Abstract

Infection with Mycobacterium tuberculosis in humans results in active disease in ∼10% of immune-competent individuals, with the most-severe clinical manifestations observed when the bacilli infect the central nervous system (CNS). Here, we use a rabbit model of tuberculous meningitis to evaluate the severity of disease caused by the M. tuberculosis clinical isolates CDC1551, a highly immunogenic strain, and HN878 or W4, 2 members of the W/Beijing family of strains. Compared with infection with CDC1551, CNS infection with HN878 or W4 resulted in higher bacillary loads in the cerebrospinal fluid and brain, increased dissemination of bacilli to other organs, persistent levels of tumor necrosis factor–α, higher leukocytosis, and more-severe clinical manifestations. This pathogenic process is associated with the production by HN878 of a polyketide synthase–derived phenolic glycolipid (PGL), as demonstrated by reduced virulence in rabbits infected with an HN878 mutant disrupted in the pks1-15 gene, which is required for PGL synthesis

Tuberculous meningitis (TBM), the most severe form of Mycobacterium tuberculosis infection in humans, predominantly affects children and is associated with up to 50% mortality [1–4 ]. It is not clear what determines dissemination of infected phagocytes across the blood-brain barrier (BBB) or when symptomatic central nervous system (CNS) disease develops. Past studies of the breach of the BBB and induction of TBM by intravenous injection of mycobacteria have been unsuccessful [5, 6]. We therefore established and characterized an alternative rabbit model of TBM [7]. In our model, mycobacteria are inoculated directly into the cisterna magna of rabbits, and the course of infection and host response are monitored. Using this model, we have shown that tumor necrosis factor (TNF)–α is a determinant of both immunity and pathogenicity in the CNS [8]

The ability of the host to recognize invading bacilli and to mount a protective immune response against M. tuberculosis determines the progression of infection and severity of disease [9]. Phenotypic diversity among clinical isolates of M. tuberculosis has been shown to contribute to differential induction of host immunity [10–13 ]. Strain CDC1551 has been associated with 5 cases of active disease, with a large percentage (72%) of the contacts converting to highly positive purified protein derivative of tuberculin skin test results [10]. Strain HN878, a member of the W/Beijing family, caused 60 cases of tuberculosis (TB) during 3 outbreaks in Texas between 1995 and 1998 [13]. Another W/Beijing strain, W4, has caused extensive disease in New Jersey [14]. In our mouse aerosol infection model, CDC1551 induced a rapid and robust cytokine response in the lungs, compared with that induced by the laboratory strains H37Rv and HN878. Well-organized granulomas with high levels of TNF-α, interleukin (IL)–6, IL-10, IL-12, and interferon (IFN)–γ mRNA were observed sooner in the lungs of mice infected with CDC1551 [11]. Infection with HN878 induced weak T cell proliferation and IFN-γ production by spleen and lymph node cells [13], in association with early death (hypervirulence) in infected mice. In rabbits infected by aerosol with Erdman, H37Rv, or CDC1551, the third strain appeared least virulent, requiring the highest number of inhaled bacilli to form 1 grossly visible pulmonary tubercle [15]

Recently, a highly bioactive lipid—a polyketide synthase (PKS)–derived phenolic glycolipid (PGL)—produced by HN878 and W4 but not by CDC1551 or H37Rv was identified and characterized [16, 17]. The PGL-deficient mutant of HN878, constructed by disrupting the pks1-15 gene cluster, was found to be more immunogenic and caused delayed death of the infected mice [16]. Here, we use the rabbit TBM model to study infection with CDC1551, HN878, W4, or H37Rv, and the PGL-deficient mutant HN878pks1-15::hyg is compared with the parental strain

Materials and Methods

Clinical isolates ofM. tuberculosisThe M. tuberculosis strains studied were as follows: (1) CDC1551, provided by T. M. Shinnick, Centers for Disease Control and Prevention, Atlanta, GA; (2) HN878, provided by J. M. Musser, Houston, TX [18]; (3) W4, provided by B. N. Kreiswirth, Public Health Research Institute, Newark, NJ; (4) HN878pks1-15::hyg a mutant strain of HN878 with a pks1-15 gene cluster disrupted by insertion of a hygromycin resistance gene, provided by M. B. Reed, Rockville, MD; and (5) H37Rv (TMC no. 102; Trudeau Institute). All procedures were performed in a biosafety level 3 laboratory

Genotypic differentiation ofM. tuberculosis strains: CDC1551 versus HN878 and HN878 versus HN878pks1-15::hygTo differentiate CDC1551 and HN878, a multiplex polymerase chain reaction (PCR) based on the strain-specific insertions of IS6110 elements in 2 M. tuberculosis chromosomal regions—mmpL4 and oriC—was used. Chromosomal DNA was isolated from single colonies grown on 7H11 agar plates by incubation in Tris-EDTA buffer at 80°C, boiling (5 min), and centrifugation at 15,366 g. Four primers from the GeneAmp PCR System 270 were used under the following conditions: 94°C for 5 min; 35 cycles of 94°C for 20 s, 62°C for 10 s, and 72°C for 15 s; and 72°C for 2 min. The primer pairs used were as follows: E-F (5′-AGCTTCTCTTGGCCCTCC-3′) and E-R (5′-ATCCCGATGGTGATAGTC-3′), which amplify a 639-bp fragment in HN878; and A4-R (5′-GTTCATCATTGACCTCTGAC-3′) and dnaA-F (5′-CCGTCAGCGCTCCAAGCGC-3′), which amplify a 394-bp fragment in CDC1551. To differentiate HN878 from HN878pks1-15::hyg a multiplex PCR was performed using the primers pks 5F (5′-GTTGATCGCAGCGATCCCCACACCC-3′), pks 7R (5′-ACGTTGCATGTGGATGAGCCTTCC-3′), hyg 3F (5′-GAGCCTGCGGAACGACCAGGAATT-3′), and hyg 1R (5′-CCAGCAGCGTGTCCACGTCCGGCA-3′). DNA from the mutant strain yields 2 PCR products, of 950 and 580 bp, whereas the parent strain, HN878, yields only the 950-bp product. A total of 573 M. tuberculosis colonies were analyzed (326 for the CDC1551 vs. HN878 comparison and 247 for the HN878 vs. HN878pks1-15::hyg comparison) (see figure A1 in the Appendix, which is not available in the print edition of the JournalAppendix, which appears only in the electronic edition of the Journal)

