Cell type- and stimulus-specific mechanisms for post-transcriptional control of neutrophil chemokine gene expression

Review on chemokine gene expression and its control by multiple post-transcriptional mechanisms that exhibit differential cell type utilization and stimulus dependency.

Abstract

mRNAs encoding inflammatory chemokines that recruit neutrophils frequently exhibit short half-lives that serve to limit their expression under inappropriate conditions but are often prolonged to ensure adequate levels during inflammatory response. Extracellular stimuli that modulate the stability of such mRNAs may be the same as the transcriptional activator, as is the case with TLR ligands, or may cooperate with independent transcriptional stimuli, as with IL-17, which extends the half-life of TNF-induced transcripts. These different stimuli engage independent signaling pathways that target different instability mechanisms distinguished by dependence on different regulatory nucleotide sequence motifs within the 3′UTRs, which involve that action of different mRNA-binding proteins. The selective use of these pathways by different stimuli and in distinct cell populations provides the potential for tailoring of chemokine expression patterns to meet specific needs in different pathophysiologic circumstances.

Introduction

Acute inflammatory response involves two major actions. The first is orchestration of the movement of leukocyte cell populations from their origin in the bone marrow or peripheral lymphoid tissue to the site of injury and the second, the instruction of these cells upon their arrival to engage appropriate functions ranging from antimicrobial activity to wound healing. The nature of the functional response reflects the temporal order of infiltration by different leukocyte subsets; the polymorphonuclear granulocyte or neutrophil is usually the first to arrive and provides multiple activities, including regulation of subsequent cell recruitment [1, 2]. The movement or trafficking of inflammatory leukocytes is thus a critical point for regulating the magnitude and nature of the response to injury. This process involves multiple steps and the action of many functionally distinct gene products but the chemoattractant cytokines or chemokines are feature players [3, 4]. As the inflammatory response in the acute and chronic format can have a markedly deleterious consequence, it is essential that this process be tightly regulated [5].

There are four chemokine families whose members exhibit selectivity for specific cell populations (mostly leukocytes) based on receptor use, and these subsets are distinguished by the position of the first cysteine residues in the mature protein sequence, designated as CC, CXC, CX3C, and C [3, 4]. Neutrophils are recruited by a subset of the CXC family that contains a signature 3-aa sequence (ELR) preceding the CXC motif, which is essential for recognition by the CXCR1 (in humans only) and CXCR2 receptors. In humans, there are seven members of this family (CXCL1–3 and 5–8), whereas in the mouse, there are only three (CXCL1, -2, and -5). There are modest sequence differences among the members (e.g., CXCL1–3 have nearly 90% aa sequence identity), and they appear to exhibit a high degree of functional redundancy [3, 6]. Nevertheless, multiple family members often expressed at a single inflammatory site.

In this context, the regulation of the expression of neutrophil-directed ELR-CXC chemokine family members shows substantial cell type and stimulus specificity. Importantly, both non-leukocyte and leukocyte cell populations can be significant sources of these chemokines, and epithelial and stromal cells that are resident within tissues are often the earliest source of chemokine expression [7–11]. The nature of the stimuli varies between tissues and with different forms of injury, but proinflammatory cytokines, including IL-1α, TNF, and IL-17, are all well-recognized as important inducers of chemokine gene expression [12–17]. Leukocytes and particularly myeloid cells, including neutrophils and macrophages, are also important contributors and often express a broader spectrum of chemokines that also target other leukocytic cell populations [12, 14]. Myeloid cells exhibit heightened sensitivity for microbial products through recognition of PAMPs by TLRs, and stimulation through TLRs is a potent inducer of chemokine expression. Interestingly, chemokine expression in nonmyeloid and myeloid cell populations appears to be controlled through distinct mechanisms [18].

