The GAIT system: a gatekeeper of inflammatory gene expression

Translational control by the GAIT pathway

Monocytes and macrophages provide an early line of defense against infection or injury. They are activated by cytokines to produce and secrete proteins and small molecules that can kill invading microorganisms. Overexpression of these toxic agents can be detrimental to host cells and tissues, contributing to chronic inflammatory disease. Therefore, host responses must be strictly controlled to limit the expression of injurious molecules. The ‘resolution of inflammation’ is an active program of regulated gene expression by which pro-inflammatory processes are restricted after neutralization or elimination of the initial insult [8]. Thus monocytes or macrophages, and their agonists, for example IFN-γ, can exhibit either pro- or anti-inflammatory responses depending on the context [9,10]. Early studies from our laboratory showed marked transcriptional induction of ceruloplasmin (Cp), an acute phase inflammatory protein, in human monocytes upon IFN-γ treatment [11]. However, Cp synthesis, as measured by pulse-labeling, was terminated almost completely after ~16 h, despite the presence of abundant Cp mRNA, a result indicative of translational inhibition. Total protein synthesis was unaltered, indicating transcript-selective inhibition. The inhibition was cell-type specific and observed only in myeloid cells (and related cell lines) [12]. The shift of Cp mRNA from the polyribosomal to non-polyribosomal pool indicated that inhibition occurred at the level of translation initiation.

Analysis of the Cp 3′UTR GAIT element

An RNA gel shift assay demonstrated that the Cp 3′UTR binds an IFN-γ-inducible protein/complex, termed the GAIT complex [13]. The functionality of the Cp 3′UTR was verified by in vitro translation; ligation of the 3′UTR downstream of a luciferase reporter conferred the translational silencing response to the heterologous transcript in rabbit reticulocyte lysates supplemented with lysate from IFN-γ-treated U937 cells. From these analyses, a minimal, 29-nt Cp 3′UTR GAIT RNA element was identified. A secondary structure consisting of a stem-loop with an asymmetric internal bulge was predicted by Mfold, an RNA-folding algorithm (http://mfold.bioinfo.rpi.edu/) [14], and was confirmed by mutation analysis (Figure 1). Most loop residues could be mutated without loss of binding or silencing activity; however, two residues in the bulge loop, A84 and U87, are essential, and mutation of either completely inactivated the element. Similar functional GAIT elements containing a stem-loop with an internal bulge were identified in the 3′UTR of multiple transcripts (Figure 1; see below).

Elucidation of the heterotetrameric GAIT complex

RNA affinity purification and the yeast three-hybrid approach [15] identified the four constituent proteins of the GAIT complex that binds the Cp GAIT RNA element but not an inactive mutant element. This complex is comprised of glutamyl-prolyl tRNA synthetase (EPRS), NS1-associated protein 1 (NSAP1, also known as SYNCRIP or heterogenous nuclear ribonucleoprotein [hnRNP] Q1), glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and ribosomal protein L13a [16,17]. The GAIT complex assembles in two distinct steps. In the first step, which occurs ~2 h after IFN-γ treatment, EPRS is released from its residence in the tRNA multisynthetase complex (MSC) and joins NSAP1 to form an inactive pre-GAIT complex that neither binds target mRNAs nor inhibits translation (Figure 2). About 12–14 h later, ribosomal protein L13a is released from the 60S ribosomal subunit and joins GAPDH and the pre-GAIT complex to form the functional GAIT complex. Elucidation of the mechanisms by which protein constituents are inducibly released from ‘stable’ macromolecular complexes remains a particularly exciting challenge [18] (Box 1).

Phosphorylation-dependent release of EPRS from the tRNA MSC

EPRS has a crucial role in the GAIT pathway because it recognizes and directly binds GAIT element-bearing target mRNAs [17]. EPRS is a unique, bifunctional tRNA synthetase that consists of two catalytic cores, which catalyze the ligation of Glu and Pro to cognate tRNAs [19,20]. The catalytic domains are joined by a noncatalytic linker containing three tandem WHEP-TRS (or WHEP) domains (named after three tRNA synthetases that contain them, i.e. Trp(W)RS, His(H)RS and EPRS). The function of WHEP domains is unknown, but they bind RNA and protein with low affinity and specificity [19,20]. The presence of EPRS WHEP domains in nearly all metazoan species examined to date suggests an important function that has been retained for half a billion years or more [21,22]. The absence of WHEP domains in the separate ERS and PRS enzymes in bacteria and yeast indicates that they are not essential for synthetase activity but instead might be involved in a noncanonical function of the enzyme.

