Introduction
Toll-like receptors (TLR) represent the best-known family of mammalian innate immune defense systems. These receptors recognize extracytoplasmic pathogen-associated structures including components of the bacterial cell wall (for a review, see ref. 1). The cytoplasmic Toll–interleukin-1 (IL-1) receptor (TIR) domain of TLRs mediates homotypic interactions with TIR domain–containing adaptor proteins, such as MyD88, TRIF (also called TICAM-1), TRAM or Mal (also known as TIRAP). In most cases, receptor-adaptor interactions lead to the recruitment and activation of kinases of the IL-1R-associated kinase (IRAK) family, which positively or negatively regulate a complex signaling cascade leading to the activation of NF-
B and mitogen-activated protein kinases (MAPK) and to the production of defense molecules such as inflammatory cytokines. Fourteen years have passed since the discovery of the first member of the IRAK family, IRAK1 (ref. 2). Although this molecule was first described in IL-1 signaling, it was then defined as an important effector of most TLRs, with the notable exception of TLR3. To date, four members (IRAK1, IRAK2, IRAK-M and IRAK4) have been characterized in mammals, and mice deficient in three of the four respective genes have also been described, with the exception being IRAK2 (ref. 3). In this issue of Nature Immunology, Kawagoe et al. now report the generation and in-depth analysis of Irak2-
/-
mice4.
Since the discovery of IRAK2 in 1997, its exact contribution in TLR and IL-1R signaling has remained elusive. IRAK2 has been shown to interact with Traf6 and MyD88 and to activate the transcription factor NF-B, pointing to a strong analogy with IRAK1 not only in structure but also in function5. Subsequently, IRAK2 was reported to bind Mal, an essential adaptor for TLR2- and TLR4-mediated, MyD88-dependent NF-B activation6, 7. More recently, siRNA-mediated knockdown of IRAK2 in human cell lines showed that this molecule participates in NF-B responses to multiple TLRs, including TLR3 (ref. 8). Hence, although IRAK2 has always been proposed to act as a positive regulator of TLR signaling, there remained several unresolved issues. Which TLR signals through this kinase? What are its in vivo functions? What are the relative contributions and potential redundancy of IRAK1 and IRAK2?
To gain insight into the in vivo function of IRAK2, Kawagoe et al.4 use a gene-targeting approach to generate a complete knockout of the corresponding gene in mice. In vivo analysis of lipopolysaccharide (LPS) and CpG responses demonstrated that IRAK2 participates in septic shock mediated by the respective receptors of LPS and CpG, namely TLR4 and TLR9. Furthermore, a careful analysis in primary macrophages showed that IRAK2 was required to mount an optimal response to most TLRs and also to IL-1. Mechanistically, this can be explained by TLR-triggered, IRAK2-dependent sustained NF-B activation, whereas IRAK2-deficient cells show normal MAPK and early-phase NF-B activation (probably due to the presence of IRAK1; see Fig. 1 and below). At the mRNA level, the induced expression of cyclooxygenase 2 and tumor necrosis factor at early time points is unaltered, but at later time points their expression is impaired. At the protein level, the absence of IRAK2 results in lower total production of some inflammatory cytokines, including tumor necrosis factor and IL-6. Together these results reveal the importance of IRAK2 in TLR responses.
Figure 1: Contributions of IRAK molecules to TLR signaling.
According to Kawagoe et al.4, TLR signaling, exemplified in this study for TLR2, uses IRAK1 and IRAK2 for early NF-B and MAPK responses, whereas IRAK2 remains activated at later time points. IRAK4 most likely governs the phosphorylation (p) of both IRAK1 and IRAK2, which leads to autophosphorylation in the case of IRAK1. Whereas IRAK1 and IRAK2 become polyubiquitinated in a Lys63-dependent manner, IRAK1 is then degraded in an unknown, but probably proteasome-independent, way as reported in IL-1 signaling13. In contrast, IRAK2 and IRAK4 amounts remain unchanged, hence allowing late responses to occur. Although Lys63-mediated IRAK1 polyubiquitination results in recruitment of the ubiquitin receptor NEMO, a prerequisite for NF-B activation, it is still unknown whether IRAK2 polyubiquitination serves identical functions. These events are orchestrated by TIR adaptor molecules, such as Mal and MyD88 in the case of TLR2 signaling, and by TRAF6, which also becomes polyubiquitinated upon TLR triggering. For simplicity, this model does not include the contribution of the TLR-induced short MyD88 (MyD88s, in mice), of IRAK-M, which dampens TLR responses, and of the TAK1-TAB2-TAB3 machinery implicated in TLR signaling. DD, death domain; KD, kinase domain; TIR, Toll–interleukin-1 receptor.
