Combination of host susceptibility and Mycobacterium tuberculosis virulence define gene expression profile in the host

Authors

  • Martin Beisiegel,

    1. Department of Immunology; Max Planck Institute for Infection Biology, Berlin, Germany
    Search for more papers by this author
  • Hans-Joachim Mollenkopf,

    1. Department of Immunology; Max Planck Institute for Infection Biology, Berlin, Germany
    Search for more papers by this author
  • Karin Hahnke,

    1. Department of Immunology; Max Planck Institute for Infection Biology, Berlin, Germany
    Search for more papers by this author
  • Markus Koch,

    1. Department of Immunology; Max Planck Institute for Infection Biology, Berlin, Germany
    Search for more papers by this author
  • Isabell Dietrich,

    1. Department of Immunology; Max Planck Institute for Infection Biology, Berlin, Germany
    Current affiliation:
    1. Bayer Schering Pharma AG; Global Drug Discovery, Target Discovery-Transgenic & Specialized in vivo Pharmacology, Berlin, Germany Isabell Dietrich, Charité Universitätsmedizin Berlin, Charité Campus Virchow-Klinikum; Pediatric Oncology and Hematology Clinic, Augustenburger Platz 1, 13353 Berlin, Germany
    Search for more papers by this author
  • Stephen T. Reece,

    1. Department of Immunology; Max Planck Institute for Infection Biology, Berlin, Germany
    Search for more papers by this author
  • Stefan H. E. Kaufmann

    Corresponding author
    1. Department of Immunology; Max Planck Institute for Infection Biology, Berlin, Germany
    • Max Planck Institute for Infection Biology, Chariteplatz 1, 10117 Berlin, Germany. Fax: +49-30-28460-501

    Search for more papers by this author

Abstract

Progression and outcome of tuberculosis is governed by extensive crosstalk between pathogen and host. Analyses of global changes in gene expression during immune response to infection with Mycobacterium tuberculosis (M.tb) can help identify molecular markers of disease state and progression. Global distribution of M.tb strains with different degrees of virulence and drug resistance, especially for the immunocompromised host, make closer analyses of host responses more pressing than ever. Here, we describe global transcriptional responses of inducible nitric oxide synthase-deficient (iNOS–/–) and WT mice infected with two related M.tb strains of markedly different virulence, namely the M.tb laboratory strains H37Rv and H37Ra. Both hosts exhibited highly similar resistance to infection with H37Ra. In contrast, iNOS–/– mice rapidly succumbed to H37Rv, whereas WT mice developed chronic course of disease. By differential analyses, virulence-specific changes in global host gene expression were analyzed to identify molecular markers characteristic for chronic versus acute infection. We identified several markers unique for different stages of disease progression and not previously associated with virulence-specific host responses in tuberculosis.

Introduction

The host's fight against tuberculosis (TB) is complex, involving innate and adaptive immune responses, which are highly intertwined 1. The murine model of TB has helped to identify several key factors, which determine course and outcome of TB 2. For example, mutant mouse strains lacking genes encoding IFN-γ 2 or inducible nitric oxide synthase (iNOS) 3, 4 severely suffer from exacerbated TB and fail to control infection with virulent strains of Mycobacterium tuberculosis (M.tb) in lungs. However, our understanding of the processes involved in acute and chronic inflammation during this devastating disease is still limited.

To further complicate matters, several lineages of M.tb have been described and evidence is accumulating for lineage- and strain-specific virulence 5.

Global gene expression analyses during the course of infection have been successfully exploited to characterize immune mechanisms and specific markers of disease progression and outcome. Most studies focus on gene expression patterns in comparison to healthy lung tissue or cell culture systems. Although such approaches successfully determine factors important during the course of infection, factors unique to M.tb infection and TB disease outcome remain elusive.

Using global gene expression, we investigated host responses to infection with the virulent M.tb strain H37Rv in comparison to the attenuated strain H37Ra. Both strains were isolated in 1934 and derived from the same clinical isolate taken in 1905 6, but differ profoundly in virulence. Critical to theses strain differences are mutations in the gene regulator protein PhoP 7, which affects many gene functions in H37Ra. Infection with this strain causes mild inflammatory responses 3, 4 and its attenuation is highly evident in face of marked immunodeficiency. In mice, stimulation with IFN-γ results in the induction of iNOS resulting in production of nitric oxide (NO) and antimicrobial reactive nitrogen intermediates 3. Although H37Rv infection is detrimental in iNOS–/– mice, which suffer from acute inflammation within the first 30 days of infection, they are able to fight infection with H37Ra to a similar degree as WT mice 3, 4. Inhibition of iNOS in human alveolar Mϕ causes reduced antimycobacterial activity 8, patients with active TB exhale abundant NO 9 and iNOS-producing cells are detected in granuloma biopsies 10, suggesting a role of iNOS in human TB as well.