EPRS phosphorylation at one or more Ser residues is the earliest IFN-γ-inducible event in activating the GAIT translational silencing pathway. Phosphorylation induces EPRS release from the MSC and facilitates binding to cytoplasmic NSAP1; Ser kinase inhibitors prevent EPRS release and its interaction with NSAP1 [17]. Earlier studies showed that tRNA synthetase phosphorylation has little or no effect on synthetase activity [23], supporting a potential role for post-translational modifications in regulating synthetase noncanonical functions.

EPRS–target mRNA binding is subject to negative and positive regulation

EPRS binds the GAIT RNA element but, intriguingly, is unable to do so until ~14 h after its release from the MSC, and a domain analysis of EPRS binding activities provided an important clue as to why. Deletion analysis showed that the upstream pair of WHEP repeats (R1 and R2) in the EPRS linker directed phosphorylation-independent, high-affinity binding to Cp GAIT element RNA [24]. A similar analysis revealed that NSAP1 also bound EPRS in the linker domain; however, binding was phosphorylation-dependent and required the downstream pair of WHEP repeats (R2 and R3). Thus NSAP1 and GAIT element RNA share overlapping binding sites on EPRS, suggesting that NSAP1 might inhibit EPRS–RNA binding. Surface plasmon resonance (SPR) experiments with recombinant, purified GAIT constituent proteins revealed that high-affinity binding of phosphorylated EPRS to GAIT element RNA was completely blocked by pre-incubation with NSAP1. NSAP1 did not bind non-phosphorylated EPRS or inhibit its binding to RNA. Thus, EPRS phosphorylation is not required for binding to the GAIT RNA but is essential for NSAP1-mediated negative regulation.

The presence of NSAP1 inhibitory activity raised a question, namely, how is inhibition overcome 16 h after IFN-γ treatment? The restoration of RNA binding by the remaining components of the GAIT complex was investigated. Addition of either GAPDH or phosphorylated L13a (P-L13a) was ineffective; however, simultaneous addition of both proteins restored EPRS binding to GAIT element RNA even in the presence of NSAP1 [24]. Comparable reconstitution of translational inhibition activity in vitro established that the functional GAIT complex requires only these four proteins [24].

The mechanism underlying regulation of the GAIT complex was investigated by measuring Förster resonance energy transfer (FRET) between NSAP1 and phosphorylated EPRS linker, both fluorescently-labeled at their N-termini. Addition of phospho-L13a and GAPDH induced a conformational shift that increased the distance between NSAP1 and the N-terminus of EPRS linker by ~9 Å [24], a shift that might be sufficient to move NSAP1 away from the upstream WHEP domains and permit RNA binding. We expect that elucidation of the specific EPRS phosphorylation sites and detailed structural analysis of GAIT constituents and their interaction sites will provide additional insight into the conformational switches that regulate GAIT complex activity.

Phosphorylation of ribosomal protein L13a and release from the 60S ribosomal subunit

Phosphorylation of L13a ~12–16 h post-IFN-γ treatment is obligatory for L13a release from the large, 60S ribosomal subunit and is a crucial rate-determining event in GAIT pathway activation [16,25]. Remarkably, nearly the entire cellular complement of L13a exits the ribosome without detectable inhibition of global protein synthesis [26]. Mass spectrometry and mutational studies revealed that IFN-γ-inducible L13a phosphorylation occurs only on Ser⁷⁷ [25] in an RXXS/T consensus kinase recognition motif [27–29]. In vitro and in vivo screening studies identified zipper-interacting protein kinase (ZIPK also known as death-associated protein kinase-3 or DAPK3), a member of the DAPK family, as the proximal kinase for L13a phosphorylation [25]. Another member of the same family, DAPK1 (referred to as DAPK) was identified as an upstream activator of ZIPK, indicating a mini DAPK–ZIPK signaling cascade, consistent with reports that DAPK can activate ZIPK [30].

