Identification of two classes of TLR4-induced genes

We first tested the hypothesis that induction of LPS tolerance would selectively inhibit expression of pro-inflammatory genes, whereas genes encoding antimicrobial effectors would remain inducible. Consistent with previous studies, induction of the pro-inflammatory cytokine interleukin-6 (IL-6) was abolished after an initial LPS stimulation, and this state of LPS tolerance persisted for 24–48 h (Fig. 1a, b)⁸. However, an antimicrobial gene, Cnlp (cathelicidin-related antimicrobial peptide; also called Camp), remained inducible in tolerant macrophages under the same conditions (Fig. 1c). We next performed a microarray analysis comparing unstimulated macrophages, naive macrophages stimulated with LPS, or tolerant macrophages re-stimulated with LPS (Supplementary Table 1). As expected, several hundred genes induced during the first stimulation were either not re-induced or induced to a much lesser degree by a second stimulation at the 24-h time point, when macrophages exhibit maximal tolerance (Fig. 1d). Notably, a second group, also comprising several hundred genes induced during the first LPS stimulation, was induced at equal or greater levels after the stimulation of tolerant macrophages (Fig. 1d). We have categorized these genes into two classes: class 'tolerizeable' (genes not inducible in tolerant macrophages, class T) and class 'non-tolerizeable' (genes inducible in tolerant macrophages, class NT). We verified these distinct expression patterns by analysing at least ten representative genes with known function from each category by quantitative polymerase chain reaction (qPCR) (Fig. 1e). Although both T and NT classes contained genes from multiple functional categories (Supplementary Fig. 1 and Supplementary Table 2), the biological significance of differential regulation is most obvious for pro-inflammatory (class T) and antimicrobial (class NT) genes. Therefore, for further analyses we chose several genes that belonged to these functional groups and other select genes that displayed robust regulation.

Figure 1: Identification of class T and class NT genes.

a, BMMs were left untreated (naive, N) or stimulated with 100 ng ml^-1 LPS for 24 h (tolerant, T), washed with PBS and given media (N, T) or 10 ng ml^-1 LPS (N+L, T+L). Cells were assayed at time points post-stimulation. b, BMMs were pretreated with 100 ng ml^-1 LPS for the indicated times and restimulated with 10 ng ml^-1 LPS for 24 h. Supernatants were analysed for IL-6 by ELISA. c–e, BMMs were stimulated as described in a. RNA was harvested after 4 h and analysed by Affymetrix genechip (d) or by RT-qPCR (c, e). f, BMMs were stimulated with 100 ng ml^-1 LPS+/-Dex for 6 h and analysed by RT–qPCR. g, BMMs were stimulated with (black squares) or without (white squares) 100 ng ml^-1 LPS, washed at 24 h and restimulated with 10 ng ml^-1 LPS. RNA was analysed by RT–qPCR at indicated times. h, BMMs were stimulated as described in a and analysed by ChIP (RNA Pol II). b, c, e–h, Data are representative of 2 or more independent experiments. Data show mean  s.e.m. from triplicate values. d, Data represents the average expression of two independent experiments.

High resolution image and legend (143K)

The glucocorticoid receptor (GR) is known to negatively regulate pro-inflammatory genes^{17, 18}. We found that the GR agonist dexamethasone selectively inhibited the induction of several class T genes, but did not inhibit, and in some cases enhanced, the induction of several class NT genes (Fig. 1f and Supplementary Fig. 2a). This differential sensitivity to GR further supports the biological significance of distinction between the two classes of LPS-induced genes. One mechanism by which GR exerts gene-specific effects is the differential use of IRF3 as a co-activator for NF-B-dependent transcription¹⁷. We found, however, that most genes in both classes were dependent on IRF3 to varying degrees (data not shown).

We next asked whether the two classes of genes were regulated at the level of transcription. First, we analysed expression kinetics of several class T and class NT genes. Although Il6 (class T) was not re-inducible in tolerant macrophages, Fpr1 (formyl peptide receptor 1; class NT) was induced to a greater extent and with faster kinetics in tolerant cells than in naive cells (Fig. 1g). Several other NT genes also appeared to be 'primed' by the first stimulus (Supplementary Fig. 2b). We measured the stability of messenger RNAs of several T and NT genes and found that several mRNAs from both classes had an overlapping range of stabilities that did not correlate with their expression in tolerant macrophages (Supplementary Fig. 3a). Finally, we performed a chromatin immunoprecipitation (ChIP) assay for RNA polymerase II (Pol II). Although Pol II was inducibly recruited to both classes of promoters in naive macrophages, in tolerant macrophages Pol II was only recruited to the class NT promoters and this occurred with faster kinetics than in naive macrophages (Fig. 1h). These results demonstrate that class NT genes, unlike class T genes, are indeed transcriptionally inducible in tolerant macrophages.

