Main

Inflammation is a fundamental adaptation to the loss of cellular and tissue homeostasis with many important physiological roles, including host defence, tissue remodelling and repair, and the regulation of metabolism1,2,3. The complexity of the inflammatory response requires that its many functional programmes are controlled coordinately in some situations but independently in others. This is achieved through multiple mechanisms that operate at different levels, including alterations in the composition of immune cells in tissues, changes in cell responsiveness to inflammatory stimuli, regulation of signalling pathways and control at the level of gene expression. So, the mechanisms that regulate inflammatory responses can be divided into cell-specific, signal-specific and gene-specific mechanisms. Cell-specific mechanisms operate at the level of different cell types, and include regulation of their recruitment and activation. Signal-specific mechanisms operate at the level of signalling pathways: for example, by terminating activation of the key transcription factor nuclear factor-κB (NF-κB) as a part of a negative feedback mechanism. Finally, gene-specific mechanisms operate at the level of individual genes and gene subsets. For example, interleukin-10 (IL-10) and many nuclear receptors negatively regulate the transcription of specific subsets of inflammatory genes. As such, gene-specific mechanisms are particularly well suited to provide functional specificity in an inflammatory response. Here, we review the recent progress in understanding the transcriptional regulation of the inflammatory response. In addition, we discuss the key factors and molecular mechanisms that regulate inflammatory gene expression, with the objective of defining some general principles that govern this important physiological process.

Macrophages are crucial mediators of the inflammatory response, and Toll-like receptors (TLRs) are the best-characterized inducers of acute inflammation. Therefore, much of this Review focuses on the transcriptional regulation of inflammation by TLR ligands — and in particular by lipopolysaccharide (LPS) — in macrophages. After only a few hours of LPS stimulation, the expression of several hundred genes is induced (and repressed) in macrophages4,5. This is a complex transcriptional response, consisting of multiple gene sets that encode functional programmes controlling cell migration, tissue repair and remodelling, antimicrobial defence, phagocytosis, metabolic reprogramming and the regulation of adaptive immune responses. These gene sets, or transcriptional modules, are often coordinately regulated by dedicated transcription factors. This feature of the transcriptional response enables autonomous control of individual transcriptional modules, because the transcriptional regulators that control their expression can be differentially regulated by positive and negative signals (Box 1).

Several examples of transcriptional regulators that control distinct transcriptional modules are currently known, both within and outside of the LPS-induced transcriptional response. These include class II transactivator (CIITA), which is a master regulator of genes involved in the MHC class II-restricted pathway of antigen processing and presentation6; interferon (IFN)-stimulated gene factor 3 (ISGF3), which controls the expression of type I IFN-induced antiviral genes; sterol regulatory element binding protein 2 (SREBP2), which controls the expression of genes involved in cholesterol biosynthesis7; and several stress-induced transcriptional regulators (such as N-ethylmaleimide-sensitive factor 2 for cellular stress induced by reactive oxygen species8, hypoxia-inducible factor 1α for the hypoxic response9, X-box-binding protein 1 for the unfolded protein response10 and aryl hydrocarbon receptor for the xenobiotic-induced response11). In each case, the genes that comprise a given transcriptional module are functionally related, which explains the requirement for their coordinated control by dedicated transcriptional regulators. Importantly, the full repertoire of TLR-induced transcriptional modules is currently unknown, as are the transcriptional master regulators of these modules. Nevertheless, the concept of transcriptional modules is useful when considering the heterogeneity of a complex transcriptional response, such as that induced by LPS in macrophages.

Here, we first review what is known regarding the regulation of the LPS-induced transcriptional response by transcription factors, chromatin modifications and transcriptional co-regulators in macrophages, and the relative roles of each of these components in the inflammatory response. Then, we describe the mechanisms by which various signalling pathways modulate this transcriptional programme in distinct biological contexts. Finally, we emphasize the modular nature of the transcriptional control of inflammation as being central to its physiological regulation and therapeutic manipulation.

Transcription factors

Induction of the LPS-dependent transcriptional response in macrophages is orchestrated by many transcription factors, consistent with the complexity of the response. These transcription factors can be divided into three categories on the basis of their mode of activation and function. This classification is not intended to demarcate mutually exclusive groups of transcription factors, but to illustrate general principles regarding their mechanisms of action and their role in the control of various inducible transcriptional modules in macrophages.

The first category (class I) consists of transcription factors that are constitutively expressed by many cell types and that are activated by signal-dependent post-translational modifications. In most cases, these transcription factors are retained in the cytoplasm in the basal state and their signal-dependent activation involves their nuclear translocation. This class is the best characterized of the three categories of transcription factors and it includes proteins that are known to have important roles in inflammation, such as NF-κB, IFN-regulatory factors (IRFs) and cAMP-responsive-element-binding protein 1 (CREB1). The genes that are induced most rapidly by LPS stimulation (the so called primary response genes) are regulated by these transcription factors (Fig. 1).