IRAK4 is known as the master IRAK member, in that its absence strongly impairs TLR- and IL-1-mediated signaling and innate immune defense, whereas the absence of IRAK1 only slightly affects host responses, as exemplified by only partially altered NF-B and MAPK activation9, 10, 11. When IRAK2-deficient mice are crossed with mice lacking IRAK1, the resulting doubly deficient mice have a profound defect in TLR responses. This defect is demonstrated both in vivo, by a complete resistance to LPS- or CpG-induced septic shock (a feature also seen in IRAK4 deficiency10), and also in macrophages, which show strongly impaired cytokine production in response to most TLRs except TLR3. That MAPK and NF-B activation at early time points is potently affected in the combined absence of IRAK1 and IRAK2 suggests that both proteins operate redundantly in the initial response, whereas IRAK2 continues to work at later time points, when IRAK1 disappears (Fig. 1). Furthermore, this scenario can be extended to IL-1 signaling, as IRAK1 IRAK2 doubly deficient mouse embryonic fibroblasts show impaired production of IL-6 upon IL-1 stimulation as compared to cells derived from wild-type or even single-knockout mice. Collectively, these results show that both kinases cooperate to mount appropriate innate immune and inflammatory reactions.
One interesting yet complex feature seen in studying the IRAK family (and also the related receptor-interacting protein (RIP) kinase family) relates to the understanding of the role of their catalytic activity, if it exists. According to well-defined rules, the analysis of the primary amino acid sequence of a kinase domain is highly predictive of the activity or inactivity of the corresponding kinase12. Each kinase domain of the entire kinase superfamily (approximately 500 proteins both in human and in mouse) can be divided into 12 subdomains (I–XI, with VI being divided into VIa and VIb). Within these subdomains, certain crucial amino acids are invariant in active kinases. First, a lysine in subdomain II is required for ATP binding. Second, subdomain VIb contains the catalytic loop, composed of a HisArgAsp (HRD) motif, in which the aspartic acid acts as a proton acceptor for the substrate's hydroxyl group during phosphotransfer. Third, the aspartic acid in the highly conserved AspPheGly (DFG) motif of subdomain VII chelates Mg2+ for a correct positioning of the ATP 
-phosphate. Hence, upon identification of these crucial amino acids, kinase activity can be predicted for any kinase. When analyzing the kinase domains of the IRAK members, it becomes evident that both IRAK1 and IRAK4 are catalytically active kinases, whereas IRAK2 and IRAK-M are predicted to be inactive3. In the case of IRAK2, both DFG and HRD motifs are missing, which strongly suggests that IRAK2 is catalytically inactive.
Hence, the most surprising hypothesis offered by Kawagoe et al.4 is that IRAK2 is enzymatically active. They inferred this from an experiment in which IRAK2-deficient macrophages were retrovirally reconstituted with either wild-type IRAK2 or a point mutant containing a lysine-to-alanine substitution (of the lysine that binds ATP in active kinases, as mentioned above). Whereas wild-type IRAK2 expression reconstituted TLR2 responses, the point mutant, surprisingly, did not confer responsiveness to a TLR2 ligand. Although the authors conclude that IRAK2 bears kinase activities that are indispensable for TLR responses, another possible explanation is that this particular mutation somehow abolishes crucial interactions between IRAK2 and yet another TLR signaling molecule, such as IRAK4. This alternative scenario is supported by data showing that the point mutant IRAK2 is not phosphorylated upon TLR2 triggering, a situation also observed for wild-type IRAK2 in IRAK4-deficient cells. Importantly, if this scenario holds true, the broader consequences are that, in the kinome area, the mutation of a single amino acid (such as the lysine that binds ATP) of a particular kinase might not be sufficient to correlate an effect to a loss of kinase activity. Rather, one should carefully perform different independent point mutations of crucial amino acids within different kinase subdomains to ensure that an observed effect reflects abolished kinase activity and not merely a change in conformation that prevents interactions with signaling partners.
Over recent years, research on IRAK proteins has greatly enhanced our understanding of TLR and IL-1 signaling in innate immunity and inflammation. This study highlights the importance of analyzing the function of an IRAK molecule not as an isolated entity, but rather as a piece connected to a network of other IRAK proteins. Although it was already established that IRAK4 controls IRAK1 activity, the partly redundant but also sequential roles of IRAK1 and IRAK2 are now better understood. By further crossing IRAK-deficient mice, it will certainly be feasible and interesting to understand possible additional cross-talks, for example between the long-lasting positive regulator IRAK2 and the TLR-inducible, negative-regulator IRAK-M proteins.
