Materials and Methods

M. tuberculosis strains. Thirty-two isolates were obtained from the collection of TB strains generated from the population-based molecular epidemiology study performed in Arkansas from 1996 to 2003. Two clinical isolates, TB282 and TB284, from an epidemiologic study of TB transmission in homeless shelters in Los Angeles were also selected [5]. Isolate TB282 is strain 210. Strain designation was blinded in all experiments.

Isolates were grown for 7 days, at 37°C, in Middlebrook 7H9 broth supplemented with Bacto Middlebrook ADC Enrichment (Difco), in 5% CO₂, with daily agitation. From these cultures, stocks of 1×10⁸ bacilli/mL were prepared and stored at −70°C until used. All procedures involving viable organisms were performed in a biosafety cabinet in a biosafety level 3 laboratory.

THP-1 cells. THP-1 cells, a human monocyticlike cell line, were grown in RPMI 1640 (Gibco BRL) supplemented with 10% fetal calf serum, 25 mmol/L HEPES, and 2 mmol/L glutamine, at 37°C, in 5% CO₂. Cells were subcultured every third day, at an initial density of 2×10⁵ cells/mL. THP-1 cells were differentiated and activated as adherent macrophages by the addition of 100 nmol/L PMA (Sigma) and recombinant human interferon (IFN)-γ (100 U/mL, specific activity 2×10⁷ U/mg; Endogen) for 1–3 days, at 37°C, in 5% CO₂, until used in experiments.

Measurement of M. tuberculosis association and intracellular growth. Adherent and activated THP-1 cells (2×10⁵ cells/well) were washed twice in a binding medium (138 mmol/L NaCl, 8.1 mmol/L Na₂HPO₄, 1.5 mmol/L KH₂PO₄, 2.7 mmol/L KCl, 0.6 mmol/L CaCl₂, 1 mmol/L MgCl₂, 5.5 mmol/L d-glucose) and acclimated for 10 min before the addition of M. tuberculosis at a ratio of 50 mycobacteria to 1 macrophage. Tubercle bacilli and macrophages were gently rocked for 1 h at 37°C, in 5% CO₂, and then incubated for 2 h. After incubation, monolayers were washed twice with PBS to remove extracellular bacteria. Fresh complete medium was added at 24-h intervals, for 7 days after infection. On days 0–7, cell viability was assessed by trypan blue exclusion staining; this involved a 10-min treatment with 0.25% trypsin, to detach macrophages, followed by vital staining, to estimate the number and viability of cells.

For the binding experiments, 8-well Nunc Lab-Tek Chamber Slides were used according to a method described elsewhere [6]. Binding was assessed at 3 h after infection and at 24 h after infection. For calculations of the mean ± SD number of bacilli bound to macrophages, we multiplied the percentage of THP-1 cells associated with organisms by the total number of THP-1 cells used in the experiment and then divided the number of organisms recovered in the lysate at 3 h after infection by this figure.

To determine the number of M. tuberculosis colony-forming units, supernatants were aspirated and monolayers were lysed with 0.5% Triton X. Both cell lysates and supernatants were rigorously mixed by use of an ultrasonic waterbath, were serially diluted, and were plated, in triplicate, on Middlebrook 7H11 agar plates. The number of colonies was counted after 3–4 weeks of incubation. Measurements of colony-forming units were taken at 3 h after infection (on day 0) and at 1–7 days after infection.

Cytokine determinations. Supernatants from solutions containing infected macrophages were aspirated after macrophages were associated with the strains for 3 h (on day 0) and at days 1–7 after infection, were frozen at −70°C, and then were assayed by use of commercial ELISA kits, according to the manufacturer's instructions. Tumor necrosis factor (TNF)-α and interleukin (IL)-10 were not detectable in the supernatant when M. tuberculosis was absent.

Statistical analysis. In at least 3 separate experiments, macrophages were infected with each isolate. The mean ± SD was determined for each measurement. Differences between isolates from clusters and unique isolates were analyzed by unpaired t test. To determine differences between measurements from individual isolates, analysis of variance was performed; P ⩽ .05 was considered significant. Correlations between growth rates and cytokine values were calculated by use of Pearson's correlation coefficient.