In this study, we focus on global profiles of the interplay between host susceptibility and M.tb virulence, which result in differential course of M.tb infection and outcome of TB. We interrogated dynamic gene expression patterns during establishment of severe infection with virulent H37Rv in resistant WT mice and highly susceptible iNOS–/– mice in comparison to mild infections with attenuated H37Ra in both mouse strains. Hence, gene products differentially expressed in inflammatory processes of acute versus chronic infection of varying severity in TB could be defined.

We discuss in more detail a selected number of differentially regulated gene products in the context of their functional role in TB. These include cytokines, chemokines and their respective receptors, molecules of innate immunity such as complement factors and pentraxin (PTX), molecules of adoptive cellular immunity such as SOCS gene products and effector molecules such as those involved in oxidative burst and iron transport. Our investigations emphasize the complexity of mechanisms underlying infection, pathogenesis and host protection which dictate disease in TB.

Results and discussion

Disease progression and global transcriptional responses after M.tb infection

Consistent with previous studies 4, we observed exponential growth of bacteria in lungs during the first 30 days after low-dose aerosol infection with either H37Ra or H37Rv organisms (Fig. 1). Counts of H37Ra organisms declined in WT and iNOS–/– lungs thereafter, whereas H37Rv organisms continued to multiply exponentially in iNOS–/– lungs, resulting in the death of the animals within a few days. In contrast, WT mice entered a chronic phase of infection with elevated but stable counts of H37Rv organisms in the lungs.

Figure 1.

M.tb burdens and global gene expression changes in WT and iNOS–/– mice infected with strains H37Rv and H37Ra. M.tb burdens as CFU in lungs of C57BL/6 (WT) mice (A) and inducible knockout (iNOS–/–) mice (B) after aerosol infection with M.tb strains H37Rv (Rv) or H37Ra (Ra). Gray arrows indicate analyses of gene expression in whole lungs either over time (d14 versus d30) or in a comparison of H37Rv- to H37Ra-specific gene expression at d14 or d30 p.i. Numbers indicate genes associated with known functions and/or diseases, numbers in parentheses indicate unique ORF. Changes in gene expression over time are also displayed for H37Rv-specific, H37Ra-specific and shared gene expressions. Crosses indicate moribund animals. CFU are displayed as medians with interquartile ranges of groups of five mice.

Global gene expression analyses were performed using RNA extracted from lungs at day (d) 14 post-infection (p.i.) and d30 p.i. with Whole Mouse Genome DNA microarrays. For vigorous analysis, only open reading frame (ORF) displaying twofold or higher changes in gene expression (p-value<0.05) were considered for further analyses as differentially regulated ORF.

We observed dynamic changes in expression patterns in lungs over time in both mouse strains after infection with either H37Ra or H37Rv. Hence, transit from acute to chronic infection with M.tb in WT mice and failure to control M.tb infection in iNOS–/– mice could be interrogated on a transcriptional level. After infecting WT mice with H37Rv, 3098 ORF were regulated over time, whereas 6362 ORF did so after infection with H37Ra (Fig. 1A). In iNOS–/– lungs, 3511 ORF were regulated over time in H37Rv-infected lungs, whereas H37Ra infection resulted in 6950 regulated ORF over time (Fig. 1B).

Numbers of genes with known functions among regulated ORF (Fig. 1) were determined using the Ingenuity Pathway Analysis platform (IPA7, Ingenuity Systems®). Numbers of genes with known functions reflected the tendencies observed for numbers of regulated ORF in WT (Fig. 1A) and iNOS–/– mice (Fig. 1B).

Further analyses included virulence-specific changes in gene expression over time by comparing H37Rv- and H37Ra-induced gene regulation in infected lungs. In WT mice, 548 genes displayed H37Rv-specific changes in expression, whereas 1455 genes displayed equal tendencies in expression changes over time in H37Rv- and H37Ra-infected mice. H37Ra-specific changes in expression were evident for 2324 genes (Fig. 1A). In iNOS–/– mice, 1251 genes displayed H37Rv-specific and 3115 genes displayed H37Ra-specific changes in expression, whereas 888 genes shared similar changes in gene expression over time (Fig. 1B). To further refine analyses, differential gene regulation in H37Rv- and H37Ra-infected lungs was investigated at d14 p.i. and d30 p.i. (Supporting Information Table 1). In WT lungs, 782 ORF were differentially regulated at d14 p.i., whereas 2345 regulated ORF were detected at d30 p.i. (Fig. 1A). In iNOS–/– lungs, 2063 regulated ORF were determined at d14 p.i. and 5006 regulated ORF at d30 p.i. (Fig. 1B). Data sets of H37Rv-specific changes in gene expression over time were compared against H37Rv-specific gene regulation at d14 p.i. (early infection phase) and d30 p.i. (established infection phase). Thus, comprehensive insights into virulence-associated genes were obtained emphasizing (i) the adaptability of the host response to differential virulence of M.tb and (ii) the compensatory capacity of the host to immunodeficiency in TB (Table 1).