Transcription from ELR-CXC genes remains very low (but not zero) in resting cell populations (leukocytic as well as nonleukocytic) and can be increased significantly and rapidly in response to stimulation with a variety of proinflammatory stimuli, including cytokines and PAMPs [12, 14, 19]. It is now widely recognized, however, that post-transcriptional control of mRNA half-life is particularly important in determining the magnitude and duration of neutrophil-specific chemokine gene expression [18, 20]. Cytokine and growth factor mRNAs have long been known to exhibit a reduced half-life that is dependent on the presence of adenine uridine rich elements (AREs) within the 3′UTR of the mRNA [21, 22]. These sequences confer instability through interaction with RNA-binding proteins and/or miRNAs with appropriate sequence recognition specificity [23, 24]. The rapid degradation and removal of messages encoding gene products with inflammatory function serve multiple end points, including prevention of inappropriate expression under noninjury conditions and restoration of background expression upon resolution of an inflammatory response [23, 25, 26]. The short half-life can, however, severely limit expression of the product, even during transcriptional bursts, but this limitation is often overcome by transient inactivation of the instability mechanism in response to extracellular stimuli [20, 23, 24, 27].

Although the concepts of ARE-mediated instability and the stimulus-induced stabilization of such messages are well-recognized, the mechanistic diversity or heterogeneity that is operative in such controls is poorly understood. In this regard, it is apparent that not all ARE-containing mRNAs are coordinately regulated and further, that the same mRNA sequence may exhibit distinct decay kinetics when examined in different cell populations or in response to different stimuli [18, 27]. In the following sections, several distinct stimulus-sensitive instability mechanisms governing the expression of the CXCL1 mRNA will be presented using specific experimental settings that illustrate cell type (myeloid vs. epithelial/stromal) and stimulus (TLR ligands vs. IL-17) differences.

MECHANISTIC DETERMINANTS OF mRNA INSTABILITY

Sequence analysis of mRNAs encoding short-lived cytokine mRNAs, such as GM-CSF and TNF, identified the importance of AREs based on their ability to confer instability to otherwise stable messages [21, 22]. ARE sequences are, however, structurally heterogeneous and include the well-characterized five nucleotide element AUUUA, as well as regions enriched in AU residues but without the specific pentamer motifs [28]. Interestingly, the frequency of AREs containing some version of the AUUUA motif within the human genome is surprisingly high, suggesting that these sequences may be broadly important determinants of gene expression [29].

AREs, as well as other sequences that determine half-life, mediate their effects via recognition by RNA-binding proteins [23, 24, 30, 31]. Multiple RNA-binding proteins exhibiting recognition specificity for AREs have been identified over the last 15 years [32–37]. Perhaps the best-characterized is the zinc finger protein TTP, which has been shown to interact with multiple ARE-containing mRNAs, resulting in enhanced decay as a consequence of recruiting components of the RNA degradation machinery (deadenylases, decapping enzymes, exonucleases) to the messenger ribonucleoprotein complex [38–40]. Mice lacking the TTP gene exhibit a systemic inflammatory syndrome that results from disregulation of TNF gene expression as a result of enhanced TNF mRNA stability [26]. Interestingly, a similar phenotype is observed in mice, in which the TNF gene has been replaced with a mutant version lacking the ARE motif [41].

As mentioned earlier, instability mechanisms can interfere with expression of the targeted genes and are frequently inactivated, at least transiently, to enable accumulation of a sufficient message to generate the needed protein products. It is worth considering that such mRNA stabilization is often a response to the same stimuli that promote gene transcription and that this can preclude detection of instability and stabilization (i.e., it is difficult to observe mRNA instability when mRNA is only detected in the presence of the stabilizing stimulus). Ligands for the TIR family, such as IL-1 and LPS, have been shown by a number of laboratories to increase the stability of ARE-containing cytokine mRNAs [42–46], although the rate of decay and the sensitivity to stimulus-induced stabilization vary substantially, suggesting mechanistic heterogeneity [47, 48]. It is worth noting that not all proinflammatory cytokines can initiate the mRNA stabilization response pathway; in particular, treatment of epithelial and stromal cells with TNF can promote transcription of ELR-CXC family chemokine mRNAs but does not prolong their half-life and as such, is by itself a relatively weak inducer [12, 43, 49]. However, other stimuli, notably IL-17, can cooperate with TNF by promoting the stabilization mechanism [49]. As discussed below, the RNA sequences, RNA-binding proteins, and signaling pathways used by different agents are often distinct.

