In general, in eukaryotic cells, initiation can be controlled at multiple levels. The eIF4BP proteins (4E-BP) interact with eIF4E, preventing its interaction with eIF4G and, therefore, inhibiting translation initiation (7). 4E-BP are regulated at multiple phosphorylation sites, often by the mammalian target of rapamycin (mTOR), which reduces 4E-BP interactions with eIF4E and enhances translation initiation (8). In addition, the cyclin-dependent kinase 2 (CDK2) phosphorylates 4E-BP on Thr70 leading to its release from eIF4E (9).

Eukaryotic translation initiation factor 4E (eIF4E) can be phosphorylated on Ser209, which decreases its affinity for the 5′-cap structure and, therefore, inhibits translation (10, 11). eIF4E is phosphorylated by the mitogen-activated protein kinase (MAPK) signal-integrating kinases Mnk1 and Mnk2 (MKNK1 and 2), which are downstream targets of the MAPK (ERK and p38) (10, 11). Also among the eIF4 family, eIF4B is phosphorylated by p70S6K and p90S6K (RSK) at Ser422, which increases its interaction with eIF3, enhancing translation initiation (12).

When the inactive form of eIF2, eIF2-GDP, leaves the 40S during initiation, it must be recharged with GTP for continuous translation initiation. Then, eIF2-GDP is converted to eIF2-GTP by eIF2B, which is a guanine-nucleotide-exchange factor (GEF). eIF2B is phosphorylated at multiple sites that includes two residues phosphorylated by casein kinase 2 (CSNK2A1) that are required for eIF2B/eIF2 interactions, eIF2 recycling and translation initiation (13). eIF2 is phosphorylated at Ser51 by as many as four kinases, all of which inhibit the eIF2–eIF2B interaction, demonstrating a critical role in protein synthesis (14).

The delivery of aa-tRNA required for elongation is driven by the hydrolysis of eEF1A-GTP to eEF1A-GDP. Thus, the GEF eEF1B acts on eEF1A-GDP as eIF2B does on eIF2-GDP. eEF1 is also targeted by a variety of kinases, including PKC, CSNK2, and cyclin-dependent kinase 1 (CDK1), but the role of the phosphorylation states of these elongation factors remains uncertain (15).

Phosphorylation by eEF2 kinase on Thr56 impairs eEF2’s ability to bind to the 40S subunit of the ribosome (16). Thr56 phosphorylation is enhanced if Ser595 is previously phosphorylated (17). The eEF2 kinase activity is calcium/calmodulin-dependent. Its activation after Ca²⁺ flux leads to the attenuation of elongation. Of note, increased eEF2 kinase activity may provide mRNA with poor translation initiation efficiency a greater chance of being synthesized (18). eEF2 kinase is itself regulated at multiple phosphorylation sites, typically by the mammalian target of rapamycin complex 1 (mTORC1) that reduces its activity (19). AMP-activated protein kinase (AMPK) and the MAPK can also phosphorylate eEF2 kinase leading to translation enhancement (18). Conversely, cAMP/PKA signaling pathway phosphorylates Ser500 (20), rending eEF2 activity independent of Ca²⁺ ions and activating the kinase. Figure 2 summarizes these different signaling events and control points.