Figure 1: Lipopolysaccharide (LPS)-induced primary and secondary response genes are regulated by three categories of transcription factors.
figure 1

The first category (class I) consists of transcription factors that are activated post-translationally by Toll-like receptor (TLR) signalling, often at the step of nuclear translocation. Examples include nuclear factor-κB (NF-κB) and interferon-regulatory factor (IRF) proteins. These transcription factors control the induction of the primary response genes. The second category (class II) is comprised of transcription factors that are induced during the primary response, such as CCAAT/enhancer-binding protein-δ (C/EBPδ), which control induction of the secondary response genes. A third category of transcription factors (class III), which includes PU.1, C/EBPβ, runt-related transcription factor 1 (RUNX1) and IRF8, is not directly targeted by pro-inflammatory signals but is induced during macrophage differentiation, and has a key role in specifying macrophage-specific patterns of inducible gene expression. ATF3, activating transcription factor 3.

Full size image

There are multiple mechanisms that quickly terminate the activation of NF-κB and IRFs; for example, inhibitor of NF-κB-α (IκBα) exports NF-κB from the nucleus and/or facilitates its removal from the promoters of target genes12,13. However, positive feed-forward mechanisms might ensure the sustained activation of these transcription factors and their participation in subsequent waves of gene induction; for example, the production of tumour necrosis factor (TNF) triggered by LPS stimulation seems to be crucial for autocrine signalling and induction of a second wave of NF-κB activation14,15.

The second category of transcription factors (class II) are synthesized de novo after LPS stimulation. It is estimated that approximately 50 proteins make up this group, many of which have poorly defined functions in the context of the LPS-induced transcriptional response. These transcription factors regulate subsequent waves of gene expression after the primary response genes, and they can do so over a prolonged period of time5. The activity of these transcription factors is often subject to positive feedback control, and because these proteins are transcriptionally upregulated, a general principle here seems to be transcriptional autoregulation. For example, the amplification of the LPS-induced transcriptional response by CCAAT/enhancer-binding protein-δ (C/EBPδ) requires its autoinduction16. The stable upregulation of expression of these transcription factors could also enable the reprogramming of macrophage functions. For this reason, we speculate that some transcription factors in this category might function as master regulators of distinct functional modules.

The third category of transcription factors (class III) consists of lineage-specific transcriptional regulators, the expression of which is turned on during macrophage differentiation. Notable members of this group include PU.1 (also known as SPI1) and C/EBPβ, as well as runt-related transcription factor 1 (RUNX1) and IRF8 (Refs 17,18). Although none of these transcription factors is exclusive to macrophages (for example, PU.1 is also expressed by B cells), they are induced during macrophage differentiation and their combinatorial expression specifies the macrophage phenotype. These proteins turn on constitutively expressed genes in macrophages, remodel chromatin at inducible genes and silence genes that are associated with alternative cell fates. In mature macrophages, these transcription factors mediate cell type-specific responses to inflammatory signals and other stimuli, presumably by conferring a permissive chromatin state on macrophage-specific inducible genes. A unique mode of action of at least some of the transcription factors in this category is the organization of cell type-specific, higher order chromatin structure or chromosomal domains. In particular, RUNX1 has been shown to 'anchor' specific genomic loci to the nuclear matrix (the architectural scaffold of the nucleus) to assemble domains of active or inactive chromatin19 that could determine the macrophage-specific patterns of active, silenced and inducible genes on a global scale.

The transcription factors of the three categories mentioned above do not act independently, but function coordinately to control the LPS-induced transcriptional response. Using a systems biology approach, Aderem and colleagues have identified some of the regulatory circuits that are shedding light on the logic of this combinatorial control4,16,20. They have found, using global kinetic profiling, that LPS-induced gene expression in macrophages can be divided into a limited number of distinct patterns of gene induction (and repression). Using this approach, combined with motif scanning of the promoters of these genes in silico, they can define sets of genes that seem to be coordinately regulated and the transcription factors that are likely to control their expression. In this way, they showed that the sustained expression of several inflammatory genes is mediated by a transcriptional network that consists of three transcription factors: NF-κB, which functions as the activator; activating transcription factor 3 (ATF3), which functions as the inducible repressor; and C/EBPδ, which functions as the inducible activator and 'amplifier'16. These observations also illustrate how combinatorial control by multiple transcription factors enables NF-κB, which is required for the induction of most of the LPS-induced genes, to engage in module-specific regulation of inflammatory gene expression.

In addition, the temporal characteristics of transcription factor activation (and attenuation) might be essential for determining the repertoire and specificity of a complex transcriptional response. A series of elegant studies by Hoffmann and colleagues have shown that perturbations in the dynamics of NF-κB activation can result in profound, qualitative changes in gene expression15,21. This is associated not only with stimulus-intrinsic alterations in inducible gene expression, but also results in the modification of transcriptional responses triggered by other signalling pathways that activate NF-κB22,23.

So, the complex transcriptional programme induced in macrophages after LPS stimulation is a product of the coordinated action of the three categories of transcription factors described above (Table 1).