Table 1. Selected transcriptional markers associated with virulence of M.tba)
inline image
inline image
inline image
inline image
inline image

Transcriptional markers of virulence during disease progression

Host–pathogen interaction

Pulmonary infection with M.tb elicits profound innate and adaptive immune responses 1. The role of pathogen recognition through TLR, especially through TLR2, has been implicated in TB 11. In our analyses, TLR2 induction was most prominent in iNOS–/– mice regardless of M.tb virulence and remained most elevated in H37Rv-infected lungs (Table 1). Induction of TLR-6 transcripts was virulence specific for both mouse strains and TLR-1 levels increased only in iNOS–/– lungs.

Components and receptors of the complement system were most prominently upregulated in H37Rv-infected iNOS–/– lungs at d30 p.i. and expression of complement component C3, reported to promote phagocytosis by alveolar Mϕ 12, was downregulated during H37Ra infection over time. Complement-facilitated uptake of M.tb evades stimulation of the oxidative burst and hence allows quiescent entry into phagocytes 13. In addition, many C-type lectins such as DC-SIGN, mannose receptor and Dectin-1, recognize specific components of M.tb 14. In our study, only Mincle and MDL-1, both C-type II lectins, exhibited a modestly increased, virulence-specific expression. Two NK cell-associated lectins, Clec1a and Clec1b, together with the mannose receptor Mrc-2, were downregulated over time and exhibited increased expression at d30 p.i. in H37Ra-infected lungs. Formyl peptide receptors and genes associated with LPS-recognition such as CD14, CD6, MST1R and LPS-induced TNF-α factor (LITAF) were strongly induced in H37Rv-infected iNOS–/– lungs. LITAF is connected to MyD88-dependent signalling 15, implying a possible role of LITAF in highly susceptible MyD88-deficient mice during M.tb infection. Elucidation of the role of LITAF in TB could shed further light on MyD88-dependent processes in TB.

Several Fc receptors and genes involved in antigen presentation via the MHC-I and MHC-II pathway were H37Rv-specifically induced in iNOS–/– mice. Only modestly increased transcription of Fc receptors and genes of the MHC-I antigen presentation pathway were detected. Transcription of the inhibitory Fc receptor FcGR2B was induced in WT and iNOS–/– mice which is associated with susceptibility to M.tb 16. Since activating Fc receptors were induced in a similar or stronger fashion, inhibitory FcGR2B signalling effects could be compensated for on a cellular level.

Leukocyte immunoglobulin-like receptor A5, regulated by TNF-α, IL-10 and IFN-γ, is associated with rheumatoid arthritis and production of proinflammatory cytokines 17. It was highly expressed in an H37Ra-dependent fashion in iNOS–/– lungs. The role of leukocyte immunoglobulin-like receptor A5 in resistance against H37Ra in these mice remains to be ascertained.

Proteins with CARD domains were strongly transcribed in H37Rv-infected iNOS–/– mice and moderately in WT mice. The critical role of these intracellular receptors and adaptor proteins of the TLR signalling pathway in innate immune defense against invading bacteria and their role in TB are currently emerging 1, 18.

The long PTX, PTX3 is a circulating pattern recognition receptor that acts as an opsonin and facilitates phagocytosis 19. We detected strong, virulence-specific induction of PTX3 transcription exclusively in iNOS–/– mice. Levels of PTX3 are linked to severity of inflammatory syndrome after infection in humans and this receptor has been suggested as a direct indicator of inflammatory processes with apparent selectivity for different pathogens 20. Moreover, elevated serum levels of PTX3 and IP-10 (CXCL-10) have been associated with TB treatment failure 21.

Adaptive immune response

The essential role of potent Th1-type CD4+ T-cells in TB is well accepted 1 and differences in gene expression of CD4 were not evident in our study. However, expression of the gene encoding CD8 increased in a virulence-specific manner, an observation which has been linked to increased bacterial burden 22. Among the cathepsin-encoding genes detected in our study, CATG, CATC and CATW transcript levels were elevated in H37Rv-infected iNOS–/– lungs at d30 p.i. and were only induced in iNOS–/– mice with the exception of CATC. Cathepsin C is involved in granzyme activation and cytotoxic T-cell-mediated killing 23 while CATW is also associated with CD8+ T cells and NK cells without displaying an essential role in cytotoxicity 24. Expression of perforin and genes encoding the granzymes A, B, F and K were detected in our study. The role of granzyme A, granzyme B and perforin in the induction of cell death by cytotoxic lymphocytes is well established 25. Since CD8+ T cells and NK cells produce and secrete granzymes in response to infection 1, differences in H37Ra- and H37Rv-dependent transcription levels could be the result of the differential contributions of NK and CD8+ T cells during infection with M.tb of different degrees of virulence.

