf serum (GIBCO/BRL) until infection. These cells were infected by incubation with either bacteria or beads for 4 hr, were washed three times with Hanks’ balanced salt solution, and were cultured for an additional 20 hr before RNA harvest. During and after infection, cells were maintained in 1% low-endotoxin fetal calf serum as previously described (18). Alveolar macrophages were obtained from consenting human donors. Bronchoalveolar lavage cells were harvested by bronchoscopy and plated at a density of 500,000 cells/ml. Alveolar macrophages were purified by adherence to plastic after growing 4 days in RPMI medium 1640 with 10% fetal calf serum. Cells were infected as described above except that cultures of M. tuberculosis H37Rv and BCG were passaged through a 25-gauge needle before addition to macrophages.

RNA was isolated using the guanidinium isothiocyanate/CsCl method (22). Five micrograms of poly(A)⁺ RNA isolated by oligo(dT) chromatography was used to construct a cDNA library from BCG-infected J774. The Superscript cDNA kit (GIBCO/BRL) was used to generate first-strand cDNA using oligo(dT). Second-strand cDNA was synthesized using a modification of the protocol of Gubler and Hoffman (23). The resulting cDNA was ligated into λgt10 using EcoRI–NotI linkers. Recombinant phage were selected and amplified using standard techniques. The cDNA library prepared from BCG-infected J774 cells contained 1.1 × 10⁶ independent recombinants with an average insert size of greater than 1 kb. Of six random phage isolates, all contained inserts.

A nonradioactive detection method, the Genius System (Boehringer Mannheim) was used for library screening. Preparation of digoxigenin-labeled cDNA probes from poly(A)⁺ RNA, hybridization, and detection with Lumi-Phos 530 or NBT/X-phosphate were used as per the manufacturer’s protocol. Plaques were screened by incubating duplicate filters with digoxigenin-labeled cDNA from E. coli or BCG-infected macrophages.

Ten micrograms of total cellular RNA was run per lane of a 1% glyoxal/dimethyl sulfoxide agarose gel and transferred to nylon membranes (Amersham). Hybridization, washing, and probe stripping were as directed by the membrane manufacturer. Probe fragments isolated from agarose gels by using DEAE membrane (Schleicher & Schuell) or by using silica adherence (Geneclean II, SUN Bioscience) were labeled by random priming (PrimeIt II, Stratagene). Relative signals were quantitated with a phosphoimager (Fuji) and were normalized to the actin signals.

One microgram of total RNA was reverse-transcribed using Superscript reagents (GIBCO/BRL). The PCR used 1/10th of the reverse-transcribed cDNA in a 100-μl volume, and included 10 μl of 10× Taq buffer, 4 μl of 5 mM dNTPs, 50 pmol of each primer, 1 μl of competitive template, and 1 μl (5 units) of Taq DNA polymerase (Perkin–Elmer). The reaction was incubated for 35 cycles of 94°C for 30 sec, 50°C for 30 sec, and 72°C for 1 min. The competitive template was generated by insertion of a 333-bp DNA fragment from an NdeI digest of the yeast RPB1 gene to the unique NdeI site of the human osteopontin gene. The primers used were 5′-CACCTGTGCCATACCAGTTAAACAG-3′ and 5′-CATGGCTGTGAAATTCATGGCTGTG-3′. These primers allowed amplification of an 830-bp cDNA fragment and a 1,163-bp fragment from the competitive template. No product was seen when these primers were used to amplify this fragment from 1 μg of purified genomic DNA. Commercially available actin primers were used (Stratagene).

