Distinct TLR- and NLR-Mediated Transcriptional Responses to an Intracellular Pathogen

How the innate immune system tailors specific responses to diverse microbial infections is not well understood. Cells use a limited number of host receptors and signaling pathways to both discriminate among extracellular and intracellular microbes, and also to generate responses commensurate to each threat. Here, we have addressed these questions by using DNA microarrays to monitor the macrophage transcriptional response to the intracellular bacterial pathogen Listeria monocytogenes. By utilizing combinations of host and bacterial mutants, we have defined the host transcriptional responses to vacuolar and cytosolic bacteria. These compartment-specific host responses induced significantly different sets of target genes, despite activating similar transcription factors. Vacuolar signaling was entirely MyD88-dependent, and induced the transcription of pro-inflammatory cytokines. The IRF3-dependent cytosolic response induced a distinct set of target genes, including IFNβ. Many of these cytosolic response genes were induced by secreted cytokines, so we further identified those host genes induced independent of secondary signaling. The host response to cytosolic bacteria was reconstituted by the cytosolic delivery of L. monocytogenes genomic DNA, but we observed an amplification of this response by NOD2 signaling in response to MDP. Correspondingly, the induction of IFNβ was reduced in nod2 −/− macrophages during infection with either L. monocytogenes or Mycobacterium tuberculosis. Combinatorial control of IFNβ induction by recognition of both DNA and MDP may highlight a mechanism by which the innate immune system integrates the responses to multiple ligands presented in the cytosol by intracellular pathogens.


Introduction
As sentinels of the immune system, macrophages must be able to determine the nature and scope of microbial threats to mount appropriate transcriptional responses [1,2]. Macrophages need to discriminate not only viral from bacterial infection, but also extracellular and possibly killed microbes from intracellular and replicating pathogens [3]. Cells receive information regarding infection using a limited number of pattern recognition receptors (PRRs) to sense conserved motifs presented by microbes [1,[4][5][6]. Two major classes of PRRs include membrane-bound Toll-like receptors (TLRs) and soluble, cytosolic NOD-like receptors (NLRs). TLRs monitor the extracellular environment and phagolysosomal compartments, and recognize pathogen associated molecular patterns (PAMPs) that include lipopolysaccharide, flagella, CpG DNA, and bacterial lipoprotein [7]. NLRs complement this host defense by providing surveillance of the cytosol. The nucleotide-binding and oligomerization domain containing (NOD) proteins, for which this class of receptors is named, recognize cell wall fragments from both Gram-negative and Gram-positive bacteria [8][9][10]. NOD1 recognizes a specific peptidoglycan (PGN) fragment containing diaminopimelic acid, while NOD2 recognizes a muramyl dipeptide (MDP) fragment of PGN. The MDA5 and RIG-I NLRs detect cytosolic dsRNA [11], while DAI detects cytosolic dsDNA [12][13][14].
How cells initiate a threat-specific transcriptional response is poorly understood, as many PRRs, in response to different stimuli, utilize the MyD88 and TRIF signaling adaptor molecules to activate the same transcription factors [15]. For instance, activation of both TLRs and NLRs causes degradation of the repressor IjB, thereby freeing the transcription factor NFjB to enter the nucleus and bind to target promoters of genes important for host defense. Some target specificity is generated by the phosphorylation and activation of the IRF3 transcription factor by only a subset of PRRs, including those that recognize nucleic acids. The induction of certain host genes, including the Type I interferons a (IFNa) and b (IFNb), requires both NFjB and IRF3 [16][17][18]. Secreted Type I interferons then induce many additional genes by secondary signaling through the Type I interferon receptor (IFNAR) [15].
L. monocytogenes is a ubiquitous Gram-positive intracellular bacterium that can cause serious illness in pregnant women and immunocompromised individuals [19], and is ideal for the study of host innate immune responses. Mutants defective in precise stages of the intracellular life cycle have been isolated, and in vitro infection of primary bone marrowderived macrophages allows dissection of host signaling pathways. Approximately 30 minutes after initial phagocytic uptake into a membrane-bound vacuole, bacteria escape to the host cytosol by perforating the vacuolar membrane, using the pore-forming toxin listeriolysin O (LLO, encoded by the gene hly) [20]. Once in the host cytosol, L. monocytogenes replicates robustly, and uses a system of actin-based motility to spread from the initially infected macrophage to colonize neighboring cells [21]. Both heat-killed and hlyÀ L. monocytogenes induce inflammatory cytokines, but in non-activated macrophages only wild-type (WT) bacteria that are able to access the host cytosol induce Type I interferons [22][23][24].
Although the L. monocytogenes Type I interferon-stimulating ligand has not been conclusively identified, evidence suggests that bacterial DNA possesses IFNb-inducing activity and might be the relevant PAMP [12,25]. PGN fragments have also been shown to induce host transcriptional responses [3,8,[26][27][28], and MDP is present both in digested L. monocytogenes PGN fragments [29,30] and in polymer-linked form in intact PGN [31]. The role of MDP during the host response to L. monocytogenes is not clear, however [3,24,32].
