Phosphodiesterase 4 Inhibition Reduces Innate Immunity and Improves Isoniazid Clearance of Mycobacterium tuberculosis in the Lungs of Infected Mice

Tuberculosis (TB) caused by Mycobacterium tuberculosis (Mtb) is one of the leading infectious disease causes of morbidity and mortality worldwide. Though current antibiotic regimens can cure the disease, treatment requires at least six months of drug therapy. One reason for the long duration of therapy is that the currently available TB drugs were selected for their ability to kill replicating organisms and are less effective against subpopulations of non-replicating persistent bacilli. Evidence from in vitro models of Mtb growth and mouse infection studies suggests that host immunity may provide some of the environmental cues that drive Mtb towards non-replicating persistence. We hypothesized that selective modulation of the host immune response to modify the environmental pressure on the bacilli may result in better bacterial clearance during TB treatment. For this proof of principal study, we compared bacillary clearance from the lungs of Mtb-infected mice treated with the anti-TB drug isoniazid (INH) in the presence and absence of an immunomodulatory phosphodiesterase 4 inhibitor (PDE4i), CC-3052. The effects of CC-3052 on host global gene expression, induction of cytokines, and T cell activation in the lungs of infected mice were evaluated. We show that CC-3052 modulates the innate immune response without causing generalized immune suppression. Immune modulation combined with INH treatment improved bacillary clearance and resulted in smaller granulomas and less lung pathology, compared to treatment with INH alone. This novel strategy of combining anti-TB drugs with an immune modulating molecule, if applied appropriately to patients, may shorten the duration of TB treatment and improve clinical outcome.


Introduction
Tuberculosis (TB) remains a major global public health problem: Approximately 2 billion people are currently infected, with over 9 million new TB cases and 1.3 million deaths occurring each year [1]. Although the majority of individuals infected by Mycobacterium tuberculosis (Mtb) do not develop primary active disease, many remain latently infected with a potential for reactivation later in life [2]. Reactivation of latent TB forms a large potential source of new infections. TB chemotherapy is generally efficacious, but current drug regimens require a minimum of six months to achieve cure. Moreover, treatment is often associated with drug toxicities and drug-drug interactions, especially in HIV co-infected patients receiving antiretroviral therapy [3]. If treatment is interrupted or inadequate, the risks of relapse and/or development of drug resistance are high. The availability of effective shorter drug regimens for treatment of TB would greatly reduce these risks and would contribute significantly to improving TB control.
The long duration of therapy required for cure has been attributed to the presence of phenotypically heterogeneous populations of bacilli in infected tissues, including subpopulations of organisms that are not responsive to antibiotics [4][5][6][7]. These bacillary subpopulations are thought to be in a slow or nonreplicating state, rendering them less responsive to the action of anti-TB drugs, many of which were selected for the ability to kill actively growing bacilli [2,8]. Non-replicating persistence of Mtb has been associated with various environmental conditions present in infected host tissues, such as low oxygen tension and limited nutrient availability [9][10][11]. In addition, several studies have implicated reactive nitrogen intermediates (RNI), including nitric oxide (NO), produced by activated macrophages, as an important environmental cue directing the shift of Mtb towards dormancy [12][13][14].
Thus, the host immune response may, to some extent, hinder TB treatment by driving a subpopulation of the bacilli into an altered metabolic state, in which they are less responsive to antibiotic killing. Some evidence to support this hypothesis is available from clinical trials using immune modulators as adjunctive therapy in treatment of pulmonary TB [15,16]. Several reports have described accelerated sputum culture conversion in pulmonary TB patients receiving adjunctive corticosteroids together with antibiotics in comparison with patients who received antibiotic treatment alone [17][18][19]. In one recent study, adjunctive etanercept (soluble TNF-receptor) resulted in reduced time to sputum culture conversion and improved clinical signs in HIVinfected TB patients [20].
To investigate whether immune pressure can affect the responsiveness of Mtb to isoniazid (INH) and to test whether immunity can be modified to improve treatment efficacy, we used the mouse Mtb aerosol infection model and evaluated the effectiveness of INH treatment in the presence and absence of an immune modulator. For these experiments, we used a novel phosphodiesterase 4 inhibitor (PDE4i) to modulate the host immune response. PDE4 is a member of a diverse family of enzymes, including 11 distinct isoforms that hydrolyze cyclic AMP (cAMP) or cyclic GMP (cGMP) [21]. The distribution of PDE isoforms varies among different tissues and cell types, facilitating the selective inhibition of specific isoforms, as a means of targeting specific cell types and/or activities. cAMP is an important intracellular second messenger which, at increased levels, has anti-inflammatory and tissue protective effects. The specific PDE4i compound used in this study was CC-3052, which increases intracellular levels of cAMP, leading to down-regulation of TNF-a and other inflammatory cytokines in monocyte/macrophages [22][23][24]. CC-3052 is water soluble and has been shown to be nontoxic, non-mutagenic and non-teratogenic [23]. The drug demonstrated little or no measurable effect on T cell activation [25,26], which is consistent with the finding that activation of T cells is regulated by the isozyme PDE7 but not PDE4 [27]. Thus, CC-3052 would not be expected to cause generalized immune suppression. We hypothesized that immune modulation with CC-3052 would alter the intracellular environment within the infected macrophage, so that a greater proportion of the bacilli would remain in a more metabolically active state and would retain their responsiveness to INH. The effects of CC-3052 on host global gene expression, induction of cytokines, and T cell activation in the lungs of infected mice were evaluated. Our results suggest that selective modulation of the innate immune response with a PDE4i can favorably alter the kinetics of INH-mediated bacillary killing and may enhance the efficacy of TB drug therapy.