AUUUA/TTP-DEPENDENT INSTABILITY AND TLR-MEDIATED STABILIZATION

The instability of the CXC chemokine mRNAs and the ability to modulate their half-life by extracellular stimulation were recognized early, and these messages have served as models for analysis of the control of mRNA stability during TIR family ligand-induced inflammatory responses [12, 13, 42, 50, 51]. Studies to define the responsible sequence motifs that confer instability were focused on the 3′UTR and particularly, the ARE regions using human IL-8 and mouse CXCL1(KC) mRNAs as examples. In both cases, the clustered collections of AUUUA elements were identified and shown to be critical determinants for instability [42, 43, 51]. However, unlike regulatory sequences governing the control of transcription, the regions of the 3′UTR are substantially longer (often >100 nucleotides) and exhibit complexity that appears to result from distinct motifs [52].

The identification of the regulatory sequences responsible for cytoplasmic mRNA decay led to identification of multiple RNA-binding proteins with appropriate sequence recognition specificity. Interestingly, the discovery of TTP as a mediator of ARE-containing mRNA instability was not a result of such efforts but rather, was based on the phenotype of TTP-deficient mice [26]. It is now clear that TTP is responsible for regulating expression levels of multiple cytokine mRNAs via its ability to promote rapid degradation, and sensitive targets include TNF, GM-CSF, IL-10, and CXCL1(KC) [33, 53–55]. TTP has been shown to recognize AUUUA motifs with high specificity [56, 57], and the seven AUUUA sequence motifs in CXCL1 mRNA 3′UTR are responsible for its sensitivity to TTP-dependent degradation, as mutation of these sites abrogates message instability [55]. Indeed, it is apparent that TTP is the major determinant for instability of CXCL1 mRNA in LPS-treated macrophages, as the mRNA remains completely stable in cells from TTP-deficient mice [55].

The modulation of chemokine mRNA decay rates by proinflammatory stimuli, particularly ligands for members of the TIR family, was reported by several laboratories as early as 1991 and confirmed independently, shortly thereafter [12, 13]. Multiple studies have provided more insight into features of this response with emphasis on definition of specific sequence requirements, identification of critical RNA-binding proteins, and characterization of key signal transduction events [42–44, 52, 58–61] (see Fig. 1). Following ligand occupancy, TIRs recruit one or more adaptor proteins, including Myd88, TIR domain-containing adaptor protein, TRIF, and/or TRIF-related adaptor molecule, which share the TIR intracellular protein-interaction domain [62, 63]. Interestingly, of all TIRs, only TLR3, the receptor that exclusively uses the TRIF adaptor for signaling, is unable to initiate the signal leading to mRNA stabilization, whereas TLR2, -4, and -9 are all fully competent to induce NF-κB for transcriptional activation, as well as the stabilization of chemokine mRNAs [46]. Myd88 also contains a second protein-interaction domain that serves to recruit members of the IRAK family, particularly IRAK1, -2, and -4 [64–66]. There is a differential requirement for kinase activity in the various IRAK activities, and the kinase function and phosphorylation state of the proteins appear to provide coupling to distinct end points [64, 67, 68]. Indeed, in mice, in which the IRAK4 gene has been substituted with a kinase-inactive version, signaling to the activation of mRNA stabilization in myeloid cells in response to multiple TIR ligands is compromised [68]. The signal pathway extends downstream of IRAKs through TRAF6, ultimately leading to activation of the IKKs and canonical NF-κB, as well as to the MAPK cascade [66, 67, 69–71].

Figure 1. TLR signaling controls transcription and mRNA half-life.