In this study, we have comprehensively determined the macrophage transcriptional responses to L. monocytogenes using DNA microarrays. Using macrophages deficient for defined host signaling pathways, and bacteria residing in different subcellular localizations, we have delineated distinct host responses to vacuolar and cytosolic bacteria, and addressed the mechanisms underlying the specificity of these responses. We have additionally determined the direct transcriptional targets of host NLR signaling in response to cytosolic bacteria. These genes are co-regulated with IFNb, and are uniquely induced in infected cells, as their induction is independent of any secondary signaling. These primary targets have the potential to modify host signaling, and may therefore critically impact the host response to infection. Cytosolic delivery of purified bacterial genomic DNA reconstituted the host response to cytosolic L. monocytogenes. This response was synergistically amplified by NOD2 signaling in response to MDP. We find a similar role for NOD2 signaling in the host response to both L. monocytogenes and M. tuberculosis, and this may represent a mechanism by which cells integrate multiple PRR signals to accurately identify bacteria able to access the host cytosol.

Identification of Distinct Macrophage Transcriptional Responses to Vacuolar and Cytosolic L. monocytogenes
We first determined the global transcriptional response of WT and myd88 À/À macrophages infected with WT bacteria and hlyÀ L. monocytogenes, using high-density oligonucleotide microarrays [33]. To identify genes induced by vacuolar bacteria, the transcriptional response of WT macrophages after 180 minutes of infection with hlyÀ L. monocytogenes was subject to Significance Analysis of Microarrays (SAM) [34] (see Materials and Methods), and the resulting genes further selected to identify those with at least a 4-fold change in abundance. Using this approach, we identified 252 macrophage genes that met these criteria, which we defined as the ''Vacuolar Response'' of macrophages to L. monocytogenes ( Figure 1A, ''Vacuolar Response''; for full target gene list, see Dataset S1). This class of genes included many proinflammatory cytokines and chemokines, such as Interleukins 1a (IL1a) and 1b (IL1b), tumor necrosis factor (TNF), KC, and MIP2. Strikingly, none of the 252 genes of the Vacuolar Response were significantly induced in myd88 À/À macrophages ( Figure S1). IL1b is hereafter used as a representative target gene of the Vacuolar Response ( Figure 1B, top panel).
To identify genes induced by cytosolic bacteria, the transcriptional response of myd88 À/À macrophages after 180 minutes of infection with WT L. monocytogenes was subject to SAM, and the resulting genes further selected to identify those with at least a 4-fold change in abundance. Using this approach, we identified 106 macrophage genes, which we defined as the ''Cytosolic Response'' of macrophages to L. monocytogenes ( Figure 1A, ''Cytosolic Response''; for full target gene list, see Dataset S2). These genes were strongly induced starting at 2 hours post-infection, by which point WT bacteria were replicating robustly in the cytosol of infected cells. Among the genes most highly induced by the Cytosolic Response were Type I interferons, including IFNb and multiple IFNa genes, and many known interferon-regulated genes. IFNb is hereafter used as a representative target gene of the Cytosolic Response ( Figure 1B, middle panel).
For 27 genes we observed induction by both the Vacuolar Response and the Cytosolic Response, as these genes were induced both in WT macrophages infected with hlyÀ L. monocytogenes and in myd88 À/À macrophages infected with WT L. monocytogenes. IL6 is hereafter used as a representative target gene of this class ( Figure 1B, bottom panel).

Identification of the Primary Cytosolic Response
Cytosolic Response targets included Type I interferons and Interleukin 6 (IL6), both of which induce the transcription of many additional genes in neighboring cells [15,35,36]. Five additional analyses were performed to identify genes directly induced by the host response to cytosolic bacteria, and not by secondary signaling. First, we required significant induction in infected ifnar À/À macrophages, which cannot respond to secreted Type I interferons. Second, induction in infected ifnar À/À macrophages must not have been significantly less than that in infected WT macrophages. These two filters removed genes significantly induced as a result of secondary

Author Summary
Macrophages are critical cells of the innate immune system, contributing to immediate and robust defense against microbial infections. Macrophages detect pathogens using host receptors located on the cell surface, in phagosomal vacuoles, and in the cytosol. While fundamental to innate immunity, it is not clear if these different receptors merely provide redundant mechanisms for sensing microbial infection, or if instead they induce distinct gene expression programs that may allow for threat-specific host responses. We addressed this question by dissecting the macrophage transcriptional responses to the model intracellular bacterial pathogen Listeria monocytogenes. Using genetic and genomic approaches, we found that the macrophage response to L. monocytogenes trapped in phagosomal compartments was distinct and separable from the response to live bacteria replicating in the host cytosol. The macrophage response to cytosolic bacteria was recapitulated by bacterial nucleic acid and cell wall fragments, and induced surprisingly few primary response genes. These findings highlight a mechanism by which the innate immune system may specifically sense intracellular bacteria, as the macrophage response to Mycobacterium tuberculosis was similarly regulated. Type I interferon signaling. Third, we required significant induction in infected WT macrophages treated with the protein synthesis inhibitor cycloheximide. Fourth, induction in cycloheximide-treated infected WT macrophages must not have been significantly less than induction in untreated infected WT macrophages. These two filters removed genes significantly induced as a result of secondary signaling by any other translated and secreted cytokine. Finally, we required that genes not have also been targets of the Vacuolar Response, as these genes would be induced in uninfected cells responding to extracellular TLR ligands. Ultimately, we identified only seven genes that met these five strict criteria, which we defined as targets of the Primary Cytosolic Response. These genes included IFNb, PELI1, MYD116, TYKI, and three members of the IFIT (interferon-inducible with tetratricopeptide repeats) family ( Figure 2). None of these genes were significantly induced in irf3 À/À macrophages (data not shown). To identify genes induced by the Primary Cytosolic Response, but also induced by either secondary signaling or the Vacuolar Response, only criteria 1 and 3 (above) were used. By these relaxed criteria 20 additional genes were identified, including IL6, CXCL10, OTUD1, MDA5, IGTP, and OASL1 (for full target gene list, see Dataset S3).