Effect of CC-3052 on INH-mediated killing of Mtb in lungs of infected mice
For these studies, we selected two clinical strains of Mtb that differ in their abilities to induce an immune response in mice and in human monocytes [28]. Although the growth curves do not differ substantially in mice, CDC1551 promotes a strong, early Th1 response and is less virulent in mice, while HN878 is less immunogenic and highly virulent, causing earlier death of infected mice. Mice were infected by low dose aerosol infection with Mtb strain CDC1551 or HN878 and treated with CC-3052 from day 1 post-infection; bacillary growth, measured as colony forming units (CFU) in the lungs, was evaluated over time ( Figure 1). The numbers of bacilli in the lungs of CDC1551-infected mice treated with CC-3052 were similar to those observed in untreated control infected mice, both of which stabilized by about 28 days postinfection ( Figure 1A). Throughout the experiment (84 days), there were no significant differences in bacterial loads between CC-3052-treated and untreated mice (P = 0.882), suggesting that CC-3052 treatment of the mice did not accelerate the growth of Mtb in the lungs. INH treatment, initiated on day 14 post-infection, initially reduced the CFU efficiently in both experimental groups, but by 63 days post-infection, the bacterial loads in the lungs of mice treated with INH alone stabilized at about 2 log 10 . These results resemble the characteristic bi-phasic killing curve that was seen in early studies of INH killing in the mouse [29,30]. In contrast, the CFU in the lungs of mice co-treated with INH plus CC-3052 continued to be cleared and at 84 days were significantly lower in numbers than those in mice treated with INH alone (P = 0.016). Plating of lung homogenates from all experimental groups on INH-containing solid medium yielded no colonies, indicating that the residual bacilli in the lungs resulted from antibiotic tolerance rather than acquired resistance to INH. CC-3052 treatment alone had no impact on the kinetics of Mtb growth in broth culture (not shown), demonstrating that the drug has no direct bactericidal/static effect. Mtb-infected mice treated with CC-3052 showed no significant reduction in body weight during the course of the experiment, as compared to untreated control mice (not shown). A similar pattern of bacillary clearance was observed in mice infected with Mtb strain HN878, indicating that the effects of CC-3052 are not Mtb strain-specific ( Figure 1B). Shifting the timing of treatment also did not alter the ability of CC-3052 to impact bacillary clearance in the lungs. When CDC1551-infected mice were treated with CC-3052 and INH beginning on day 14 and day 28 post-infection, respectively, Mtb killing by INH alone did not slow until 84 days post-infection; the CFU numbers continued to decline in co-treated mice up until day 112 ( Figure 1C).
We evaluated the impact of varying the dose of CC-3052 treatment (5 mg/kg, 25 mg/kg, 100 mg/kg) on the kinetics of Mtb growth in the lungs of infected mice and found no differences in the CFU over 28 days (not shown). Importantly, administration of high dose CC-3052 (100 mg/kg), which is above the pharmacologically active dose (Celgene Corporation, unpublished data), did not adversely affect the ability of the animals to control Mtb growth in the lungs.