We next asked whether the two classes of genes are regulated by the same signals. Consistent with previous reports^{11, 12, 13, 14, 15}, we found that the activation of NF-B and mitogen-activated protein kinases (MAPK) was deficient in tolerant macrophages (Supplementary Fig. 4a). MAPK inhibitors blocked the induction of several class T and NT genes in naive macrophages (Supplementary Fig. 4b), as well as the induction of class NT genes in tolerant cells (Supplementary Fig. 4c), suggesting that the same signalling pathways contribute to class NT gene induction in naive and tolerant macrophages. Class T and NT genes were induced similarly by sub-optimal doses of LPS in naive macrophages (Supplementary Fig. 3b), suggesting that class NT genes are not intrinsically more sensitive to LPS. There was no correlation between interferon (IFN)-/ receptor dependence and class T and NT gene expression, indicating that class NT genes are not regulated through IFN-/ feedback (data not shown). Finally, class NT genes were not induced by conditioned media from LPS-stimulated macrophages, suggesting that class NT gene induction is not the result of positive feedback of any other secreted factor (Supplementary Fig. 3c).

Collectively, these results indicate that even attenuated TLR4 signalling in tolerant macrophages is sufficient for the induction of class NT genes, but not class T genes.

Distinct histone modifications at T and NT promoters

These results suggested that the two classes of LPS-induced genes are differentially regulated by gene-specific characteristics. Gene-specific regulation occurs at the level of chromatin and includes nucleosome remodelling and covalent histone modifications^{19, 20}. Histone acetylation is a positive mark associated with transcriptionally active chromatin, whereas deacetylated histones are found in closed, inactive chromatin^{21, 22, 23}. Although promoters of both classes were inducibly acetylated at histone H4 in naive macrophages, only histones at class NT promoters were re-acetylated after stimulation of tolerant macrophages (Fig. 2a). The kinetics of induction of histone acetylation at class NT promoters in tolerant macrophages mirrored gene expression: acetylation persisted at a higher level relative to unstimulated cells, and was re-induced by stimulation of tolerant cells, in some cases with faster kinetics and to a greater extent (Fig. 2b and Supplementary Fig. 2c). Thus, class T and class NT genes exhibit distinct patterns of inducible histone acetylation that correspond to their transcriptional activity.

Figure 2: Histone modifications are differentially regulated at class T and NT promoters.

Naive and tolerant BMMs (N, T) were stimulated with LPS for a, 3 h (N+L) or 1 h (T+L) or b, the indicated times and analysed by ChIP (AcH4). c–e, Naive BMMs were stimulated with LPS (N+L, white bars) or LPS+TSA (N+L, black bars). Tolerant BMMs (T+L) were prepared as above (white bars) or with TSA and stimulated with LPS+TSA (black bars). RT–qPCR (c), ELISA (d) and ChIP (AcH4) (e) were performed at 4 h (c, e) or 24 h (d). f, Naive and tolerant BMMs (N, T) were stimulated with LPS for 3 h (N+L, T+L) and analysed by ChIP (H3K4me3). g, h, Naive and tolerant BMMs were prepared with pargyline as in c and e. RT–qPCR (g) and ChIP (H3K4me3) (h) were performed at 4 h. a–h, Data are representative of 3 or more independent experiments; shown are mean  s.e.m. from triplicate values.

High resolution image and legend (153K)

We next stimulated macrophages in the presence of trichostatin A (TSA), a histone deacetylase inhibitor, and measured the induction of Il6. As others have shown previously, TSA inhibited Il6 expression in naive macrophages stimulated with LPS, presumably by affecting the acetylation of transcription factors^{24, 25}. However, inhibition of histone deacetylases during the first LPS stimulation of naive macrophages reversed silencing of Il6 and several other class T genes in tolerant macrophages (Fig. 2c, d and data not shown). TSA treatment did not affect NF-B or MAPK activation (Supplementary Fig. 5a), but did result in an increased level of histone acetylation at the Il6 promoter in tolerant macrophages (Fig. 2e). Thus, selective histone deacetylation contributes to the silencing of class T genes in tolerant macrophages.