TLR–ligand interaction intiates signaling by recruiting the TLR adaptor Myd88, leading to the assembly of a complex containing, additionally, IRAK1, IRAK4, TRAF6, and TGF-β-activated kinase 1 (TAK1). The signal bifurcates to activation of NF-κB through the IKK complex or to mRNA stabilization via the p38 MAPK pathway. MAPKAP2 phosphorylates TTP and inactivates its mRNA decay function.

The MAPK pathways and particularly, p38 MAPK are well-recognized as important in the regulation of cytokine mRNA half-life, specifically in the context of extracellular stimulus-dependent modulation [45, 58, 60, 72]. This is also true for control of mRNAs encoding neutrophil-directed chemokines, and the half-lives of IL-8 and CXCL1(KC) have been linked with p38 activation using pharmacologic inhibitors and via genetic manipulation [42, 55, 58, 73]. This is supported further in mice, where deletion of the genes encoding MAPKAP2 or MAPKAP3 results in significant diminution in the expression of CXCL1(KC) as a consequence of the inability of the stimulus (LPS) to prolong the half-life of the mRNA [72, 74]. It is evident however, that this outcome depends on the specific TLR involved; macrophages stimulated with LPS through TLR4 show enhanced stability of TNF mRNA, whereas stimulation with ligands that activate through TLR7 or TLR9 impact mRNA translation but not decay [75]. The substrate of MAKAP2/3, responsible for mediating RNA stabilization, is TTP, which is phosphorylated on two critical serine residues (S52, S178) [74, 76–78] and results in alterations in TTP functions, including nuclear/cytoplasmic localization, TTP protein half-life, P body association, and the ability of TTP to recruit deadenylase activity to the mRNA [38, 78]. This pathway provides a molecular basis for stimulus-induced modulation of mRNA half-life from receptor occupancy to effector mechanism. Although TTP certainly has a broad array of activities and has been associated with the control of multiple short-lived mRNAs, numerous questions remain concerning cell-type expression and function, as well as a detailed understanding of the biochemical mechanism.

A DISTINCT INSTABILITY MECHANISM FOR IL-17-MEDIATED mRNA STABILIZATION

Chemokines, unlike some proinflammatory cytokines, are expressed at high levels, not only by inflammatory leukocytes but also by nonhematopoietic cell populations, including epithelial and stromal cells as well. Indeed, epithelium, endothelium, and fibroblasts are often the major source for the early expression of CXCL1(KC) within injured tissues, and hence, it is appropriate to consider the mechanisms involved in such cell types and whether they are similar to those observed in myeloid cells [11]. IL-17 is known to be a potent stimulus of neutrophil mobilization and trafficking to sites of infection and has been reported by several laboratories to promote expression of chemokines and other inflammation-related gene products, at least in part by enhancing message stability [49, 79, 80]. IL-17 is produced predominantly by a subpopulation of T cells (now known as Th17) and contributes to potent proinflammatory activity associated with a number of acute and chronic inflammatory diseases, in which nonmyeloid cell populations appear to be the primary targets [81, 82]. It is noteworthy that IL-17 alone can activate NF-κB, although in comparison with multiple well-characterized agents, the magnitude of this response is modest [49, 83]. Hence, although IL-17 is not by itself a strong stimulus of CXCL1(KC) gene transcription, it can cooperate effectively with TNF via stabilization of TNF-induced chemokine mRNA transcripts [49, 79, 84].

Although IL-17 is an effective agent for inducing enhanced chemokine mRNA stability in fibroblasts and epithelial cells, it is now evident that this activity is mechanistically distinct from TLR-mediated message stabilization occurring in myeloid cells. The first indication of mechanistic heterogeneity for control of the CXCL1(KC) half-life was the finding that mRNA from a construct, in which all seven AUUUA pentamers in the 3′UTR were deleted or mutated, retained instability and stimulus sensitivity [55, 61]. This suggested that the message contained a distinct instability motif with different regulatory properties, and indeed, instability and sensitivity for stabilization in cells stimulated with IL-1 or IL-17 could be demonstrated in a fragment of ∼155 nucleotides located at the 3′end of the 3′UTR of CXCL1(KC), which does not contain AUUUA motifs [85]. Similarly, mRNAs derived from plasmid constructs that contained only the AUUUA pentamer-enriched region retained instability that was not subject to modulation by IL-17. Hence, this region contains AUUUA-independent determinants of instability and stimulus sensitivity, which operate in the nonmyeloid cell environment.