IRF3 Determines the Specificity of the Cytosolic Response
The minimal overlap between the targets of the Vacuolar Response and the Cytosolic Response indicated a high degree of target gene specificity in these responses ( Figure 1C). Stimulation of the Cytosolic Response during infection of myd88 À/À macrophages with WT L. monocytogenes resulted in the activation and nuclear translocation of both NFjB and IRF3 ( Figure   macrophages by hlyÀ L. monocytogenes are indicated as targets of the ''Vacuolar Response''. Genes identified by SAM and at least 4-fold induced in myd88 À/À macrophages by WT L. monocytogenes are indicated as targets of the ''Cytosolic Response''. For target gene lists, see Datasets S1 and S2. The determination of Vacuolar Response and Cytosolic Response target genes was from multiple arrays representing four independent experiments (e.g. four independent dishes of uninfected myd88 À/À macrophages and four independent dishes of myd88 À/À macrophages infected with WT L. monocytogenes for 180 minutes were used for Cytosolic Response determination). (B) Summary of the transcriptional responses of IL1b, IFNb, and IL6 in WT and myd88 À/À macrophages during infection with either WT or hlyÀ L. monocytogenes, at the indicated times post-infection (in minutes), determined by microarray analyses (Dataset S5). (C) Scatter plot representation of the transcriptional response of all genes defined as targets of the Vacuolar Response or the Cytosolic Response. Microarray data from myd88 À/À macrophages infected with WT L. monocytogenes (Cytosolic Response-stimulating conditions) is plotted on the Y-axis, and microarray data from WT macrophages infected with hlyÀ L. monocytogenes (Vacuolar Response-stimulating conditions) is plotted on the X-axis. Targets of only the Vacuolar Response are depicted as red squares (e.g. IL1b), targets of only the Cytosolic Response as blue triangles (e.g. IFNb), and targets of both the Vacuolar Response and Cytosolic Response as purple diamonds (e.g. IL6). doi:10.1371/journal.ppat.0040006.g001 with hlyÀ L. monocytogenes (lanes 7-12) did not lead to nuclear localization of IRF3, which is consistent with failure of the Vacuolar Response to induce IFNb. NFjB, c-Jun, and ATF2, however, were localized to the nucleus during the Vacuolar Response (lanes 10 and 12).
The induction of IFNb during infection of macrophages with WT L. monocytogenes requires IRF3 [24,37]. We further found that the majority of the Cytosolic Response is IRF3dependent, as 94% of Cytosolic Response-specific target genes (i.e. genes that are not also targets of the Vacuolar Response, such as IL6) were significantly less induced in infected irf3 À/À macrophages compared to infection of WT macrophages ( Figure 3B).
WT bacteria in the cytosol of myd88 À/À macrophages failed to induce the vast majority of Vacuolar Response genes, despite the robust nuclear localization of NFjB in these infected cells. The Cytosolic Response did not lead to feedback inhibition of Vacuolar Response target gene induction, unlike what has been reported during similar innate immune signaling in Drosophila [38]. Only 1.7% of Vacuolar Response-specific target genes showed greater induction during infection of irf3 À/À macrophages, in which , phospho-c-Jun, phospho-ATF2, and IRF3 in fractionated lysates of myd88 À/À or WT macrophages infected with WT or hlyÀ L. monocytogenes, respectively, for the indicated times (in hours). ''C'' indicates cytosolic fractions and ''N'' indicates nuclear fractions. Data is from the pooled lysates of two independent dishes. (B) Scatter plot representation of the transcriptional response of WT and irf3 À/À macrophages infected for 3 hours with WT L. monocytogenes, as determined by microarrays. Shown are the responses of Cytosolic Response-specific target genes. Y-axis values are log 2 fold change in RNA abundance in WT macrophages infected with WT L. monocytogenes. X-axis values are log 2 fold change in RNA abundance in irf3 À/À macrophages infected with WT L. monocytogenes. The superimposed dashed line has a slope ¼ 1. Spots in blue are those identified by SAM as being significantly differently induced in the two conditions, while spots in pink are not different. (C) Scatter plot representation of the transcriptional response of WT and irf3 À/À macrophages infected for 3 hours with WT L. monocytogenes, as determined by microarrays. Shown are the responses of Vacuolar Response-specific target genes. Y-axis values are log 2 fold change in RNA abundance in WT macrophages infected with WT L. monocytogenes. X-axis values are log 2 fold change in RNA abundance in irf3 À/À macrophages infected with WT L. monocytogenes. The superimposed dashed line has a slope ¼ 1. Spots in blue are those identified by SAM as being significantly differently induced in the two conditions, while spots in pink are not different. doi:10.1371/journal.ppat.0040006.g003 the Cytosolic Response was almost entirely absent, compared to infected WT macrophages ( Figure 3C).