Effect of CC-3052 on histopathology of Mtb-infected mouse lungs
Specimens of lung tissue were collected from CDC1551-infected mice treated with INH from day 14 and/or CC-3052 from day 1; these were prepared for histopathologic and immunohistologic analysis. Examination of H&E stained sections from lungs collected during the first 4 weeks of infection showed minimal differences among the different experimental groups (not shown). At 42 days post-infection, the lungs from CC-3052-treated mice showed somewhat less organized granulomas than those in control untreated mice, but the differences were not striking ( Figure 2). The lesions in the lungs of untreated mice contained small aggregates of macrophages (arrows) surrounded by large numbers of lymphoid cells (Figure 2A and B). In the lesions of these mice, CD3 + staining cells were distributed predominantly around the aggregates of macrophage and accounted for a minority of lymphoid cells seen in the granulomas ( Figure 2C). An average of 171626 CD3 + T cells were seen in each 640 field of granuloma. The macrophages contained single or small clusters of acid-fast bacilli (AFB) (Figure 2D arrows). By comparison, in the granulomas of CC-3052-treated mice, more diffuse macrophage aggregates and more evenly dispersed lymphoid cells were seen ( Figure 2E and F). The number of CD3 + T cells was similar to that seen in control untreated mouse lungs, with an average of 156622 CD3 + T cells per 640 field of granuloma ( Figure 2G). In CC-3052-treated animals too, the macrophages contained single or small clusters of AFB ( Figure 2H). In contrast, the lungs of mice treated with INH alone contained condensed small granulomas with many lymphoid cells ( Figure 2I), with few CD3 + T cells ( Figure 2J) and few AFB (not shown). CC-3052 plus INH treatment at 42 days resulted in even fewer and smaller granulomas ( Figure 2K and L), compared to those of mice treated with INH alone, and a further reduction in the numbers of visible AFB (not shown).
Morphometric analysis of the extent of the granulomatous infiltration in the lungs of infected mice showed differences in the number of granulomas per cm 2 Figure S1). The average area of the lung parenchyma occupied by granulomas was: controls, 25 Figure S1).