Table 1 Three main classes of transcription factors regulate the lipopolysaccharide-induced transcriptional programme
Full size table

Dynamic chromatin remodelling

Recent studies have highlighted an important role for chromatin in the control of inflammatory gene expression24. DNA methylation and RNA interference are other epigenetic mechanisms for regulating gene expression, and future studies will shed light on the functional interplay between these processes in the control of inflammation. Here, we discuss the role of chromatin modifications in regulating inflammatory gene expression, with a focus on covalent histone modifications (Box 2). We describe what is known about how differential histone modifications can mediate gene-specific transcriptional regulation and discuss how chromatin architecture delimits cell type and signal specificity in the transcriptional control of inflammation.

Histone modifications. Several histone modifications have been shown recently to differentially regulate subsets of LPS-induced genes and they are of particular interest because of their role in regulating specific transcriptional modules within the multicomponent transcriptional response mediated by LPS-induced signalling (Box 2). One of the first studies in this area indicated that phosphorylation of histone 3 at serine 10 (H3S10) might have a gene-specific role in NF-κB recruitment25. After LPS stimulation, the genes encoding IL-6, IL-12p40 and CC-chemokine ligand 2 (CCL2), but not TNF and CCL3, undergo H3S10 phosphorylation at their promoters. This phosphorylation event depends on the mitogen-activated protein kinase p38, and specific blockade of p38 activation inhibits H3S10 phosphorylation, NF-κB recruitment and gene induction25. It is unclear how the phosphorylation of H3S10 is coupled to NF-κB recruitment, but this model is consistent with other studies showing that H3S10 phosphorylation is associated with transcriptional activity26. Moreover, the inhibition of inflammatory gene induction by some pathogens is associated with disruption of H3S10 phosphorylation (see below).

It is not known whether H3S10 is a direct substrate of p38, but the data indicate that p38 or a p38-regulated kinase, such as mitogen- and stress-activated kinase 1 (MSK1) or MSK2, might be recruited to a subset of genes to confer H3S10-dependent transcriptional regulation. IκB kinase-α (IKKα) can also be recruited to gene promoters in a signal-dependent manner to phosphorylate H3S10 (Refs 27,28). These examples illustrate that chromatin can be a direct, albeit distal, target of signal transduction cascades, and they underscore the fundamental role of chromatin in regulating gene–environment interactions; however, it is not known what directs the promoter-specific targeting of these kinases.

Perhaps even more interesting are the inhibitory histone modifications that mark subsets of inducible genes. Ubiquitylation of H2A at lysine 119 (H2AK119) is one such modification that inhibits the basal expression level of some LPS-inducible genes — for example, CCL5, CXC-chemokine ligand 10 (CXCL10) and CXCL2, but not CXCL1 — in a macrophage cell line29. This study proposed that ubiquitylation of H2AK119 at these gene promoters is necessary to prevent recruitment of the elongation factor FACT (facilitates chromatin transcription) complex. LPS-induced signalling triggers the gene-specific recruitment of the deubiquitylating enzyme 2A-HUB (also known as DZIP3), the activity of which is necessary for H2AK119 deubiquitylation and hence gene induction29. Consistent with this model, small interfering RNA (siRNA)-mediated knockdown of 2A-HUB is sufficient to decrease the level of basal H2AK119 ubiquitylation and leads to the recruitment of FACT to gene promoters and small but significant increases in the level of basal gene expression29. This indicates that at these promoters, H2AK119 ubiquitylation might be necessary to prevent low-level, 'leaky' gene expression. It is interesting that some, but not all, inflammatory genes seem to require this additional regulatory mechanism for maintaining basal repression; perhaps these genes encode proteins with biological functions that must be particularly tightly regulated.

Another subset of LPS-inducible genes undergoes signal-dependent demethylation of trimethylated H3K27 as a prerequisite for induction. The enzyme responsible for this activity, lysine-specific demethylase 6B (KDM6B; also known as JMJD3), is itself transcriptionally upregulated after LPS stimulation, and siRNA-mediated knockdown of KDM6B inhibits demethylation and the induction of Bmp2 (bone morphogenetic protein 2) expression30. This is consistent with previous studies that have established a role for H3K27 trimethylation in maintaining gene silencing through the recruitment of Polycomb proteins. As removal of the H3K27 inhibitory modification enables gene expression during cellular differentiation31, KDM6B-dependent regulation of an LPS-inducible gene30 indicates that macrophage activation has some features that are similar to differentiation.

It is important to point out that repressive histone modifications, such as H2AK119 ubiquitylation and H3K27 trimethylation, are additional regulatory checkpoints of inducible gene expression. Stimulus-dependent induction of genes bearing these modifications requires recruitment of the appropriate chromatin-modifying enzymes; hence, these histone modifications restrict both the types of biological signal and the classes of genes that are induced. Moreover, these repressive histone modifications could be targeted by anti-inflammatory signalling pathways as a means to enforce gene-specific inhibition of the inflammatory response (see below). It should be interesting to characterize further the LPS-induced transcriptional response with respect to other inhibitory epigenetic modifications, including H3 arginine 2 dimethylation, H3K9 dimethylation and DNA methylation.