Induction of the Cytosolic Response by Cytosolic Delivery of L. monocytogenes Genomic DNA Is Amplified by Co-Delivery of MDP
To examine and dissect the Cytosolic Response, we attempted to recapitulate the host response to cytosolic bacteria by instead treating macrophages with purified bacterial ligands, using L. monocytogenes genomic DNA and synthetic MDP. When delivered directly into the cytosol of WT macrophages by transfection, L. monocytogenes DNA strongly induced the expression of IFNb ( Figure 4A, lane 2). However, co-transfection of macrophages with both L. monocytogenes genomic DNA and synthetic MDP yielded double the induction of IFNb than that observed with DNA alone (lane 4), even though MDP alone yielded minimal IFNb induction (lane 3). IFNb induction was absolutely dependent on TBK1 (lanes 6 and 8), a kinase required by all identified nucleic acid recognition receptors to activate the IRF3 transcription factor [37,39]. This defect was specific to the Cytosolic Response, as induction of Vacuolar Response target genes in response to hlyÀ L. monocytogenes was unaffected in tbk1 À/À cells (data not shown). Macrophages deficient for RIP2, an adaptor molecule required for NOD2-dependent NFjB activation [40,41], still induced IFNb in response to DNA (lane 10), but co-transfection with MDP no longer produced any additional synergistic induction (lane 12). This response was independent of both MyD88 and TRIF ( Figure S2A), and was not due to contamination of the L. monocytogenes genomic DNA with other bacterial ligands, as we observed identical results with synthetic poly(dAT-dTA) DNA ( Figure S2B). While virtually all Cytosolic Response-specific target genes were induced by transfection of macrophages with L. monocytogenes DNA, 94% of these genes were induced to an even greater magnitude by co-transfection of macrophages with both DNA and MDP ( Figure S2C). Cytosolic delivery of DNA and MDP specifically induced Cytosolic Response genes without inducing Vacuolar Response genes, as these ligands only weakly induced IL1b ( Figure 4B, compare lanes 2-4 to lane 5; Dataset S4).

Nuclear NFjB Abundance Controls Synergistic Induction of the Cytosolic Response
Western blots were used to assess the activation of Cytosolic Response transcription factors by delivery of DNA and MDP. Transfection of WT macrophages with DNA alone ( Figure 5A, lane 4) or MDP alone (lane 6) strongly activated NFjB to similar levels. When macrophages were transfected with both DNA and MDP, NFjB accumulated in the nucleus to a magnitude equal to the sum of that observed for the two ligands individually (lane 8). Similar results were observed for nuclear translocation of c-Jun (lanes 4, 6, and 8). No additive nuclear translocation of ATF2 was observed by co-delivery of both ligands (data not shown). MDP has never been found to activate IRF3 [8,24,28], consistent with our finding that MDP delivery alone does not induce IFNb ( Figure 4A, lanes 3 and 7).
To investigate the role of NFjB abundance in the synergistic induction of IFNb, WT macrophages infected with WT L. monocytogenes were treated with 10 lg/ml caffeic acid phenyl ester (CAPE), a pharmacological inhibitor of NFjB nuclear trafficking [42]. CAPE treatment of infected macrophages caused a greater than 4-fold reduction in nuclear NFjB p65 accumulation, but had no effect on either c-Jun or IRF3 nuclear localization ( Figure 5B). Compared to untreated macrophages, induction of IFNb by L. monocytogenes in CAPEtreated macrophages was reduced greater than 7-fold ( Figure  5C). Using DNA microarrays, we observed that this inhibition of NFjB nuclear translocation affected the entire Cytosolic Response, as 99% of Cytosolic Response-specific target genes were significantly less induced in infected CAPE-treated macrophages, compared to infected macrophages not CAPEtreated ( Figure S3).
Additionally, macrophages transfected with L. monocytogenes genomic DNA were treated with concentrations of CAPE ranging from 5 ng/ml to 40 ng/ml, which reduced nuclear NFjB abundance from 10%-50% compared to untreated cells ( Figure 6A, lanes 3-10). These lower concentrations of CAPE allowed precise titration of nuclear NFjB, instead of the near complete inhibition observed previously with higher doses ( Figure 5). Coincident with this reduction in nuclear NFjB abundance, induction of IFNb decreased 30%-80%. At each increase in CAPE dosage, translocation of NFjB to the nucleus was reduced and induction of IFNb declined. Similarly, macrophages transfected with either DNA alone or with both DNA and MDP were treated with a titration of CAPE. Without CAPE treatment, co-transfection of macrophages with both DNA and MDP yielded both twice the induction of IFNb and twice the relocalization of NFjB to the nucleus, compared to transfection with DNA alone ( Figure  6B To determine if L. monocytogenes release MDP during intracellular infection of macrophages, and if this is sensed by NOD2, IFNb induction in nod2 À/À macrophages infected with WT L. monocytogenes was assessed. Residual vacuolar signaling in these macrophages continued to activate NFjB, however, obscuring the contribution of NOD2 signaling to nuclear NFjB abundance ( Figure S4A). This residual TLR signaling was likely in response to the high concentrations of bacterial fragments delivered during the initial infection inoculum, and would be absent during a natural infection by a single bacterium. Therefore, to reduce NFjB activation by TLR overstimulation, macrophages were first tolerized by prior exposure to the synthetic TLR2 agonist Pam 3 CSK 4 [43] ( Figure S4B and S4C).