Induction of meningitisNew Zealand White rabbits (Covance Research Products) were used, as described elsewhere [7]. On the day of experiment, rabbits were anesthetized and immobilized. A spinal needle was used to withdraw 0.3 mL of cerebrospinal fluid (CSF) and to inject 0.2 mL (5×105 cfu) of M. tuberculosis intracisternally. After 2 h, CSF was obtained and plated onto 7H11 agar (Difco), to determine the inoculum. CSF and blood samples were obtained from rabbits weekly for 8 weeks. Half of the brain and part of the lung, liver, and spleen were used for the colony-forming unit assay. The rest of the brain, lung, liver, and spleen were fixed in 10% buffered formalin acetate (vol/vol) (Fisher Chemical) for histopathologic examination. The protocol was approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Medicine and Dentistry of New Jersey, Newark campus, and at the Public Health Research Institute

CSF samplesCSF samples were analyzed for total leukocyte counts (Coulter Electronics), differential counts (Diff-Quick; Baxter), and colony-forming units. The remaining CSF was centrifuged, and the supernatant was stored at −70°C for the TNF assay

Blood samplesHeparinized blood, collected from the ear artery, was centrifuged at 10,000 g. Plasma was separated and frozen at −70°C for the TNF assay

TNF assayTNF biological activity was evaluated using a cytotoxicity assay for murine L929 fibroblasts, as described elsewhere [7]. This assay does not discriminate between TNFs

Colony-forming unit assayColony-forming units were evaluated in the CSF and organ homogenates by plating 10-fold serial dilutions onto Middlebrook 7H11 agar (Difco)

Clinical scoring system for rabbitsTo evaluate the clinical course of CNS infection in rabbits, we developed the following scoring system: stage 0 (normal), stage 1 (hyperesthesia, head tilt, and lethargy), stage 2 (monoparesis), stage 3 (hemiparesis and recumbency), stage 4 (quadriplegia), and stage 5 (anorexia, and CNS depression progressing to moribund state and death). Rabbits at stages 4 and 5 were killed as required by our IACUC

Histopathologic analysisFormalin-fixed brains were cut transversely in serial sections from rostral to caudal, representing the fore-, mid-, and hindbrains; embedded in paraffin; sectioned; and stained with hematoxylin-eosin and Ziehl-Neelsen

Statistical analysisThe independent Student’s t test or the Mann-Whitney U test for nonparametric independent data was used for analyses. The Kruskal-Wallis test was used to determine statistical differences in the clinical manifestation scores between different strains. P<.05 was considered to be significant

Results

Bacillary load in CSF and tissues of rabbits infected intrathecally with differentM. tuberculosis clinical isolates Rabbits were infected by intracisternal inoculation of 5×105 cfu and were monitored for 8 weeks. At 2 h after infection, 4 log10 cfu/mL of CSF were detected for all 4 M. tuberculosis strains (figure 1A). By 7 days after infection, a reduction in the number of colony-forming units was noted for all strains. Thereafter, complete clearance of CDC1551 from the CSF started earlier (by 2 weeks after infection; P=.01), and, from 4 weeks after infection, no viable bacilli were detected in the CSF. In contrast, numbers of HN878 and W4 remained elevated for 8 weeks. H37Rv persisted up to 5 weeks and was then fully cleared. At 8 weeks after infection, HN878 showed the highest bacillary load in the brain, compared with the other strains (P=.02, vs. CDC1551) (figure 1B). No dissemination of CDC1551 to the lungs was noted at day 14 and day 21 after infection (data not shown) or at 8 weeks after infection (P>.001, vs. HN878 and W4). Similar results were seen in the liver (P>.001). The high loads of HN878 in the lungs suggested local replication and/or continuous dissemination from the CNS

Figure 1

Bacillary load in cerebrospinal fluid (CSF) and tissues of rabbits infected intracisternally with Mycobacterium tuberculosis. A No. of M. tuberculosis colony-forming units in the CSF of rabbits infected with CDC1551 (white diamonds) (n=10), HN878 (black triangles) (n = 12), W4 (black circles) (n=10), or H37Rv (white squares) (n=10). The clearance of CDC1551 differed significantly from that of other strains (P=.01). B Colony-forming units in the brains, lungs, and livers 8 weeks after infection of rabbits with CDC1551 (hatched bars) HN878 (black bars) W4 (gray bars) or H37Rv (white bars) (in brain, CDC1551 vs. HN878, P=.02). No dissemination of CDC1551 to the lungs (P>.001, vs. HN878 and W4) or livers (P>.001) was seen. Values are means±SEs

Figure 1

Bacillary load in cerebrospinal fluid (CSF) and tissues of rabbits infected intracisternally with Mycobacterium tuberculosis. A No. of M. tuberculosis colony-forming units in the CSF of rabbits infected with CDC1551 (white diamonds) (n=10), HN878 (black triangles) (n = 12), W4 (black circles) (n=10), or H37Rv (white squares) (n=10). The clearance of CDC1551 differed significantly from that of other strains (P=.01). B Colony-forming units in the brains, lungs, and livers 8 weeks after infection of rabbits with CDC1551 (hatched bars) HN878 (black bars) W4 (gray bars) or H37Rv (white bars) (in brain, CDC1551 vs. HN878, P=.02). No dissemination of CDC1551 to the lungs (P>.001, vs. HN878 and W4) or livers (P>.001) was seen. Values are means±SEs