Recently, using two-dimensional liquid chromatography coupled with high-resolution mass spectrometry, 6,813 proteins were identified in unstimulated human blood EOS ((3), and see article by Mosher et al., in this issue for more details). In addition, 4,802 site-specific phosphorylation events were simultaneously identified (3). Furthermore, using RNA-Seq, ~7,981 protein-coding genes expressed by unstimulated human blood EOS were identified (21). The cellular content (mRNA and protein) and phosphorylation state of the main proteins involved in the initiation, elongation, and termination of protein translation have been extracted from the published proteome and transcriptome (shown in Table 1). Notably, Table 1 shows examples of the disconnections between mRNA and protein levels, which suggests that production of certain proteins is tightly regulated at the translational level in EOS (i.e., EIF4G2, ETF1, etc.). For instance, while ratio of protein to mRNA expression generally reached ~1,000 and above, ratios for EIF4G2 and ETF1 were only 107 and 160, respectively (Table 1), suggesting marginal translation for these two transcripts in resting EOS. With the possible exception of the inhibitor of elongation, eEF2K, resting blood EOS possess all the essential proteins involved in protein translation. However, the identification of eEF2 phosphorylation on Thr56 (Table 1) suggests the existence of eEF2K activity, preventing the eEF2/40S interaction and blockade of translation elongation (16). In addition, the lack of phosphorylation of eIF2B (Table 1) suggests a possible lack of eIF2B/eIF2 interactions and reduced recycling of eIF2 into its active form (eIF2-GTP), which would dampen translation initiation (13). Conversely, in agreement with our previous report (22), 4E-BP is phosphorylated in resting EOS (Table 1). This indicates that 4E-BP does not act as a blocker of eIF4E binding to the 5′ cap in resting EOS and, therefore, other factors are responsible for restricting protein translation in resting EOS. Thus, the combined lack of eIF2B phosphorylation with the phosphorylation of eEF2 on Thr56 suggests attenuation of both initiation and elongation of protein translation in resting EOS (22).

IL-5, GM-CSF, and IL-3 are critical cytokines for EOS development and function. Each interacts with a specific α-chain receptor and a common, associated β-chain (24, 25). Not surprisingly, these receptors can generate both common and unique signals (22, 26–28). As indicated above, we have shown that 4E-BP is highly phosphorylated in resting EOS (22). After activation with IL-3, IL-5, or GM-CSF, 4E-BP phosphorylation state remains largely unaffected (22), suggesting that the increased translation induced by these cytokines is likely 4E-BP-independent. In addition to 4E-BP, we have unpublished observations indicating that EIF4B phosphorylation at Ser422 was unaffected by GM-CSF. Therefore, as for 4E-BP, eIF4B activity cannot explain the significant enhancement of translation in GM-CSF-activated cells (22). However, a slight but significant increase in the phosphorylation of eIF4B was observed in EOS activated by IL-3 for 20 h (unpublished data). This phosphorylation on Ser422 may account for the differences in translation seen in IL-3- versus GM-CSF-activated EOS (22). The signaling accounting for regulated translation in IL-3 or GM-CSF-activated EOS remains largely unstudied.

Ligation of growth factors with tyrosine kinase or G-protein coupled (GPC) receptors typically leads to phosphorylation of the membrane phospholipid, phosphatidylinositol-4,5-biphosphate (PI-4,5-P₂) into phosphatidylinositol-3,4,5-triphosphate (PIP₃), by the class I lipid kinase PI3K. This transformation into PIP3 is reversed by phosphatases such as the phosphatase and tensin homolog deleted on chromosome 10 (PTEN) and the SH2-domain containing inositol phosphatase (SHIP) (29). PIP3 drives the phosphorylation and activation of Akt (also called, PKB), via 3-phosphoinositol-dependent kinase 1 (PDK1) (30). Akt activity is also augmented by the mTORC2 complex, composed of mTOR and rictor (rapamycin-insensitive companion of mTOR) (29). Akt can in turn phosphorylate and inhibit the glycogen synthase kinase 3 (GSK3) leading to dephosphorylation and activation of eIF2B with translation initiation (31). In addition, Akt phosphorylates five sites leading to the inhibition of the GTPase activity of tuberous sclerosis complex 2 (TSC2), on the small GTPase Ras homolog enriched in brain (RHEB), which in its GTP form stimulates the kinase activity of mTORC1 (32, 33). Therefore, the activity of the mTORC1 complex, composed of mTOR, RHEB, the mTOR associated protein, LST8 (MLST8), and the regulatory-associated protein of mTOR (Raptor), is downregulated by unphosphorylated TSC2 that is derepressed by Akt kinase activity.

