In multicellular organisms, the epithelia is a contact surface with the surrounding environment and is exposed to a variety of adverse biotic (pathogenic) and abiotic (chemical) factors. Multi-layered pathways that operate on different time scales have evolved to preserve cellular integrity and elicit stress-specific response. Several stress-response programs are activated until a complete elimination of the stress is achieved. The innate immune response, which is triggered by pathogenic invasion, is rather harmful when active over a prolonged time, thus the response follows characteristic oscillatory trajectories. Here, we review different translation programs that function to precisely fine-tune the time at which various components of the innate immune response dwell between active and inactive. We discuss how different pro-inflammatory pathways are co-ordinated to temporally offset single reactions and to achieve an optimal balance between fighting pathogens and being less harmful for healthy cells.
Translational initiation: a bull's-eye for regulation
Adequate reprogramming of metabolic activities under environmental stress is crucial for cell survival. Regulation at the level of translation provides the necessary plasticity for immediate changes of protein levels and, consequently, cellular activities . Initiation is the most potent step for stress-induced regulation of translation (Figure 1A). Key control steps include eukaryotic initiation factor (eIF) 2α-phosphorylation, which inhibits 80S ribosome assembly (Figure 1B) and phosphorylation of eIF4E-binding protein, which enhances association of the 43S pre-initiation complex (PIC) with mRNA (Figure 1C). Stress-specific response is shaped by the activation of distinct kinases for each type of stress, e.g. haem-regulated inhibitor kinase (HRI) by oxidative stress, general control non-derepressible 2 (GCN2) kinase by amino acid deprivation and dsRNA-activated protein kinase (PKR) and PKR-like endoplasmic reticulum (ER) kinase (PERK) by unfolded protein stress . A common hallmark of these mechanisms is that the response is executed at the level of translation initiation. A sizeable subset of genes escapes the global kinase-dependent inhibition and translation remains active upon moderate stress exposure; most probably those genes contribute to stress-selective stress response . In addition, signalling cascades allow preferential translation of defined mRNA subsets [4,5]. Although important for maintaining the stress response, the protein products of some stress-activated genes might harm the cell or neighbouring tissues when expressed over a long period, even under persisting stress. Hence, mechanisms have evolved to regulate the expression of some genes in oscillation to preserve tissue integrity. These include pro-inflammatory genes whose expression is triggered by infection. Pathogens elicit a multi-layered response which includes repression of translation in kinase-dependent eIF2α-phosphorylation (Figure 1B) or by eIF4E-dependent activation (Figure 1C) to selectively regulate the expression of pro-inflammatory genes (Figure 1D).
Regulation of translation initiation
Several excellent reviews have been published that describe kinase-dependent stress response and the components of the single cascades activated to counteract stress [6–10]. Here we rather focus on emerging concepts regarding the translation regulation of the inflammatory response with a particular emphasis on the mechanisms of oscillation of the stress response programs between activated and deactivated, a striking paradigm of dwelling between stress-response maintenance and preserving tissue integrity.
Activating and deactivating eIF2α: dwelling between translation-competent and translation-incompetent
eIF2α is a central node in the cellular stress response network where several stress sensing pathways congregate (Figures 1A and 1B). The specificity in the response to stress is determined by the activating kinase (PKR, PERK, HRI, GCN2), each of which is triggered by different stimuli. PKR is central to viral infection (Figure 2). Viral dsRNA binds to the N-terminal RNA-binding domain of PKR, which dimerizes, autophosphorylates and phosphorylates eIF2α [11–13]. eIF2α phosphorylation decreases global protein synthesis which consequently diminishes viral replication and protects neighbouring cells.
