Human Interleukin-1 Receptor-associated Kinase-2 Is Essential for Toll-like Receptor-mediated Transcriptional and Post-transcriptional Regulation of Tumor Necrosis Factor α*

Toll-like receptors (TLRs) are pattern-recognition receptors that recognize microbial ligands and subsequently trigger intracellular signaling pathways involving transcription factors such as NFκB and MAPKs such as p38. TLR signaling can regulate both transcriptional and post-transcriptional events leading to altered gene expression and thus appropriate immune responses. The interleukin-1 receptor-associated kinase (IRAK) family comprises four kinases that regulate TLR signaling. However, the role of IRAK-2 has remained unclear, especially in human cells. Recent studies using cells from in-bred Irak2−/− mice showed that murine IRAK-2 was not required for early TLR signaling events but had a role in delayed NFκB activation and in cytokine production. IRAK-2 in mice has four splice variants, two of which are inhibitory, whereas human IRAK-2 has no splice variants. Thus IRAK-2 in mice and humans may function differently, and therefore we analyzed the role of IRAK-2 in TLR responses in primary human cells. siRNA knockdown of IRAK-2 expression in human peripheral blood mononuclear cells showed a role for human IRAK-2 in both TLR4- and TLR8-mediated early NFκB and p38 MAPK activation and in induction of TNF mRNA. These data conflict with findings from the in-bred Irak2−/− mice but concur with what has been seen in wild-derived mice for TLR2. Moreover, human IRAK-2 was required for regulating MyD88-dependent TNFα mRNA stability via the TNF 3′UTR. Collectively, these data demonstrate for the first time an essential role for IRAK-2 in primary human cells for both transcriptional and post-transcriptional TLR responses.

The innate immune system detects the presence of pathogens through a variety of pattern-recognition receptors, which recognize conserved pathogen-associated molecular patterns on invading pathogens (1). One important family of such receptors is the Toll-like receptors (TLRs), 2 of which there are 10 members in the human and 13 members in the mouse (2). For example, the pathogen-associated molecular pattern for TLR4 is LPS, TLR9 responds to CpG DNA, whereas TLR7 and TLR8 both recognize viral single strand RNA and synthetic imidazoquinoline-like molecules (3). TLRs are transmembrane proteins, which contain a cytoplasmic TIR (Toll/IL-1 receptor) domain and leucine-rich repeats located extracellularly (2). Detection of a pathogen-associated molecular pattern by a TLR results in receptor dimerization, which allows recruitment of the TIR-domain-containing adaptor proteins: MyD88 is required for all TLR signaling pathways except for TLR3 and MyD88-independent TLR4 pathway (4): MyD88-adaptor-like (Mal) is required for TLR2 and TLR4 signaling to recruit MyD88 to the receptor complex (5); TIR domain-containing adaptor inducing IFN-␤ (TRIF) is required for TLR3 signaling and for a MyD88-independent TLR4 pathway (6), while TRIFrelated adaptor molecule is also required for the MyD88-independent TLR4 pathway (7). The formation of these receptoradaptor complexes results in the activation of various signaling pathways of the innate axis of the immune response. These signaling pathways can lead to the activation of MAPKs, such as p38 and JNK, and of transcription factors such as NFB and IFN regulatory factors (2). NFB is required for the transcription of various pro-inflammatory cytokines such as IL-1 and TNF and of chemokines such as IL-8. Similarly, p38 activation results in increased expression of cytokines and chemokines due to direct phosphorylation of transcription factors and through increased stability and enhanced translation of mRNAs (8).
The serine/threonine interleukin-1 receptor-associated kinases (IRAKs) are recruited to the TLR-adaptor complexes and control downstream signaling to MAPKs and transcription factors (9), for example through the activation of the E3 ligase activity of TNF-receptor-associated factor 6 (TRAF6), which is essential for some signaling events (10). Structurally, IRAK family members share similar domains. They contain a N-terminal death domain (DD), a ProST domain, a central conserved kinase domain, and, apart from IRAK-4, a C-terminal domain (9). IRAK-1 and IRAK-4 have been characteristically described as active kinase family members (11,12), whereas IRAK-M, a negative regulator, is thought to have no kinase activity (13). IRAK-2, which was always assumed to be a pseudokinase due to the absence of an aspartate residue in the kinase domain, has recently been described as having kinase activity (14). Upon TLR stimulation, IRAK-4 interacts with both MyD88 and IRAK-2 via homotypic DD interactions, to form a complex proximal to the receptor termed the Myddosome (15). IRAK-4 is widely regarded as the most critical IRAK in MyD88-dependent TLR pathways, because IRAK-4 knock-out mice show impaired signaling to MAPKs and transcription factors for all TLR signaling pathways except TLR3 (16). In addition, patients with an inherited IRAK-4 deficiency have been shown to be susceptible to a range of bacterial infections (17,18). In contrast to IRAK-4, deletion of IRAK-1 attenuates, but does not eliminate, TLR-induced NFB, MAPK activation, and gene induction (19 -21). In addition, IRAK-1 was shown to be required for IFN regulatory factor activation (22,23).