Results

Isolate characteristics and patient epidemiology. Strains were selected from the isolate collection from the population-based molecular epidemiology study in Arkansas that began in 1996. Clinical and epidemiologic data were obtained for the corresponding patients from the Arkansas Department of Health TB Control Program. We surveyed the IS6110 RFLP patterns and identified 6 that had persisted for ≥2 of the preceding 8 years (table 1). The number of patients with isolates from the same cluster was 3–24. Patients in each cluster had isolates with identical IS6110 RFLP patterns and spoligotypes, and the patients generally were associated with a specific setting—prison, nursing home, homeless shelter—that was considered a high-risk area for TB transmission. In all but 1 cluster, most cases were epidemiologically linked. In the cluster associated with a large urban community, 3 patients had isolates with identical IS6110 fingerprints, but no epidemiological links could be established. Each cluster of isolates demonstrated a different IS6110 RFLP pattern, and the number of IS6110 bands was 10–15. The 2 clusters consisting of isolates with 15-band fingerprints had different RFLP patterns, although they had the same spoligotype. None of the strains was a member of the Beijing family, whereas strain 210 is a member of this family. Three strains from clusters found in Arkansas matched RFLP patterns listed in the National Tuberculosis Genotyping and Surveillance Network database [7] that were found in 2 other states (strain 158, Maryland and Texas; strain 60, California and Texas; strain 25, California and New Jersey). Similarly, strain 210, a cluster of isolates with 21 copies of IS6110, found in Arkansas, was also found in association with clustered cases in multiple states [8] and was ultimately discovered to be disseminated worldwide.

For comparison, we examined 6 strains, collected during the same time period, that caused disease in a single person (table 1). Selection of these strains was based on patient characteristics that indicated that the cases were infectious. Specifically, all patients had both a sputum culture positive for acid-fast bacilli and cavitary disease, were <43 years of age, and had been born in the United States (except for 1 subject, who had been born in Mexico and lived in the United States for the preceding 20 years). The patients who had strains with unique RFLP patterns resided in populated areas, and some worked in settings where they had extensive exposure to others (2 were truck drivers, 1 was a poultry-processing plant worker); the unique RFLP patterns represented by these 6 strains have not been encountered in Arkansas since their initial appearance, and a search of the National Tuberculosis Genotyping and Surveillance Network database showed that strains with these RFLP patterns were not observed at the other 6 sentinel surveillance sites in the United States (1996–2000).

Isolates TB282 and TB284, which were included as controls in all experiments, were from TB patients residing in homeless shelters in Los Angeles and were thought to differ in their ability to spread in a community [4, 5]. Isolate TB282 is synonymous with strain 210, which has been sequenced by the Institute for Genomic Research and is considered to be a hypervirulent strain (it is indistinguishable from isolate HN878 in reference [2]). In this study, TB282 is a prototype of a high-transmission strain from a cluster, and TB284 is a prototype of a unique, low-transmission strain.

Association of strains from clusters and unique strains with THP-1 cells. In previous experiments with TB282 and TB284, the percentage of THP-1 cells binding to ≥1 mycobacteria was ~60% [4], which is comparable to the observations reported by Stokes and Doxsee [6] . The extent of association with PMA- and IFN-γactivated THP-1 cells was essentially the same at 3 h after infection and at 24 h after infection. In the present study, 6 strains representing the persistent clusters and 6 unique strains were investigated for their ability to bind to THP-1 cells. At both 3 h and 24 h after infection, ~60% of the THP-1 cells had ≥1 tubercle bacilli bound to their cell surfaces and/or ingested (there was no attempt to distinguish between the 2 situations) (figure 1). In 1 experiment set, the number of tubercle bacilli bound to the THP-1 cells at 3 h after infection with 1 of the 6 strains from clusters was 3.44–4.5 (mean, 4.06), and the corresponding value for the unique strains was 3.87–5.06 (mean, 4.37); these numbers remained consistent throughout the other experiments. The difference between the 2 groups was not significant; thus, in the very early stage of infection, the THP-1 cells interacted with all strains in a similar manner.