The genes encoding CD1d, prosaposin and NKTR were strongly induced in H37Rv-infected mice, substantiating findings that implicate a role for NKT cells in TB 26. We detected strong H37Ra-related upregulation of NK cell-related genes evident in iNOS–/– lungs. The importance of NK-cells in susceptible hosts has been described and they could contribute to resistance of iNOS–/– mice to H37Ra infection 27, but it remains to be clarified why H37Rv-infected iNOS–/– mice failed to induce similar NK-cell responses and if these findings are represented on a cellular level as well.

Both M.tb strains caused strong transcriptional induction of typical Th1-type cytokines such as IFN-γ and TNF-α in WT and iNOS–/– lungs during infection. Both cytokines mediate important functions in resistance to M.tb 1 and were more strongly induced in H37Rv- than in H37Ra-infected lungs. Many IFN-γ-inducible genes such as IFI16, IFI30 (IP30), IFI47 (IRG-47), IGTP and IRGM (LRG-47) exhibited equal tendencies of gene regulation. Thus, these finding do not support a correlation between marker effector cytokines and pathogen virulence. Rather, they are compatible with the notion that Th1 cytokine production directly relates with bacterial load 22. IGTP, IRG-47 and LRG-47 have been described to play distinct roles in immune defense against protozoan and bacterial infections 28 and LRG-47 has been tightly linked to phagosomal maturation and susceptibility to TB 29.

Members of the Schlafen (Slfn) gene family are involved in T-cell activation and activation of Mϕ or myeloid cell differentiation 30. All of the Schlafen genes detected in our study were also differentially regulated during rheumatoid arthritis 31. Slfn-1 and Slfn-2 were originally implicated in fibroblast growth arrest but it seems more likely that they play a role in hematopoetic development and immune responses. The fact that Slfn-2 can be induced in Mϕ by stimulation with LPS or CpG 32 strengthens such notions. Little is known about Slfn-4, one of the most strongly induced genes in H37Rv-infected iNOS–/– mice, but involvement in lung alveolarization has been reported 33. It is thus conceivable that Slfn-4 contributes to lung repair in TB. Whether Slfn-4 plays an additional role in M.tb-induced lung inflammation remains to be investigated.

We detected elevated levels of chemokines CCL19, CCL20 and CXCL13 in lungs infected with virulent H37Rv. These chemokines have been associated with the formation of tertiary lymphoid structures in lungs during viral infections (iBALT) 34. The formation of such structures has been linked to host defence against bacterial infection 35, which suggests that they could also arise during TB and play a role during infection.

A proinflammatory Th17 response mediated by CD4+ T cells has been associated with protection against M.tb 36. The most prominent source of IL-17 during murine TB seems to be γδ T cells 37 yet mice deficient in the IL-17 receptor do not display decreased protection to H37Rv infection 37. We detected modest gene regulation associated with Th17 responses. Increased expression of IL-17A or IL-21 over time in WT mice did not correlate with increased IL-17 α or IL-21 transcript levels at d14 p.i. or d30 p.i. A combination of IL-17D with EBI3 results in formation of either IL-27, which is associated with negative regulation of Th17 differentiation 38, or of IL-35, an inhibitory cytokine associated with suppressive functions of regulatory T cells 39. We detected elevated transcription of genes encoding IL-15, associated with IL-2 responsiveness and survival of NK cells 40 and expression of Foxp3 in CD4+ T cells 41, and IL-16, linked to recruitment of CD4+ T cells, in particular regulatory T cells 42 in H37Ra-infected iNOS–/– mice at later time points during infection. IL-15 has been reported to participate in protection against M.tb infection 43. This process could contribute to the high resistance to H37Ra even in the absence of iNOS.

Expression of T-bet showed that proinflammatory Th1 responses were induced in all mice but H37Ra-dependent Gata3 expression implicates an emerging Th2 response in iNOS–/– mice. Increased expression of Th2-associated chemokines CCL17 and CCL27 44 further substantiate these findings.

We also detected H37Ra-induced B-cell-associated receptors of the TNF receptor family at d30 p.i. in WT and iNOS–/– lungs and higher expression of B220 in H37Ra-infected iNOS–/–mice. The role of B cells and antibodies during TB has recently gained renewed interest 16. Elevated IgG-related gene transcription in H37Rv-infected iNOS–/– mice emphasizes a role for these genes in severe inflammation during TB.

Innate immune response

Several genes associated with chemokine signalling, such as the chemokine receptors CXCR2, CCR1, G-CSFR, CD177 and triggering receptor on myeloid cells 1 (TREM1), are implicated in neutrophil-mediated processes during H37Rv-mediated inflammation in iNOS–/– lungs. TREM1 in humans is associated with extracellular bacterial infections and acute pneumonia rather than with M.tb 45. Strong TREM1 expression in iNOS–/– mice infected with virulent M.tb is therefore probably linked to inflammation in these mice and likely plays an important role in active TB disease in humans as well. IL-1 signalling elicits strong inflammatory responses and is directly involved in neutrophil responses to dead cells 46. We observed massive induction of IL-1-related genes during severe inflammation in H37Rv-infected iNOS–/– mice, suggesting that this cytokine orchestrates many of the inflammatory processes during TB.