Effect of PDE4 inhibition by CC-3052 treatment on host gene expression in the lungs of Mtb-infected mice
The impact of PDE4 inhibition on gene expression in the lungs of Mtb-infected mice was evaluated by Affymetrix DNA array. Mice were infected with CDC1551, with and without CC-3052 treatment initiated on day 1. Total RNA was isolated from the lungs at 14, 21, 28, and 42 days post-infection to follow temporal changes in global gene expression. At 14 days post-infection, relatively few genes (45) were differentially expressed in response to infection in the absence of PDE4i, based on Significance Analysis of Microarrays (SAM) with a cut-off of $1.5-fold differences using Partek Genomics Suite ( Table 1). The number of differentially expressed genes increased by 21 days (767) and was maximal at 28 days (1,130), declining again by 42 days (693) post-infection (Table 1). This temporal pattern is consistent with what is known about the host immune response in murine pulmonary TB infection, which involves initial recruitment of leukocytes to the site of infection and activation of the pro-inflammatory response, followed by dampening of the early response, as granulomas are established and the infection matures into a chronic steady-state [31,32]. The functional categories of host genes most affected by Mtb infection included innate immune response, T cell activation, antimicrobial response, apoptosis and metabolism, based on analysis by Pathway Express (Figure 3B), consistent with a previously published study of the murine transcriptional response to Mtb [33].
Treatment of infected mice with CC-3052 affected the expression of a accumulative total of 186 genes over the course of infection, of which 164 were down-regulated and 22 were up-regulated ($1.5 fold), as compared to no treatment (data not shown). The temporal changes in gene expression induced by CC-3052 were similar to those due to infection, with relatively minor changes at 14 days (4 genes up-regulated; 33 genes down-regulated), increasing effects at day 21 (20 up; 114 down) and day 28 (5 up; 158 down), and fewer genes differentially expressed at 42 days (15 up; 28 down) post- infection ( Table 1). The numbers of genes differentially expressed by Mtb infection alone or CC-3052 treatment at 21 and 28 days postinfection were compared using Ingenuity Pathway Analysis (IPA), which includes only annotated genes. Of the genes differentially expressed in response to CC-3052 treatment, 86.6% (116/134) and 95.7% (156/163) overlapped with those whose expression was affected by Mtb infection ( Figure 3A). Many of the genes differentially expressed by CC-3052 treatment were involved in the host innate immune response, for example: innate immunity (cd68, clec4e, clec5a, clec7a, clec12, ly86, irg1, tlr7, c1qa, c1qb, c1qc), inflammatory response (ccl3, ccl8, cxcl5, cxcl9, cxcl10, ccr5, cxcr6, ifn-c, il-1a, il-1b, tnf-a, tnfaip2, tnip3), immune regulation (aoah) and apoptosis (bub1) ( Table S1). In general, treatment with the PDE4i resulted in down-modulation of a subset of the same genes that were up-regulated by infection. The genes that were differentially expressed in response to CC-3052 treatment belonged to pathways regulating cytokine-cytokine receptor interaction, Toll-like receptor signaling, complement and coagulation cascades, graft-versus-host disease, T cell receptor signaling, and apoptosis ( Figure 3B).
Quantitative PCR (qPCR) was used to confirm the differential expression of a subset of genes associated with the proinflammatory response to Mtb infection ( Figure 3C). TNF-a, IL-1b, NOS2 and IFN-c play a central role in the activation of macrophages and are essential for host control of Mtb infection [32,34]. TNF-a, IL-1b and NOS2 are produced by macrophages in response to infection with Mtb, while IFN-c is produced by dendritic cells and various lymphoid cells, including lymphoid cells of the innate immune response. Using the same mRNA pools from the CC-3052 treated and untreated mice at 21 and 28 days postinfection, we evaluated the impact of PDE4i treatment on the expression of these genes. We found that Mtb infection significantly increased the expression of tnf-a, 2.49/3.01; ifn-c, 2.45/4.11; il-1b, 2.76/3.43; and nos2, 2.69/5.98 at 21 and 28 days post-infection respectively (differential expression $1.5 fold, P,0.05). In contrast, CC-3052 treatment resulted in decreased gene expression: tnf-a, 21.63/21.50; ifn-c, 21.64/21.54; il-1b, 21.67/ 21.58; and nos2, 21.68/21.84 at 21 and 28 days post-infection respectively ( Figure 3C). In addition, we used qPCR to compare the relative expression of a number of other genes which are known to be regulated by TNF-a, including those encoding chemokines (cxcl5, cxcl10), TNF-a induced protein 2 (tnfaip2), and argininosuccinate synthetase 1 (ass1) [35,36]. All of these genes were up-regulated in response to Mtb infection and downregulated by CC-3052 treatment at both 21 and 28 days postinfection, confirming that TNF-a is a target of PDE4 inhibition. We also examined the expression levels of a number of genes which are known to play important roles in innate immunity, including Toll-like receptor 7 (tlr7), serum amyloid A3 (saa3), immune responsive gene (irg1), and lymphocyte antigen 86 (ly86) [33,[37][38][39][40][41]. These genes were similarly up-regulated by Mtb infection and down-regulated by CC-3052 treatment ( Figure 3C).
One potential explanation for the impact of CC-3052 treatment on global gene expression in the infected mice may be that PDE4 inhibition is causing changes in the relative distributions of cell types in the lungs. To test this possibility, single cell suspensions were prepared from the lungs at 28 days post-infection and used to analyze the cellular composition of the tissues by flow cytometry.
Results of this analysis showed that treatment with CC-3052 did not affect the relative abundance of CD4 + and CD8 + T cells and monocyte/macrophages (CD14 + ) in the lungs ( Figure S2A and B). In addition, by staining the cell suspensions for intracellular IFN-c and analyzing by flow cytometry, we showed that the percentages of CD4 + and CD8 + T cells producing IFN-c in the lungs were unaffected by CC-3052 treatment ( Figure S2C).

Effect of CC-3052 treatment of mice on Mtb-induced spleen T cell activation
The impact of CC-3052 on Mtb-specific activation of CD4 + and CD8 + T cells in chronically infected mice (42 days), at the time when acquired immunity is fully established, was evaluated. Single cell suspensions prepared from spleen were stimulated ex vivo with ConA or Mtb CDC1551 for 6 days, and T cell proliferation was measured by the carboxyfluorescein succinimidyl ester (CFSE) flow cytometry assay. The capacity of CD4 + and CD8 + T cell to proliferate in response to ConA or CDC1551 stimulation was similar in spleen cells from the infected mice with or without CC-3052 treatment (Figure 4). Our observations suggest that CC-3052 does not interfere with T cell activation, consistent with published reports [25,26].