The role of IRAK-2 has been the most enigmatic of the family. Evidence of a role for IRAK-2 in TLR signaling was revealed though studies of the vaccinia virus protein A52 (24). A52, shown to be important for virus virulence (24), inhibited all TLR pathways to NFB through interacting with IRAK-2 (25). This viral targeting of IRAK-2 and not IRAK-1 by A52 to disrupt TLR-induced NFB activation (26) suggested a predominant role for IRAK-2 in NFB activation. Consistent with this, knockdown of IRAK-2 expression by siRNA in a human cell line inhibited TLR3-, TLR4-, and TLR8-induced NFB-dependent reporter gene activity (26). The recent generation of Irak2 knock-out mice showed that Irak2 Ϫ/Ϫ mice were highly resistant to LPS-and CpG-induced septic shock (14), whereas a previous study showed that the difference in mortality between wild-type and Irak1 Ϫ/Ϫ mice was only subtle (20). Interestingly, even though the Irak2 Ϫ/Ϫ mice were resistant to TLR-induced septic shock, in cells derived from these mice early TLR2 signaling events were intact, with IRAK-2 only being required for late NFB activation and not MAPK activation (14). Using independently generated Irak2 Ϫ/Ϫ mice, another group also showed for TLR7 that only late NFB activation was impaired, whereas both early and late NFB activation were normal for TLR4 stimulation (27). Thus currently there are contradictory findings on the role of IRAK-2 in NFB and p38 MAPK in a human cell line compared with murine cells from in-bred knock-out mice (28). Furthermore, nothing is known about the role of human IRAK-2 in post-transcriptional regulation. Of note, IRAK-2 in mice has four splice variants, two of which are inhibitory, whereas human IRAK-2 has no splice variants (29). Thus IRAK-2 in mice and humans may function differently.
Therefore to further elucidate and clarify the role of IRAK-2 in TLR signaling in the human system, we investigated the function of IRAK-2 in both transcriptional and post-transcriptional TLR responses in primary human cells. The results show that IRAK-2 is essential for TLR4-and TLR8-induced cytokine and chemokine mRNA and protein production in human peripheral blood mononuclear cells (PBMCs). In contrast to what was seen in murine cells (14,27), in human cells IRAK-2 was required for both NFB and p38 MAPK activation in response to TLR4 or TLR8 stimulation. Furthermore, human IRAK-2 controlled a MyD88-dependent TLR pathway to TNF␣ mRNA stability via the TNF 3ЈUTR. Thus we have revealed an essential role for IRAK-2 for both transcriptional and post-transcriptional TLR responses in primary human cells, while the particular role of IRAK-2 in regulating gene induction of TNF␣ suggests that IRAK-2 is a critical pro-inflammatory kinase in humans.

EXPERIMENTAL PROCEDURES
Cell Culture-Human PBMCs from healthy volunteers were provided by the Irish Blood Transfusion Service. PBMCs were purified from the buffy coat of heparinized whole blood preparations from healthy volunteers by density centrifugation on low-endotoxin Ficoll-Hypaque. Isolated PBMCs were washed three times in sterile PBS (137 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4 , 2 mM NaH 2 PO 4 ), counted, and seeded at a density of 1 ϫ 10 6 cells/ml in complete RPMI 1640 medium. Human embryonic kidney (HEK) 293 cells stably transfected with the IL-1R (HEK293_R1) were provided by Tularik Inc. (San Francisco, CA).
Plasmids-WT-IRAK-2-Myc and MyD88 expression plasmids were provided by M. Muzio (Mario Negri Institute, Milan, Italy). The TK-Renilla construct was purchased from Promega (Madison, WI). The empty vector control pcDNA 3.1 was purchased from Invitrogen. The TNF 3ЈUTR-luciferase reporter construct and the ␤-actin 3ЈUTR-luciferase reporter construct were kindly provided by Aihao Ding (Cornell University, New York, NY). The IL-8 3ЈUTR-luciferase reporter construct was a gift from Peter King (Birmingham Veterans Affairs Medical Centre, Birmingham, AL). Plasmids expressing K237A IRAK-2 and E528A IRAK-2 were generated in-house by site-directed mutagenesis. The TRAF6 expression plasmids FLAG-TRAF6 and FLAG-TRAF domain (dominant-negative TRAF6 amino acids 289 -522) were provided by Tularik Inc., and C70A TRAF6 and K124R TRAF6 were provided by Justin McCarthy (University College Cork, Ireland).