In tolerized macrophages infected with WT L. monocytogenes, at 4 hours post-infection nod2 À/À macrophages exhibited a greater than 2-fold reduction in IFNb induction, as compared to WT macrophages ( Figure 7A). Thus, NOD2 signaling doubled the induction of IFNb under these conditions. This exactly mirrored the effects of cytosolic co-delivery of DNA and MDP-a 2-fold amplification of IFNb induction ( Figure  4A) and a 2-fold amplification of NFjB nuclear abundance ( Figure 5A), compared to the response elicited by bacterial DNA alone. To determine if NOD2 signaling was necessary for the full induction of IFNb in response to infection with other intracellular bacteria, we infected tolerized WT and nod2 À/À macrophages with the pathogenic Gram-positive intracellular bacterium M. tuberculosis [44,45]. We observed that induction of IFNb in nod2 À/À macrophages infected with M. tuberculosis for 4 hours was less than half that observed in infected WT macrophages ( Figure 7B).

Identification of the Vacuolar and Cytosolic Responses to L. monocytogenes
In this study we have rigorously and comprehensively identified the Vacuolar and Cytosolic Responses of macrophages to infection with an intracellular bacterial pathogen. Our model system-primary bone marrow-derived macrophages of multiple genotypes infected with both WT L. monocytogenes and hlyÀ bacteria-allowed precise separation of host responses to cytosolic and vacuolar bacteria. Previous Nuclear abundance of proteins (indicated below the nuclear fraction within each blot) was normalized to the nuclear protein Lamin B, which also serves as fractionation control. For each transcription factor, abundance in the nucleus of macrophages transfected with L. monocytogenes DNA was arbitrarily set to 10.0, and all other abundances were calculated relative to this. Data was collected from two blots, with each blot using pooled lysates of two independent dishes. (B) Quantitative Western blot analysis of NFjB p65, c-Jun, and IRF3 protein distribution in fractionated lysates of WT macrophages infected with WT L. monocytogenes for the indicated times (in hours). Where indicated, cells were additionally treated with 10 lg/ml CAPE. genomic analyses of the macrophage transcriptional response to L. monocytogenes did not use myd88 À/À macrophages to fully separate TLR and NLR signaling [23].
We have found that the Vacuolar Response was entirely MyD88-dependent, suggesting that the role of the TLR adaptor TRIF may be specific to innate immune responses to Gram-negative bacteria. The Vacuolar Response controlled the transcriptional induction of many pro-inflammatory cytokines, such as IL1a, IL1b, and TNF. In contrast, the IRF3-dependent Cytosolic Response induced a distinct and significantly non-overlapping set of 106 host response genes, including Type I interferons. We have further identified the 27 targets of the Cytosolic Response that were directly induced in response to cytosolic L. monocytogenes, and did not require secondary cytokine signaling for induction. Seven of these Primary Cytosolic Response target genes were induced entirely independent of any secondary signaling, and therefore were only induced directly in infected cells. Of these 7 genes, only IFNb encodes a secreted protein, while many of the others encode potential regulators of signaling. For instance, MYD116 is a homolog of GADD34, which can form a complex with protein phosphatase 1a to dephosphorylate eIF2a, thereby inhibiting protein synthesis [46]. PELI1 is an E3 ubiquitin ligase that has been shown to modulate both TLR and IL1 signaling [47]. Consequently, the significant representation and robust induction of members of the IFIT family (IFIT1-3; induced 29.9-, 21.1-, and 13.9-fold, respectively) in this select group of genes warrants further study. Very little is known about IFIT proteins other than the potential of their tetratricopeptide repeats to mediate protein-protein interactions [48]. Despite their name, the IFIT genes responded directly to NLR signaling in response to L. monocytogenes.
We propose that these 7 targets of the Primary Cytosolic Response may provide a mechanism by which infected and uninfected cells could respond differently to the Type I interferons and other cytokines secreted during infection. Cytokines would trigger secondary signaling in all cells near the site of infection, but induction of MYD116, PELI1, TYKI, and IFIT1-3 only in infected cells might modify this secondary signaling to trigger a different response. In this manner, infected cells could be specifically re-programmed to help contain infection.