M. tuberculosis growth in THP-1 cells. To study intracellular growth of clustered and unique strains, the number of colony-forming units was determined at 3 h after infection (on day 0) and on days 1–7 after infection of the activated THP-1 cells (figure 2), and growth rates were calculated over the 7-day infection period (table 2). Growth rate was expressed as the slope (m) of the line generated when the log₁₀-cfu values were plotted over the 7-day infection period. After the macrophages were incubated with the isolates for 3 h (day 0) (figure 2), the log₁₀-cfu values for all strains were 5.51–5.84, confirming that the bacillary inoculum of each strain was comparable and that the different strains behaved similarly in the initial binding reaction. On average, between 3 h and 2 days after infection, the increase in colony-forming units was essentially the same for all strains; the rate of growth started to differ between clustered and unique strains, at 3 days after infection, and, at 7 days after infection, the mean log₁₀-cfu value for the unique strains was significantly less than that of the strains from clusters (6.44 vs. 6.81; P ⩽ .004). Table 2 shows the m of the growth rates for the 6 strains from clusters and the 6 unique strains; the unique strains had significantly lower growth rates (m, 0.193 vs. 0.281; P ⩽ .001). Interestingly, the 3 strains found in states besides Arkansas demonstrated the highest growth rates (m, 0.302–0.333) and grew slightly faster than isolate TB282 (strain 210) (m, 0.281).

Next, isolates from other patients belonging to the persistent clusters were tested. Intracellular growth rates of an additional 20 isolates were determined in a single blinded experiment. Isolates belonging to a cluster had similar growth rates (table 2), reinforcing the tenet that isolates with the same IS6110 fingerprint and from the same geographic area are in fact the same strain (clonally related).

In a previous report, we have shown that the cultures of the supernatants from solutions with infected macrophages showed no growth at any of the time points, indicating that the M. tuberculosis was not growing extracellularly [4], and this was confirmed by results for the isolates used in the present study. In addition, throughout the 7-day incubation period, the percentage of viable THP-1 cells (87%–100%) remained constant regardless of the infecting isolate.

IL-10 production by THP-1 cells infected with M. tuberculosis. IL-10 is produced by macrophages after phagocytosis of M. tuberculosis and after binding of mycobacterial lipoarabinomannan. IL-10 antagonizes the proinflammatory cytokine response by down-regulating the production of IFN-γ, TNF-α, and IL-12, thus interfering with host defense against M. tuberculosis. To evaluate the THP-1-cell response to infection with strains from clusters and with unique strains, we determined the levels of IL-10 at 3 h after infection (on day 0) and at 1, 3, 5, and 7 days after infection. As seen in figure 3, there are 2 separate time points for peak concentration of this cytokine. The strains from clusters, including control strain TB282 (data not shown), rapidly induced IL-10 after association with the macrophages, and peak levels (110–40 pg/mL) occurred at 1 day after infection. This was followed by a rapid decrease over the next 2 days and a gradual decline to very low levels by day 7 after infection. The unique strains, including control strain TB284 (data not shown), induced high levels of IL-10 between 3 and 5 days after infection, and 3 of the strains had levels of 145–65 pg/mL. The amount of IL-10 declined during the next 2 days. There was no significant difference between the amount of IL-10 produced by unique strains and clustered strains; however, the induction of IL-10 production by unique strains was delayed by 4 days. Regardless of the number of intracellular organisms, similar amounts of IL-10 were produced by the THP-1 cells. When growth rates were compared to levels of IL-10 at 1 day after infection, there was a high linear correlation (r, 0.87; P = .0057). Therefore, the ability of the strain to grow rapidly in the macrophages was highly correlated with the rapid induction of IL-10 production, when either adherence with surface receptors or phagocytosis occurred.

TNF-α production in THP-1 cells infected with M. tuberculosis. The ability of clustered and unique strains to induce THP-1 cells to secrete TNF-α was determined at 3 h after infection (on day 0) and at 1 and at 2 days after infection (figure 4). Maximal concentration of TNF-α was detected at 2 days after infection, with 5 of the unique strains (those with the slowest growth rates) eliciting the highest levels at this time. The strains with the slower growth rates induced 1216 ± 279 pg/mL of TNF-α, whereas the strains with the faster growth rates induced only 530 ± 85 pg/mL. As a group, the unique strains induced significantly higher levels of TNF-α than did the strains from clusters (P ⩽ .05). Results with isolates TB284 and TB282, prototypes of a unique and a clustered strain, respectively, coincided with those of their respective groups (figure 4). Comparison of the 7-day growth rates with the peak level of TNF-α secretion at 2 days after infection showed a negative linear correlation (r, −0.84; P = .0003). Thus, there is a strong association between the ability to grow rapidly and the ability to suppress TNF-α secretion.