Inflammatory response and clinical course of disease in rabbits with CNS infectionAfter infection with CDC1551, the leukocyte influx into the CSF was low and dropped slowly (⩽1000 cells/mm3) (figure 2A). In contrast, at 1 and 2 weeks after infection, HN878 and W4 leukocyte influx was higher and persisted longer than that of CDC1551 (P=.001 and P=.01, respectively). Leukocytosis induced by H37Rv was similar to that induced by W4. The predominant leukocytes in the CSF were monocytes and lymphocytes (>90%)

Figure 2

Inflammatory response and clinical course of disease in rabbits with meningitis. A Leukocyte density in the cerebrospinal fluid (CSF) of rabbits infected with CDC1551 (white diamonds) HN878 (black triangles) W4 (black circles) or H37Rv (white squares). HN878- and W4-induced leukocyte densities were higher than those induced by CDC1551 at 1 and 2 weeks after infection (P=.001 and P=.01, respectively). Values are means±SEs. B Tumor necrosis factor (TNF) level in the CSF of rabbits infected with CDC1551 (hatched bars) HN878 (black bars) W4 (gray bars) or H37Rv (white bars). Values are means±SEs. C Severity of signs of disease (clinical score) in rabbits infected with HN878 (dark gray area) (n=12), W4 (light gray area) (n=9), CDC1551 (n=10), or H37Rv (n=10). CDC1551 and H37Rv did not induce neurologic signs. Rabbits infected with HN878 or W4 showed more-severe progressive signs of disease than did those infected with CDC1551 and H37Rv (P<.0001). Results are expressed as the area under the curve

Figure 2

Inflammatory response and clinical course of disease in rabbits with meningitis. A Leukocyte density in the cerebrospinal fluid (CSF) of rabbits infected with CDC1551 (white diamonds) HN878 (black triangles) W4 (black circles) or H37Rv (white squares). HN878- and W4-induced leukocyte densities were higher than those induced by CDC1551 at 1 and 2 weeks after infection (P=.001 and P=.01, respectively). Values are means±SEs. B Tumor necrosis factor (TNF) level in the CSF of rabbits infected with CDC1551 (hatched bars) HN878 (black bars) W4 (gray bars) or H37Rv (white bars). Values are means±SEs. C Severity of signs of disease (clinical score) in rabbits infected with HN878 (dark gray area) (n=12), W4 (light gray area) (n=9), CDC1551 (n=10), or H37Rv (n=10). CDC1551 and H37Rv did not induce neurologic signs. Rabbits infected with HN878 or W4 showed more-severe progressive signs of disease than did those infected with CDC1551 and H37Rv (P<.0001). Results are expressed as the area under the curve

Elevated numbers of colony-forming units and CSF leukocytosis were associated with local TNF production. CDC1551 initially induced higher mean levels of TNF; however, this response waned by 4 weeks after infection and then disappeared (figure 2B). In contrast, the TNF levels induced by HN878 and W4 were lower at the start of infection and then persisted in the CSF for 8 weeks. H37Rv induced very low to undetectable levels of TNF in the CSF; plasma TNF showed similar patterns (data not shown)

The inflammatory response was correlated with the clinical manifestations of TBM. No neurologic signs were seen in rabbits infected with CDC1551 or H37Rv (figure 2C), whereas rabbits infected with HN878 or W4 demonstrated significantly worse signs of disease beginning 3 weeks after infection, including loss of coordination, pareses, and paralysis of the limbs (P<.0001)

CNS pathologic changes in infected rabbitsVery mild focal inflammation of the meninges, with no vasculitis or necrosis and minimal mononuclear leukocytic infiltration, was induced by H37Rv (figure 3A and 3B). CDC1551 induced moderate focal inflammation with no vasculitis or necrosis, some distension of the subarachnoid space, and increased mononuclear leukocytic infiltration (figure 3C and 3D). In contrast, diffuse severe meningitis, with thickening of the leptomeninges and prominent necrotizing vasculitis, was observed after infection with W4 (figure 3E and 3F). In rabbits infected with HN878, the inflammatory infiltrate was most severe, with necrotizing granulomatous meningitis, encephalitis, and vasculitis within the cortex of the brain (figure 3G and 3H). Large granulomas with central necrosis, surrounded by macrophages, lymphocytes, and scattered polymorphonuclear cells, were noted. Numerous acid-fast bacilli were found within the macrophages of the necrotizing granulomas (figure 3G, insert)

Figure 3

Histopathologic assessment of rabbits with central nervous system infection 8 weeks after infection. A and B Rabbit infected with H37Rv. Mild focal inflammation is seen within the meninges (arrowheads). There is no evidence of vasculitis (arrow); “V” indicates the vessels. C and D Rabbit infected with CDC1551. Moderate focal inflammation of the meninges is seen, with a higher number of inflammatory cells. E and F Rabbit infected with W4. Diffuse severe meningitis with distension of the subarachnoid space (arrowheads) a large number of inflammatory cells, and necrotizing vasculitis (“V”) is seen. G and H Rabbit infected with HN878. Necrotizing granulomatous meningitis and encephalitis are observed. The insert shows intracellular acid-fast bacilli at the border of the necrotic area (*). Original magnification: ×10 (A, C, E and G), ×40 (B, D, F and H), and ×200 (insert)