Effect of CC-3052 treatment on monocyte/macrophage activation
To evaluate the impact of CC-3052 on monocyte/macrophage activation, the ability of CD14+ cells to produce TNF-a in response to ex vivo stimulation was assessed by flow cytometry. Single cell suspensions from spleens of untreated and CC-3052-treated mice were collected at 21 days post-infection and stimulated ex vivo for 6 h by PPD or LPS. The percentage of CD14 + cells producing TNF-a was significantly lower in spleens from CC-3052 treated mice than those from untreated infected control mice following ex vivo stimulation with PPD (P = 0.05) or LPS (P = 0.04) ( Figure 5A and B). These data confirm that CC-3052 targets monocyte/macrophage function and support the results of the gene expression profiling, indicating that treatment of Mtb-infected mice with the PDE4i causes modulation of innate immunity. Taken together, the results of the ex vivo stimulation experiments suggest that CC-3052 modulates the innate immune response, without affecting T cell activation, indicating that the drug is not generally immune suppressive, as previously described [25,26].
Effect of anti-TNF-a antibody (TN3-19.12) on the mouse response to Mtb infection As a comparison with the selective inhibition of monocyte/ macrophage TNF-a induced by treatment of mice with CC-3052, we evaluated the impact of treatment with a monoclonal anti-TNFa neutralizing antibody (TN3-19.12) on Mtb infection of mice [42]. Mice infected with CDC1551 were treated with IgG1, TN3-19.12 and/or INH starting at 14 days post-infection. In response to treatment with TN3-19.12 alone, Mtb continued to grow logarithmically in the lungs, reaching levels of 9 log 10 CFU at 35 days postinfection, after which none of the mice in this treatment group survived ( Figure 6A). In contrast, the infection was controlled in lungs of untreated mice, achieving a steady state of about 6 log 10 bacilli from 28 days post-infection to the end of the experiment ( Figure 6A). In mice treated with INH, bacillary loads decreased similarly in both the TN3-19.12 and control IgG1 co-treated animals. In addition, treatment of mice with TN3-19.12 alone resulted in progressive weight loss. By 14 days of treatment (28 days post-infection), mice had lost more than 20% of their body weight and were moribund (not shown). The combination of high bacillary burden and severe weight loss in the TN3-19.12-treated animals was associated with rapid mortality (within 3 weeks post-initiation of treatment). In contrast, mice in the other treatment groups gained weight similarly over time. The expression levels of genes encoding TNF-a, IFN-c and NOS2 in the lungs of mice treated with TN3-19.12 at 28 days post-infection, were significantly reduced compared to the control IgG1 treated animals ( Figure 6B).

Impact of TN3-19.12 on gross appearance and histopathology of CDC1551-infected mouse lungs
The effect of TNF-a inhibition with TN3-19.12 on gross lung appearance was compared to the effect of inhibition with CC-3052. Lungs from CDC1551-infected mice treated from day 14 with IgG1 (control), CC-3052, or TN3-19.12 with or without INH  Table S1. PCR was performed with the same RNA samples in duplicate independently from 4 mice per experimental group. Results of the qPCR are represented as mean 6 SD. The gene ID numbers and gene definitions may be viewed at http://www.ncbi.nlm.nih.gov. doi:10.1371/journal.pone.0017091.g003 co-treatment were isolated at 35 days post-infection. The lungs harvested from CC-3052 treated mice were slightly enlarged compared to those from IgG1 control mice ( Figure 6C). In contrast, the lungs from TN3-19.12-treated mice were strikingly enlarged, with numerous large granulomatous lesions not seen in the IgG1 control or CC-3052-treated mouse lungs ( Figure 6C).  mice were somewhat less compact, with many macrophages and more evenly distributed lymphocytes than those in control animals ( Figure 7B and E). In contrast, mice treated with TN3-19.12 had large granulomas with many polymorphonuclear neutrophils (PMNs) ( Figure 7G) and extensive necrosis ( Figure 7H). Extremely high numbers of AFB were detected in the granulomas of these mice ( Figure 7I). By comparison, co-treatment with TN3-19.12 plus INH resulted in smaller, well-organized lung granulomas, with higher numbers of lymphocytes and few to no AFB ( Figure 7J, K, and L).