Downstream mTORC1, the TOS (target of rapamycin signaling)-containing 4E-BP and p70S6K are phosphorylated. As seen above, phosphorylated 4E-BP is inactive and allows eIF4E to bind eIF4G to initiate translation. In dividing cells, mTOR phosphorylates p70S6K at Thr389, which in turn can phosphorylate ribosomal S6 protein (RPS6), eIF4B and programmed cell death 4 (PDCD4). While the function of phosphorylated RPS6 remains largely unknown, eIF4B and PDCD4 are positive and negative regulators, respectively, of the RNA helicase, eIF4A (34). mTORC1 also downregulates the activity of the eEF2 Kinase, which then subsequently enhances the elongation step of translation by eEF2. The general protein translational capacity is also enhanced by mTOR via increased transcription (more mRNA), and stimulation of the translation of mRNAs containing a string of 5′–pyrimidines (5′TOP mRNA) (35). In addition to its activation by growth factors, mTOR also senses cellular nutrient, oxygen, and energy level (36). As its name implies, most of mTOR effects are neutralized by rapamycin. The FKBP12–Rapamycin complex quickly binds close to the kinase domain (37), leading to mTOR conformational changes, dissociation from Raptor (38, 39) and inhibition of some of mTORC1 functions (40). By binding newly produced mTOR, FKBP12–rapamycin complex also inhibits the assembly of mTORC2 (41). Rapamycin also inhibits the binding of phosphatidic acid (PA) to mTOR, reducing the stabilization of the mTORC1 and mTORC2 complexes (42). PA is synthetized during membrane phospholipid biogenesis (43), and its intracellular level modulates the amount of rapamycin required to inhibit mTOR (44). Interestingly, low doses of rapamycin inhibit mTOR-induced p70S6K phosphorylation while much higher doses are required to block mTOR-induced 4E-BP Thr37/Thr46 phosphorylation (45). As a result, other compounds that are stronger inhibitors of mTORC1 and C2 than rapamycin, such as PP242 and AZD8055, were developed.

Surprisingly, mTOR has not been studied in EOS, although its inhibitor, rapamycin has shown effects on EOS in vitro and in vivo. As shown in Table 2, resting human blood EOS express relatively little mTOR, but very high amount of FKBP12. FKBP12 is bound by both rapamycin and FK506 and is required for these drugs to exert their inhibitory effects in cells. Interestingly, nanomolar doses of FK506 strongly inhibit calcium ionophore-induced cytokine (GM-CSF) production in EOS, while micromolar doses of rapamycin does not (46). Due to the competition between rapamycin and FK506 on FKBP12, high amount of rapamycin antagonizes the FK506-mediated inhibition of cytokine production in EOS (46). However, rapamycin is more potent than FK506 in inhibiting IL-5-induced prolonged EOS survival (46). The divergence between FK506 and rapamycin has also been described in T lymphocyte and mast cells, where rapamycin modulates proliferation rather that gene expression (47, 48). Another study (49) confirmed that rapamycin reduces IL-5-induced pro-survival signaling in EOS but the effect was modest and required high doses of drug for at least 72 h. In the same study, rapamycin also partially inhibited IL-5-induced eosinophil cationic protein (ECP) release from EOS (49). In addition, mTOR has important functions during EOS differentiation as rapamycin inhibited mouse EOS differentiation downstream of IL-5 (50). This is in agreement with the dependence of T cell proliferation and differentiation on mTOR (51). Remarkably, rapamycin has no inhibitory effect on mouse EOS recruitment into the BALF after exposure to dust-mite allergen in chronic allergic models (52), suggesting that the role of mTOR signaling is confined to development and possibly survival but not cell migration.

In both human and mouse EOS, PI3K is required for a variety of functions. These include chemokine-induced EOS granule proteins release (53), platelet-activating factor (PAF)-induced chemotaxis but not LTC4 release (54). The PI3K/Akt pathway is also essential for IL-5-induced β2-integrin adhesion to bovine serum albumin (BSA) (55), and IL-5-induced guinea pig EOS mobilization from the bone marrow (56). In EOS, the PI3K/Akt pathway can be activated by fMLP or RANTES after priming with IL-5 or IL-3 (57). Prostaglandin E₂ (PGE₂) via EP4 induces PI3K/PDK1-dependent increase in Akt phosphorylation, which consequently inhibits eotaxin-induced EOS shape changes and chemotaxis (30). Therefore, the PI3K/PDK1/Akt pathway is important in EOS and regulates a variety of functions depending on its activator.