We have found that two separable pathways coordinately control the Cytosolic Response. The first pathway consists of recognition of bacterial nucleic acid by a cytosolic pattern recognition receptor, perhaps DAI, that activates the transcription factors IRF3 and NFjB. We have demonstrated that L. monocytogenes genomic DNA induced Cytosolic Response target genes, and others have also shown that L. monocytogenes DNA induces IFNb [12]. The response to both genomic DNA and live bacteria was TBK1-dependent (this study and [37]), and DAI has been shown to associate with TBK1 [14].
The second pathway consists of peptidoglycan fragment recognition by NOD2, which activates NFjB but not IRF3. We have demonstrated that the nuclear abundance of NFjB was limiting for the induction of IFNb, and our data suggest that the two pathways converge by coordinate control of NFjB nuclear abundance. First, co-delivery of bacterial DNA and synthetic MDP doubled the nuclear abundance of NFjB, and induced twice as much IFNb, compared to transfection of macrophages with DNA alone. Second, CAPE inhibited induction of IFNb, and blocked only the nuclear translocation of NFjB. Third, under two conditions yielding equivalent NFjB activation-one with DNA alone and one with DNA, MDP, and CAPE-IFNb induction was nearly identical, suggesting that control of NFjB abundance was the principal mechanism by which NOD2 amplified IFNb induction. The further activation of NFjB by NOD2 signaling explains how this pathway contributed to IFNb induction without activating IRF3.
We further found that nod2 À/À macrophages, when TLR signaling was eliminated, induced significantly less IFNb during infection with either L. monocytogenes or M. tuberculosis. We speculate that NOD2 may play a similar role in response to other IFNb-inducing intracellular bacteria whose peptidoglycan contains MDP, such as Francisella tularensis [36]. The convergence of NLR signaling pathways at the level of transcription factor abundance might allow complex signal integration in cells using a limited number of PRRs.

Specificity in Target Gene Induction by the Vacuolar and Cytosolic Responses
The Vacuolar Response and Cytosolic Response controlled the transcription of largely distinct sets of target genes. The Vacuolar Response activated NFjB, and thereby induced proinflammatory cytokines such as IL1b, but did not activate IRF3, and hence did not activate IRF3-dependent targets including IFNb. The Vacuolar Response was completely MyD88-dependent, and it has been demonstrated that a TLR-dependent but NFjB-independent remodeling of nucleosomes at the promoters of certain pro-inflammatory cytokines is required for induction [49][50][51]. Chromatin modifications are increasingly viewed as critical modulators ensuring appropriate control of inflammation [43], and may be a mechanism for determining target specificity of the Vacuolar and Cytosolic Responses. This may explain why activation of both NFjB and IRF3 during the Cytosolic Response did not induce IL1b, even though IFNb was induced. Infection of myd88 À/À macrophages with WT bacteria and the cytosolic delivery of purified ligands both bypassed TLR signaling, and therefore may not have triggered this nucleosome remodeling.
The OspF virulence factor injected by the bacterial pathogen Shigella flexneri modifies host chromatin during infection to block the activation of certain NFjB target genes [52], but we have no evidence that L. monocytogenes possesses analogous effector molecules. Instead, by escaping the vacuole, L. monocytogenes may avoid induction of inflammation by taking advantage of the inherent Cytosolic Response target specificity. Previous studies in which MDP was observed to induce pro-inflammatory cytokines used nonprimary, non-immune system cells, and often assessed host response as late as 24 hours after MDP delivery, all of which could have resulted in significantly altered chromatin states [8,9,[53][54][55].

The Role of NOD2 in Bacterial Pathogenesis
The role of NOD2 in innate immunity is controversial [56,57]. A previous study found that NOD2 was involved in the mouse innate immune response to L. monocytogenes, but only during intragastric infection [32]. This in vivo infection model may be particularly important for understanding the pathogenesis of food-borne bacteria such as L. monocytogenes. In certain cells of the gut epithelium, through which L. monocytogenes must pass in an oral model of infection, expression of many TLRs is naturally downregulated, possibly to reduce inappropriate responses to commensal flora [58]. This is consistent with our finding that NOD2-mediated synergistic induction of IFNb in response to intracellular bacterial infection was only manifest when TLR signaling was pharmacologically suppressed, and may also explain the results of a previous study in which rip2 À/À macrophages with intact TLR signaling had no defect in their response to L. monocytogenes [37]. Curiously, while irf3 À/À and ifnar À/À mice infected intravenously with L. monocytogenes were more resistant to infection [59][60][61], nod2 À/À mice infected intragastrically were more susceptible [32]. Given our results showing that infected nod2 À/À macrophages induced less IFNb than infected WT macrophages, the role of Type I interferons in intracellular bacterial pathogenesis may differ depending on the in vivo model used.

Materials and Methods
For more detailed versions of many of the following methods, please see: http://microbiology.berkeley.edu/ppmicroarrayhybridprotoc. htm.