Discussion
We have shown that modulation of the immune response in CDC1551 or HN878-infected mice with the PDE4i CC-3052 improves INH-mediated Mtb killing in the lungs. This was demonstrated by the clearance of Mtb in INH plus CC-3052 co-treated animals, which continued to decline up to the end of the infection, in contrast with clearance during the first 2 months of treatment in INH only treated mice, where Mtb killing reached a plateau before the end of the experiment. Thus, while the rate of bacillary clearance was similar in both treatment groups, the extended duration of killing in the CC-3052 co-treated mice resulted in a difference of approximately 2 log 10 in CFU, as compared to INH treatment alone, at later time points. The continued clearance of bacilli is similar to the impact of adding pyrazinamide (PZA), which is more active against non-replicating bacilli, to an INH monotherapy regimen, as described in previous studies on treatment of Mtb-infected mice [30,43]. Although it is not clear whether CC-3052 co-treatment can improve the efficacy of multidrug TB therapy, the use of a host-targeted drug has a considerable benefit because Mtb cannot develop genotypic resistance to a drug that targets host function. In our study, we did not evaluate the ability of co-treatment with CC-3052 plus INH to achieve sterilization of the bacilli. Larger and more extensive experiments to test whether these mice can reactivate the infection upon discontinuation of treatment and general immune suppression studies are needed to answer this question.
Treatment with CC-3052 alone had no impact on Mtb growth in vitro or in vivo, indicating that the drug itself is not bactericidal/ static and does not enhance bacillary replication. Rather, we suggest that down-modulation of the host innate immune response may have dampened some environmental signals that normally drive a subset of bacilli into a less metabolically active state, in which they are not responsive to INH-mediated killing. This would explain why the enhanced efficacy of INH-mediated killing by CC-3052 was manifested only at later time points, when the rate of bacillary clearance mediated by INH alone slowed down, resulting in the bi-phasic killing curve. This hypothesis must now be confirmed in experiments to evaluate the physiologic state of the bacilli in the lungs of mice treated with CC-3052.
The results of the host gene expression studies showed that treatment with CC-3052 resulted in down-regulation of a subset of immune response genes that were up-regulated in response to Mtb infection. These included many genes that are involved in innate immunity and inflammation. This result is consistent with our observations that T cell activation was unaffected in mice treated with CC-3052 alone, and that these mice were fully capable of controlling the infection by 3-4 weeks, the time at which the adaptive immune response is optimally induced. It was interesting to find that the expression of nos2, which is responsible for macrophage production of NO, was significantly lower in the lungs of mice treated with CC-3052 at 21, 28 and 42 days postinfection. As NO has been implicated in the induction of Mtb nonreplicating persistence [12,13,31], it is intriguing to speculate that reduction of NO by CC-3052 treatment may partially explain the improved bacillary clearance. This possibility is supported by the recent identification of a novel antimicrobial primarily active against non-replicating organisms, which targets Mtb dihydrolipamide acyltransferase (DlaT), an enzyme necessary for resistance to RNI-induced stress [44]. It is also of interest to note that Mtb secretes endogenous cAMP, shown to increase the levels of host TNF-a production [45]. Agarwal et al suggest that their findings may provide a mechanism by which Mtb itself may contribute towards eliciting granuloma formation as part of its survival program.
Small molecule PDE4i penetrate the plasma membrane and act to increase cAMP levels within the cells [22], reducing production of TNF-a and other innate immune mediators within the macrophage. Thus, treatment with CC-3052 resulted in limited effects on the innate immune response, rather than generalized immune suppression. In contrast, neutralization of TNF-a using a monoclonal antibody (TN3-19.12) resulted in immune suppression leading to loss of control of bacillary growth in the lungs of infected mice, with extensively increased pathology and greatly accelerated mortality. Anti-TNF-a monoclonal antibodies would not be expected to penetrate the plasma membrane, but rather act extracellular to bind and inactivate any available TNF-a. Since TNF-a is essential for orchestrating both the innate and acquired immune response, the complete neutralization of TNF-a would be expected to lead to a generalized immune suppression in the animal as seen here. Previous studies have shown that depletion of TNF-a by treatment of mice with anti-TNF-a antibodies leads to reactivation of latent Mtb infection [46,47]. Similarly, treatment of patients with inflammatory disorders with the anti-TNF-a monoclonal antibody Infliximab has led to increased incidence of reactivation TB [48].
We also found that the combined effects of CC-3052 and INH resulted in a reduction in lung inflammation and accelerated resolution of lung pathology, as compared with mice treated with INH alone. TB is a striking example of TNF-a as a ''double-edged sword''. TNF-a plays an essential role in controlling Mtb infection and disease severity, but this cytokine can also cause chronic inflammation leading to tissue damage [49,50]. TB patients initiating antibiotic chemotherapy can experience a transiently increased inflammation and worsening of disease in association with an early burst in circulating levels of TNF-a [51][52][53]. From a clinical perspective, the limited impact of CC-3052 on TNF-a production by monocyte/macrophages is an advantage. The normal activation of T cells, which requires some TNF-a, ensured the control of bacillary growth via the contribution of the acquired immune response to protective immunity in CC-3052-treated animals. Thus, use of a PDE4i has enabled us to achieve this critical balance between limiting the negative effects of TNF-a, while retaining sufficient immunity to control infection. Recently, the multiple functions of TNF-a in granuloma formation and control of infection and disease pathogenesis were addressed using a computer-generated model, which predicted that a critical balance of TNF-a bioavailability must be maintained in order to achieve reduction of inflammation while retaining resistance to infection and microbial disease [54]. We propose that CC-3052 is a molecule that modulates TNF-a production, without adversely disrupting this critical balance, and can improve antibioticmediated bacillary clearance. Further studies are needed to evaluate the impact of PDE4 inhibition in a more clinically relevant TB regimen of multidrug therapy.