Marked differences were detected in the transcription of CXCL- and CCL-type chemokines. Many of these chemokines attract leukocytes to sites of inflammation and thereby orchestrate the inflammatory host response. CXCL-type chemokines, especially CXCL1 (KC), CXCL2, CXCL3 and CXCL5 (LIX), all of which signal through the CXCR2 receptor, were among the most highly induced genes. The gene for CXCL2 (MIP-2) was one of the genes with highest increases in transcription over time and transcriptional differences later during H37Rv infection were induced exclusively in iNOS–/– mice. Previous studies on iNOS-dependent control of MIP-2 transcription suggest that neutrophil recruitment to the lungs is specific for iNOS–/– mice and correlates with induction of massive cell death and impaired control of MIP-2 signalling 47. Further research is needed to determine whether interference with CXCR2-mediated signalling could ameliorate severe inflammation in TB.

Chemokines of the C-C motif (CCL) family were also highly induced during inflammation in iNOS–/– mice. Especially, MIP-1α (CCL3), MIP1-β (CCL4) and RANTES (CCL5) exhibited abundant transcription in H37Rv-infected iNOS–/– lungs. Previous studies have shown that RANTES and MIP-1 are expressed in a virulence-specific manner in vivo 48. CCL3 and CCL4 are implicated in neutrophil recruitment 49. CCL4, GM-CSF, hypoxia inducible factor 1, CFLAR 50 and an MIF act as neutrophil survival factors 51.

In a different study, RANTES and IP-10 (CXCL10) enhanced the ability of DC to recruit T cells thereby inducing effective mycobacterial killing 52. Interestingly, higher IP-10 transcription at d30 p.i. was only evident in H37Rv-infected WT lungs and not in iNOS–/– lungs. Our data link IP-10 transcription to infection with M.tb in general.

Induction of genes in H37Rv-infected iNOS–/– mice with relation to oxidative burst (phagocyte oxidase phox, myeloperoxidase (MPO), xanthine dehydrogenase (XDH) 53, lymphocyte cytosolic protein 2 LCP2 54) and protection against oxidative stress (superoxide dismutases, peroxireduxins, metallothioneins and glutathione reductase) or other neutrophil functions such as plastin (LCP1) 55 emphasize that neutrophils and oxidative processes are prominent in H37Rv-mediated inflammation. MMP, especially MMP3 56, MMP8 and MMP9 57 are involved in neutrophil-mediated inflammatory processes 56, 57. These MMP were strongly induced during H37Rv-mediated inflammation in iNOS–/– lungs as well. Transcript levels of metallopeptidase inhibitor 1 (TIMP1) were also elevated in iNOS–/– lungs infected with H37Rv. TIMP1 and TIMP2 are specific inhibitors for MMP1, MMP2 and MMP9 and ratios of MMP9:TIMP1 have been associated with TB pleural effusions 58.

Many genes of the serin protease inhibitor (SERPIN) family were elevated in H37Rv-infected iNOS–/– lungs. They convey a multitude of functions 59 ranging from inhibition of thrombin (SERPINE1), of extracellular Cathepsin G (SERPINA3) or of complement components (SERPING1) to protection against TNF-α-induced cell death (SERPINB2) 60 or protection against necrotic cell death due to oxidative stress 61. The neutrophil elastase inhibitor SERPINA1 and intracellular Cathepsin G inhibitor SERPINB4 are associated with neutrophil functions. These inhibitors were only elevated in H37Ra-infected WT lungs later during infection. Cathepsin G is a major content of neutrophils and is involved in bacterial killing but is also able to activate TNF-α and IL-1 62. Our observations substantiate the important roles of neutrophils in the inflammatory response to M.tb infection in susceptible mouse strains, including iNOS–/– mice 3.

Genes of the ADAM family of metalloproteinases were strongly induced in H37Rv-infected lungs. They play diverse roles in immune and inflammatory responses due to their characteristic activities as membrane-bound proteases with protein-shedding activities. ADAM8 mediates VCAM-1 shedding and modulates leukocyte extravasation during inflammation 63. ADAMTSL4 interacts with cathepsin B and has been reported to increase TNF-α-mediated apoptosis 64. ADAMTS2 is highly induced in glucocorticoid-treated alveolar Mϕ and is involved in collagen processing 65. ADAMTS8 is induced during differentiation of monocytes to Mϕ and associates with Mϕ-rich areas in arteriosclerotic plaques 66. The role of this diverse gene family in TB remains to be elucidated.