Trimethylation of histone H3 at lysine 4 (H3K4me3) also marks transcriptionally active genes^{26, 27, 28, 29}. We found that H3K4 trimethylation was induced in naive macrophages at both classes of promoters (Fig. 2f). Interestingly, following LPS stimulation of tolerant macrophages, this modification was rapidly and selectively lost at class T promoters, but was maintained at class NT promoters (see Fig. 2f, and below). Treatment of macrophages with pargyline, an inhibitor of H3K4 demethylase LSD1 (ref. 30), prevented Il6 silencing in tolerant macrophages (Fig. 2g), and maintained H3K4me3 levels at the Il6 promoter (Fig. 2h). Pargyline treatment did not affect NF-B and MAPK activation by LPS (Supplementary Fig. 5b). Thus two types of positive histone modifications, H4 acetylation and H3K4 trimethylation, are selectively lost at the class T promoters; inhibiting their loss prevents silencing of the class T genes in tolerant macrophages.

Nucleosome remodelling at T and NT promoters

We next examined the recruitment of two ATP-dependent chromatin remodelling complexes, Brg1 and Mi-2, to class T and class NT promoters in naive and tolerant macrophages²⁰. Following LPS stimulation, Brg1 was recruited to both classes of promoters in naive macrophages, but was only recruited to class NT promoters in tolerant macrophages (Fig. 3a). Similar to the recruitment of Pol II, Brg1 was recruited with faster kinetics to class NT promoters in tolerant macrophages (Fig. 3a, Fpr1). As shown previously, Mi-2 was recruited simultaneously with Brg1 in all cases (data not shown)³¹.

Figure 3: Chromatin remodelling is differentially regulated at class T and NT promoters.

a, Naive and tolerant BMMs were left untreated (N, T) or were stimulated with LPS (N+L, T+L) for 3 h. Cells were fixed with DMA/formaldehyde and analysed by ChIP (Brg1). b, BMMs were stimulated as in a for 3 h (Il6 (N+L, T+L), Fpr1 (T+L)) or 8 h (Fpr1 (N+L)) and analysed by REA/LM-PCR. Input (Inp.) and experimental (Exp.) amplification products are shown. a, b, Data are representative of 3 or more independent experiments; shown are mean  s.e.m. from triplicate values.

High resolution image and legend (50K)

Consistent with the pattern of Brg1 recruitment, Il6 and Fpr1 promoters became nuclease accessible in naive macrophages stimulated with LPS. However, following stimulation of tolerant macrophages, Il6 was inaccessible, whereas Fpr1 was highly accessible at even earlier time points than in naive macrophages (Fig. 3b). Interestingly, the accessibility of the Fpr1 promoter was differentially regulated in tolerant macrophages at two different sites in the promoter, with one site displaying a stable increase in accessibility in tolerant macrophages in the absence of second stimulation (Fig. 3b, Fpr1 TSS). These data demonstrate that in tolerant macrophages chromatin remodelling is inhibited at class T promoters, whereas class NT promoters are both stably and inducibly remodelled.

TLR4-induced transcripts regulate T and NT genes

We next investigated whether gene products induced by LPS in naive macrophages are required for the silencing of class T genes and the priming of class NT genes in tolerant macrophages. The transcription elongation inhibitor DRB was used to block LPS-induced transcription in naive macrophages. As expected, DRB inhibited the transcription of several class T genes in LPS-stimulated naive macrophages (Fig. 4a, b). Interestingly, DRB treatment of naive macrophages completely prevented silencing of Il6 and several other class T genes in tolerant macrophages (Fig. 4b, c). Accordingly, LPS stimulation led to the recruitment of Pol II, NF-B and Brg1 to the Il6 promoter in DRB pre-treated, but not in untreated, tolerant macrophages, and this recruitment occurred with faster kinetics than in naive macrophages (Fig. 4d). Similarly, we found increased H4 acetylation and sustained levels of H3K4me3 at the Il6 and Lipg promoters in DRB-treated macrophages (Fig. 4e). Thus, DRB treatment of naive macrophages results in a reversal of tolerance in LPS-pretreated macrophages, as determined by the induction of positive histone modifications (H4 acetylation and H3K4 trimethylation), the recruitment of transcription factors and nucleosome remodelling complexes, and finally, the recruitment of the Pol II complex and transcriptional induction of class T genes.