Sensing pathogenic stress
In parallel, a transcriptional program is activated to fight pathogens. Viral RNAs are also recognized by the endosomal toll-like receptors (TLR3) which activate a cascade culminating at different key transcription factors (TFs), such as interferon (IFN), regulatory factor 3 (IRF3) and nuclear factor kappa-light-enhancer of activated B-cells (NF-κB) and induce transcription of pro-inflammatory cytokines (e.g. IFNs and interleukins (ILs); ) (Figure 2). Despite increased NF-κB-dependent transcription of ILs (specifically, IL6 and IL8), their translation is first initiated upon reactivation (i.e. de-phosphorylation) of eIF2α . Among other targets, activating TF 4 (ATF4) induces activation of growth arrest and DNA-damage-inducible 34 (GADD34; Figure 2), a phosphatase required for eIF2α dephosphorylation and its reactivation . Under eIF2α limiting conditions, both the stress-related TF ATF4 and GADD34 are preferentially translated; their translation is controlled through upstream ORFs . GADD34-reactivated eIF2α mediates translation initiation of pro-inflammatory cytokines to overcome infection (Figure 2) [17–19]. Under prolonged stress exposure, the cells alternate between these two modes, translation-deficient (i.e. eIF2α phosphorylated) and translation-competent (i.e. eIF2α dephosphorylated) (Figure 1E), thus decreasing the harmful effect of continuous expression of pro-inflammatory cytokines.
In addition, to guarantee the stringency of the response to pathogens a second feedback loop is activated through PKR-dependent IFN-γ translation . The pseudoknot in the 5′-UTR of IFN-γ mRNAs serves as a PKR docking site; PKR is auto-activated and inactivates any proximal eIF2α. Consequently, IFN-γ mRNA translation decreases and reduces the overall titre of IFN-γ. The amount of IFN-γ is precisely regulated to avoid its harmful overexpression : intensive translation is replaced by down-regulated synthesis staggered to the activation/deactivation cycles of eIF2α (Figure 1E).
In addition to PKR activation, some pathogens induce PERK, which senses ER stress and also shuts down translation by phosphorylating eIF2α  (Figure 2). For example, during Pseudomonas aeruginosa infection, epithelia face high concentrations of quorum sensing-molecule, homoserine lactone, which misregulates Ca2+ concentration in the ER . The function of many Ca2+-dependent ER chaperones, including calnexin and calreticulin, is compromised. Consequently, the amount of misfolded proteins increases, which in turn activates PERK .
mTOR goes for the TOP
TLRs at the cell surface are crucial sensors of the mammalian firewall against microbial infection. TLRs recognize bacterial cell wall and viral surface components to detect infection. TLR-activation induces the specific expression of pro-inflammatory cytokines (IL6, IL8, tumour necrosis factor alpha (TNFα)) mainly at translational level via the mechanistic target of rapamycin (mTOR)-dependent phosphorylation of 4E-BP (Figure 1C) and ribosomal protein S6 kinase (S6K) which phosphorylates ribosomal protein S6 . Additionally, mRNAs that contain 5′-terminal oligopyrimidine tracts (5′TOP) are preferentially translated during bacterial infection to support the high synthesis of pro-inflammatory cytokines . In contrast, viral infection (e.g., mosquito-transmitted bunya virus, Rift Valley fever virus) leads to reduced translation of 5′TOP mRNAs upon mTOR-dependent 4E-BP phosphorylation . Viruses snatch the cap-structure including the 5′TOP to enhance translation of their own mRNAs . The cell counteracts the viral propagation by preferential degradation of mRNAs bearing 5′TOP motifs . Notably, proteins participating in the translation machinery (e.g. ribosomal proteins, translation factors) also bear 5′TOP motifs. 5′TOP mRNA degradation reduces the amount of the translation machinery components together with newly 5′TOP-decorated viral mRNA which consequently represses viral replication . However, various pathogens elicit different reactions: Rift Valley fever virus infection reduces activated protein kinase B (PKB or Akt) activity in a TLR-dependent fashion and down-regulates 5′TOP mRNA translation, whereas the increased translation during microbial infection appears to be conferred by IL-receptor activity [23,24].
Similar to IFN-γ translation, mTOR-dependent eIF4E activation of pro-inflammatory cytokines also exhibits oscillatory behaviour to achieve a balance between positive and harmful effect. mTOR activity reduces NF-κB signalling by phosphorylating the p65 (TF p65 or NF-κB p65) subunit and consequently diminishes transcription of pro-inflammatory genes . Thus, mTOR reduces pro-inflammatory signalling and plays a role in suppressing inflammation.