Sequence motifs that confer instability and sensitivity to IL-17 for stabilization showed no sensitivity to the action of TTP [85]. Moreover, using embryo fibroblasts from mice deficient in the TTP gene, CXCL1(KC) mRNA retained significant instability and could be stabilized in response to IL-17. Hence, TTP is not necessary for instability or IL-17-mediated stabilization of this mRNA. It remains possible, however, that TTP-mediated decay of mRNAs bearing AUUUA target sequences may be sensitive to IL-17 in cells that express IL-17R. Another ARE-binding protein—KSRP—has been reported to contribute to the instability of IL-8 mRNA, and its stimulus sensitivity is distinct from that operating through the TTP pathway [86]. Although KSRP can bind the KC mRNA and may impact through the AUUUA motif, it is not responsible for the IL-17-sensitive instability mechanism [85].

The IL-17R is composed of two chains (IL-17RA and IL-17RC), and signaling depends on an intracellular protein domain termed SEFIR, which has some structural similarity to TLRs [87, 88] (see Fig. 2). Although it is now clear that TIR adaptors, such as Myd88 and TRIF, do not participate in IL-17 signal transduction, the SEFIR domain does serve as a docking location for an adaptor protein known as Act1 or TRAF3-interacting protein, which also contains a SEFIR domain [83, 89]. Act1 has several important structural features, among which, are two TRAF-interaction motifs, and IL-17 is known to induce the activation of NF-κB through a pathway that requires TRAF6 [49, 83, 89]. As TRAF6 appears to be important in signaling from TIRs for mRNA stabilization, it seems reasonable that it might also be important in IL-17-mediated mRNA stabilization. Surprisingly, however, TRAF6 does not appear to participate in the IL-17-induced pathway, at least in nonmyeloid cells, as IL-17 can promote enhanced stability of CXCL1(KC) mRNA even in embryo fibroblasts from mice deficient in TRAF6 [84]. Hence, the signaling pathway from the IL-17R bifurcates above TRAF6 with one pathway linking to the activation of NF-κB through TRAF6 and a second TRAF6-independent pathway connecting with mRNA stabilization. Responses to IL-1α in some epithelial cells appear to be independent of TRAF6 as well, suggesting that mechanistic heterogeneity can derive from ligand–receptor and cell-type differences [84].

Figure 2. TNF and IL-17 cooperate to promote transcription and stabilization of target mRNAs.

TNF provides strong transcriptional gene induction through activation of NF-κB via TNFR-associated death domain (TRADD), TRAF2 or -5, receptor-interacting protein (RIP), and IKK. IL-17 can also couple to NF-κB activation via Act1 and TRAF6. The signal from IL-17R can be shunted to mRNA stabilization via the action of IKKε-mediated phosphorylation of Act1, promoting a complex with TRAF5 or TRAF2 and SF2/ASF, which binds CXCL1 mRNA to promote instability, and this may be disrupted following recruitment of SF2/ASF into the complex with TRAFs and Act1.

Recent work has identified additional features of the IL-17 signaling pathway, which link IL-17R with mRNA stabilization [90, 91], and is illustrated further in Fig. 2. Although TRAF6 is not required, TRAF2 or TRAF5 appear to be essential. Cells deficient in both factors cannot stabilize mRNA in response to IL-17. The requirement for both is reminiscent of their redundancy in the TNFR signaling pathway; however, it is possible that TRAF5 is the physiologically relevant component. This is suggested by selective effects of dominant interfering versions of TRAF5 versus TRAF2 on an IL-17-induced CXCL1 mRNA half-life. TRAF2 and/or TRAF5 also interact with Act1, and this interaction is dependent on phosphorylation of Act1 via IKKε on serine 311 [91]. In IKKε-deficient embryo fibroblasts, the CXCL1 mRNA half-life is not sensitive to IL-17, and the interaction of Act1 with TRAF2/TRAF5 is also lost. Moreover, IL-17-mediated neutrophil infiltration into the lungs of IKKε–/– mice is reduced compared with that seen in WT mice. It is noteworthy, however, that the Act1/TRAF6 interaction is not disrupted by the IKKε deficiency.