Cell culture and bacteria. Macrophages were derived from the bone marrow of mice over 7 days in media composed of DMEM, 2 mM glutamine, 1 mM pyruvate, 10% CSF from 3T3 cells, 20% heat inactivated FBS, and penicillin-streptomycin. For all experiments, macrophages were grown in identical media without penicillinstreptomycin. For infections, WT 10403S (DP-L184) and hlyÀ (DP-L2161) L. monocytogenes were grown to mid-log in BHI media at 30 8C with shaking. Bacteria were then PBS-washed, resuspended in PBS at a normalized OD 600 ¼ 1.2, and added to macrophages at a 1:50 volume:volume ratio, resulting in . 75% of macrophages infected with at least one bacterium. For experiments with ''low MOI'' infection, and experiments involving Pam 3 CSK 4 -pre-treated macrophages, resuspended bacteria were added at a 1:1000 ratio. At 30 minutes post-infection macrophages were washed 3 times with fresh pre-warmed media, and at 60 minutes post-infection gentamicin was added to a final concentration of 50 lg/ml. For experiments containing a 30 minute post-infection time point, macrophages were instead washed at 20 minutes and gentamicin added at 30 minutes. CAPE (Calbiochem) (or EtOH for mock-treated controls) was added 60 minutes post-infection. For tolerization of cells, Pam 3 CSK 4 (Invivogen) was added at a final concentration of 100 ng/ml, 24 hours prior to the subsequent experiment. Where indicated, cycloheximide (or water for mock-treated controls) was added at a final concentration of 10 lg/ml 30 minutes before infection with bacteria, and added back after the washes at t ¼ 30 minutes. This treatment reduces macrophage translation by 94.5% (determined by 35 S-Met incorporation, data not shown).
For infections with M. tuberculosis, macrophages were infected with the Erdman strain as previously described [45]. Briefly, M. tuberculosis cultures were washed with PBS and resuspended in DMEM supplemented with 10% horse serum. Pam 3 CSK 4 -pre-treated macrophages were incubated with bacteria in DMEM þ 10% horse serum for 2 hours at an MOI of 10. Cells were then washed three times with PBS and returned to macrophage media.
Mice. All macrophages were from mice in the C57BL/6 genetic background, including femurs from knockout mice (see Acknowledgments).
Macrophage RNA preparation. Macrophage RNA was isolated with the Ambion RNAqueous kit (Applied Biosystems) according to the manufacturer's protocol, after first treating the cells with Ambion RNAlater (Applied Biosystems). For microarray experiments, RNA was amplified to generate amplified RNA (aRNA) using the Ambion Amino-Allyl Message Amp II aRNA Amplification Kit (Applied Biosystems) according to the manufacturer's protocol. For qPCR experiments, RNA was DNase treated with the Ambion TURBO DNAfree kit (Applied Biosystems).
Microarrays. Microarrays were printed at the UCSF Center for Advanced Technology, using the MEEBO 70-mer oligonucleotide set (Illumina; for more details see http://alizadehlab.stanford.edu/). Microarray probes were generated by coupling aRNA to Cy dye monofunctional NHS esters (Amersham) using 100 mM sodium bicarbonate (pH 9.0) and 50% DMSO. 5 lg of Cy5-coupled sample and 5 lg Cy3coupled reference (generated by pooling an equal volume of every sample used in a given set of arrays) were hybridized to MEEBO microarrays at 63 8C for 2 days. After washing, arrays were scanned on a GenePix 4000B scanner (Molecular Devices). Arrays were gridded with SpotReader software (Niles Scientific). Acquisition of data was performed using the GenePix Pro 6 software package (Molecular Devices). Data was normalized by first using stringent criteria to identify a subset of features of highest quality, and then calculating a normalization factor such that the ratio of medians of the Cy5 and Cy3 values for these features ¼ 1. This ratio was then applied to all the features. Features not meeting minimum criteria to assure quality were removed from the datasets. These criteria are available upon request. Hierarchical Pearson clustering and other analyses were performed with the Acuity 4 software package (Molecular Devices). The determination of Vacuolar Response and Cytosolic Response target genes was each from multiple arrays representing four independent experiments (e.g. 4 independent dishes of uninfected myd88 À/À macrophages and 4 independent dishes of myd88 À/À macrophages infected with WT L. monocytogenes for 180 minutes were used for Cytosolic Response determination), and other determinations were from 2-4 independent experiments. Fold change in RNA abundance is relative to uninfected samples. SAM analysis (Stanford University) [34] was performed with two-class unpaired designs to identify genes that were differently expressed in infected versus uninfected macrophages. For initial target gene discovery (defining the Vacuolar Response and Cytosolic Response), the false discovery rate (FDR) was set to 1%. For subsequent SAM analyses the FDR was set to 10%. Information linked to each unique Oligo ID can be found at http://meebo.ucsf.edu:8080/meebo/meeboInfo.jsp?oligoid¼(insert Oligo ID here). The GEO accession number for the primary array data is GSE8104. Accession numbers for genes of the Primary Cytosolic Response are as follows: NM_010510 (IFNb), NM_008331 (IFIT1), NM_008332 (IFIT2), NM_010501 (IFIT3), NM_020557 (TYKI), NM_030015 (PELI1), NM_008654 (MYD116), NM_031168 (IL6), BC030067 (CXCL10), XM_130015 (OTUD1), NM_027835 (MDA5), NM_018738 (IGTP), and NM_145209 (OASL1). See Dataset S3 for more information on the Primary Cytosolic Response target genes.