The MAPK (ERK and p38) signaling pathways are involved in most of cellular functions, including differentiation and proliferation. ERK1, ERK2, p38α, and p38β are coded by four different genes (MAPK3, MAPK1, MAPK14, and MAPK11). Following intracellular or extracellular activation, the MAP kinase kinase kinases (MEKK) are activated, leading to phosphorylation of MAP kinase kinases (MEK) and, finally, MAPK are phosphorylated (58). ERK1/2 alone possess more than 150 substrates involved in a large variety of cell functions, including transcription, cell death, autophagy metabolism, and translation (59). Among the kinases activated by ERK or p38 are kinases involved in protein translation, including p90S6K (RSK), the MAPK-interacting kinases (Mnk), and the MAPK-activated protein kinase 2 (MK2) (2). The latter has an important role in 3′UTR directed, mRNA binding protein-dependent translation. P90S6K are activated by ERK signaling that can then phosphorylate TSC2 at Ser1798, activating mTORC1 and protein synthesis (60, 61). Of note, ERK may also directly phosphorylate and inhibit TSC2, leading to increased mTORC1 activity (62). Like p70S6K, p90S6K also phosphorylates both eIF4B and eEF2 kinase, which enhances eIF4B/eIF3 interactions and eEF2 function and, consequently, protein initiation and elongation (12, 23). While Mnk2 activity is thought to be constitutive, Mnk1 phosphorylation and activation can be triggered downstream ERK and p38 leading to eIF4E phosphorylation at Ser209 (63). Although this phosphorylation inhibits eIF4E binding to the 5′-cap, it may also control the translation of specific mRNAs (63).

Mitogen-activated protein kinases have important roles in many critical events, including EOS survival, migration, adhesion, production of inflammatory mediators, and degranulation. In EOS, ERK and p38 are phosphorylated and active following stimulation with a variety of mediators, including the β-chain cytokines (IL-3, IL-5, and GM-CSF), chemokines, fMLP, the PAF, and matrix proteins (26, 28, 53, 57, 64–71). Table 3 shows the expression levels of ERK, and their downstream targets, RSK1–3 (p90S6K), all of which are phosphorylated at a detectable level in resting cells. However, EOS contain little Mnk1/2 (Table 3), suggesting that the MAPK activation likely does not regulate protein translation via eIF4E phosphorylation (Figure 2); and despite its phosphorylation, the low level of Mnk2 probably have little effect on eIF4E phosphorylation. Consistent with MAPK activation, upstream MEK and MEKK were also phosphorylated at multiple sites in circulating EOS (Table 3). These data suggest that such cells are not truly resting but have been partially activated or primed either in vivo or during isolation.

Semaphorin-7A mRNA level is relatively high in resting cells and changes only slightly in activated blood EOS. However, its translation remains almost undetectable despite GM-CSF activation (22). Surprisingly, despite similar mRNA levels, the translation rate for semaphorin-7A is more than 10-fold higher in IL-3- versus GM-CSF-activated EOS (22). Consistent with increased translation, semaphorin-7A mRNA was enriched in polyribosome fractions following IL-3 compared to GM-CSF (22). Of note, TOP mRNAs (EEF1A1 and PABP) were not enriched in the polyribosome fraction by IL-3, suggesting unique and highly selective signaling from IL-3 receptor to the translational machinery.

Freshly purified blood EOSs possess surface semaphorin-7A, which tends to decrease overtime during the first 20 h of cell culture (unpublished data). Activation with IL-5 or GM-CSF maintains or slightly increases surface semaphorin-7A over this same time span (27). On the other hand, over a broad range of doses, IL-3 significantly increased surface semaphorin-7A expression (27). Interestingly, IL-3-induced semaphorin-7A translation occurred more than 6 h after activation (unpublished data), suggesting that considerable signaling and possibly the translation of accessory proteins precedes semaphorin-7A translation initiation.