Figure 3

Histopathologic assessment of rabbits with central nervous system infection 8 weeks after infection. A and B Rabbit infected with H37Rv. Mild focal inflammation is seen within the meninges (arrowheads). There is no evidence of vasculitis (arrow); “V” indicates the vessels. C and D Rabbit infected with CDC1551. Moderate focal inflammation of the meninges is seen, with a higher number of inflammatory cells. E and F Rabbit infected with W4. Diffuse severe meningitis with distension of the subarachnoid space (arrowheads) a large number of inflammatory cells, and necrotizing vasculitis (“V”) is seen. G and H Rabbit infected with HN878. Necrotizing granulomatous meningitis and encephalitis are observed. The insert shows intracellular acid-fast bacilli at the border of the necrotic area (*). Original magnification: ×10 (A, C, E and G), ×40 (B, D, F and H), and ×200 (insert)

Inflammatory response and clinical course of disease in rabbits infected intrathecally with HN878 or HN878pks1-15::hyg Next, rabbits were infected with HN878 or the PGL-deficient mutant, HN878pks1-15::hyg and were monitored for 8 weeks. By 1 week after infection, we observed a significant difference between the numbers of colony-forming units of HN878 and HN878pks1-15::hyg in CSF (P=.001) (figure 4A). Clearance of HN878pks1-15::hyg progressed faster, and no bacilli were detected from 5 to 8 weeks after infection. CSF leukocytosis induced by HN878pks1-15::hyg was significantly lower throughout the experiment (P=.01) (figure 4B). No TNF was detected in the CSF or plasma of rabbits infected with HN878pks1-15::hyg; fewer neurologic signs and a significantly milder course of disease were noted (figure 4C). Only 1 rabbit infected with HN878pks1-15::hyg developed paresis of the hind limbs during week 8. Histopathologic examination of brain and meninges of rabbits infected with HN878pks1-15::hyg revealed mild focal meningeal inflammation with no vasculitis (data not shown). Both the inflammation and pathologic changes induced by HN878pks1-15::hyg were less severe, compared with that induced by HN878, even though >4 log10 cfu were observed in the brain (figure 4D)

Figure 4

Inflammatory response and bacillary load in the central nervous system of rabbits infected intrathecally with Mycobacterium tuberculosis strain HN878 or HN878pks1-15::hyg. A No. of M. tuberculosis colony-forming units in the cerebrospinal fluid (CSF) of rabbits infected with HN878pks1-15::hyg (white triangles) (n=9) or HN878 (black triangles) (n=9). Clearance of HN878pks1-15::hyg is much faster than that of HN878 (P=.001) at week 1 after infection. B Leukocyte density in the CSF of rabbits infected with HN878 (dark gray bars) or HN878pks1-15::hyg (checkered bars). HN878-induced leukocyte densities were higher than those induced by HN878 pks1-15::hyg throughout the experiment (P=.01). C Severity of signs of disease (clinical score) in rabbits infected with HN878 (dark gray area) (n=9) or HN878pks1-15::hyg (white area) (n=9). Rabbits infected with HN878pks1-15::hyg showed fewer signs of disease than did those infected with HN878 (P<.0001). Results are expressed as the area under the curve. D Colony-forming units in the brain at 2, 4, and 8 weeks after infection with HN878 (dark gray bars) or HN878pks1-15::hyg (checkered bars). Values are means±SEs

Figure 4

Inflammatory response and bacillary load in the central nervous system of rabbits infected intrathecally with Mycobacterium tuberculosis strain HN878 or HN878pks1-15::hyg. A No. of M. tuberculosis colony-forming units in the cerebrospinal fluid (CSF) of rabbits infected with HN878pks1-15::hyg (white triangles) (n=9) or HN878 (black triangles) (n=9). Clearance of HN878pks1-15::hyg is much faster than that of HN878 (P=.001) at week 1 after infection. B Leukocyte density in the CSF of rabbits infected with HN878 (dark gray bars) or HN878pks1-15::hyg (checkered bars). HN878-induced leukocyte densities were higher than those induced by HN878 pks1-15::hyg throughout the experiment (P=.01). C Severity of signs of disease (clinical score) in rabbits infected with HN878 (dark gray area) (n=9) or HN878pks1-15::hyg (white area) (n=9). Rabbits infected with HN878pks1-15::hyg showed fewer signs of disease than did those infected with HN878 (P<.0001). Results are expressed as the area under the curve. D Colony-forming units in the brain at 2, 4, and 8 weeks after infection with HN878 (dark gray bars) or HN878pks1-15::hyg (checkered bars). Values are means±SEs

When rabbits were infected with a mixture (1:1) of HN878 and HN878pks1-15::hyg viable bacilli persisted in the CSF throughout the experiment (figure 5A). PCR analyses of individual colonies showed that, beginning 2–4 weeks after infection, the predominant strain was HN878, representing >73% of the total colony-forming units in the CSF. In the brain and lungs, HN878 represented 75% and 89% of the total colony-forming units, respectively. Similarly, infection of rabbits with a 1:1 mixture of HN878 and CDC1551 resulted in better control of the CDC1551 strain (figure 5B). In the CSF, brain, and lungs, the predominant strain by 4 weeks after infection was HN878 (100% of colony-forming units in the lung)

Figure 5

Bacillary load in the central nervous system and lungs of rabbits infected intrathecally for 4 weeks with a mixture of Mycobacterium tuberculosis strains HN878 (n=4) and HN878pks1-15::hyg (n=4) (A) or strains HN878 (n=8) and CDC1551 (n=8) (B). Upper panels No. of colony-forming units in the cerebrospinal fluid (CSF), based on polymerase chain reaction analyses of isolates. Middle panels Ratio (%) of HN878 (dark gray bars) versus HN878pks1-15::hyg or CDC1551 (checkered bars) in the CSF at each time point. Lower panels Ratio (%) of HN878 (dark gray bars) vs. HN878pks1-15::hyg or CDC1551 (checkered bars) in the brains and lungs of rabbits at 4 weeks after infection. No CDC1551 was detected in the lungs. Values are means±SEs