Antimycobacterial mechanisms and lysosomal functions

The role of vitamin D(3)-dependent gene expression in both WT and iNOS/– mice during TB has been established 67. Cathelicidin is an antimicrobial peptide expressed in response to TLR2 stimuli and vitamin D treatment in a process also involving Gp91(phox)/NADPH oxidase (NOX2, Table 1 (IV)) 67. Among vitamin D receptor (VDR)-regulated genes, cathelicidin plays a critical role during TB 68. Induction of the murine homolog (CAMP) was only modest in our study and no differential induction in response to M.tb of different virulence was observed, suggesting that other genes under the control of VDR could influence the host response. The availability of iron is a critical factor for M.tb survival during infection. Lactotransferrin gene levels were strongly elevated in severely inflamed iNOS–/– and WT lungs as part of the inflammatory response. Sequestration of iron from the bacteria through competitive binding to lactotransferrin limits growth of mycobacteria, which are highly dependent on iron availability 13.

Overexpression of the epithelial form of NOS3 in the endothelium of mice results in reduced leukocyte infiltration 69. Higher transcript levels of NOS3 were only evident in iNOS–/– lungs infected with H37Rv. Enzymes controlling the availability of iNOS substrate Arginine (ARG1, ARG2 and ASS1) exhibited elevated transcription in H37Rv-infected lungs but we did not detect induction of iNOS in WT mice. Many genes involved in oxidative processes were induced in highly inflamed lungs (discussed in innate immune response) and are implicated in defence against M.tb. Mice deficient in iNOS and Phox fail to perform a profound oxidative burst after infection and do not show increased susceptibility to H37Ra infection 3, implying that other factors are sufficient to control infection with H37Ra.

Secretory leukocyte peptidase inhibitor (SLPI) is secreted by bronchial and alveolar epithelial cells and alveolar Mϕ 70. Recombinant mouse SLPI effectively inhibited in vitro growth of M.tb through disruption of the mycobacterial cell wall structure 70. SLPI is among genes with highest virulence-specific transcription levels at d30 p.i. in iNOS–/– mice. The role of SLPI during TB and its influence on inflammatory processes during M.tb infection should therefore be further investigated.

Immunosuppression

Among the genes of the TGF family, increased transcription of Tgf-β 1 (TGFB1) was detected later in H37Rv infection of WT and iNOS–/– lungs. Accordingly, genes induced by TGFB1 (TGIF, TGFBI and TGFB1I1) were also upregulated. Upregulated Smad3 mediates pro-fibrotic activities of TGF-β 71. Therefore, TGF-β is involved as both a critical cytokine of immunosuppression and contributes to structural and cellular composition of lung granulomas 72.

Transcription of IDO in H37Rv-infected WT lungs was only modestly enhanced at d30 p.i., whereas gene transcription of tryptophanyl-tRNA synthetase (WARS) was strongly induced in iNOS–/– mice and modestly in WT mice. Immunomodulatory functions of IDO and tryptophan catabolites 73 in TB imply an important role for these genes.

SOCS proteins are involved in negative feedback loops controlling innate and adaptive immune responses 74. With the exception of SOCS1, they were induced in an H37Ra-dependent fashion (SOCS6 and SOCS7). Other genes associated with negative regulation of the immune response were expressed in an H37Rv-dependent fashion. CD47-signalling downregulates IL-12 responsiveness and inhibits DC activation 75. Lysosomal multispanning membrane protein 5 modulates surface expression of the T-cell receptor 76. Mϕ scavenger receptor 1 is suggested to play a role in LPS-induced production of IL-10 77. Leukocyte immunoglobulin-like receptor B4 has been reported to downregulate DC activity to prevent excessive T-cell activation 78. Chemokine-binding protein 2 participates in the resolution of the cutaneous inflammatory response 79.

Many cell death-related genes were differentially regulated during the course of M.tb infection. Among them were genes of the Bcl-2 family (Bcl-3 and Bcl-6), and genes of the Fas and TRAIL pathways. The importance of apoptosis and its manipulation by M.tb becomes increasingly evident 80, 81. Moreover, H37Rv, but not H37Ra, inhibits apoptosis in vitro 82 and causes exacerbated necrosis in Mϕ in vitro 83. Apoptosis sensitizer gene BTG1, synergistically enhanced by LPS and IFN-γ, has been implicated in activation-induced apoptosis 84.

Concomitant increases in transcript levels of the apoptosis facilitator Bim and the survival genes BCL3 85 and Mcl-1 86 in H37Rv-infected iNOS–/– lungs combined with lower TRAIL transcription than in H37Ra-infected iNOS–/– lungs 85 point to active involvement of the mitochondrial apoptotic pathway in M.tb infection. Enhanced levels of caspase 9 during M.tb infection have been linked to necrotic rather than apoptotic cell death 87. CD61, involved in Mϕ-induced cell death of neutrophils 88, was downregulated over time during H37Ra infection and increased levels were detected in H37Rv-infected iNOS–/– lungs. Prolonged neutrophil survival can result in secondary necrosis, especially if Mϕ are unable to clear apoptotic cells efficiently 89. Closer analysis of this process and more detailed investigation of gene functions connected to apoptosis and necrosis could elucidate how these processes contribute to inflammation and influence the course of infection in TB.