Along with RL13A (74), RPS6 is one of the rare ribosomal proteins that is phosphorylated following cellular stimulation in eukaryotic cells (75, 76). In stromal cells, RPS6 phosphorylation is directly controlled by the kinases p70S6K1 and p70S6K2, downstream of mTOR (77). In knock-in mice, genetically modified at RPS6 phospho-sites, aggregate protein synthesis was decreased in liver and embryonic fibroblasts (78). Other studies have suggested that RPS6 phosphorylation facilitated more efficient 40S ribosomal subunit assembly (79). This idea is supported by structural and biochemical data demonstrating that phosphorylated RPS6 is located at the interface between the small and the large ribosomal subunits near the tRNA-binding sites (80), and is enriched in polyribosomes (75). The correlation of RPS6 phosphorylation with cell division during mitogenic activation suggests that RPS6 participates in translation control in dividing cells (81). However, the role of phosphorylated RPS6 in non-dividing cells, such as EOS, remains unexplored.

We found that all β-chain cytokines strongly induced RPS6 phosphorylation at Ser235 and Ser236. However, while RPS6 phosphorylation persisted for only 1–4 h in EOS culture with IL-5 or GM-CSF, IL-3 induced continuous RPS6 phosphorylation for as long as IL-3 remained present in the culture medium (22). Of note, this unique feature of IL-3 to prolong RPS6 phosphorylation has also been observed in basophils (82, 83). Anti-IL-3 neutralization rapidly reversed RPS6 phosphorylation indicating that constant presence of IL-3 was required and that signaling was likely driven by a labile secondary messenger following IL-3 activation (22). Interestingly, the relatively rapid RPS6 dephosphorylation in GM-CSF-activated EOS was phosphatase 1 (PP1)-dependent, although total PP1 activity in cell lysates was the same in GM-CSF- and IL-3-activated EOS (22). This suggests that PP1 activity toward RPS6 may be negatively regulated only in IL-3-activated but not in GM-CSF-activated EOS. Of note, a 23 KDa ribosomal-associated inhibitor of PP1, termed ribosomal-associated inhibitor of phosphatase 1 (RIPP1) has been identified but remains incompletely described (84, 85).

As mentioned above, RPS6 can be phosphorylated downstream of the PI3K/Akt/mTOR/p70S6K pathway (77). However, in EOS neither rapamycin, PI3K nor p70S6K inhibitors prevented IL-3-induced RPS6 phosphorylation (22). On the contrary, p90S6K (RSK) inhibitors significantly reduced IL-3-induced, RPS6 phosphorylation on both Ser235 and Ser236 (22). GM-CSF activation of p90S6K peaked after 10 min, and p90S6K was already largely dephosphorylated by 1 h (22). Conversely, progressive phosphorylation of p90S6K occurred after IL-3, peaking, at 16–20 h and still detectable until IL-3 was removed or neutralized in the culture medium (22). P90S6K was the first RPS6-phosphorylating kinase described in Xenopus oocytes (86), but has since been implicated in cell proliferation and survival (87). P90S6K includes three isoforms (RSK1, 2, and 3), all with inducible phosphorylation-dependent activity and similar functions. P90S6K phosphorylation is downstream of ERK and phosphorylated p90S6K has been found associated with polyribosomes (88). Phosphorylation on Thr573 is sequentially followed by Thr359, Ser363, and finally Ser380. All four sites are strongly phosphorylated following IL-3-activated EOS (22). Ultimately 3′–phosphoinositol-dependent kinase-1 (PDK1) phosphorylates Ser221 leading to maximal p90S6K activation (89). In addition to RPS6, p90S6K also phosphorylates eIF4B and GSK3 (12, 90). Phosphorylated eIF4B interacts with eIF3A, enhancing translation initiation (91). P90S6K inactivates GSK3, which would in turn dephosphorylate and activate eIF2B, thus promoting eIF2 recycling and increasing translation initiation [(90); Figure 2]. While the dephosphorylation of eIF2B possibly occurs via changes in PP1 activity (90), differential activation of p90S6K by the different β-chain cytokines was not accompanied by changes in PP1 activity (22), suggesting that IL-3-induced and prolonged p90S6K activation does not affect translation via the GSK3/PP1/eIF2B pathway. As proposed above, the β-chain cytokines could differentially regulate a ribosomal specific PP1 regulatory protein (85).