Similar to 4E-BP, eIF4E activity is also regulated by phosphorylation by several kinases, e.g. mTOR . The specificity of eIF4E for 5′TOP mRNAs is phosphorylation-dependent [27–29], although the exact mechanism of activating the innate immune response upon eIF4E phosphorylation remains elusive. The observed effects are polarized between boosting translation of chemokines upon eIF4E activation  and potentiating the antiviral response through decreased translation of NF-κB inhibitor (IκB) and consequently the increased activity of NF-κB  (Figure 2).
Sequester and release: a way to commence and silence an immune response
Stress-induced inhibition of translation initiation, particularly by phosphorylation of eIF2α, triggers formation of stress granules (SG) . Many of the mRNAs sequestered in SGs have AU-rich elements (AREs) in the UTRs suggested to be the binding platform of scaffold proteins participating in the assembly of SGs, e.g. cytotoxic granule-associated RNA binding protein, TIA-1 and TIA-1 related protein, TIAR . Notably, TIA-1 and TIAR confer a robust decrease in translational activity of two pro-inflammatory proteins, TNF-α and cyclooxygenase-2 which bear AREs and induce early reactions to pathogens [32–34]. Intriguingly, phosphorylated eIF4E, which together with other initiation factors  resides in SGs under stress, also exhibits an affinity to AREs and binds them . Upon stress relief, phosphorylated eIF4E reallocates mRNAs from the granules to the polysomes, most probably by recognizing their AREs . Thus, translation commences concomitantly .
The amount and composition of SGs and processing bodies (or p-bodies, which are discrete loci of mRNA decay or temporarily sequestering mRNAs away from translation) differ among the subsets of functional macrophages, such as M1 and M2 macrophages . M1 macrophages are involved in acute infection, whereas M2 macrophages engage in tissue remodelling and resolution of inflammation . This change is in part attributable to differential expression of p-body assembling proteins Dcp1 and EDC4, which also regulate the amount of synthesized IL6 protein whereas its mRNA remains constant . In addition, M1 macrophages form fewer SGs in response to stress and confer relatively high translation activity under stress . Hence, differential SG and p-body composition, which differently regulate the translation of inflammatory proteins, impart functional variation in different subsets of macrophages .
GAITing pro-inflammation into silence
In response to viral infections, signalling cascades foster transcription of IFNs through activation by NF-κB or IRF3 TFs (Figure 2). Among them is IFN-γ that enhances pro-inflammatory response via activation of macrophage and attraction of natural killer cells. IFN-γ influences translation on a much broader level in mTOR-dependent fashion . Genes associated with metabolic processes together with pro-inflammatory genes exhibit the largest changes in expression level upon IFN-γ stimulation . Prolonged IFN-γ activity however shuts down the synthesis of a defined set of mRNAs, including chemokine ligand 22 (CCL22), vascular endothelial growth factor A (VEGFA), ceruloplasmin, zipper-interacting protein kinase (ZIPK) and death-associated protein kinase 1 (DAPK). This is achieved through a series of temporally defined phosphorylation events. It is initiated by release of glutamyl–prolyl–tRNA–synthetase (EPRS) from the tRNA multisynthetase complex . Free, phosphorylated EPRS binds cytosolic NS1-associated protein 1 (NSAP), forming an inactive pre-IFN-γ-activated inhibitor of translation (pre-GAIT) complex after short-time exposure to IFN-γ  (Figure 1D). IFN-γ-dependent phosphorylation releases the ribosomal protein of the large subunit (RPL13a) from the ribosome via the DAPK–ZIPK signalling cascade; RPL13a then associates with glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and activates the pre-GAIT complex after a long-time exposure to IFN-γ [41,42] (Figure 1D).
The GAIT complex blocks initiation by binding to eIF4G through RPL13a. This binding competes out eIF3 from eIF4G–eIF3  and prevents one of the crucial interactions for association of PIC (Figure 1A). The GAIT complex does not execute global translation but rather specifically reduces initiation rates of a subset of mRNAs which bear a conserved secondary structure, a discontinuous stem loop with an asymmetrical bulge in their 3′-UTRs, named GAIT sequence, which is recognized by EPRS . The GAIT-binding motif is enriched in genes associated with inflammation: out of 52 human genes carrying a GAIT sequence in their 3′-UTR, 19 associated with inflammation .