Chromatin architecture. Chromatin remodelling32 can also regulate the dynamic induction of inflammatory gene expression. This activity is mediated by chromatin remodelling complexes, which use ATP to 'slide' nucleosomes relative to DNA or to alter nucleosome–DNA contacts, thereby modulating the accessibility of chromatin-associated DNA to transcriptional regulators33. Chromatin remodelling complexes are generally thought to be regulated mainly at the level of recruitment to target gene promoters. However, a recent study showed that the chromatin remodelling complex SWI–SNF (switching-defective–sucrose non-fermenting; also known as BAF) can also be regulated after recruitment to a target gene by a Ca2+–calmodulin-dependent signal in LPS-stimulated macrophages34.

Smale and colleagues35 have recently shown that the requirements for chromatin remodelling differ between LPS-inducible primary and secondary response genes (Fig. 2). The induction of secondary response genes requires de novo protein synthesis, whereas the upregulation of expression of primary response genes does not (although this does not preclude the regulation of primary response genes in the secondary phase of the transcriptional response)35. Chromatin remodelling by SWI–SNF is necessary for the induction of secondary response genes (and the delayed primary response genes), but not for the early primary response genes35. Indeed, subsequent studies showed that whereas the promoters of secondary response genes undergo LPS-dependent H3K4 trimethylation and H3 acetylation, the promoters of SWI–SNF-independent primary response genes have high basal levels of these active histone modifications, even before LPS stimulation32,36,37. Importantly, the presence of active chromatin correlates closely with promoter GC content, which indicates that the underlying DNA sequence at promoters with CpG islands, which are found in many SWI–SNF-independent primary response genes, could direct chromatin remodelling to the active state37. This is supported by the observations that these CpG islands are associated with low nucleosome density and do not assemble stable nucleosomes in vitro37, and that these promoters also bear active chromatin even in embryonic stem cells32,37, which indicates that chromatin remodelling is not a consequence of cellular differentiation. Previous studies38,39 have indicated a role for chromatin in restricting the accessibility of transcription factors to their target genes, and these studies expand on this concept, showing that active chromatin might enable more promiscuous induction of GC-rich promoters of primary response genes, whereas inactive chromatin at the promoters of secondary response genes restricts gene induction to a limited range of stimuli in specialized cell types32,37. Along these lines, it is noteworthy that GC-rich primary response genes, but not most secondary response genes, are regulated by the transcriptional co-repressors nuclear receptor co-repressor (NCOR) and REST co-repressor (RCOR; also known as CoREST)32 (see below), perhaps to counter the active chromatin that is associated with these genes in the basal state. It will be interesting to further characterize primary and secondary response genes with respect to other features of chromatin, and to understand the consequences of these chromatin modifications for gene induction.

Figure 2: Two distinct modes for regulating inducible genes.
figure 2

a | At primary response genes that have promoters containing CpG islands, Toll-like receptor (TLR) signalling induces a switch from basal gene transcription (mediated by serine 5 (Ser5)-phosphorylated RNA polymerase II (Pol II)) to the stimulus-dependent production of mature, processed transcripts, which depends on the recruitment of positive transcription elongation factor b (P-TEFb). b | By contrast, secondary response genes require chromatin remodelling as a prerequisite for transcription factor binding and the recruitment of histone-modifying enzymes and the general transcription initiation machinery. BRG1, BRM/SWI2-related gene 1 (also known as SMARCA4); HAT, histone acetyltransferase; IRF3, interferon-regulatory factor 3; NF-κB, nuclear factor-κB; SP1, specificity protein 1.

Full size image

Further analyses of the GC-rich primary response genes have uncovered additional surprises (Fig. 2). In the basal state, these genes (but not secondary response genes) are transcribed at low levels by RNA polymerase II (Pol II) that is phosphorylated on serine 5 (Ser5P) but not on Ser2 of its carboxy terminal domain. Consistent with a crucial role of Ser2P in recruiting the RNA processing and splicing machinery, these basal transcripts of the primary response genes are extensively elongated but are not spliced32. LPS-induced signalling recruits positive transcription elongation factor b (P-TEFb), which phosphorylates Pol II on Ser2 and results in the generation of high levels of mature, spliced transcripts of the primary response genes. So, an important checkpoint for inducible gene expression seems to involve signal-dependent engagement of P-TEFb, conversion of Ser5P Pol II to the Ser2P form, and the production of mature, spliced transcripts of primary response genes32 (Box 3).

The organization of inducible transcriptional responses into a primary and a secondary component was first described for the mitogen-induced response40, and it is probably generally applicable to all other inducible stimuli. It will be interesting to see whether the different mechanisms described here for the regulation of CpG-associated primary response genes and secondary response genes will also apply to other inducible gene expression programmes.