Reporter Gene Assays-HEK293_R1 cells were seeded at 1 ϫ 10 5 cells/ml and 24 h later were transfected with 60 ng of either TNF, IL-8, or ␤-actin 3ЈUTR reporter constructs. The TK-Renilla construct (20 ng) was included to normalize the data for transfection efficiency. Where indicated, 5-25 ng of expression plasmids for MyD88, IRAK-2, or TRAF6 expression plasmids were co-transfected with reporters. The total amount of DNA transfected was kept constant at 230 ng by the addition of pcDNA3.1. Luciferase activity was measured 24 h following transfection. All transfections were done in triplicate, and data are expressed as fold induction (mean Ϯ S.D.) relative to control levels for a representative experiment of a minimum of three separate experiments.
siRNA Gene Silencing-siRNA duplexes targeting the IRAK2 gene targeted the sequence 5Ј-CCAGATCATCCTGAACTG-GAA-3Ј, as previously described (26). Non-silencing siRNA was used as a control (Qiagen). PBMCs were seeded at 1 ϫ 10 6 cells/ml (200 l per well) for cytokine analysis in 96-well plates and transfected with siRNA 1 h later. 5 pmol of siRNA diluted in serum-free RPMI was used per transfection. Each well received a solution containing 0.2 l of Lipofectamine 2000 (Invitrogen), 0.5 l siRNA, and 49.3 l of serum-free medium per transfection. An identical second transfection of siRNA with Lipo-fectamine 2000 was carried out 24 h later. After a further 24 h, cells were stimulated with either 100 ng/ml LPS or 2.5 g/ml CL075 and finally harvested 24 h later. For detection of IB␣ or p-p38, PBMCs were seeded at 2 ϫ 10 6 cells/ml in 6-well plates (2 ml per well). 50 pmol of siRNA diluted in serum-free RPMI and 2 l of Lipofectamine 2000 were used per transfection, using a protocol identical to that used for the 96-well plates. To examine the effect of IRAK-2 siRNA on MyD88-induced TNF␣ 3ЈUTR or the IL-8 3ЈUTR, HEK_293 R1 cells were seeded at 1 ϫ 10 5 cells/ml (200 l per well). Cells were transfected with 5 pmol of siRNA using 0.5 l of Lipofectamine 2000 24 h later, followed by transfection with MyD88 and either TNF␣ 3ЈUTR, IL-8 3ЈUTR, or ␤-actin 3ЈUTR plasmids and a second siRNA transfection 24 h later. After a further 24 h, cells were harvested and lysed, and the relative luciferase activity was measured.
Cytokine Analysis by ELISA-PBMCs were seeded at 1 ϫ 10 6 cells/ml in a 96-well plate (200 l/well) 1 h prior to siRNA treatment. The supernatants were collected and assayed for IL-8 or TNF␣ by ELISA (R&D Systems, Minneapolis, MN).
Measurement of mRNA Induction and Stability-PBMCs were seeded at 1 ϫ 10 6 cells/ml in 12-well plates (1 ml per well). Cells were transfected 1 h later with 25 pmol of siRNA and Lipofectamine 2000. Each well received a solution containing 1 l of Lipofectamine 2000, 2.5 l of siRNA, and 246.5 l of serum-free medium per transfection Cells were transfected with an identical siRNA treatment 24 h later. After another 24 h, cells were stimulated with either 100 ng/ml LPS or 2.5 g/ml CL075 for the indicated times. Cells were harvested, and RNA was isolated using High Pure Isolation kits from Roche Applied Science (Burgess Hill, UK). Induction of mRNA in PBMCs was assayed by quantitative real-time PCR using GoTaq qPCR Master Mix (Promega) and normalized to ␤-actin mRNA according to the manufacturer's instructions. To examine the stability of mRNA, PBMCs were seeded at 1 ϫ 10 6 cells/ ml. Cells were transfected with siRNA using 0.2 l of Lipofectamine 2000 per transfection and subsequently transfected with siRNA again 24 h later. After another 24 h, cells were stimulated with either 100 ng/ml LPS or 2.5 g/ml CL075 for 2 h followed by treatment of 5 g/ml Actinomycin D (Act D, Sigma-Aldrich) and either LPS or CL075. Cells were harvested, and RNA analysis was carried out as described above. The primers used were as follows: TNF␣, 5Ј-GAA CCC CGA GTG ACA AGC CTG-3Ј and 5Ј-TCA GCT CCA CGC CAT TTG  CCA-3Ј; IL8, 5Ј-CTC TGT GTG AAG GTG CAG TTT TG-3Ј  and 5Ј-AAG CTT TAC AAT AAT TTC TGT GGT-3Ј; and ␤-actin, 5Ј-CGC GAG AAG ATG ACC CAG ATC-3Ј and 5Ј-GCC AGA GGC GT CAG GGA TA-3Ј.