Furthermore, depletion of RPL13a causes a significant reduction in ribosomal RNA (rRNA) methylation and reduces cap-independent initiation mediated by internal ribosomal entry site (IRES) elements . Many viral genes, whose translation is mostly initiated in IRES-dependent fashion, are not expressed, consequently reducing the progression of infection. The reduced rRNA methylation through RPL13a is however dispensable for the ca nonical cap-dependent translation .
Importantly, the activity of GAIT is precisely regulated through a feedback loop sensing the amount of the GAIT-activators DAPK and ZIPK. Many inflammatory genes, but also mRNAs of proteins participating in the GAIT complex assembly (e.g. DAPK and ZIPK) carry a GAIT sequence . IFN-γ exposure results in translational silencing and decrease DAPK and ZIPK, which in turn reduces RPL13a phosphorylation and inactivates the GAIT complex. Consequently, after an inflammatory boost upon pathogenic infection, cells shut down IFN-γ-dependent genes and reach a basal state of inflammatory proteins (Figure 1E). If the infection is not overcome, a second round of IFN-γ signalling can be initiated to eliminate remaining pathogens (Figure 1E). The precise orchestration of alternating GAIT activation and inactivation enables establishing an optimal balance to fight pathogens while avoiding harmful long-lasting activation of the inflammatory response.
In conditions of severe inflammation however, the oscillation of GAIT between activated and inactivated states can be modified to prolong translation of pro-inflammatory genes and extend the innate immune response. The two-step mechanism of GAIT complex formation provides an additional regulatory loop to prolong the phase of GAIT-activation. Cells facing IFN-γ in the context of inflammation (e.g. arteriosclerosis, Alzheimer's disease) employ this auxiliary loop. Inflamed tissues induce S-nitrosylation and oxidation of lipids and proteins, such as low-density lipoprotein (LDL) . GAPDH is S-nitrosylated upon LDL oxidation which in turn reduces the affinity of GAPDH for phosphorylated RPL13a . Hence, free RPL13a is efficiently degraded and GAIT complex formation is reduced, which in turn prolongs the production of inflammatory proteins. Only after the removal of oxidized LDL, GAPDH is able to bind RPL13a and form an active GAIT complex. Hence, this secondary regulation loop provides an extended time window for synthesis of inflammatory genes to efficiently fight infections before translation is finally shut down [41,48].
Two decades of research have revealed an extraordinary complexity of translational regulation that fine tunes the cellular response to pathogenic invasion and damage. Key compounds in inflammatory signalling are under translational control, with consequences ranging from a global decrease of protein synthesis to transcript-specific enhancement or inhibition of translation. Fine-tuning the innate immune response by regulating the activity of crucial components at the level of translation bears many advantages as opposed to their transcriptional regulation. This includes fast and selective synthesis of signalling and pathogen-attacking molecules. Yet, fundamental questions remain to be addressed to fully understand the network of processes that control inflammation and infection. What are the kinetics of activation/deactivation of crucial regulatory components? Do different pathogens activate the innate immune response with similar kinetics? Does chronic infection render these mechanisms insensitive or hypersensitive? These are key area of study that will enable us to gain a more accurate understanding of the integration and cross-talk between various pathways that activate the innate immune response. This knowledge is needed to understand different aspects specific for each pathology but also may reveal potential targets for suppressing or enhancing the immune response.
This work was supported by the Deutsche Forschungsgemeinschaft [grant number FOR 1805] to Z.I.
4E binding protein
5′-terminal oligopyrimidine tracts
activating transcription factor 4
death-associated protein kinase 1
eukaryotic initiation factor
growth arrest and DNA-damage-inducible 34
IFN-γ-activated inhibitor of translation
general control non-derepressible 2 kinase
haem-regulated inhibitor kinase
internal ribosomal entry site
interferon regulatory factor 3
mechanistic target of rapamycin
nuclear factor kappa-light-enhancer of activated B-cells
PKR–like endoplasmic reticulum kinase
43S pre-initiation complex
dsRNA-activated protein kinase
ribosomal protein of the large subunit 13a
ribosomal protein kinase
tumour necrosis factor
zipper-interacting protein kinase
Translation UK 2015: Held at the University of Aberdeen, U.K., 7–9 July 2015.