Figure 4: Transcription of new genes contributes to the tolerant signature.

a, Naive BMMs were left untreated (N), or were stimulated after 24 h with LPS (N+L, white bars) or LPS+DRB (N+L, black bars). Tolerant BMMs were prepared with or without DRB, washed with PBS, then left untreated (T) or stimulated with LPS (T+L). In all panels, white and black bars indicate stimulation in the absence or presence of DRB, respectively. b, f, RT-qPCR was performed 4 h after stimulation. c, Supernatants were analysed by ELISA for IL-6 24 h after stimulation. d, ChIP analysis was performed at 1 h (T+L1) or 3 h (N+L, T+L3) for RNA Pol II, p65 and Brg1. e, g, ChIP was performed at 1 h (T+L1), 2 h (T+L2) or 3 h (N+L, T+L3) for AcH4 or H3K4me3. b–g, Data are representative of 3 or more independent experiments; shown are mean  s.e.m. from triplicate values.

High resolution image and legend (161K)

Because negative transcriptional regulators induced by the first LPS stimulation seem to be responsible for silencing class T genes, we wondered if positive regulators similarly induced are responsible for priming class NT genes in tolerant macrophages. Therefore, we examined the effect of DRB on induction of class NT genes. Similarly to class T genes, class NT genes were inhibited by DRB treatment of naive macrophages stimulated with LPS (Fig. 4f). However, in contrast to the DRB-mediated reversal of silencing observed at class T genes, DRB pre-treatment prevented priming of class NT genes in tolerant macrophages. Both the magnitude of their transcriptional induction and the delayed kinetics of H3K4 trimethylation were similar to that seen in LPS-stimulated naive macrophages (Fig. 4f, g).

DRB treatment did not affect NF-B and MAPK signalling in naive and tolerant macrophages or the induction of negative regulators of TLR signalling, such as Irak-M (Supplementary Fig. 5c, d). These controls confirmed that the effect of DRB was gene-specific, rather than signal-specific. In addition, these experiments demonstrate that reduced signalling in tolerant macrophages is sufficient for class T induction in the absence of negative regulators, illustrating that permissive chromatin structure, not signal strength or specificity, is the determining factor for differential gene induction.

Altered transcriptional requirements of NT genes

As discussed above, class NT genes are induced with enhanced kinetics and magnitude in tolerant macrophages, indicating that their transcriptional requirements change as a result of the first exposure to LPS (Fig. 1g). One characteristic of inducible genes that correlates with the kinetics of induction is the requirement for protein synthesis³¹. Primary response genes are not dependent on protein synthesis and are induced within 2 h of stimulation. Secondary response genes are induced with later kinetics and their transcription depends on protein synthesis. Both class T and class NT genes contained primary and secondary genes. However, the change in kinetics of induction of class NT genes in naive and tolerant macrophages raised the possibility that those class NT genes that are secondary in naive macrophages may be converted into primary genes in tolerant macrophages. We therefore chose several secondary class NT genes and measured their induction in naive and tolerant macrophages in the presence of cycloheximide (CHX), a protein synthesis inhibitor. We found that although these genes were inhibited by CHX in naive macrophages, CHX treatment had no effect on the transcription of class NT genes in tolerant macrophages (Fig. 5a). Furthermore, Pol II recruitment to promoters of secondary class NT genes was CHX sensitive in naive macrophages, but CHX insensitive in tolerant macrophages (Fig. 5b). Similarly, the accessibility of the Fpr1 promoter (class NT) was dependent on new protein synthesis in naive, but not tolerant, macrophages (Fig. 5c). These results indicate that secondary class NT genes are indeed converted into primary genes as a result of macrophage 'priming' by the first stimulation with LPS.