Figure 5

Bacillary load in the central nervous system and lungs of rabbits infected intrathecally for 4 weeks with a mixture of Mycobacterium tuberculosis strains HN878 (n=4) and HN878pks1-15::hyg (n=4) (A) or strains HN878 (n=8) and CDC1551 (n=8) (B). Upper panels No. of colony-forming units in the cerebrospinal fluid (CSF), based on polymerase chain reaction analyses of isolates. Middle panels Ratio (%) of HN878 (dark gray bars) versus HN878pks1-15::hyg or CDC1551 (checkered bars) in the CSF at each time point. Lower panels Ratio (%) of HN878 (dark gray bars) vs. HN878pks1-15::hyg or CDC1551 (checkered bars) in the brains and lungs of rabbits at 4 weeks after infection. No CDC1551 was detected in the lungs. Values are means±SEs

Discussion

We have demonstrated that M. tuberculosis clinical isolates are differentially virulent (as defined by severity of disease and/or bacillary load in infected tissues) in the rabbit. We have shown that W4 and HN878, both of which are members of the W/Beijing family of strains, are more virulent than CDC1551 and the control laboratory strain H37Rv. Infection with HN878 or W4 resulted in a persistent bacillary load in the CSF and higher bacillary loads in the brain; a propensity to disseminate to other tissues, such as lung and liver; and prolonged inflammation. Rabbits infected with HN878 or W4 developed severe progressive meningitis with loss of coordination and limb paralyses by 3 weeks after infection. The hypervirulence of HN878 appeared to be related to PGL, which is produced by this strain and W4 but not by CDC1551 or H37Rv [16]. Rabbits infected with HN878pks1-15::hyg which does not produce PGL, showed reduced pathologic changes and attenuated clinical manifestations that were more similar to those induced by CDC1551. Interestingly, rabbits infected with CDC1551, H37Rv, or HN878pks1-15::hyg did not show any (or showed only very limited) neurologic signs, in spite of the presence of a significant bacillary load in the brain

Differences between the virulence of HN878 and that of the mutant HN878pks1-15::hyg strain have recently been reported in the mouse infection model [16]. Infection with HN878pks1-15::hyg was found to be associated with increased survival of the mice and improved host Th1 immunity, compared with that in mice infected with HN878 [16]. However, there was no reduction in the bacillary load in the lungs of mice infected with HN878pks1-15::hyg. Differences between the various animal models often make it difficult to directly compare results from one model to another. However, the different models do provide alternative and complementary approaches to the study of TB. Our model of TBM does not fully reflect the natural history of the infection in humans. However, the model does resemble human TBM clinically and histopathologically and facilitates the study of bacterial factors that promote inflammation, which is especially prominent in the CNS. Mice are more resistant to M. tuberculosis infection than are humans and show no typical lung pathologic changes [19]. However, mice are ideal for the study of the host immune response to the bacilli. Guinea pigs are much more susceptible to M. tuberculosis infection than are humans, and even a single organism kills every animal [20]. Lung granulomas in these animals develop necrosis, as do some granulomas in human lungs. Nonhuman primates mimic human disease very well, probably better than any other model [21], but they are very expensive, and biocontainment facilities for the study of monkeys are few and limited in their capacity. In vitro human monocyte infection, although clearly very reductionist, provides the opportunity to perform selected functional assays and molecular analyses of signaling and gene expression after infection with different strains. To better understand TB, we must select the most relevant model for each experimental question pursued. Taken together, the results we have obtained using different infection models are consistent with the idea that HN878 is more virulent than CDC1551 and that this property is, at least in part, dependent on the production of PGL, a lipid that appears to subvert the activation of the host protective immune response [11, 13, 16, 22]

The results of our mixed-infection experiments suggest that, during the early response to low-dose CSF M. tuberculosis infection, the host cellular response to each strain develops and progresses independently (figure 5)—that is, macrophages infected with HN878 were not activated by their intracellular bacilli, and these macrophages remained permissive to the infection. In contrast, macrophages infected with CDC1551 or HN878pks1-15::hyg were activated by their intracellular bacilli, and these macrophages controlled the growth of the organisms. In addition, macrophages infected with HN878 that left the subarachnoid space and migrated into the brain supported the growth of the bacilli. Macrophages infected with CDC1551 or HN878pks1-15::hyg controlled the growth of the bacilli in the brain. Similarly, macrophages infected with HN878 breached the BBB and seeded the lung more efficiently than did macrophages infected with CDC1551 or the mutant strain. Thus, the innate immune response to infection with one strain did not appear to affect the innate response to the other strain during coinfection. Since we examined only the early events during infection in these studies, we cannot predict the nature of the acquired immune response to mixed infection

Clearly, TNF-α is required for protection against M. tuberculosis [23, 24]. The most clinically relevant example of this requirement is the observation that neutralization of TNF-α by anti-TNF monoclonal antibodies in patients with rheumatoid arthritis can induce reactivation of latent TB [25–27 ]. Our recent studies in mice and in human monocytes in vitro have characterized the immunologic determinants associated with an efficient Th1 protective response against M. tuberculosis infection. In addition to TNF-α, other cytokines and chemokines, including IL-12, macrophage inflammatory protein–3α, chemokine receptor CCR-7, IL-6, and Fas-ligand, appear to be required for a protective response. These mediators are induced suboptimally during the early interaction of HN878 with human or mouse macrophages, compared with the interaction of CDC1551 or HN878pks1-15::hyg with the cells [16, 22]. In the present study, too, rabbits infected with CDC1551 demonstrated an early peak in TNF levels in the CSF and plasma, followed by accelerated clearance of the bacilli from the CSF. However, the presence of TNF was not sufficient to provide protection, as was demonstrated by the fact that both HN878 and W4 induced TNF production in the CSF but were not cleared. Since reagents are not yet available for the measurement of other immune-regulatory mediators in the rabbit, we have not determined their levels. However, we predict that the other cytokines and chemokines associated with optimal Th1 protective responses are not efficiently induced in the rabbit in response to HN878 and W4 infection