Mechanism of L13a-mediated translational inhibition

Binding of proteins (or microRNAs) to 3′UTRs influences translation initiation at the distant 5′UTR. The ‘closed-loop’ model of mRNA translation provides a crucial clue. According to this model, an interaction between the 3′UTR and 5′UTR improves translation efficiency by ensuring that ribosomes released at the 3′ end after the completion of protein synthesis are close to the 5′ end to enable initiation of a new round of synthesis. In this model, the transcripts are ‘circularized’ by interaction of eukaryotic initiation factor 4G (eIF4G) at the 5′ terminus with poly(A)-binding protein (PABP), which binds the poly(A) tail at the 3′ terminus (Figure 2) [31]. Initial studies in the GAIT system sought to understand whether or not the GAIT complex interrupted transcript closure, thereby inhibiting protein synthesis. Despite multiple approaches, the GAIT complex was not found to disrupt any interactions at the termini [32]. Subsequent studies of the alternative hypothesis, that the GAIT complex does not impede circularization but rather depends on it, revealed that PABP depletion prevented translational silencing of a reporter transcript containing the GAIT element. Therefore, transcript circularization not only promotes efficient translation but also can be essential for transcript-selective translational control. The requirement for a poly(A) tail has been shown for translational inhibition by at least one microRNA, but the generality of this requirement for transcript closure has not been established [33,34].

Multiple mechanisms of transcript-selective translation inhibition have been shown, including inhibition of eIF4F (cap-binding complex) activity, recruitment of the 43S pre-initiation complex, scanning of the 43S complex to the initiation codon, and 60S ribosomal subunit joining [2,3,6]. L13a was considered a strong candidate for the translational repressor because its phosphorylation and release from the ribosome temporally coincided with translational silencing of target mRNAs. Indeed, P-L13a targets eIF4G, the component of the 5′ cap-binding, translation-initiation complex that makes initial contact with the ribosome via interaction with eIF3 of the 43S translation preinitiation complex (which contains the 40S small ribosomal subunit) (Figure 2). Its large size, central positioning and manifold interactions render eIF4G a conspicuous target for translational control. Studies with the eIF4G-specific protease HRV-2A (human rhinovirus-2A) complemented by deletion mapping revealed that P-L13a binds eIF4G at its eIF3-binding site [35]. This interaction blocks recruitment of the eIF3-containing 43S ribosomal complex to GAIT element-bearing mRNAs, thereby blocking translation initiation. Inactivation or degradation of eIF4G are well-established RNA-independent mechanisms to achieve global translation inhibition, for example during cellular stress [35]. However, the RNA-dependent eIF4G–P-L13a interaction provides a unique mechanism for transcript-selective translation control.

Pattern analysis identifies a GAIT element in the 3′UTR of VEGF-A and other mRNAs

PatSearch (http://www.ba.itb.cnr.it/BIG/PatSearch/), a folding energy-independent, pattern analysis algorithm, was used to search a 3′UTR database containing more than 200 000 sequences for a consensus GAIT element pattern, based on sequence and secondary structure of the human Cp GAIT element [41]. The algorithm found 55 human mRNAs with putative 3′UTR GAIT elements; only two mRNAs were predicted to contain 5′UTR GAIT elements, indicating marked positional specificity [39]. A Gene Ontology (GO)-based literature search [42] indicated that many putative GAIT-element-bearing transcripts were associated with the inflammatory response or induced by inflammatory agonists [39].