Upstream, p90S6K phosphorylation is known to be regulated by the MAPK and particularly by ERK1/2 (92). Consistent with these data, a selective inhibitor of both MEK1 and MEK2 (U0126) added 3 h after IL-3, blocked the phosphorylation of p90S6K on Ser380 and RPS6 in EOS in culture. Another MAPK, p38, has also been implicated as a potential activator of p90S6K in dendritic cells (93). However, a p38 inhibitor (SB203580) had no effect on p90S6K phosphorylation in IL-3-activated EOS.

In addition to semaphorin-7A, we have more recently shown that the low-affinity IgG receptors, FCGR2B and FCGR2C (CD32B and CD32C) were upregulated at the translational level by IL-3, in a p90S6K-dependent manner (94). Therefore, we have so far identified two transcripts whose translation is exclusively enhanced by the prolonged effect of IL-3 through ERK/p90S6K signaling. MS proteomic analysis of EOS treated with IL-3 with and without ERK inhibitors will yield insight into the identity of other similarly regulated mRNA.

RNA-binding proteins (RBP) regulate all aspects of RNA metabolism, including biogenesis, cellular localization and transport, stability, and translation. With the emergence of high throughput screening and quantitative proteomics, several hundred (approximately 500) potential RBP have been identified (95). Given their obvious importance, enormous effort has been directed to expand our knowledge on how RNA-protein interactions determine RNA function and cell fate. It bears reiterating that mRNA is not a rod but a complex 3-dimensional shape. As such, RBP can interact with mRNA via structure, sequence or structure, and sequence elements. A simple example is 5′-cap binding protein eIF4E. A more complex example is PABP, which interacts with poly A tails, a combination of sequence (Poly A) and structure. The iron-response binding protein (IRE-BP) interacts with a sequence presented on a stem-loop and bulge (96). Alterations in the size of the loop, the distance between the loop and bulge or the loop sequence ablates binding. Given these levels of target specificity, some RBPs will no doubt be successfully targeted with therapeutics to treat human disease.

Once transcribed from genomic DNA, newly produced pre-mRNAs are immediately covered by a number of nuclear RBP to protect from degradation by nucleases, guide splicing and prepare for cytoplasmic transport. As mature mRNA are translocated, the inventory of bound proteins are often replaced with a new set of RBP that determine intracellular location, define degradation rates as well as translatability (see above) in the cytoplasm. In response to extrinsic and intrinsic stimuli, free and bound RBP are subject to post-translational modifications (PTM) (e.g., phosphorylation, ubiquitination, acetylation, and methylation) that may induce conformational changes and alter the association between RBP and target mRNA (97, 98). Depending on stimulus and cell type, the modified RBP may associate with or dissociate from mRNA, affecting the transcript stability as well as its translation, clearly affecting protein production. RBP bind to RNA via a variety of domains, including the so-called RNA-recognition motif (RRM), zinc finger motives, K-homology domains (KH), RGG boxes, and DEAD/DEAH boxes (97). Often more than one binding domain are present allowing simultaneous interactions with multiple mRNAs, multiple sites within one mRNA target or between specific mRNA sequences and organelles such as ribosomes or stress granules. RBP can also form higher order structures through protein–protein interactions either as homo- or heterodimers/trimers, etc. As a rule, RBP that interact with 5′ or 3′ ends of mRNA often regulate translation initiation (e.g., translation initiation factors and their partners; PABP) while those that bind to coding regions can affect translation, localization, or mRNA decay (e.g., IRE-BP). 3′ UTR RBP (e.g., AUF1, HuR, TTP, TIA-1, TIAR, FMRP, PTB, KSRP, hnRNPs, nucleolin, and CUGBP) are most often involved in mRNA localization and decay (99).