Chemicals and reagents
All chemicals were purchased from Sigma-Aldrich (St. Louis, MO), unless otherwise stated. PDE4 inhibitor CC-3052 was supplied by Celgene Corporation (Summit, NJ) and prepared as a suspension in sterile 0.2% carboxylmethylcellulose solution (CMC, Sigma). The purified monoclonal anti-TNF-a antibody (TN3-19.12) and isotype control hamster IgG1 antibody were provided by Dr. Robert D. Schreiber (Washington University in St. Louis, MO) [42].

Infection of mice and drug treatment
Clinical Mtb strains CDC1551 and HN878 were grown to midlog phase in Middlebrook 7H9 broth (Difco, Detroit, MI) containing 10% OADC enrichment (Beckton Dickinson, Franklin Lakes, NJ), 0.2% glycerol, and 0.05% Tween 80. Animals were housed in a Biosafety Level 3 pathogen-free facility. In brief, 8 to 10 week-old female B6D2F1 mice (Jackson Laboratory, Bar Harbor, ME) were infected with low dose (approximately 2 log 10 cfu) of Mtb using an aerosol exposure system (CH Technology, Inc., Westwood, NJ) as described previously [55]. CC-3052 (25 mg/kg) was administered 5 days per week by gavage, and INH (50 mg/kg) was delivered ad libitum in drinking water. Neutralizing anti-TNF-a antibody TN3-19.12 (500 mg/mouse) and IgG1 isotype control (500 mg/mouse) were suspended in 300 ml of phosphate-buffered saline and administered by intraperitoneal injection once a week.

CFU and body weight
The numbers of viable mycobacteria were evaluated using colony forming units (CFU) assay. For quantitation of initial inoculum, lungs were isolated from mice 3 hrs post-infection; lung homogenates were plated on 7H11 Middlebrook agar plates, incubated at 37uC for 21 days, and colonies counted. Bacterial loads are presented as the mean value in log 10 6 standard deviation (SD). Body weight was measured every 3 days and expressed in percent (%) change to body weight at baseline (day 0). The Student's t-test was used to determine statistical significance between groups of data.

Histopathology, immunohistology and morphometric analysis
Segments of lung tissue were fixed in 10% buffered formalin (Sigma-Aldrich) and paraffin-embedded. Sections were stained with Hematoxylin-Eosin (H&E) or Ziehl-Neelsen (ZN) acid-fast stain for evaluation of pathology and mycobacterial load, respectively. All photographs of the H&E and ZN stained sections were taken and analyzed using Nikon FX-35DX and microphot-FX. For morphometric analysis of granulomas in mouse lungs, H&E stained lung sections (n = 3-6 per group) were scanned with a PathScan Enabler IV scanner (Meyer Instruments, Houston, TX). SigmaScan Pro 5 software was used to count the number and size of lesions. The extent of lung involvement was calculated using three parameters: the average size of granulomas, the number of lesions per cm 2 tissue, and the percent of the lung sections occupied by granulomas. Morphometric analysis was carried out by an independent single investigator blinded to the source of tissues. Immunohistochemistry staining was done with polyclonal anti-mouse CD3 antibody (Biocare Medical, Concord, CA). The number of CD3 + T cells was counted in 10640 fields by 2 independent investigators blinded to tissue sources.

DNA array and quantitative real time PCR
Global expression analysis of four independent arrays per group was performed using Affymetrix mouse genome ST 1.0 genes chips. For analysis, uninfected mice (baseline), untreated mice, CC-3052treated mice at 14, 21, 28 and 42 days after infection (n = 4 per group) were euthanized. RNA from mouse lungs was isolated with Trizol (Invitrogen), DNase I treated, and further purified using RNeasy miniprep kit (Qiagen). Microarray experiments were performed once or twice and four independent arrays per group were analyzed. Data reduction and analysis of comparison of treatment groups was normalized, and filtered expression files were analyzed using Partek Genomic Suite. SAM (significance analysis of microarray) tool was used for adjusting the P values from a comparison test based on the number of tests performed. Differentially expressed genes were selected with a cutoff value of 61.5 fold change (P,0.05). Gene ontology analysis was performed using Pathway Express (Wayne State University, Detroit, MI). Reverse transcription of total lung RNA for the qPCR was completed with a first-strand cDNA synthesis kit by oligo (dT) priming (Invitrogen), according to the manufacturer's protocol. Real time PCR was performed using the Invitrogen SYBR GreenER qPCR universal module, according to the manufacturer. Primers specific to genes of interest were designed as recommended by PrimerBank (www. primerbank.org). Quantified, purified and diluted PCR product was used to generate external standard curves for each primer pair. Cycle number values were converted to copy number using these curves post-amplification. qPCR of glyceraldehyde-3-phosphate dehydrogenase (gapdh) was always run in parallel with that of the other genes of interest and levels of gene expression measured were normalized to the housekeeping gene gapdh. The complete gene lists are accessible through GEO database, NCBI accession number GSE25313.