Figure 5: Class NT genes have different transcriptional requirements in naive and tolerant macrophages.

a, Naive BMMs were stimulated with LPS (white bars) or LPS+CHX (black bars) for 6 h. Tolerant BMMs were left untreated (grey bars) or stimulated with LPS (white bars) or LPS+CHX (black bars) for 2 h. RT–qPCR was performed. b, c, BMMs were stimulated as in a, but cells were analysed by ChIP (RNA Pol II) (b) or REA/LM-PCR (c) after 3 h. a–c, Data are representative of 2 or more independent experiments; a, b, shown are mean  s.e.m. from triplicate values.

High resolution image and legend (74K)

Discussion

TLR activation induces expression of hundreds of genes in macrophages. These fall into multiple functional categories, including inflammatory cytokines, chemokines, antimicrobial proteins and peptides, tissue-repair and coagulation factors, and metabolic regulators. We reasoned that different components of the TLR-induced response should have different regulatory requirements that reflect their functions. We show here that TLR4-induced genes fall into two classes. On repeated exposure to LPS, one class of genes (class T, including inflammatory cytokines) is transiently silenced to prevent pathology associated with excessive inflammation. The second class of genes (class NT) includes antimicrobial effectors, which remain inducible to protect the host from infection.

Although our initial hypothesis was based on the functional distinction between pro-inflammatory and antimicrobial genes, a more comprehensive gene expression analysis revealed that other functional groups of LPS-induced genes belong to these two categories (Supplementary Tables 1, 2; Supplementary Fig. 1). Indeed, several groups have reported examples of anti-inflammatory gene induction in tolerant macrophages ^{32, 33, 34, 35, 36}. In all cases, the classification of genes as class T or NT probably depends on whether persistent inducibility of a gene would be deleterious or advantageous, and thus reflects a general principle linking gene regulation with the function of the encoded products. Because both classes of genes are induced by the same receptor, their differential regulation occurs by gene-specific rather than signal-specific mechanisms, through chromatin modifications at the level of individual promoters. We speculate that similar mechanisms operate in other multi-component transcriptional response programmes.

Several mechanisms regulate the TLR-induced inflammatory response and these collectively contribute to LPS tolerance in vitro and in vivo. Our results demonstrate that selective and transient silencing of inflammatory genes at the level of chromatin plays a critical role in LPS tolerance.

We show that gene products (presumably transcriptional regulators) induced by the first LPS stimulation of naive macrophages are required for the silencing of class T genes and the priming (enhanced inducibility) of class NT genes in tolerant macrophages (Fig. 6). In contrast to the long-term epigenetic memory that propagates inheritance of gene expression through cell divisions, this is an example of transient gene silencing in terminally differentiated macrophages. Stable gene silencing is associated with histone H3 methylation at H3K9 and H3K27 (ref. 37). We did not detect these modifications at class T genes in tolerant macrophages (data not shown), suggesting that this type of transient gene silencing occurs by a different mechanism.

Figure 6: Model for gene-specific regulation of class T and NT genes.

Following LPS stimulation of naive macrophages, class T and NT promoters exhibit transcription factor recruitment, increased histone acetylation, H3K4 trimethylation, and chromatin remodelling. In tolerant macrophages, H3K4 trimethylation is high at both classes of promoters, and class NT promoters exhibit increased levels of histone acetylation and accessibility. Following stimulation with LPS, class T promoters remain deacetylated and inaccessible, whereas class NT promoters become even more acetylated and accessible, this time with faster kinetics. TLR4-induced negative and positive factors contribute to the silencing of class T promoters and the priming of class NT promoters in tolerant macrophages.

High resolution image and legend (46K)

We found that following the initial exposure to LPS, class NT genes become modified so that their induction by a second LPS stimulation occurs with faster kinetics and an increased magnitude. This transcriptional memory may constitute an adaptive component of the innate immune response; as many NT genes encode antimicrobial effectors, this enhanced response would increase the efficiency of innate host defence.

A recent study found that primary and secondary genes differ in their requirement for inducible nucleosome remodelling mediated by Brg1 (ref. 31). Primary response gene products recruit Brg1 to the promoters of secondary response genes, which explains why secondary response genes require protein synthesis³¹. We found that the transcriptional requirements for class NT genes change during LPS stimulation, such that secondary NT genes are converted into primary genes in tolerant macrophages. This conversion may reflect either inducible recruitment of pre-made primary gene products to the promoters, or persistent changes in chromatin structure initiated by primary gene products during the first stimulation. Regardless, this finding explains why the induction of class NT genes is qualitatively and quantitatively different in naive and tolerant macrophages.