Tissue sections were obtained from the Gaensler Lung Archive at the Boston University Pulmonary Center and the Boston Veterans Affairs Hospital. These specimens are paraffin-embedded tissue processed using standard clinical techniques. Immunohistochemical staining was done on a Ventanna ES automated stainer (Ventanna, Tucson, AZ). Sections 5 μm thick were baked for 1 hr at 60°C onto positively charged slides and then deparaffinized with xylene and hydrated in graded alcohol washes to water. Slides were then processed with Ventanna’s protease I reagent for 4 min and incubated with primary antibodies for 32 min at 42°C. The manufacturer’s protocol was used for the diaminobenzidine kit (osteopontin) and the fast red kit (CD68). Images were photographed with Kodak Ektachrome or Ektar film and scanned into Adobe Photoshop to create composite figures.

mAbs MPIIIB10₁, specific for rat and human osteopontin, and QH1, specific for quail vascular endothelial cells, were obtained from the Developmental Studies Hybridoma Bank (Iowa City, Iowa) and used at 1:200 dilution of hybridoma supernatant (400 and 500 ng/ml, respectively). mAb KP-1 (24) recognizing human CD68 was obtained from Dako.

The effect of bacterial infection on macrophage gene expression was investigated by differential screening of cDNA libraries. The murine macrophage cell line J774 was used in these initial experiments because previous work has established its usefulness for studying interactions between mycobacteria and macrophages (18, 19). Differential screening of libraries has the advantage of detecting smaller changes in relative mRNA levels than subtractive hybridization.

The screening strategy is summarized in Fig. 1. The J774 cells were exposed to latex beads, or E. coli, or M. bovis BCG, and recombinant cDNA libraries and probes were prepared from the macrophage poly(A)⁺ mRNA. The phage library representing mRNAs of macrophages infected with BCG was plated with host bacteria, and duplicate filters were placed on the plaques. These filters then were hybridized with cDNA probes prepared from J774 cells infected with either E. coli or BCG, and plaques with different signal strengths were isolated. This screening strategy was designed to identify genes whose expression differed when macrophages were exposed to BCG versus E. coli. The library constructed with mRNA from BCG-infected J774 cells was screened to maximize the chance of identifying genes whose expression was specifically increased by BCG.

Approximately 730,000 recombinant phage were screened by probing duplicate filters with labeled cDNA prepared from J774 cells that had phagocytized BCG or E. coli. Three clones were isolated that produced stronger signals with cDNA from macrophages infected with E. coli. Sequencing of their insert DNAs revealed that one cDNA clone encoded macrophage inflammatory protein 1-α (MIP-1α) (25) and two encoded ferritin (26). Ten clones were isolated that produced stronger signals with cDNA from macrophages infected with BCG. All of these encoded osteopontin (27).

To confirm that the three genes represented by the cDNA clones identified by the screening process were indeed differentially expressed, radiolabeled insert DNAs from the clones were used to probe macrophage mRNA immobilized on nylon filters (Fig. 2A). The mRNAs of all three genes accumulated to higher levels when macrophages phagocytized the two bacterial species versus the latex beads.

Osteopontin mRNA accumulated to similar levels in macrophages exposed to latex beads or E. coli, but accumulated to at least 5-fold higher levels after infection of macrophages with BCG (Fig. 2A). There were similar increases in osteopontin mRNA levels in macrophages that had phagocytized live and heat-killed BCG (Fig. 2B). As shown in Fig. 2C, neither 7H9 medium nor supernatant from a BCG culture was capable of eliciting the increase in osteopontin expression caused by BCG organisms. Thus, osteopontin mRNA levels increase substantially in murine macrophages that phagocytize BCG.

We investigated whether increased osteopontin gene expression occurs in primary human cells in response to phagocytosis of virulent M. tuberculosis. Human alveolar macrophages were exposed to latex beads, BCG, and M. tuberculosis strain H37Rv. To determine precisely the number of osteopontin mRNA molecules present per cell before and after infection, a reverse transcriptase-PCR assay was used with varying levels of competing template. The results demonstrate that the levels of osteopontin mRNA were higher in alveolar macrophages infected with BCG (Fig. 3, lane B) and virulent M. tuberculosis (Fig. 3, lane Rv) than in macrophages that had phagocytized latex beads (Fig. 3, lane L). The human alveolar macrophages that had phagocytized latex beads contained approximately 1 molecule of osteopontin mRNA per cell. In contrast, the alveolar macrophages that had phagocytized BCG and M. tuberculosis had about 10 molecules of osteopontin mRNA per cell. This magnitude of increase in osteopontin mRNA after exposure to mycobacteria was similar to that observed with the murine macrophage model.