Even in the absence of immunologic reagents, the rabbit CNS infection model is useful for the study of pathogenesis and disease, because of the excessive sensitivity of the brain to inflammatory and/or immune perturbations. Although TNF-α is protective against M. tuberculosis infection, it is known to be detrimental to the host when present in excess and for a long time, especially in the brain [8]. The clinical signs observed after intrathecal infection with W4 or HN878 were most likely a result of tissue damage in the brain, associated with prolonged bacillary load, leukocytosis, and persistent TNF production. One of the activities of TNF is to induce coagulation [28] and formation of thrombi, leading to vascular occlusions and necrosis. In the present study, histopathologic examination of the brains of rabbits infected with HN878 or W4 demonstrated directly the impairment of the vasculature. A possible mechanism leading to endarteritis and tissue damage is TNF-induced production of nitric oxide synthase, nitric oxide, and free radicals [29–32 ]. It is also known that proinflammatory cytokines in the CNS alter the BBB and cause enhanced adhesion-molecule expression on the microvasculature [33]. This activation of endothelial cells can lead to production of proinflammatory cytokines, with an influx of monocytes and T cells into the tissues and intensification of the endothelial damage [34, 35]. Injection of TNF-α directly into the neural parenchyma in rats induces formation of inflammatory infiltrates around the injection site [36, 37]. In addition, recombinant human TNF-α injected into the cisterna magna of rabbits causes a reduction in cerebral oxygen uptake and lower cerebral blood flow [29]. When capillaries are presensitized by mycobacterial products, even very small amounts of TNF-α can produce deleterious effects [38]. These activities of TNF may, in part, drive the pathogenesis of TBM, since the most common reason for the cranial nerve palsies, pareses, and paralyses is occlusions of large and small vessels. Since the major sources of TNF in the CNS are newly recruited blood monocytes and resident microglial cells, it is likely that these cells are more sensitive and/or responsive to certain mycobacterial components, especially in the rabbit

In the present study, the PGL-deficient strain HN878pks1-15::hyg did not induce any TNF production in the CSF and caused very limited signs of disease. Thus, this mutant was more like CDC1551 than the parental HN878 strain. Our results demonstrate clearly that M. tuberculosis PGL is involved in the induction of CNS inflammation and tissue damage in infected rabbits and contributes significantly to M. tuberculosis virulence. However, the contribution of PGL notwithstanding, our results and those of others suggest that the virulence of M. tuberculosis is not determined by a single microbial gene but is the outcome of a complex, dynamic interaction between host and microbial properties [39]. For example, a number of recent studies have suggested that other surface lipids, including the sulfolipid SL1 [40, 41], lipoarabinomannan [42], the 19-kDa lipoprotein [43], and others [44], contribute to virulence of mycobacteria

Polymorphisms among M. tuberculosis strains are more extensive than initially anticipated. Several differences between the genomes of the laboratory strain M. tuberculosis H37Rv, which has a long history of passage, and the clinical isolate CDC1551 were recently reported [45]. The genome of another M. tuberculosis clinical isolate (strain 210), which has an IS6110-based restriction fragment length polymorphism pattern identical to that of HN878 [22], is in the process of being sequenced at The Institute of Genome Research, and early results suggest that it differs significantly from the genome of CDC1551 [46]. Many of these genomic differences might contribute to differences in immunogenicity and pathogenicity among M. tuberculosis clinical isolates. In the future, the ability to combine genomic information with pathogenesis studies employing diverse clinical strains, such as those described in the present article, will enable investigators to continue to unravel the molecular basis of M. tuberculosis virulence

Acknowledgments

We thank Nackmoon Sung for preparation of the graphs, Suresh Kumar and Paul Kim for help with the histological figures, Dr. Dorothy Fallows for valuable suggestions, and Sabrina Dalton for secretarial assistance