Statistical Analysis-Statistical analysis was carried out using a paired Student's t test.

IRAK-2 Is Required for TLR4-and TLR8-mediated Cytokine
Induction in Primary Human Cells-We have previously showed that knockdown of IRAK-2 expression by siRNA inhibited human TLR4-and TLR8-dependent NFB reporter gene activity in HEK293 cells (26). In contrast to this, reported data from murine Irak2 Ϫ/Ϫ cells has shown that IRAK-2 is dispensable for early NFB and p38 MAPK activation (14,27). Thus, here we wanted to assess the requirement for human IRAK-2 in early TLR signaling in the more physiological setting of primary human PBMCs. Firstly the efficacy of siRNA oligonucleotides that target IRAK2 in PBMCs was confirmed by examining their effect on endogenous IRAK-2 protein levels. Fig. 1A shows that endogenous IRAK-2 protein levels were almost undetectable in PBMCs treated with IRAK-2 siRNA, relative to the control siRNA-treated samples, either in the presence or absence of stimulation with CL075 (3M002). CL075 is a thiazoloquinolone derivative agonist of TLR7 and TLR8, which is known to preferentially activate TLR8 in PBMCs (3). PBMCs are known to respond to both LPS and CL075, via TLR4 and TLR8, respectively, leading to the production of inflammatory cytokines (30). When PBMCs were treated with CL075, the production of IL-8 in response to CL075 was reduced to basal levels in the cells treated with IRAK-2 siRNA relative to those treated with the control siRNA (Fig. 1B). CL075-induced TNF␣ production was also significantly inhibited in PBMCs treated with IRAK-2 siRNA compared with control siRNA-treated samples (Fig.  1C), thus demonstrating for the first time to our knowledge a role for IRAK-2 in TLR8 responses in primary human cells. IRAK-2 was also required for LPS/TLR4-mediated IL-8 production in PBMCs (Fig. 1D). Furthermore TLR4-mediated TNF␣ release from PBMCs was also significantly inhibited by treatment of cells with IRAK-2 siRNA (Fig. 1E). In addition, both CL075-and LPS-induced IL-6 production was significantly reduced in PBMCs (data not shown). Thus IRAK-2 has a critical role in cytokine production in primary human cells, for both endosomal (TLR8) and non-endosomal (TLR4) TLR pathways.
IRAK-2 Is Essential for TLR-mediated NFB and p38 MAPK Activation in Primary Human Cells-Our data from primary human cells showed a requirement for IRAK-2 in TLR-induced cytokine induction. Previous work showed that TLR-dependent cytokine induction from macrophages from in-bred Irak2 Ϫ/Ϫ mice was also impaired, although NFB and MAPK activation were largely unaffected (14,27,28). In contrast, in macrophages from a wild-derived mouse strain, RNA interference targeting Irak2 did inhibit TLR-induced NFB and p38 MAPK activation (31). Thus the mechanistic basis for the requirement for IRAK-2 for cytokine induction in the more physiological setting of primary human cells was not obvious, and therefore we next explored the role of IRAK-2 in TLRinduced signaling pathways in PBMCs.
Treatment of PBMCs with CL075 led to NFB activation within 5 min of stimulation, as measured by degradation of the inhibitory NFB subunit, IB␣ ( Fig. 2A). When cells were pretreated with IRAK-2 siRNA, TLR8-dependent degradation of IB␣ was prevented at every time point tested (5-30 min) compared with cells treated with control siRNA (Fig. 2A). Thus this demonstrated for the first time a role for IRAK-2 in early TLR8induced NFB activation in primary human cells, which is not the case for TLR7 in murine cells (27). A role for IRAK-2 in TLR4-mediated NFB activation was also observed in PBMCs, because IRAK-2 siRNA caused reduced IB␣ degradation after 20 or 30 min LPS stimulation, compared with cells treated with control siRNA (Fig 2B). Again, this contrasts with what was observed in macrophages from mice lacking IRAK-2, where IRAK-2 had no role in either early or late TLR4-induced NFB activation (28).
We next determined whether IRAK-2 was involved in p38 MAPK activation in TLR-stimulated primary human cells. Upon stimulation of PBMCs with LPS or CL075 we detected phosphorylation of p38 after 5 min, sustained until at least 20 min (Fig. 2, C and D). Crucially, when PBMCs were treated with siRNA targeting IRAK-2 prior to stimulation a dramatic reduction of p38 phosphorylation was observed relative to control siRNA-treated cells (Fig. 2, C and D). Together these results reveal a critical and novel role for IRAK-2 in early signaling events for TLR pathways in primary human cells.