Collectively, these results indicate that gene products induced by LPS in naive macrophages differentially modify chromatin at class T and class NT promoters to silence the former and to prime the latter for their differential regulation by a second LPS stimulation.

Safe manipulation of the innate immune response and inflammation has been problematic because most known therapeutic agents inhibit the induction of both antimicrobial effectors and inflammatory cytokines. Here we have shown that the two types of TLR-induced responses can be dissociated and are differentially regulated, suggesting the existence of novel targets for selective control of inflammatory and antimicrobial responses.

Methods summary

Mice

C57BL/6 mice were obtained from Jackson Laboratories. Mice were maintained at the animal facility of Yale University School of Medicine and used at 8–12 weeks of age.

Reagents

LPS, 5,6-dichloro-1--d-ribofuranosylbenzimidazole (DRB), polymyxin B sulphate (PB), trichostatin A (TSA), dexamethosone (Dex), actinomycin D (ActD), pargyline (Parg) and cycloheximide (CHX) were purchased from Sigma. All restriction enzymes were purchased from New England BioLabs. Dimethyl adipimidate (DMA) was purchased from Pierce. p38 and JNK inhibitors were purchased from Calbiochem. Recombinant mouse IL-6 was purchased from R&D Systems. Paired antibodies for IL-6 and antibodies to phosphorylated p38, ERK, and JNK were purchased from BD Biosciences. Antibodies to H4-Ac (06-866), H3K4me3 (07-473) were purchased from Upstate Biotechnologies. Antibodies to RNA polymerase II (N-20) (sc-899X), p65 (C-20) (sc-372X), and IkB were purchased from Santa Cruz Biotechnologies. Antibody to Grp94 was purchased from Stressgen. Antibodies to Brg1 (J1) and Mi-2 were a kind gift from T. Chi (Yale Univ.) and S.T. Smale (UCLA), respectively.

Bone marrow macrophage (BMM) cultures

Bone marrow progenitors were harvested from mice and cultured for 7 days on Petri dishes in M-CSF supplemented RPMI-1640. Cells were lifted with cold PBS and replated on tissue-culture treated plates. On day 8, macrophages were left untreated (naive, N) or stimulated with 100 ng ml^-1 LPS for 24 h (tolerant, T), washed twice with warm PBS and given fresh media (N, T) or 10 ng ml^-1 LPS (N+L, T+L). Where indicated, BMM were treated with Dex (1 M), ActD (10 g ml^-1), TSA (50 nM), DRB (15 M), Parg (3 M), PB (50 g ml^-1) and CHX (100 g ml^-1). The p38/JNK inhibitor cocktail included PD98059 (45 M), SB203580 (2.5 M), SB202190 (0.5 M), PD169136 (5 M) and JNK inhibitor II (500 nM).

ELISA

Supernatants were collected 24 h after stimulation and IL-6 was detected with paired antibodies, using recombinant protein to generate a standard curve. Antibody binding was detected by streptavidin-horseradish peroxidase (Zymed) and developed with o-phenylenediamine dihydrochloride (Sigma).

Reverse transcription and quantitative PCR (RT–qPCR)

Total RNA from BMM was isolated with RNA-bee reagent (Tel-Test). Total RNA was reverse transcribed with an oligo (dT) primer using Superscript reverse transcriptase III (Gibco BRL). Complementary DNA was analysed in triplicate by qPCR amplification using SYBR Green QPCR Master Mix (Qiagen) on the MX3000P QPCR System (Stratagene). The PCR amplification conditions were: 95 °C (15 min), 45 cycles of 94 °C (30 s), 58 °C (30 s) and 72 °C (1 min). Primer pairs were designed to amplify mRNA-specific fragments and unique products were tested by melt-curve analysis. Data was analysed by comparative quantification using MXPro software with naive values set as calibrator and expression normalized to hypoxanthine guanine phosphoribosyl transferase 1 (Hprt). Data are represented as the fold induction over naive (unstimulated).