Appendix

References

1.
Farinha
NJ
Razali
KA
Holzel
H
Morgan
G
Novelli
VM
Tuberculosis of the central nervous system in children: a 20‐year survey
J Infect
 , 
2000
, vol. 
41
 (pg. 
61
-
8
)
2.
Paganini
H
Gonzalez
F
Santander
C
Casimir
L
Berberian
G
Rosanova
MT
Tuberculous meningitis in children: clinical features and outcome in 40 cases
Scand J Infect Dis
 , 
2000
, vol. 
32
 (pg. 
41
-
5
)
3.
Sinner
SW
Tunkel
AR
Approach to the diagnosis and management of tuberculous meningitis
Curr Infect Dis Rep
 , 
2002
, vol. 
4
 (pg. 
324
-
31
)
4.
Schoeman
J
Wait
J
Burger
M
, et al.  . 
Long‐term follow up of childhood tuberculous meningitis
Dev Med Child Neurol
 , 
2002
, vol. 
44
 (pg. 
522
-
6
)
5.
Rich
AR
McCordock
HA
Pathogenesis of tuberculous meningitis
Bull Johns Hopkins Hosp
 , 
1933
, vol. 
52
 (pg. 
5
-
37
)
6.
MacGregor
AR
Green
CA
Tuberculosis of the central nervous system with special reference to tuberculous meningitis
J Pathol Bacteriol
 , 
1937
, vol. 
45
 (pg. 
613
-
45
)
7.
Tsenova
L
Sokol
K
Freedman
VH
Kaplan
G
A combination of thalidomide plus antibiotics protects rabbits from mycobacterial meningitis‐associated death
J Infect Dis
 , 
1998
, vol. 
177
 (pg. 
1563
-
72
)
8.
Tsenova
L
Bergtold
A
Freedman
VH
Young
RA
Kaplan
G
Tumor necrosis factor α is a determinant of pathogenesis and disease progression in mycobacterial infection in the central nervous system
Proc Natl Acad Sci USA
 , 
1999
, vol. 
96
 (pg. 
5657
-
62
)
9.
Flynn
JL
Ernst
JD
Immune responses in tuberculosis
Curr Opin Immunol
 , 
2000
, vol. 
12
 (pg. 
432
-
6
)
10.
Valway
SE
Sanchez
MP
Shinnick
TF
, et al.  . 
An outbreak involving extensive transmission of a virulent strain of Mycobacterium tuberculosis
N Engl J Med
 , 
1998
, vol. 
338
 (pg. 
633
-
9
)
11.
Manca
C
Tsenova
L
Barry
CE
3rd
, et al.  . 
Mycobacterium tuberculosis CDC1551 induces a more vigorous host response in vivo and in vitro, but is not more virulent than other clinical isolates
J Immunol
 , 
1999
, vol. 
162
 (pg. 
6740
-
6
)
12.
Bishai
WR
Dannenberg
AM
Jr
Parrish
N
, et al.  . 
Virulence of Mycobacterium tuberculosis CDC1551 and H37Rv in rabbits evaluated by Lurie’s pulmonary tubercle count method
Infect Immun
 , 
1999
, vol. 
67
 (pg. 
4931
-
4
)
13.
Manca
C
Tsenova
L
Bergtold
A
, et al.  . 
Virulence of a Mycobacterium tuberculosis clinical isolate in mice is determined by failure to induce Th1 type immunity and is associated with induction of IFN‐α/β
Proc Natl Acad Sci USA
 , 
2001
, vol. 
98
 (pg. 
5752
-
7
)
14.
Bifani
PJ
Mathema
B
Liu
Z
, et al.  . 
Identification of a W variant outbreak of Mycobacterium tuberculosis via population‐based molecular epidemiology
JAMA
 , 
1999
, vol. 
282
 (pg. 
2321
-
7
)
15.
Manabe
YC
Dannenberg
AM
Jr
Tyagi
SK
, et al.  . 
Different strains of Mycobacterium tuberculosis cause various spectrums of disease in the rabbit model of tuberculosis
Infect Immun
 , 
2003
, vol. 
71
 (pg. 
6004
-
11
)
16.
Reed
MB
Domenech
P
Manca
C
, et al.  . 
A glycolipid of hypervirulent tuberculosis strains that inhibits the innate immune response
Nature
 , 
2004
, vol. 
431
 (pg. 
84
-
7
)
17.
Constant
P
Perez
E
Malaga
W
, et al.  . 
Role of the pks15/1 gene in the biosynthesis of phenolglycolipids in the Mycobacterium tuberculosis complex: evidence that all strains synthesize glycosylated p‐hydroxybenzoic methly esters and that strains devoid of phenolglycolipids harbor a frameshift mutation in the pks15/1 gene
J Biol Chem
 , 
2002
, vol. 
277
 (pg. 
38148
-
58
)
18.
Sreevatsan
S
Pan
X
Stockbauer
KE
, et al.  . 
Restricted structural gene polymorphism in the Mycobacterium tuberculosis complex indicates evolutionarily recent global dissemination
Proc Natl Acad Sci USA
 , 
1997
, vol. 
94
 (pg. 
9869
-
74
)
19.
Orme
IM
Bloom
BR
Mouse model for tuberculosis
Tuberculosis: pathogenesis, protection, and control
 , 
1994
Washington, DC
ASM Press
(pg. 
113
-
34
)
20.
McMurray
DN
Bloom
BR
Guinea pig model of tuberculosis
Tuberculosis: pathogenesis, protection, and control
 , 
1994
Washington, DC
ASM Press
(pg. 
135
-
47
)
21.
Flynn
JL
Capuano
SV
Croix
D
, et al.  . 
Non‐human primates: a model for tuberculosis research
Tuberculosis (Edinb)
 , 
2003
, vol. 
83
 (pg. 
116
-
8
)
22.
Manca
C
Reed
MB
Freeman
S
, et al.  . 
Differential monocyte activation underlies strain‐specific Mycobacterium tuberculosis pathogenesis
Infect Immun
 , 
2004
, vol. 
72
 (pg. 
5511
-
4
)
23.
Flynn
JL
Goldstein
MM
Chan
J
, et al.  . 
Tumor necrosis factor‐alpha is required in the protective immune response against Mycobacterium tuberculosis in mice
Immunity
 , 
1995
, vol. 
2
 (pg. 
561
-
72
)
24.
Cooper
AM
Flynn
JL
The protective immune response to Mycobacterium tuberculosis
Curr Opin Immunol
 , 
1995
, vol. 