IRAK-2 Is Required for TLR-mediated TNF␣ and IL-8 mRNA
Induction in PBMCs-The role of IRAK-2 in TLR4-and TLR8dependent mRNA induction was next examined. Because both NFB and p38 are involved in the regulation of induction of IL-8 and TNF␣ mRNA, it was hypothesized that IRAK-2 would have a role in regulating IL-8 and TNF␣ mRNA levels. A time course of LPS-induced TNF␣ transcript in PBMCs indicated that TNF␣ mRNA peaked at 2 h after stimulation (Fig. 3A). When cells were treated with IRAK-2 siRNA, there was a significant reduction in the levels of LPS-induced TNF␣ mRNA at 2 h compared with cells treated with control siRNA (Fig. 3A). This was again in contrast to previous work in murine Irak2 Ϫ/Ϫ macrophages where LPS-induced TNF␣ mRNA was not impaired at any time point examined (27,28). Furthermore, in PBMCs, not only was LPS-induced TNF␣ mRNA induction IRAK-2-dependent, but a 70% reduction of LPS-induced IL-8 mRNA was observed in the presence of IRAK-2 siRNA (Fig.  3B). Similarly, CL075-dependent IL-8 and TNF␣ mRNA were also significantly IRAK-2-dependent (Fig. 3B).
IRAK-2 Is Required for the Stabilization of TNF␣ but Not IL-8 mRNA in Primary Human Cells-IRAK-4 and IRAK-1 have been shown to regulate stability of some transcripts (27,32,33), while the role of human IRAK-2 in regulating mRNA stability is untested. Of note, mechanisms are known to exist for TLRs to increase the stability of both IL-8 and TNF␣ mRNA, although the signaling molecules involved have not been fully defined (34 -36). Given that reduction of IRAK-2 expression by siRNA reduced the total amount of IL-8 and TNF␣ mRNA detected after either TLR4 or TLR8 stimulation, the potential role of IRAK-2 in the stabilization of these mRNA transcripts was investigated. For this, PBMCs were initially transfected with either control or IRAK-2 siRNA, and then stimulated for 2 h with LPS or CL075 followed by treatment with Act D. The effect of reduced IRAK-2 expression on the rate of mRNA decay was then determined. This showed that for both LPS-and CL075induced IL-8 mRNA, there was no difference between the rate of decay of IL-8 mRNA in the control siRNA-treated cells compared with those treated with IRAK-2 siRNA (Fig. 4, A and B), suggesting that IRAK-2 is not required for the TLR pathway that induces stability of IL-8 mRNA. In contrast to the lack of effect on IL-8 mRNA decay, in the presence of IRAK-2 siRNA, the decay of TNF␣ mRNA was accelerated compared with cells treated with the control siRNA. This was observed for both TLR8-induced TNF␣ mRNA (Fig. 4C) and TLR4-induced TNF␣ mRNA (Fig. 4D), but was more marked, and statistically significant for TLR4 (Fig. 4D). It is known that p38 is one of the post-transcriptional regulators of TNF␣ production, and one of the mechanisms whereby p38 regulates LPS-induced TNF␣ is by stabilizing the mRNA (34,35). Therefore we determined the effect of the p38 inhibitor, SB202190, on LPS-induced TNF␣ mRNA stability, in comparison to the effect observed with IRAK-2 siRNA. As expected, SB202190 caused accelerated  decay of LPS-induced TNF␣ mRNA (Fig. 4E) but actually to a lesser extent than IRAK-2 siRNA (Fig. 4D).
Thus IRAK-2 controls a TLR-dependent pathway to TNF␣ mRNA stability in primary human cells. This can be at least partially explained by the requirement for IRAK-2 for TLRinduced p38 activation. The role of human IRAK-2 in mediating TNF␣ mRNA stability in response to LPS is in contrast to the lack of requirement for murine IRAK-2 in this process, as demonstrated in Irak2 Ϫ/Ϫ macrophages (27).
IRAK-2 Controls a MyD88-dependent Pathway to mRNA Stability via the TNF␣ 3ЈUTR-To further explore mechanistically how IRAK-2 regulates the stability of TNF␣ mRNA and not IL-8 mRNA, we examined the potential role of the 3ЈUTR of both mRNA transcripts. Some cytokine mRNAs, including TNF␣ and IL-8, are rendered unstable due to the presence of adenosine-and uridine-rich elements (AREs) in the 3ЈUTR (37). For TLR signaling, a MyD88-dependent pathway, which would operate for both TLR4 and TLR8, has been shown to mediate an mRNA-stabilizing signal that converges on some cytokine 3ЈUTRs (38). To assay the role of MyD88 and IRAK-2 in mRNA stability pathways converging on cytokine 3ЈUTRs, luciferase reporters under the control of the TNF␣ 3ЈUTR, the IL-8 3ЈUTR, or as a control the ␤-actin 3ЈUTR were transfected into HEK293 cells.