The connection between TRAF2/TRAF5 and chemokine mRNA half-life is accomplished via an RNA-binding protein known predominantly for its role in conventional and alternative splicing, termed SF2 or ASF [92, 93]. The identification of SF2/ASF in the IL-17 pathway was based on coimmunoprecipitation with TRAF2/TRAF5. Importantly, SF2/ASF appears in the complex that contains Act1 and the TRAFs and is induced in response to IL-17 stimulation. SF2/ASF has the capacity to promote the decay of CXCL1(KC) mRNA and shows sequence selectivity, as the IL-17-sensitive motif, but not AUUUA-containing motifs, exhibits selective sensitivity. SF2/ASF can bind to such mRNAs, and the magnitude of binding is strongly diminished in cells stimulated with IL-17. Hence, the connection from receptor to effector stage for IL-17-mediated changes in mRNA half-life differs markedly from that for the TLR-driven response in macrophages and yet, can impact the same mRNA, albeit through distinct sequence motifs.

PERSPECTIVES AND CONCLUSIONS

Much emphasis has been placed on the importance of transcription as the principle regulatory point in determining patterns of inflammatory gene expression, and it is not our intention to discredit this concept. Of course all gene expression must initiate with transcription, and for chemokine genes in particular, there is abundant evidence supporting stimulus-induced modulation of their transcription [14, 19]. That being said, it is increasingly clear that this represents only the beginning of the process, and hence, additional effort should be focused on sequential steps that include splicing, polyadenylation, nuclear-cytoplasmic transport, cytoplasmic localization, and translation. Moreover, each of these steps may be targets of signal transduction pathways that modulate their function and ultimately impact the magnitude and/or duration of expression for specific genes. One of the major remaining hurdles in the biology of multicellular organisms is the mechanistic basis for the tremendous specificity in cellular behavior, and pursuit of understanding the details of post-transcriptional processes will contribute to its solution.

In this context, the emerging knowledge of mRNA decay and its modulation by extracellular stimuli during physiologic and pathophysiologic responses indicate that there is likely to be substantial, functional diversity of process. Although the heterogeneity of ARE sequences within the genome would suggest the existence of multiple mechanisms operating on distinct subsets of mRNAs, there has been some sense that AU-rich sequence control represents a single process. There are now, however, several good examples that illustrate different activities linked to distinct sequence motifs and of course, the RNA-binding proteins that recognize them [52, 85, 94, 95]. Equally important is the concept that binding alone may not be the primary determinant of action, and additional features, including conformational changes on RNA–protein interaction, modification of protein–protein interaction, and the presence of other proteins within the ribonucleoprotein particle, seem certain to contribute [38]. The subcellular localization or concentration of process-specific components (e.g., P bodies and stress granules) and the participation of miRNA species and machinery in the process are additional variables likely to have significant impact [96, 97]. Finally, it is also clear that these mechanisms will operate with distinct cell-type specificity, reflecting the individual cellular proteome. The examples of chemokine behavior in distinct cells and in response to different cellular stimuli outlined in this review provide clear examples of this potential. The number of independent, functionally redundant, neutrophil-specific chemokine genes in the human genome raises the issue of the evolutionary basis for their maintenance. Although perhaps not the only contributing feature, it is interesting to speculate that this may reflect the value of encoding distinct patterns of regulatory potential into different members of a multigene family to enable tailoring expression of common gene function to specific physiologic circumstances.

AUTHORSHIP

All authors contributed to the review of literature and interpretation of data described within this article.