The increase in osteopontin mRNA observed in infected alveolar macrophages predicts that osteopontin protein levels should increase in human tissues infected by M. tuberculosis. We therefore investigated whether the presence of osteopontin protein in patient tissues correlated with the histopathology of tuberculosis. Tissue sections from normal and diseased lungs were stained with a mAb specific for osteopontin, MPIIIB10₁, with a mAb specific for the macrophage antigen CD68 (24), or with an isotype control, QH 1 (28). The MPIIIB10₁ mAb has been used extensively to identify osteopontin by immunohistochemistry (29, 30) and immunoprecipitation (31, 32).

Osteopontin was readily identified throughout specimens of tuberculous lung. The architecture of a tuberculosis lesion with necrosis (N) and with a lymphoid aggregate (L) above alveolar air spaces was evident in sections treated with the isotype control antibody and the hematoxylin counterstain (Fig. 4A). No staining (brown color) of this specimen was observed with the isotype control (Fig. 4A). However, when an adjacent serial section was probed with the MPIIIB10₁ antibody, a strong osteopontin signal was detected in the inflammatory border surrounding the necrosis, and immediately adjacent to and within the areas of caseating necrosis (Fig. 4B). At higher power, macrophages were identified in a serial section with anti-CD68 antibody and the fast red reagent (Fig. 4C). These cells also expressed osteopontin (Fig. 4D). Lymphocytes identified at low power and at high power in Fig. 4 A and C (L) demonstrated intense osteopontin signal singly and in aggregates (Fig. 4 B and D). Giant cells expressed CD68 (Fig. 4E) but did not universally stain for osteopontin; in general, epithelioid giant cells expressed osteopontin (Fig. 4F). Bronchiolar epithelium (Fig. 4G) and small blood vessels (Fig. 4H and below columnar epithelium in Fig. 4G) stained with the MPIIIB10₁ antibody though they have not been known to express osteopontin in normal adult tissue (33–36). In addition, osteopontin appeared in layers near caseation (Fig. 4B) and was intimately associated with fibroblast-like cells in a number of views (Fig. 4 F, G, H). In summary, osteopontin was identified in a variety of cell types within the tuberculosis specimens.

Specimens from normal lung, an acute inflammatory condition (bacterial bronchopneumonia), and two chronic inflammatory conditions (silicosis and granulation tissue of a pilonidal cyst) were studied to assess the variety of situations in which osteopontin could be detected. Normal lung did not show significant staining for osteopontin either in alveolar networks (Fig. 5A) or in airways (not shown). The bronchopneumonia specimen had scattered cells of macrophage and lymphocyte morphology that produced some signal with MPIIIB10₁ (Fig. 5B). Lymphoid aggregations associated with the pneumonia were also positive for osteopontin (not shown). The two chronic inflammatory conditions showed dichotomous results for the presence of osteopontin. Silicosis is a noninfectious granulomatous disease (37). The silicotic nodule with its characteristic fibrosis showed an abundance of macrophages that stained with CD68 (Fig. 5C). These cells, and presumably some of the extracellular matrix, showed heavy signal for osteopontin on immunohistochemical staining (Fig. 5D). Granulation tissue is another pathologic process characterized by mononuclear cell infiltration but is distinct from granulomatous inflammation (37). While the granulation tissue of the pilonidal cyst showed a large aggregation of giant cells and numerous macrophages based on CD68 staining (Fig. 5E), there was no osteopontin detected (Fig. 5F). Thus, only tuberculosis and silicosis, the two granulomatous conditions, expressed high levels of osteopontin in the tissue pathology.