As expected, ectopic expression of MyD88 activated mRNA-stabilizing pathways leading to increased production of luciferase protein expressed from the reporters under the control of the TNF␣ and IL-8, but not the ␤-actin 3ЈUTR (Fig. 5A). Using IRAK-2 siRNA, we examined whether there was a role for IRAK-2 on this TLR-MyD88 -3ЈUTR pathway. In cells that were treated with control siRNA, MyD88-induced 3ЈUTR reporter activity was barely affected (Fig. 5, B-D). Similarly, IRAK-2 siRNA did not affect the ␤-actin 3ЈUTR (Fig. 5D). However, in cells treated with IRAK-2 siRNA, there was a very sig-

Role of IRAK-2 in TLR Signaling in Primary Human Cells
nificant decrease in the MyD88-dependent activation of TNF␣ 3ЈUTR compared with cells treated with control siRNA (Fig.  5B). This result demonstrates that IRAK-2 participates in the MyD88-dependent pathway that regulates the TNF␣ 3ЈUTR and further confirmed the role of IRAK-2 in TNF␣ mRNA stability. Significantly, and in contrast to this, MyD88-dependent induction of the IL-8 3ЈUTR was not reduced in the presence of IRAK-2 siRNA (Fig. 5C), confirming the lack of a role for IRAK-2 in IL-8 mRNA stabilization (Fig. 4).
Consistent with the requirement for IRAK-2 in the ability of MyD88 to stimulate the TNF␣ 3ЈUTR, expression of WT IRAK-2 dose dependently induced the TNF␣ (but not the ␤-actin) 3ЈUTR reporter (Fig. 5, E and F).
Key Residues of IRAK-2 and TRAF6 Involved in 3ЈUTR-mediated TNF␣ mRNA Stabilization-Having established that IRAK-2 is required for TNF␣ mRNA stabilization, and to begin to ascertain which activities of IRAK-2 are required for this function, we examined the role of key IRAK-2 residues in the regulation of the TNF␣ 3ЈUTR using the reporter system. Although long assumed to be a pseudokinase, recently it has been proposed that an invariant lysine residue located in the ATP-binding pocket of IRAK-2 (at Lys-237) is necessary for IRAK-2 to act as an active kinase (14,39). Here we found that a K237A mutant of human IRAK-2 no longer activated NFB or p38, nor induced cytokine responses (data not shown). Furthermore, this residue was also required for induction of the TNF␣ 3ЈUTR (Fig. 5G).
We have previously described another IRAK-2 residue, Glu-528, as critical for the ability of IRAK-2 to stimulate TRAF6-dependent ubiquitination activity, and an E528A IRAK-2 mutant was unable to activate NFB or p38, or to induce cytokine production (26). Here, we also found that this residue was essential for induction of the TNF␣ 3ЈUTR by IRAK-2 (Fig. 5G). This suggested that TRAF6 may also be involved in the MyD88-IRAK2-TNF 3ЈUTR pathway. Consistent with this notion, WT TRAF6 could strongly induce the TNF␣ 3ЈUTR reporter, in comparison to a dominant-negative TRAF6 (with residues of 1-289 deleted, Fig. 5H). Previously it has been shown that a key cysteine residue (Cys-70) in the RING-domain of TRAF6 is critical for its E3 ligase activity (10). When this cysteine is mutated to an alanine (C70A), the ability of TRAF6 to activate NFB is impaired (10). Here, Cys-70 was also required for induction of the TNF␣ 3ЈUTR pathway, because the TRAF6 C70A mutant no longer induced the TNF␣ 3ЈUTR reporter (Fig. 5H). Finally we sought to identify whether TRAF6 autoubiquitination is required for its function in this pathway. A lysine at residue 124 in TRAF6 has been shown to be the predominant site for TRAF6 autoubiquitination, and there have been conflicting reports as to whether TRAF6 autoubiquitination at this site is required for TRAF6 functions (10,40). Here, a K124R mutant of TRAF6 activated the TNF␣ 3ЈUTR reporter to a level similar to WT TRAF6 (Fig. 5H). These data suggest that the ability of IRAK-2 to stimulate TRAF6 E3 ligase activity, but not TRAF6 autoubiquitination, is required for the MyD88-IRAK2-TRAF6-TNF␣ 3ЈUTR pathway.
Collectively, these data demonstrate for the first time a critical role in primary human cells for IRAK-2 in multiple TLR responses, namely NFB activation, p38 activation, early cytokine mRNA induction, and stabilization of TNF␣ mRNA via the 3ЈUTR.