Vascular endothelial growth factor A (VEGF-A), a hallmark inflammatory protein that induces angiogenesis, was among the candidate mRNAs. Mfold, which calculates the optimal thermodynamic folding structure of RNA [43], revealed a bipartite stem-loop structure in the VEGF-A 3′UTR (nt 358–386) similar to the Cp GAIT element [13]. A series of experiments confirmed the authenticity of the VEGF-A GAIT element. The element conferred translational silencing to a heterologous reporter RNA in an in vitro translation system [39]. IFN-γ shifted VEGF-A mRNA from ribosome-rich, polysomal fractions to a ribosome-poor, ribonucleoprotein fraction. An RNA electrophoretic mobility shift assay (EMSA) ‘supershift’ experiment demonstrated an interaction with the four GAIT proteins. Moreover, suppression was not observed in L13a-deficient cells. In monocytic cells, IFN-γ induced VEGF-A protein expression after 8 h, followed by a return to the basal level after 24 h despite abundant VEGF-A mRNA, indicating delayed translational repression of endogenous VEGF-A. This translational inhibition decreased angiogenic activity of monocytic cells, as shown by reduced formation of capillary-like tubes by endothelial cells cultured in Matrigel^™ [39].

Analysis of the kinase cascade (DAPK–ZIPK) responsible for L13a phosphorylation showed that expression and activity of both DAPK and ZIPK plummeted ~16 h after IFN-γ treatment, despite robust mRNA expression, suggesting the possibility of translational control [25]. The PatSearch algorithm revealed a candidate GAIT element in the human DAPK mRNA 3′UTR but not in ZIPK. Secondary structure analysis of the ZIPK 3′UTR by Mfold revealed a split stem-loop structure with some resemblance to the Cp and VEGF-A GAIT element structures (Figure 1). The crucial ‘A’ and ‘U’ nucleotides were present in the bulge of the DAPK GAIT element, as in Cp and VEGF-A, but were missing from the ZIPK GAIT element. Functionality of both the DAPK and ZIPK elements was shown by translational silencing of a heterologous reporter in vitro, by RNA EMSA and by supershift analysis [25]. Transfection of antisense morpholino oligomers complementary to the GAIT elements of both kinases prevented their translational silencing, presumably by blocking GAIT complex binding. The morpholinos maintained long-term expression of the kinases as well as L13a in its active, phosphorylated state. The possibility of restricting macrophage inflammatory activity using morpholinos targeted against DAPK and ZIPK presents an exciting application with potential therapeutic benefit.

Translational silencing of DAPK and ZIPK can prevent the accumulation of these potentially injurious, pro-apoptotic kinases [30,44]. More importantly, DAPK, ZIPK and L13a form a negative-feedback module in which L13a phosphorylation activates the GAIT complex, represses kinase translation and returns L13a to the non-phosphorylated form, thereby restoring the cell to the basal state, and permitting re-stimulation by IFN-γ [25]. This network module could exemplify a principle of parsimonious pathway evolution in which the principal function of an existing system, in this case translational silencing of a regulon of inflammatory transcripts, is co-opted to downregulate the initial stimulus. Likewise, negative-feedback regulation of translation has been observed for the tyrosine kinase c-Src; binding of its phosphorylated substrate, hnRNP K, to the cSrc mRNA 3′UTR inhibits expression during erythroid maturation [45,46]. Thus, inhibition of kinase mRNA translation by binding of a phosphorylated substrate to the 3′UTR might represent a new ‘network motif’ in gene expression [47].

Multiple chemokine ligand and receptor mRNAs are GAIT pathway targets

Genome-wide microarray analysis of polysome-bound mRNAs isolated from IFN-γ-activated monocytic cells identified a family of mRNAs encoding multiple chemokine ligands and receptors as candidate GAIT pathway targets [40]. Structural prediction supported this finding. The presence of functional GAIT elements in transcripts encoding chemokine ligand and receptors CCL22, CCR3, CCR4 and CCR6, and also in APOL2 (apolipoprotein L 2), was validated by their ability to repress the translation of a heterologous reporter RNA in L13a-dependent fashion [40]. Importantly, overexpression of chemokines and their receptors is a major contributor to the pathogenesis of multiple inflammatory conditions, for example atherosclerosis and neuroinflammation [48,49]. Despite the expanding cohort of GAIT-targeted transcripts, consensus on the secondary structure of the element has not yet been achieved, except that all form stems split by their asymmetric bulge loops. Likewise, information is not yet available on the relative binding affinity of the GAIT elements in these transcripts for the GAIT complex, and it remains unknown whether a silencing hierarchy exists whereby certain transcripts are more effectively silenced than others.