Co-regulators of the inflammatory response

Co-regulators are transcriptional regulators that, unlike transcription factors, lack DNA-binding specificity and must be recruited to their target genes through other mechanisms. Because many co-activators and co-repressors (co-regulators that activate and inhibit gene expression, respectively) function in the LPS-induced transcriptional response, here we focus on only a few of the most important or interesting examples.

Co-activators. Inflammatory gene induction by transcription factors such as NF-κB depends on co-activator proteins. These transcriptional regulators can promote inflammatory gene expression in multiple ways. Many co-activators have histone-modifying activities and can remodel chromatin at target genes to promote gene induction. For example, CBP–p300-mediated histone acetylation can be coupled to the recruitment of SWI–SNF and other factors41, whereas PCAF (p300–CBP-associated factor) and GCN5 (general control of amino-acid synthesis 5) histone acetyltransferases can direct transcription elongation factors to target genes32. Other co-activators lack intrinsic enzymatic activity and might promote the assembly of a transactivating complex. Yet another example of a co-activator function is provided by the IκB family member IκBζ. This protein is transcriptionally induced by TLR signalling and promotes the exchange of inhibitory p50 (also known as NF-κB1) homodimers for transcriptionally active p50–p65 (also known as RELA) heterodimers on some target gene promoters42.

Intriguingly, recent studies indicate that p65 and IRF3 have additional roles in the LPS-induced transcriptional programme as co-activators as well as DNA-binding transcription factors. Leung et al. showed that IRF3-dependent regulation of Ccl2 expression requires IRF3 functioning as a co-activator for p65 (Ref. 43). Consistent with this finding, Glass and colleagues showed that the glucocorticoid receptor inhibits a subset of LPS-inducible genes by blocking the co-activator function of IRF3. In addition, p65 and IRF3 were shown to interact, and the glucocorticoid receptor could disrupt this binding44. Conversely, p65 was suggested to be a co-activator for IRF3 and to facilitate IRF3 binding to DNA45.

Why do some but not all p65-regulated genes engage IRF3 as a co-activator? It is possible that subtle differences in the NF-κB-binding site in a gene can induce conformational changes in p65 in an allosteric manner, thereby conferring specificity of co-activator binding43,46. p65 and IRF3 might have evolved mutually co-activating functions because of their parallel roles in many aspects of the inflammatory response, and the combined transactivation potential of the two factors could result in increased inducible transcription of target genes. It would be interesting to determine if p65 and IRF3 can co-activate other transcription factors in the regulation of additional functional programmes.

Co-repressors. NCOR and the closely related protein SMRT (silencing mediator of retinoid and thyroid receptors) have emerged as important regulators of inflammatory gene expression. The NCOR and SMRT multiprotein co-repressor complexes contain histone deacetylases (HDACs) and potentially other activities for inhibiting gene expression, and their stimulus-dependent 'dismissal' from the promoters of inflammatory genes (a phenomenon known as derepression) is a prerequisite for the inducible expression of these genes. NCOR is directed to many inflammatory genes, in part by the transcription factor JUN (also known as AP1), and its clearance is triggered by signal-induced exchange of JUN homodimers for transcriptionally active JUN–FOS heterodimers47,48. By contrast, SMRT is recruited to gene promoters by the transcriptional repressor TEL (translocation–ETS–leukaemia)47. Interestingly, some inflammatory genes are regulated by both NCOR and SMRT co-repressor complexes, indicating that this group of genes can be regulated by a greater diversity of signals47,49. In this regard, it is interesting that another co-repressor complex, RCOR, is also associated with inflammatory genes and is cleared from target promoters after LPS-induced signalling32, which indicates that it might also provide a stimulus-specific regulatory checkpoint. Importantly, whereas LPS stimulation dismisses NCOR from target genes before their induction, a counteracting activity is provided by nuclear receptor-mediated stabilization of NCOR (Fig. 3 and see below).

Figure 3: Control of inflammatory gene expression by co-activators and co-repressors.
figure 3

In the basal state, co-repressors such as nuclear receptor co-repressor (NCOR), silencing mediator of retinoid and thyroid receptors (SMRT) and REST co-repressor (RCOR) are recruited to target promoters by various transcription factors, where they counter inflammatory gene expression by inhibiting histone acetylation (and possibly also other activating histone modifications such as H3K4 trimethylation). Toll-like receptor (TLR) signalling and other pro-inflammatory signals induce the exchange of co-repressors for co-activators on target promoters, resulting in the activation of gene expression. Nuclear receptors such as peroxisome proliferator-activated receptor-γ (PPARγ), glucocorticoid receptor and liver X receptors (LXRs) constitute an important class of anti-inflammatory regulators, which block inflammation in part by inhibiting this exchange. In the best studied case of NCOR, inflammatory signals trigger its proteasomal degradation, and this can in turn be inhibited by PPARγ and LXRs. HAT, histone acetyltransferase; HDAC, histone deacetylase; Pol II, RNA polymerase II.