DISCUSSION
Elucidating the roles of IRAK family members in IL-1 and TLR signaling has been an intense subject of research in the past decade, however IRAK-2 has remained the most enigmatic family member. Prior to this study, we demonstrated that IRAK-2 is required for TLR3-, TLR4-, and TLR8-mediated NFB-dependent reporter gene activation in a human cell line (26). Here for the first time we provide a comprehensive analysis of the role of human IRAK-2 in primary cells and demonstrate a critical requirement for it in both early signaling events and in post-transcriptional regulation for both a membranebound TLR (TLR4) and an endosomal TLR (TLR8).
Recent studies using cells from Irak2 Ϫ/Ϫ mice showed that early NFB activation is normal for TLR2, TLR7, and TLR9, although delayed activation is impaired, whereas for TLR4, there is no defect in NFB (14,27,28). However, in the primary human cells studied here, reduction of IRAK-2 expression by siRNA led to impaired NFB activation for both TLR4 and TLR8. Human IRAK-2 was also shown to be more important than the murine form for p38 MAPK activation, because, in contrast to the lack of a role for IRAK-2 reported in the Irak2 Ϫ/Ϫ studies for TLR2, TLR4, and TLR7 (14,27), in PBMCs IRAK-2 was required for p38 MAPK activation in both the TLR4 and TLR8 pathways. Further differences in the murine and human system emerged upon examining cytokine induction and post-transcriptional regulation: we showed here that human IRAK-2 regulated TNF␣ mRNA induction and stability in response to LPS (and to a TLR8 ligand), whereas lack of murine IRAK-2 did not affect TNF␣ mRNA induction or stability after LPS treatment but did impair translation of the mRNA (27). Overall then, although IRAK-2 has now been shown to be required for TLR-induced cytokine production in both human and murine cells, there seems to be a more subtle requirement for IRAK-2 in TLR signaling in the murine system.
Murine and human IRAK-2 show 67% sequence identity (41). They also share the same domain structure and are highly conserved in their DD and kinase domains (41) and are there- fore unlikely to be significantly different structurally (41). However, differences between murine and human IRAK-2 are apparent at the level of splicing. There are four splice variants of murine IRAK-2 (IRAK-2a, -2b, -2c, and -2d), and different functions have been attributed to the different isoforms, because IRAK-2c and IRAK-2d have been proposed as having a negative role in TLR signaling, whereas IRAK-2a is a positive regulator of signaling (29,31). In contrast, there is no evidence of splice variants for human IRAK-2, so that the human protein is most like murine IRAK-2a, the positive regulator. Intriguingly, a recent study examining the innate immune response of wild-derived mice versus classic in-bred strains revealed a more important role for IRAK-2 in the wild-derived mice in TLR signaling events, compared with the in-bred mice, which was attributed to differential expression of IRAK-2 splice variants between the two strains (31). The classic in-bred strain, C57BL/ 6J, was shown to express high levels of the inhibitory isoform, IRAK-2c, which inhibits the pro-inflammatory isoform IRAK-2a (29,31). This may explain the more dominant role of IRAK-1, rather than IRAK-2, in early signaling events in inbred mice (31). However, as time progresses, expression of the inhibitory isoform (IRAK-2c) decreased (31), which likely allows IRAK-2a, which is no longer being inhibited, to function in a pro-inflammatory manner. Thus IRAK-2a in in-bred mice is responsible for late and sustained NFB activation. Wildderived mice differ significantly from experimental models as the genetic diversity of wild-derived mice has arisen in an evolutionary context, and these mice display a higher degree of polymorphisms. In the wild-derived strain MOLF/Ei, a natural mutation occurs in the promoter of IRAK-2c leading to significantly less IRAK-2c being expressed, and thus the positive function of IRAK-2a predominates. Because humans only express the IRAK-2a isoform, the wild-derived mice are a better model for human IRAK-2 function than in-bred mice. Indeed, siRNA targeting IRAK-2a in macrophages from wild-derived mice showed that IRAK-2a is required for early activation of NFB and p38 MAPK by TLR2 (31). Thus TLR-induced p38 and NFB responses are similarly impaired by IRAK-2 siRNA in MOLF/Ei macrophages and human PBMCs.