Microarray analysis

Total RNA was isolated with RNA-bee reagent (Tel-Test) and purified with RNEasy Kit (Qiagen). Two biological replicates were performed for each experimental condition. Sample preparation and hybridization to Affymetrix Mouse Genome 430 2.0 arrays were performed at the Yale W.M. Keck facility. Briefly, target cDNA generated from each sample was biotinylated, hybridized, and stained as per manufacturer's recommendation using an Affymetrix GeneChip Instrument System. Arrays were scanned on an Affymetrix GeneChip scanner 3000 according to Affymetrix standard protocols (GeneChip Expression Analysis Technical Manual, Affymetrix, 2004). Data was processed using Affymetrix Microarray Suite version 5.0, scaled to a target intensity of 500. Raw and normalized data have been submitted to the GEO database (http://www.ncbi.nlm.nih.gov/geo/), accession number GSE7348. Data were further analysed using GeneSpring (Silicongenetics). Analysed expression data are presented in Supplementary Table 1. Probesets present on only one of the 2 arrays for each condition were excluded, as were probesets with a detection call of 'absent' or 'marginal' in all three conditions. The signal intensity for each probeset was averaged over the two arrays, and probesets with a signal intensity ratio of less than 2 in stimulated versus unstimulated macrophages were excluded. Class T genes are defined as genes induced in naive macrophages stimulated with LPS and downregulated more than 3-fold in tolerant macrophages stimulated with LPS (that is, (N+L)/(T+L) > 3). Class NT genes are defined as genes induced in naive macrophages stimulated with LPS and expressed at equal or higher levels in tolerant macrophages stimulated with LPS (that is, (N+L)/(T+L)  1). Data are displayed as 'fold N+L' (signal N+L/signal N), 'fold T+L' (signal T+L/signal N), 'fold N+L/fold T+L' (class T) and 'fold T+L/Fold N+L' (class NT).

Restriction enzyme accessibility (REA) and ligation mediated (LM)-PCR

Isolated cell nuclei were incubated with limiting amounts of a restriction enzyme (AflII(il-6); Sac(fpr1TSS); Mse(fpr15')) at 37 °C for 10 min, and then treated with proteinase K overnight ('Experimental' digest). DNA was purified and digested to completion at an upstream site (Nhe(il-6); Mse(fpr1TSS); Alu(fpr15')) ('Input' digest). Digested fragments were ligated to complementary annealed linkers and amplified using SYBR Green QPCR Master Mix, a 5' primer derived from the linker, and a 3' gene-specific primer. The touchdown PCR amplification conditions were: 95 °C (15 min), 10 cycles of 94 °C (15 s), 70 °C (30 s) (-1°C per cycle), 72 °C (2 min), then 18 cycles of 94 °C (15 s), 60 °C (30 s), 72 °C (2 min). This reaction was further amplified for 2–3 cycles with a ³²P-labelled nested 3' gene-specific primer. The PCR products were run out on a 6% polyacrylamide gel and visualized by autoradiography. The ratio of the experimental digest (Exp.) to the input digest (Inp.) reflects the degree of accessibility at the experimental restriction site.

Chromatin immunoprecipitation (ChIP)

10  10⁶ BMM were stimulated, washed with PBS, and either fixed with 1% formaldehyde for 5 min at 37 °C (Pol II, AcH4, H3K4me3) or fixed with 25 mM DMA for 90 min and then fixed with 1% formaldehyde for 5 min at room temperature (Brg1, p65, Pol II (Fig. 5c only), AcH4 (Fig. 5d only), H3K4me3 (Fig. 5d, f only)). Formaldehyde fixation was stopped with the addition of 1.25 M glycine. Fixed cells were sonicated for either 3 min of 0.5on/0.5off pulses at setting 1, or 2 min 7 s of 0.5on/0.5off pulses at setting 5, to obtain fragments ranging from 200 to 700 bp in size. Sonicates were diluted 5 and incubated with antibody with rotating overnight. Protein A/G beads (Upstate Biotechnologies) were added for 3 h and collected beads were washed extensively. Protein–DNA complexes were eluted from the beads and treated with 200 mM NaCl to reverse cross-links and proteinase K to digest proteins. Recovered DNA was phenol:chloroform extracted and precipitated with isopropanol and glycogen. Immunoprecipitated DNA and input DNA were amplified with gene-specific and -globin (Hbb) primers by qPCR, using input DNA to generate a standard curve. ChIP data is represented as %input(gene-specific)/% input(-globin), except in Fig. 1e, where it is represented as %input.