Full size image

With respect to the control of inflammatory gene expression by these co-repressors, several interesting questions remain. What is the repertoire of genes that are regulated by each co-repressor? Why are some genes controlled by more than one co-repressor? How do these co-repressors mediate repression, and do they have distinct modes of repression? How are they recruited in the basal state and dismissed in a signal-dependent manner?

Negative regulation of inflammatory genes

The induction of an inflammatory response is essential for host defence during infection, but timely resolution is also important to limit the detrimental effects of inflammation, particularly when it is inappropriately sustained or increased. For this reason, acute inflammation leads to the upregulation of expression of many negative regulators of inflammation. These negative regulators fall into two main categories: signal-specific regulators and gene-specific regulators. The first category consists of regulators that inhibit signal transduction by TLRs and other inflammatory pathways, and includes A20, ST2, IL-1R-associated kinase M (IRAKM) and suppressor of cytokine signalling (SOCS) proteins50. Although these proteins inhibit inflammatory signalling through various mechanisms, they all function proximal to the receptor, and so are expected to block gene induction by that receptor in a global manner. The second category includes transcriptional repressors or other negative regulators that function to modulate gene expression.

There are two types of transcriptional negative regulator: basal repressors and inducible repressors. Basal repressors are constitutively expressed and are important for the basal repression of many inflammatory genes. For example, homodimers of the NF-κB family member p50 can function as transcriptional repressors that are exchanged for transcriptionally active p65–p50 heterodimers as a result of signalling induced by LPS51. JUN49, SIRT6 (silent mating type information regulation 2 homologue 6)52 and co-repressors such as NCOR47,49, SMRT47 and RCOR32 have also been shown to mediate basal repression of a subset of LPS-inducible inflammatory genes (Fig. 3). By contrast, inducible repressors are normally expressed only at low levels or not at all, but are transcriptionally induced by LPS-induced signalling, indicating that they are part of a negative feedback mechanism that limits the inflammatory response. Importantly, inducible repressors block the expression of secondary response genes, whereas basal repressors inhibit the expression of primary response genes associated with CpG islands. These primary response genes might require repression in the basal state to prevent low-level constitutive expression, whereas the inaccessible nature of the chromatin at secondary response genes is generally refractory to basal gene expression due, in part, to the occlusion of binding sites for transcription factors. Therefore, it is probable that different modes of transcriptional repression can operate to inhibit primary and secondary response genes.

LPS-inducible negative feedback loops. Members of the IκB family are noteworthy examples of inducible negative regulators. IκBα, which was the first identified member of this family, mainly inhibits the expression of NF-κB-dependent genes on a global level, but IκBNS and B cell lymphoma 3 (BCL-3) limit inflammation in a gene-specific manner53,54,55. IκBNS and BCL-3 are transcriptionally induced by LPS stimulation, and their genetic deletion results in the enhanced induction of inflammatory genes and increased susceptibility to endotoxic shock54,56. Both proteins mediate the negative regulation of inflammation by modulating the exchange of active NF-κB dimers for their inactive counterparts at target gene promoters. Interestingly, IκBNS and BCL-3 seem to control distinct sets of genes — IL-6, IL-12p40 and IL-18 for IκBNS, and TNF, IL-10 and IL-1β for BCL-3 — although the basis of this specificity remains to be determined54,55.

ATF3 is another transcriptional negative regulator that can be induced by LPS stimulation. Similarly to IκBNS and BCL-3, loss of ATF3 leads to hyperinduction of a subset of LPS-inducible genes and increased susceptibility to endotoxic shock20. ATF3-mediated transcriptional inhibition can be reversed by HDAC inhibitors, which indicates that at least one key aspect of the function of ATF3 is to recruit HDACs to target genes20. Moreover, ATF3 is a member of the CREB family of basic leucine zipper transcription factors and has been shown to form a regulatory circuit with NF-κB and C/EBPδ at some LPS-inducible genes16. The transcriptional programme induced by LPS is undoubtedly controlled by many other transcriptional repressors, the identity and mechanisms of function of which remain poorly characterized and warrant further study.

Finally, it is worth noting that many of the same proteins that function in the LPS-induced negative feedback loops are also upregulated by other pathways that inhibit inflammatory gene expression. For example, IL-10, which is an important anti-inflammatory cytokine for many cells of the immune system including macrophages, can also induce the expression of both IκBNS and BCL-3 (Refs 57,58). Interestingly, in colonic macrophages, IL-10-mediated induction of IκBNS expression seems to be of particular importance in inhibiting the production of IL-6 and other inflammatory responses57. This is shown by the development of colitis in IκBNS-deficient mice, similarly to IL-10-deficient mice54. In other settings, however, BCL-3 or other IL-10-inducible negative regulators might be more important for mediating IL-10-dependent suppression of inflammation.

Negative regulation by anti-inflammatory pathways. Inflammatory gene induction is subject to negative regulation by a large number of pathways and mechanisms that are important to limit the pathophysiological consequences of excessive inflammation. These include anti-inflammatory cytokines (such as IL-10 and transforming growth factor-β (TGFβ)), nuclear hormone receptors (including, but not limited to, glucocorticoid receptors, liver X receptors (LXRs), peroxisome proliferator-activated receptors (PPARs) and vitamin D receptor) and cAMP.