Discussion

The present study demonstrates that the activated THP-1-cell model is reliable and highly reproducible for determination of mycobacterial growth rates and cytokine levels after infection with various M. tuberculosis strains. In addition, clinical strains in persistent RFLP clusters have significantly faster growth rates than do clinical strains with unique RFLP patterns. Remarkably, the strains that have been found in states besides Arkansas— and thus potentially widely disseminated in the United States— demonstrate phenotypic properties similar to those of the hypervirulent strain 210. Together, these findings confirm the hypothesis that strains at an advantage for either transmission or establishment of disease in humans can be readily distinguished from other clinical strains, on the basis of phenotypic differences in the activated THP-1-cell system. This is the first report of phenotypic markers—such as growth rate, rapid induction of IL-10, and suppression of TNF-α—being associated with an epidemiologically relevant phenotype. The significance of this phenotype is further supported by the recent observation that the principal sigma factor sigA, which may regulate the expression of virulence genes, is up-regulated in macrophagegrown isolates of strain 210, in contrast to what occurs in other clinical strains [9].

The cytokine-inducing capacity of a strain appears to be a property of its growth phenotype. TNF-α secretion peaked at 2 days after infection with each strain. Interestingly, infection of the THP-1 cells with the unique strains resulted in induction of significantly greater amounts of TNF-α, compared with infection with the strains in clusters. The TNF-α-induction profile of the unique strains coincides with that reported by Giacomine et al. [10], who showed that, in monocyte-derived macrophages infected with M. tuberculosis H37Rv, TNF-α was secreted rapidly and at high levels (peak at 48 h, 1500 pg/mL) in a sustained fashion [10]. Similar levels of TNF-α in humanderived macrophages were observed by Zhang et al. [5] at 24 h after infection by each of 6 isolates of strain 210 and each of 6 small-cluster or unique clinical isolates, but there were no differences in the levels of TNF-α secretion induced by these isolates [5]. Manca et al. showed that significantly higher levels of TNF-α were elicited from human monocytes after infection with M. tuberculosis strain CDC1551 when compared to H37Rv, with the peak concentration at 24 h [1]. CDC1551 caused a large number of tuberculin skin test conversions in a community outbreak but was not responsible for a persistent cluster of cases. Early expression of TNF-α in the lungs of CDC1551–infected mice was associated with an early granulomatous response and may have contributed to increased survival rates in the CDC1551-infected mice [1]. The ability to induce TNFa is a property of the “slow-growth” phenotype. Most likely, the course of infection with this phenotype is modulated by the rapid and robust TNF-α response, which restricts mycobacterial replication; in contrast, the “rapid-growth” phenotype suppresses TNF-α secretion, likely through induction of a high level of IL-10 production.

The initial control of M. tuberculosis infection is dependent on macrophage activation, including the production of reactive oxygen and nitrogen intermediates and TNF-α by these cells. Few cytokines are known to suppress macrophage function, yet IL-10 has antiinflammatory activities that specifically deactivate this type of cell [11, 12]. Not only has IL-10 been shown to suppress the type-1 cytokine response to M. tuberculosis infection [13, 14] but also it down-regulates release of TNF-α from macrophages, and its inhibitory effect is dependent on its concentration [11, 12]. In the study by Zhang et al. [5], peak values of IL-10 induced by isolates of strain 210 and by 6 small-cluster or unique clinical isolates were observed at 24 h after infection, with the values being essentially the same. However, secretion of IL-10 in THP-1 cells infected with unique strains was delayed for 5 days after infection. This suggests that the early interaction between persistent cluster strains and THP-1 cells, during association or phagocytosis, triggers or blocks a mechanism which results in a rapid antiinflammatory response from the infected cells, thereby offering an increased chance of survival of persistent cluster strains.

It will be important to further explore the virulence of these strains in the murine model of TB. We predict that, compared with unique strains, strains from persistent clusters will grow to higher bacillary concentrations in the tissues and will produce accelerated mortality in the mice, thus confirming that strains from persistent clusters have greater virulence. This assumption is based on the study by Manca et al., in which clinical isolate HN878 (which is a 210 strain) was shown to be hypervirulent because, compared with other clinical isolates, it caused the unusually early death of infected immunocompetent mice and failed to stimulate Th1-type immunity [2]. In addition to the assessment of the virulence of these strains in mice, it will be worthwhile to identify the receptors involved in binding and the mechanism(s) by which these strains alter cell-signaling systems, which provides strains from clusters with a survival advantage over other isolates.