Previous studies have revealed a requirement for the IRAK family in post-transcriptional regulation of cytokine induction. Here we show for the first time a critical role for human IRAK-2 in post-transcriptional regulation, in that IRAK-2 was required for TLR-mediated stabilization of TNF␣ mRNA in primary human cells. TNF␣, a key pro-inflammatory cytokine and mediator of inflammatory and autoimmune disease, is subject to tightly controlled regulation both at the transcriptional and translational level, and p38 MAPK has a key role in this (36,37). Here, depletion of IRAK-2 by siRNA caused more rapid degradation of TNF␣ mRNA than that observed for inhibition of p38, thus highlighting the importance of IRAK-2 in the TNF␣ mRNA stability pathway. Therefore, although p38 activation was shown to be IRAK-2-dependent, IRAK-2 may also have further p38-independent mechanisms of stabilizing TNF␣ mRNA.
An MyD88-dependent pathway, which would operate for both TLR4 and TLR8 in human cells, has previously been shown to mediate both IL-8 and TNF mRNA stability (35,38).
Here, IRAK-2 was found to be required for this MyD88-dependent pathway to TNF␣ mRNA stability for both TLR4 and TLR8, but not to be required for IL-8 mRNA stability. This suggests that either different MyD88 complexes exist that differentially regulate mRNA stability or the pathways regulating TNF␣ and IL-8 mRNA stability bifurcate downstream of MyD88, with IRAK-2 at least one step removed from MyD88. It is possible that different Myddosomes exist that regulate different post-transcriptional events. For example, because IRAK-1 has been implicated in regulating mRNA stability of the mouse chemokine KC (keratinocyte-derived chemokine), it may be that a MyD88-IRAK-4-IRAK-1 Myddosome regulates murine KC and/or human IL-8 mRNA stability, whereas a MyD88-IRAK-4-IRAK-2 Myddosome regulates TNF␣ mRNA stability (15). Consistent with this idea, murine IRAK-4 has been shown to be required for both KC and TNF␣ mRNA stability (33).
IRAK-2 was shown to be required for MyD88-dependent stabilization of mRNA via the TNF␣ 3ЈUTR but not the IL-8 3ЈUTR. Although both TNF␣ and IL-8 mRNA are known to be regulated via AREs within their 3ЈUTRs, these AREs are subject to differential post-transcriptional control depending on which ARE-binding proteins target specific AREs (34,42,43). Although the TNF␣ and IL-8 3ЈUTRs both contain multiple AREs, their AREs vary in terms of abundance, sequence, and relative orientation to each other (AREsite, Universität Wien, available on-line). Thus it is possible that IRAK-2 modulates the activity of specific ARE-binding proteins that target the TNF␣ but not the IL-8 3ЈUTRs.
How exactly IRAK-2 regulates the mRNA stability pathway converging on the TNF␣ 3ЈUTR remains to be determined. IRAK-2, like all other IRAK family members, contains a functional ATP-binding pocket with an invariant lysine residue (Lys-237) in the protein kinase subdomain. It has been recently proposed that this residue is required for IRAK-2 to act as an active kinase (14,39), although previously IRAK-2 was assumed to be a pseudokinase. Here we found that Lys-237 was required for the ability of IRAK-2 to stimulate the TNF␣ 3ЈUTR. Thus, similar to the case with IRAK-4 (33), the kinase activity of IRAK-2 may be required for post-transcriptional regulation of the TNF␣ transcript.
It is known that TRAF6 plays a critical role in TLR-induced NFB and MAPK pathways, and in previous work we have demonstrated that IRAK-2 is critical for the formation of polyubiquitin chains associated with TRAF6, an activity that requires the IRAK-2 Glu-528 residue (26). However, the role of TRAF6 in mRNA stabilization is less well defined and will require further investigation. It has been previously reported that TRAF6 is not required for IL-1-induced stabilization of KC and MIP-2 mRNA (32). Here we provided circumstantial evidence that TRAF6 is required for the ability of IRAK-2 to stabilize TNF␣ mRNA, because, IRAK-2 E528A no longer stimulated the TNF␣ 3ЈUTR reporter. Furthermore, although WT TRAF6 stimulated the TNF␣ 3ЈUTR reporter, a TRAF6 mutant (C70A) impaired in its E3 ligase activity was not active, whereas a TRAF6 mutant with a mutated autoubiquitination site (K124R) was. Notwithstanding the need for further experiments to confirm a role for TRAF6, the data suggest the existence of an MyD88-IRAK-2-TRAF6-dependent pathway con-verging on the TNF␣ 3ЈUTR, which is dependent on the kinase activity of IRAK-2, and its ability to stimulate TRAF6 E3 ligase activity, but which does not require TRAF6 autoubiquitination.
In conclusion we have revealed a previously unrecognized role for IRAK-2 in early TLR signaling events and in post-transcriptional regulation of TNF␣ in primary human cells, which differs from its role in in-bred mice. Given that the dysregulation of TLR-mediated NFB and p38 activity, and of TNF␣ production, is implicated in many autoimmune and inflammatory diseases, IRAK-2 likely represents an attractive therapeutic target in human disease.