Consistent with the crucial anti-inflammatory role of IL-10, mice lacking this cytokine mount exaggerated inflammatory responses to infection and in septic shock59 and develop spontaneous colitis due to perturbed host–commensal homeostasis in the intestines59. An important mechanism by which IL-10 inhibits inflammation is at the level of transcription, as indicated by the ability of cycloheximide treatment to block IL-10-mediated inhibition of primary response genes60. As mentioned above, IL-10 induces the expression of several negative regulators that mediate gene-specific repression of the LPS-induced inflammatory response. IκBNS and BCL-3 disrupt NF-κB-mediated transcription, but the function of other IL-10-induced proteins and whether they modify chromatin in a gene-specific manner are not clear. cAMP is another negative regulator of inflammation61. It is activated by G-protein-coupled receptors for glucagon, acetylcholine, adenosine and many other biological signals. cAMP works mainly by activating protein kinase A (PKA), but the exact mechanism of cAMP-mediated inhibition of inflammatory gene expression is not known.

Nuclear receptors are another major class of transcriptional negative regulators of inflammation. These include glucocorticoid receptors, PPARs and LXRs. These proteins therefore integrate the control of inflammation with various important physiological functions, such as metabolism62. Nuclear receptors are thought to inhibit inflammation by at least two distinct mechanisms. Activated nuclear receptors can directly induce gene expression programmes that are anti-inflammatory. For example, the activation of glucocorticoid receptor increases the uptake of apoptotic cells by macrophages concurrent with the inhibition of inflammatory signalling63. Nuclear receptors can also inhibit inflammation directly, in a gene-specific manner. For example, PPARδ activation results in disassociation of the BCL-6 co-repressor from the Ccl2 promoter, thereby enabling recruitment of BCL-6 to inflammatory target genes to mediate transcriptional inhibition64. Alternatively, activated nuclear receptors can be recruited to some inflammatory genes where they inhibit the clearance of the co-repressors NCOR and SMRT, a process known as transrepression62. Interestingly, transrepression by LXRs and PPARγ results in the impaired induction of distinct subsets of inflammatory genes, but it is not known what dictates the differential gene-specific recruitment of LXRs and PPARγ62. The multitude of mechanisms by which nuclear receptors can inhibit inflammatory gene expression underscores the importance of transcriptional control of inflammation by metabolism and other physiological processes.

SIRT proteins, which are nicotinamide adenine dinucleotide (NAD)-dependent deacetylases of the class III HDAC family, have also been implicated recently in the transcriptional control of inflammatory genes. Unlike its yeast homologue Sir2 and the class I and class II HDACs, SIRT1 targets transcription factors and co-activators for deacetylation; SIRT1 counters p300-mediated acetylation of NF-κB p65 in its transactivation domain, which leads to a block of p65-dependent gene induction that is independent of DNA binding65. SIRT6 also inhibits NF-κB activity, but in this context NF-κB itself is not the target; instead, SIRT6 deacetylates H3K9 at the promoters of some NF-κB-regulated genes66. Acetylated H3K9 is closely associated with transcriptional activation across the genome, and SIRT6-mediated deacetylation of H3K9 represses both basal and stimulus-dependent gene induction66. These studies raise the possibility that SIRT1 and SIRT6 (and perhaps other members of the family) might couple energy status (in the form of NAD sensing) to the control of inflammation.

Finally, HES and HEY are two other LPS-inducible transcriptional repressors that function to attenuate the induction of a subset of inflammatory responses67. In an intriguing example of signalling pathway crosstalk, LPS-dependent upregulation of expression of HES and HEY by macrophages requires concomitant activation of the Notch pathway and the downstream transcriptional activator recombination-signal-binding protein-J (RBP-J). Moreover, IFNγ signalling blocks HES and HEY upregulation, illustrating a novel mechanism for the well-established 'priming' effect of IFNγ on LPS-stimulated macrophages67. Global transcriptional profiling to define the repertoire of HES- and HEY-repressed genes should help to define the physiological contexts in which Notch signalling inhibits inflammatory gene expression.

Pathogen-mediated chromatin remodelling. Pathogens can modulate inflammation at multiple levels. General inflammatory signalling pathways are a common target of pathogen virulence factors, but in addition, pathogens have evolved mechanisms to inhibit inflammation in a module-specific manner. For example, Mycobacterium tuberculosis subverts the IFNγ response in macrophages without disrupting the JAK–STAT (Janus kinase–signal transducer and activator of transcription) pathway, and it does so in a gene-specific manner that targets a subset of IFNγ-regulated genes68. In addition, several recent studies have highlighted how pathogens can modify chromatin at inflammatory genes69. Inhibition of TNF (but not IL-10) production by infection with Toxoplasma gondii is associated with decreased H3 acetylation and H3S10 phosphorylation, and decreased recruitment of p65 to the Tnf promoter70. Interestingly, several other pathogens also downregulate H3S10 phosphorylation, presumably for the purpose of blocking expression of a subset of inflammatory genes71. It is not known why or how different pathogens have converged on this particular chromatin modification to subvert the host inflammatory response. p38 is a common target of pathogen virulence factors, so it is possible that the downregulation of H3S10 phosphorylation is a consequence of p38 inhibition. It will be interesting to see if pathogens use other chromatin remodelling activities to subvert specific functional programmes of the host immune response.