Single tissue cell preparation
Single cell suspensions from lung tissue were prepared by incubating minced tissue with collagenase A (1 mg/ml; Boehringer-Mannheim) and DNase I (25-50 U/ml; Sigma-Aldrich) for 1 hour at 37uC. The digested lung was further disrupted by gentle teasing through a cell strainer (BD Bioscience). Spleens were disrupted by teasing the tissue directly through a cell strainer without digestion. Red blood cells were lysed with ACK buffer (GIBCO), washed and resuspended in RPMI 1640 medium supplemented with 10% fetal calf serum (FCS), penicillin/ streptomycin (50 U/50 mg/ml; R10). Total number of cells was determined by microscopic counting using a hemocytometer. Spleen cells were cultured at 37uC at a density of 2.5610 5 /100 ml in 96-well U-bottom plates. The cells were stimulated in R10 with Purified Protein Derivative (PPD, 10 mg/ml final; Staten Serum Institute) or lipopolysaccharide (LPS, 1 mg/ml final; Sigma) and incubated at 37uC for 6 hrs, 24 hrs or 6 days. At each time point, culture supernatant was removed, filtered using 0.22 mm lowbinding microfuge filter tubes (Millipore, Billerica, MA) and stored at 280uC for cytokine assays. Isolated cells were frozen at 280uC in R10 with 10% DMSO for flow cytometric analysis.
Flow cytometric analysis of cell surface markers and intracellular cytokine assay Spleen or lungs were removed from CC-3052 treated or untreated infected mice at the indicated time points and processed as described [56]. Briefly, cells were stained with FITC-anti-CD14, PE-anti-CD4, PercP-anti-CD3, and APC-anti-CD8 monoclonal antibodies (BD Bioscience). For detection of intracellular TNF-a, spleen cells, cultured in the presence of PPD (10 mg/ml final) for 6 hrs, were stained with PE-anti-IFN-c, ACP-anti-TNF-a, FITCanti-CD4, PercP-anti-CD3, APC-anti-CD8 and FITC-anti-CD14 (BD Biosciences, San Jose, CA). Intracellular IFN-c in lung cells was measure without prior in vitro stimulation. Data acquisition and analysis were performed using FACSCalibur (BD Bioscience) and FlowJo software (Tree Star, Ashland, OR) respectively.

Proliferation assay
Spleen cells isolated as described above were stained with carboxyfluorescein succinimidyl ester dye (CFSE; Invitrogen, Carlsbad, CA), according to manufacturer's instructions. 2.5610 5 CFSE-labeled cells/100 ml were plated in 96-well Ubottom plates, and stimulated for 6 days with ConA (5 mg/ml final concentration; Sigma) or heat-killed, sonicated Mtb CDC1551 at a multiplicity of infection of 10 to 1 (MOI 10:1), or left unstimulated. Cells were fixed and stored at 280uC. Cells were stained for surface markers with the following mAbs: PE-anti-CD4, APC-anti-CD8 and PerCP-anti-CD3 (BD Bioscience). Data acquisition and analysis were performed using FACSCalibur (BD Bioscience) and FlowJo software (Tree Star, Ashland, OR). Proliferation was measured as a reduction in CFSE fluorescence intensity.

Statistical analysis
All data are presented as the mean 6 standard deviation (SD). Statistical significance was determined using SAM (significance analysis microarrays) and Student's t-test. A value of P,0.05 was considered significant.

Accession numbers
Detailed information for the genes/proteins from this study can be found at NCBI GenBank database www.ncbi.nih.gov/genbank. The gene name and ID numbers in this publication are listed in Table S1. The ID numbers of TNF-a and IFN-c shown in this publication are NP_038721 and AAI19061, respectively. Table S1 Differentially expressed mouse genes by CC-3052 treatment. Gene expression was analyzed from total RNA of uninfected and Mtb-infected mouse lungs, treated or not treated with CC-3052, at 28 days post-infection. Results are representative of 4 independent arrays per group and presented as relative gene expression. Fold change ($1.5 fold) of differentially expressed genes from arrays was statistically significant based on SAM analysis (P,0.05). The gene name and ID numbers are available at www.ncbi.nih.gov/genbank. (TIF)