qPCR. DNase-treated macrophage RNA was reverse transcribed and subject to quantitative PCR using DyNAmo SYBR Green 2-step qRT-PCR reagents (NEB/Finnzyme) and was performed on an Mx3000P machine and analyzed using MxPro software (Stratagene). The sequences of gene-specific primers are as follows: tggcattgttaccaactgggacg (5' b-actin), gcttctctttgatgtcacgcacg (3' b-actin), gcactgggtggaatgagactattg (5' IFNb), ttctgaggcatcaactgacaggtc (3' IFNb), gacctgttctttgaagttgacgg (5' IL1b), and tgtcgttgcttggttctccttg (3' IL1b). All RNA abundances are normalized to b-actin, and fold induction is relative to mock-infected samples. Data shown is the mean of at least three, and in most cases four, independent experiments (except for the data shown in Figure 5C, which represents the average of two independent experiments), and error bars represent the standard error of the mean. Statistical analyses (t-Tests with equal variances assumed) were performed using the Analysis Package Add-In feature of Microsoft Excel, and p-values are noted in figure legends.
Cell fractionation and Western blots. Macrophages were fractionated using NE-PER reagents (Pierce) supplemented with HALT protease inhibitors (Pierce) following the manufacturer's protocol. Protein concentration was determined using BCA reagents (Pierce) and equal masses of protein were run on 10% NuPAGE bis-tris gels (Invitrogen). Blots were probed with the following primary antibodies, all from Santa Cruz Biotech: anti-NFjB p65 (sc-372X), anti-IRF3 (sc-9082X), anti phospho-c-Jun (sc-16312X), anti-phospho-ATF2 (sc-8398X), and anti-Lamin B (sc-6217). Protein blots were probed with a secondary antibody covalently attached to an infrared emitting fluorphor (IRdye-680 and IRdye-800, Li-Cor Biosciences). Blots were scanned using the Odyssey Infrared Imaging System (Li-Cor Biosciences), and quantitated using the accompanying software package. Relative protein abundances were normalized to Lamin B. Data was collected from two blots, with each blot using pooled lysates of two independent dishes (four dishes total, except for the data shown in Figure 2A, which is from two independent dishes). The nonquantitative Western blot shown in Figure 2A was instead probed with an HRP-coupled secondary antibody, and developed with the ECL Plus Western Blot Detection System (GE Healthcare).
Transfections. Macrophages were transfected using FuGene 6 Transfection Reagent (Roche) according to the manufacturer's protocol. L. monocytogenes genomic DNA was prepared by manual disruption of mid-log WT 10403S bacteria using glass beads and phenol:chloroform. DNA was extracted 3 times with phenol:chloroform, treated with RNases A and H, and extracted again with phenol:chloroform. DNA was then digested to completion using EcoRI and BamHI, yielding fragments averaging 1-5 kb, extracted with phenol:chloroform, and precipitated. All phenol:chloroform extractions were done using PhaseLock gel (Eppendorf). DNA was dissolved in pyrogen-free water and used at a final concentration of 250 ng/ml. MDP (Calbiochem) was used at final concentration of 100 lg/ml. Poly(dAT-dTA) double-stranded nucleic acid (Amersham) was dissolved in 100 mM NaCl and used at a final concentration of 250 ng/ ml. All transfection experiments were done for 6 hours. CAPE (Calbiochem) (or EtOH for mock-treated controls) was added at the indicated concentration 180 minutes post-infection. Cycloheximide (or water for mock-treated controls) was added at a final concentration of 10 lg/ml 180 minutes after initial ligand transfection.

Supporting Information
Dataset S1. Found at doi: 10 4 We previously observed that WT macrophages infected with WT L. monocytogenes exhibited persistent TLR-mediated signaling, even at later time points. From our microarray analyses, we found that L. monocytogenes induced IL1b exclusively by MyD88-dependent signaling pathways (summarized in Figure 1B). However, IL1b was still strongly induced at later time points in WT macrophages infected with WT L. monocytogenes, when the bacteria were in the cytosol. This suggested that these cells continued TLR signaling from the vacuole even after the bacteria had escaped from this compartment. As TLR signaling strongly activated NFjB (Figure 2A), we considered it likely that this residual TLR signaling masked any contribution by NOD2 signaling to nuclear NFjB. Indeed, we found that both WT and nod2 À/À macrophages infected with WT L. monocytogenes induced IFNb to nearly identical levels ( Figure S4A). To remove residual TLR signaling in these cells we treated macrophages with Pam 3 CSK 4 , a synthetic TLR2 agonist. Pretreatment for 24 hours with Pam 3 CSK 4 had previously been shown to ''tolerize'' macrophages, causing their TLR signaling to dampen after transitory stimulation [43]. Compared to non-tolerized macrophages, Pam 3 CSK 4 -treated WT macrophages almost fully ceased the induction of IL1b 2 hours after infection with WT L. monocytogenes ( Figure S4B). These tolerized macrophages were still capable of appropriate TLR-mediated signaling, as we observed robust and sustained induction of IL1b during infection of Pam 3 CSK 4 -treated WT macrophages with hlyÀ L. monocytogenes, which remain in the vacuole ( Figure S4C).