7
 (pg. 
512
-
6
)
25.
Centers for Disease Control and Prevention
Tuberculosis associated with blocking agents against tumor necrosis factor‐alpha—California, 2002–2003
MMWR Morb Mortal Wkly Rep
 , 
2004
, vol. 
53
 (pg. 
683
-
6
)
26.
Keane
J
Gershon
S
Wise
RP
, et al.  . 
Tuberculosis associated with infliximab, a tumor necrosis factor α‐neutralizing agent
N Engl J Med
 , 
2001
, vol. 
345
 (pg. 
1098
-
104
)
27.
Mohan
AK
Cote
TR
Block
JA
Manadan
AM
Siegel
JN
Braun
MM
Tuberculosis following the use of etanercept, a tumor necrosis factor inhibitor
Clin Infect Dis
 , 
2004
, vol. 
39
 (pg. 
295
-
9
)
28.
Bevilacqua
MP
Pober
JS
Majeau
GR
Fiers
W
Cotran
RS
Gimbrone
MA
Jr
Recombinant tumor necrosis factor induces procoagulant activity in cultured human vascular endothelium: characterization and comparison with the actions of interleukin 1
Proc Natl Acad Sci USA
 , 
1986
, vol. 
83
 (pg. 
4533
-
7
)
29.
Tureen
J
Effect of recombinant human tumor necrosis factor‐alpha on cerebral oxygen uptake, cerebrospinal fluid lactate, and cerebral blood flow in the rabbit: role of nitric oxide
J Clin Invest
 , 
1995
, vol. 
95
 (pg. 
1086
-
91
)
30.
Bernard
C
Tedgui
A
Cytokine network and the vessel wall. Insights into septic shock pathogenesis
Eur Cytokine Netw
 , 
1992
, vol. 
3
 (pg. 
19
-
33
)
31.
Yoshizumi
M
Perrella
MA
Burnett
JC
Jr
Lee
ME
Tumor necrosis factor downregulates an endothelial nitric oxide synthase mRNA by shortening its half‐life
Circ Res
 , 
1993
, vol. 
73
 (pg. 
205
-
9
)
32.
Licinio
J
Prolo
P
McCann
SM
Wong
ML
Brain iNOS: current understanding and clinical implications
Mol Med Today
 , 
1999
, vol. 
5
 (pg. 
225
-
32
)
33.
Hickey
WF
Leukocyte traffic in the central nervous system: the participants and their roles
Semin Immunol
 , 
1999
, vol. 
11
 (pg. 
125
-
37
)
34.
Siren
AL
McCarron
R
Wang
L
, et al.  . 
Proinflammatory cytokine expression contributes to brain injury provoked by chronic monocyte activation
Mol Med
 , 
2001
, vol. 
7
 (pg. 
219
-
29
)
35.
Harper
L
Savage
CO
Leukocyte‐endothelial interactions in antineutrophil cytoplasmic antibody‐associated systemic vasculitis
Rheum Dis Clin North Am
 , 
2001
, vol. 
27
 (pg. 
887
-
903
)
36.
Willenborg
DO
Simmons
RD
Tamatani
T
Miyasaka
M
ICAM‐1‐dependent pathway is not critically involved in the inflammatory process of autoimmune encephalomyelitis or in cytokine‐induced inflammation of the central nervous system
J Neuroimmunol
 , 
1993
, vol. 
45
 (pg. 
147
-
54
)
37.
Simmons
RD
Willenborg
DO
Direct injection of cytokines into the spinal cord causes autoimmune encephalomyelitis‐like inflammation
J Neurol Sci
 , 
1990
, vol. 
100
 (pg. 
37
-
42
)
38.
Rook
GA
Attiyah
RA
Foley
N
The role of cytokines in the immunopathology of tuberculosis, and the regulation of agalactosyl IgG
Lymphokine Res
 , 
1989
, vol. 
8
 (pg. 
323
-
8
)
39.
Casadevall
A
Pirofski
L
Host‐pathogen interactions: the attributes of virulence
J Infect Dis
 , 
2001
, vol. 
184
 (pg. 
337
-
44
)
40.
Domenech
P
Reed
MB
Dowd
CS
Manca
C
Kaplan
G
Barry
CE
3rd
The role of MmpL8 in sulfatide biogenesis and virulence of M. tuberculosis
J Biol Chem
 , 
2004
, vol. 
279
 (pg. 
21257
-
65
)
41.
Converse
SE
Mougous
JD
Leavell
MD
Leary
JA
Bertozzi
CR
Cox
JS
MmpL8 is required for sulfolipid‐1 biosynthesis and Mycobacterium tuberculosis virulence
Proc Natl Acad Sci USA
 , 
2003
, vol. 
100
 (pg. 
6121
-
6
)
42.
Chan
J
Fan
XD
Hunter
SW
Brennan
PJ
Bloom
BR
Lipoarabinomannan, a possible virulence factor involved in persistence of Mycobacterium tuberculosis within macrophages
Infect Immun
 , 
1991
, vol. 
59
 (pg. 
1755
-
61
)
43.
Noss
EH
Pai
RK
Sellati
TJ
, et al.  . 
Toll‐like receptor 2‐dependent inhibition of macrophage class II MHC expression and antigen processing by 19‐kDa lipoprotein of Mycobacterium tuberculosis
J Immunol
 , 
2001
, vol. 
167
 (pg. 
910
-
8
)
44.
Smith
I
Mycobacterium tuberculosis pathogenesis and molecular determinants of virulence
Clin Microbiol Rev
 , 
2003
, vol. 
16
 (pg. 
463
-
96
)
45.
Fleischmann
RD
Alland
D
Eisen
JA
, et al.  . 
Whole‐genome comparison of Mycobacterium tuberculosis clinical and laboratory strains
J Bacteriol
 , 
2002
, vol. 
184
 (pg. 
5479
-
90
)
46.
Gutacker
MM
Smoot
JC
Migliaccio
CA
, et al.  . 
Genome‐wide analysis of synonymous single nucleotide polymorphisms in Mycobacterium tuberculosis complex organisms: resolution of genetic relationships among closely related microbial strains
Genetics
 , 
2002
, vol. 
162
 (pg. 
1533
-
43
)
Presented in part: Keystone Symposia—Tuberculosis: Integrating Host and Pathogen Biology, Taos, New Mexico, 25–30 January 2003 (abstract 008)
Financial support: National Institutes of Health (grant AI 054338 to G.K.); Stony Wold Foundation (research grant)