Concluding remarks

Our global gene expression profiling as function of M.tb virulence and host susceptibility have produced a plethora of differentially expressed gene products in experimental TB in mice. It is worth noting that the genes of the different components of the host response were found to be differentially regulated. Our studies elucidate a unique signature of cytokines and chemokines and their receptors. They not only include a number of intracellular and extracellular effector molecules, including complement factors, enzymes involved in oxidative burst, lactotransferrin and PTX, but also regulatory components such as members of the SERPIN family. Similarly, both metalloproteinases and their inhibitors were differentially expressed. At the level of adoptive cellular immunity, numerous cytokines, notably the Th1 marker cytokines IFN-γ and TNF-α were regulated although at the same time SOCS gene products involved in inhibition of immune responses were upregulated as well. Hence, as a general conclusion from our analyses, single molecule investigations in complex diseases such as TB should be replaced by global analyses of complex networks since infection, pathogenesis and protection depend on numerous stimulatory and inhibitory signals, which determine the outcome of TB disease.

Global gene expression analyses of infected tissues are often confronted with the fact that changes in gene expression can be caused by a multitude of confounding factors. Therefore, unbiased analyses of the dynamic interplay between host and pathogen during infection are difficult to perform. Although this is also true for our current investigation, we believe that interrogating gene expression profiles in our matrix-based approach represent a first attempt toward characterization of the host response to M.tb in a framework encompassing time, pathogen virulence and host susceptibility. Hence, our study emphasizes the need for integrating multiple components into TB research due to the complexity of the disease rather than addressing unique phenomena in isolation. We chose the established pair of laboratory M.tb strains H37Rv and H37Ra to facilitate analyses of virulence-associated responses since these strains are closely related but differ highly in their degree of virulence. On the host side, iNOS–/– mice were chosen as a model of severe deficiency with distinct effects on TB.

We believe that phenotypic consequences of iNOS deficiency are more limited than those of other known gene defects underlying TB susceptibility although NO conveys several signalling functions. We identified numerous genes that were downregulated in response to infection with attenuated M.tb but upregulated to infection with virulent M.tb. Among them, we consider VDR, schlafen 4 (SLFN4) and XDH as particularly interesting targets for future investigations. Although a strong role for VDR signalling in human TB has already been established 68, associations of XDH and SLFN4 with TB have not been described yet. Experimental M.tb infection of KO mice deficient in these genes would be a logical first step toward their functional elucidation. Furthermore, virulence-specific induction of IDO and WARS links tryptophan catabolism of the host with M.tb virulence 90. Finally, the strong induction of PTX3 in the susceptible host to virulent M.tb substantiates findings that associate this gene with susceptibility to TB in humans 91. Further research into the role of PTX3 and tryptophan catabolites in TB and their potential use as biomarkers appears apt.

Our investigations also extended knowledge about known susceptibility genes in TB such as members of the IFN-γ and TNF-α signalling pathways, which affect multiple downstream events in host defence, pathogenesis and inflammation. Our study emphasizes the importance of the Th1 cell response. Yet, neither induction nor inhibition of Th1 immunity could be nailed down as underlying cause for virulence-associated differences in TB regardless of underlying host susceptibility.

In the search for robust biomarkers in human TB, chemokine and cytokine profiles are of central interest. In our investigation, we identified a unique chemokine signalling pattern in response to infection with virulent M.tb as major contributing factor for disease outcome.

Profound transcriptional induction of CXCL1-3 and CCL3-5 link susceptible hosts with virulence of M.tb and suggests that these chemokines are promising candidates for biomarkers. Notably, CXCR2 signalling – in particular through CXCL2 – was distinctly involved in susceptible hosts infected with virulent M.tb. CXCR2 has been linked to recruitment of NK cells 92 and neutrophils 93 in lung inflammation and responsiveness to CXCR2 is affected in human TB 94.

Although the role of neutrophils in protection and pathogenesis in TB remains controversial, these effector cells clearly contribute to inflammation in acute TB 95, 96. The involvement of neutrophil-mediated lung injury could participate in the formation of necrotic and caseous granulomas 97. Finally, neutrophils could play a role in unique stages of human TB, such as the immune restoration inflammatory syndrome (IRIS) prominent in M.tb/HIV-coinfected patients undergoing antiretroviral therapy 98. Expression of several survival factors during inflammation implies secondary necrosis of neutrophils 89 with concomitant deregulated signalling through CXCR2 47 as a critical mechanism of severe inflammation. It remains to be determined to which degree these processes are involved in human disease progression such as reactivating granulomas or acute inflammation associated with TB.

Currently, information about mechanisms underlying IRIS is scarce but participation of aberrant chemokine responses appears likely. Elucidation of mediators involved in IRIS could provide relevant information for the design of biomarkers predicting IRIS and form the basis for rational immune intervention strategies. A protective role of NK cells in human TB has been proposed 99. In our investigations, susceptible mice exhibited profound transcriptional downregulation of genes tightly associated with NK cells in response to virulent M.tb 92. This finding encourages further research into the contribution of NK cells to TB progression, diagnosis and intervention.