Additional control mechanisms

Recent studies have provided evidence of a potential role for long non-coding RNAs (lncRNAs) in regulating inflammatory gene expression72,73. The role of microRNAs in this process has been recently reviewed74 and will not be addressed here. Apart from their role in X chromosome inactivation and imprinting, the function of lncRNAs is not well understood. Broadly speaking, lncRNAs have been described that direct the expression of specific genomic loci (regulation in trans); alternatively, the process rather than the product of lncRNA transcription has been shown to inhibit or activate gene expression (regulation in cis)75. An example of regulation in cis is the requirement for the signal-dependent transcription of a lncRNA for the induction of the closely juxtaposed lysozyme gene73. Lysozyme is an antimicrobial enzyme that hydrolyses bacterial cell wall peptidoglycan, the expression of which is induced during macrophage differentiation. In the basal state, the insulator protein CCCTC-binding factor (CTCF) is bound to regulatory elements of the lysozyme gene, but after LPS-induced signalling, transcription of a lncRNA through this regulatory region of the lysozyme gene displaces CTCF and leads to local chromatin remodelling and induction of the lysozyme gene. Insulator proteins can block the interactions between regulatory regions of a gene76, so the displacement of CTCF might enable the distal enhancer of the lysozyme gene to interact, in a signal-dependent manner, with the promoter73. In light of this, a recent study identified ∼1,600 highly conserved lncRNAs in mammalian cells, of which more than 20 can be induced by LPS stimulation of bone marrow-derived dendritic cells72. Some of these inducible lncRNAs might regulate other genes in cis, and it will be interesting to understand what categories of LPS-inducible genes are subject to this type of control.

Previous studies have shown that many pro-inflammatory cytokine and chemokine genes are also regulated post-transcriptionally. These genes encode mRNAs with AU-rich elements (AREs) in the 3′ untranslated regions, which are recognized by a network of ARE-binding proteins that control mRNA metabolism by distinct mechanisms, including regulation of translation and mRNA decay. So, the maximal induction of TNF, IL-1β and IL-6 by LPS requires inactivation of ARE-mediated mRNA destabilization in a manner that depends on the p38 mitogen-activated protein kinase pathway77,78. Finally, the endonuclease ZC3H12A was recently shown to regulate the stability of inducible gene transcripts79. In its absence, macrophages produce increased levels of certain pro-inflammatory cytokines such as IL-6 and IL-12, and mice that lack ZC3H12A develop inflammation and severe immune diseases that lead to early death. ZC3H12A is thought to target a distinct repertoire of mRNAs, independently of ARE recognition, further underscoring the role of mRNA transcript stability in regulating the inflammatory response.

Conclusions and future perspectives

The LPS-inducible transcriptional programme has served as an excellent model for understanding the transcriptional control of inflammation. Many questions remain that would be of interest to address. First, we have focused on several pathways that negatively regulate inflammation to illustrate some general principles. But these are only a small fraction of the total repertoire of physiological signals that inhibit inflammation, underscoring the necessity of being able to fine-tune inflammation in a module-specific manner. Future studies that address the mechanisms of inhibition by these different signals will be important. For example, what sets of genes and functional programmes are inhibited by IL-10 compared with TGFβ? Are there transcriptional repressors that are shared by multiple negative regulatory pathways? Of the several repressors that are activated or induced by a particular negative regulator of inflammation, what are the relative contributions of individual repressors and how do they differ in distinct cell types and physiological settings? Pathways that positively modulate inflammation undoubtedly exist but are not discussed here, although some of them (for example, IFNγ in host defence and other cellular stressors) must control inflammation in a module-specific manner, through transcriptional mechanisms that would be of great interest to understand.

In addition, we have discussed how transcriptional modules encode functional programmes of the inflammatory response, and it will be important to further define these modules and the mechanisms that enable their autonomous regulation. The identification of module-specific transcriptional regulators and chromatin modifications is important not only for understanding the organization of this transcriptional programme, but would also enable the specific manipulation of various components of the inflammatory response.

Finally, there is a growing awareness that many of the most prevalent human diseases are associated with pathophysiological chronic inflammation. This type of inflammation is persistent and long-lasting, and is associated with self-amplifying loops that maintain its expression. Given the role of chromatin in regulating both dynamic and stable patterns of gene expression, chronic inflammation is probably associated with a reprogramming of inflammatory gene expression that is mediated by alterations to chromatin. Therefore, it will be important to determine whether chromatin dysregulation underlies chronic inflammation in many disease settings.