It is well known that many rapidly inducible mRNA coding for pro-inflammatory cytokines and oncoproteins are very short-lived. Inevitably, these mRNA contain cis-acting sequences into their 3′-UTR (100). The best-characterized instability determinant is composed of adenosine-uridine (AU)-rich element (ARE) repeats that are found in 3′-UTR of GM-CSF, IL-3, IL-5, IL-2, IFN-γ, and TNF-α and other cytokine mRNA. The life-span of ARE mRNA are regulated by a subset of binding proteins (AUBPs) that preferentially target the ARE and stabilize or further destabilize the transcripts. To date, approximately 20 AUBP have been identified. EOS express 7 AUBP (AUF1, hnRNP C, YB-1, nucleolin, TIA-1, HuR, and BRF1) (3) and their role in the regulation of mRNA stability has been demonstrated by many studies (101–105). In response to an exogenous pro-survival signal, Y-box binding protein 1 (YB1) and heterogeneous nuclear ribonucleoprotein C (hnRNP C) became associated with, while heterogeneous nuclear ribonucleoprotein D (hnRNP D or AUF1) dissociated from the ARE of GM-CSF mRNA (101, 106). These interactions were accompanied by the multiple phosphorylation of AUF1 (Ser83, Ser87, and Thr91) likely by ERK, CK1, GSK3β, or PKA (103, 107–109). Presumably, phosphorylation reduced the affinity of AUF1 for the ARE. AUF1 also undergoes post-transcriptional, alternative splicing events (110), yielding four AUF1 mRNAs and isoform variants (p37, p40, p42, and p45), all of which are detectable in human EOS (106). Thus, the regulatory control by AUF1 isoforms appears to be highly complex and includes their potential to form heterodimers (111) with a different affinity for ARE containing mRNAs (p37 > p42 > p45 > p40) (112, 113). While AUF1 has additional functions, its best-characterized function is to accelerate the decay of associated ARE-rich mRNAs. The p37 isoform has been shown to interact with the exosome in EOS (103) and exhibit the greatest destabilizing activity toward ARE-containing mRNAs compared to other isoforms (114).

mRNA turnover is often linked to translation (115) and the role of AUBP in RNA translation has been extensively studied in many systems. Similar mechanisms may occur in EOS although no direct evidence has yet been published. As mentioned above, PI3K/Akt/mTOR and MAPK cascades are the major signaling pathways that control global RNA translation upstream of the ribosomal machinery. These pathways have also been linked to AUBP-mediated mRNA decay in many cell types. EOS possess all translational machinery (Table 1) and can activate those kinase pathways when stimulated with various agonists (fMLP, RANTES, eotaxin, IL-5, IL-3, and PGE2) (30, 57). For example, ERK is activated by hyaluronic acid, IL-3 or IL-5, and likely drives AUF1 phosphorylation (102, 103), impacting the translation and decay of multiple ARE mRNA, including GM-CSF. These data suggest that ARE mRNA will be subject to translation control in EOS. This has been investigated by analysis of transfected mRNA, which has revealed striking differences in protein production despite similar cytosolic steady state levels of coding mRNA (116). Below, we discuss well-defined AUBP and their potential roles in target mRNA translation in EOS.

We have described above how RBP regulate mRNA stability and translatability, particularly of GM-CSF mRNA in EOS. GM-CSF plays a pivotal role in the modulation of EOS differentiation, function, and survival. The cytokine is upregulated in eosinophilic diseases and a major contributor to enhancing EOS survival in the lungs of patients during active asthma (158). Recombinant GM-CSF promotes EOS survival about five times as potently as equal concentrations of IL-5 (159). In asthmatics, GM-CSF is produced by a wide spectrum of cell types, including lung epithelial cells, lymphocytes, alveolar macrophages, EOS, endothelial cells, and fibroblasts. As EOS typically increase by 20-fold in the lung within a few days of an allergen challenge (160), autocrine GM-CSF is an important source in order to support survival. The level of endogenous GM-CSF in BAL fluid is low (161–163) compared to IL-5 in both mice and humans (164). However, intranasal delivery of Adeno-GM-CSF to the airways of OVA-sensitized mice resulted in sustained accumulation of various inflammatory cell types, most noticeably EOS, in the lung for more than 2 weeks post OVA aerosol challenge (165, 166). Conversely, neutralization of endogenous GM-CSF during aeroallergen exposure significantly inhibited eosinophilic inflammation and airway hyper-responsiveness. This suggests that small amount of endogenous GM-CSF can significantly contribute to the development and persistence of eosinophilic airway inflammation. In vitro, purified peripheral blood EOS synthesize small amounts (~1 pg/ml) of anti-apoptotic GM-CSF (167, 168), after stimulation with a variety of factors (fibronectin, hyaluronic acid, TNF-α, IL-3, IL-5, IL-15, integrins, IFN-γ, calcium ionophore, cross-linking of cell surface molecules) (67, 102, 167, 169–174). Activation-induced survival was blocked by the addition of neutralizing anti-GM-CSF even 2 days after the initiation of culture, indicating that the cells continuously release low levels of GM-CSF on which survival depends (103, 172–174). Similarly, the majority of BAL EOS obtained 48 h after segmental allergen challenge died in vitro at 6 days in the presence of neutralizing anti-GM-CSF. Both in situ hybridization (tissue EOS) (175) and qPCR (purified EOS) (67, 172, 173) analyses have demonstrated that increased GM-CSF mRNA was associated with GM-CSF protein secretion and prolonged EOS survival.