In contrast to CXCR2-associated chemokine signalling, transcriptions of other chemokines − especially IP-10, Mig and IP-9 − were discretely elevated in competent mice in response to infection with virulent M.tb. Early induction of Mig has been linked to granuloma formation 100. Hence, prompt and profound induction of Mig transcription in competent WT lungs in response to virulent M.tb infection as observed here points to this chemokine as contributing factor to formation of effective granuloma formation. Furthermore, our findings substantiate the proposed potential role of IP-10 as a biomarker in human TB. Recently, a protective role for IP-10 in human TB has been strengthened by genetic analysis, linking IP-10 transactivation to protection against TB 101. We observed early and profound induction of IP-10 transcription in competent WT mice as compared with compromised iNOS–/– mice which was sustained during M.tb infection. Hence, our study not only supports a role of IP-10 in protection against TB, but also emphasizes that the chemokine pattern to M.tb infection serves as attractive basis for a biosignature predicting TB disease outcome.

Elucidation of the role in human TB of candidates identified here will provide guidelines for better understanding of the mechanisms underlying pathogenesis and protection which determine severity of active TB disease and can contribute to the design of novel diagnostic and intervention measures in human TB.

Materials and methods

Mice and M.tb infection

C57BL/6 (WT) mice were purchased from Charles River Laboratories and NOS2tm1/lau (iNOS–/–) mice from Jackson Laboratories. Mice were bred in our facilities and kept under specific pathogen-free conditions with food and water provided ad libidum. All animal experiments were conducted according to German animal protection law. M.tb H37Ra (ATCC 25177) and M.tb H37Rv (obtained as a kind gift from William Jacobs, AECOM, New York) infection stocks were grown to mid-log phase from low-passage seed lots (Middlebrook 7H9 medium, ADC suppl., 0.05% Tween 80) and an aliquot of 1 mL was frozen at –80°C until use.

Mice were infected by aerosol with ∼200 CFU of M.tb H37Rv and ∼2000 CFU of H37Ra using an aerosol chamber (Glas-Col). Bacterial counts of inocula and in homogenates of lungs, spleens and livers at various timepoints post-infection (p.i.) were determined as CFU by plating serial dilutions of whole organ homogenates in PBS/0.05% Tween 80 on Middlebrook 7H11-OADC ampicillin plates and incubating cultures at 37°C for 3–4 wk.

RNA preparation and microarray hybridization and quantitative real-time PCR quantification

Total RNA was isolated from pooled lung samples homogenized in TRIzol™ reagent (Invitrogen) according to the manufacturer's specifications. DNase I (Invitrogen)-treated RNA was purified (RNeasy, Qiagen) and cDNA was synthesized (Superscript™ III, Invitrogen).

Using oligo-dT-T7, 4 μg total RNA were reverse transcribed to incorporate Cyanine-3-CTP or Cyanine-5-CTP (NEB Life Science Products) into cRNA in a fluorescent linear amplification reaction (Agilent Technologies) according to the supplier's protocol. After precipitation, purification and quantification, 1.25 μg of labeled cRNA was mixed, fragmented, hybridized to Whole Mouse Genome 44 k microarrays (Agilent Technologies), according to the supplier's protocol and scanned at 5 μm resolution with a DNA microarray laser scanner (Agilent Technologies). Features were extracted with an image analysis tool (Version A4.045, Agilent Technologies) with default settings. To ensure statistical relevance, each cRNA was synthesized in both colors and hybridizations were repeated with swapped colors. Quantitative real time PCR was performed using custom-made TaqMan low-density arrays according to the manufacturer's specifications (Applied Bioscience).

Data analysis

Microarray data were analyzed on the Rosetta Inpharmatics platform (Resolver Built 3.0.0.3.22). Features detected as differentially regulated with a p-value ≤0.05 and at least twofold differences in signal intensity were included in data sets. Sorted data sets (Spotfire DXP™, TIBCO and Excel, Microsoft) were analyzed using Ingenuity Pathways Analysis (Ingenuity® Systems, www.ingenuity.com) for functional analyses using GenBank Accession numbers as unique identifiers. Genes from the data set that were associated with biological functions and/or disease in the Ingenuity Pathways Knowledge Base were considered for analysis. Fischer's exact test was used to calculate a p-value determining probability that each biological function and/or disease assigned to that data set is a result of chance alone. A complete data set of genes associated with biological functions and/or disease is found in the Supporting Information.

Acknowledgements

S. H. E. Kaufmann acknowledges support from the German Bundesministerium für Bildung und Forschung (BMBF) Kompetenz Netzwerk “PathoGenoMikPlus” and from the EU FP7 project TB-VIR (HEALTH-F3-2008-200973). The authors thank M. L. Grossman for critical reading of the manuscript.

Conflict of interest: The authors declare no financial or commercial conflict of interest.

Ancillary