TPI ASM8 is a drug in development, targeting the common β-chain receptor for all IL-5, GM-CSF, and IL-3, in the form of RNA-targeted inhaled oligonucleotide antisense phosphorotioates (184). Although TPI ASM8 seems to be well tolerated and leads to some reduction of EOS and eosinophilic hematopoietic progenitor (CD34⁺IL5R⁺), other alternative therapeutic targets more specific to each of the 3 cytokine should be developed.

So far, we have identified semaphorin-7A and FCGR2B/C (CD32) as specific genes exclusively responding to IL-3 activation via prolonged ERK/p90S6K signaling. It is probable that additional genes are similarly regulated at a translation level by IL-3/ERK/p90S6K. Likely other candidate genes may share specific mRNA cis-elements whose identity may be inferred by homology searches among IL-3 upregulated mRNA. We started analyzing how semaphorin-7A or FCGR2 affect EOS function. We found that IL-3-activated EOS adhere to the only known semaphorin-7A ligand, plexin-C1 (27). Plexin-C1 is expressed by many cell types, including lymphocytes, monocytes, dendritic cells, and neutrophils (185), and has an important role in the migration of these inflammatory cells. Plexin-C1 is also expressed by stromal cells (186), which could facilitate migration or activate EOS in fibrotic tissue. Interestingly, IL-3-activated EOS migration on plexin-C1 was largely resistant to semaphorin-7A blockade while neutralizing anti-αMβ2 integrin were far more inhibitory (187). Migration in the absence of chemotaxis indicates that a haptotaxis process is operative for plexin-C1- or periostin-mediated migration (187, 188). Semphorin-7A signaling may also skew fibroblasts toward a pro-fibrotic, more mesenchymal phenotype (27, 189, 190), although we recently demonstrated anti-fibrotic functions for endogenous semaphorin-7A expressed by lung fibroblasts (191).

The upregulation of CD32 by IL-3 on EOS has a profound impact on EOS function. EOS-driven pathology in tissue requires both EOS migration from circulating blood to the site of inflammation and the release (degranulation) of preformed toxic proteins and mediators of the inflammation. IL-3-activated EOS strongly degranulate on heat-aggregated (HA)-IgG, with extrusion of ~25% of their total cellular eosinophil-derived neurotoxin (EDN) in 6 h (94) compared to less than 10% after IL-5 (94). Degranulation on HA-IgG was CD32-dependent (94). Thus, IL-3 and its downstream intracellular effectors may be potential therapeutic targets to limit EOS degranulation and EOS-driven pathologies. The use of anti-IL-5 therapies on patients with severe eosinophilic asthma has reduced asthma exacerbations and blood eosinophilia (192, 193), see other article by Nair in this issue. However, airway EOS are still present and active despite treatment (194, 195). This partially reflects loss of the surface IL-5 receptor expression by airway EOS (22, 196, 197). Conversely, IL-3 and the surface IL-3 receptor are upregulated and highly expressed on airway EOS (27, 198). Therefore, combined targeting of the IL-3 and IL-5 pathways may provide additive or synergistic benefits.