Sequential Autophosphorylation Steps in the Interleukin-1 Receptor-associated Kinase-1 Regulate its Availability as an Adapter in Interleukin-1 Signaling*

The interleukin-1 receptor-associated kinase 1 (IRAK-1) is an important adapter in the signaling complex of the Toll/interleukin-1 (IL-1) receptor family. Formation of the signaling IL-1 receptor complex results in the activation and hyperphosphorylation of IRAK-1, which leads to a pronounced shift of its apparent molecular mass in gel electrophoresis. Presently, the individual residues phosphorylated in IRAK-1 and the consequences for IRAK-1 function are unknown. We define sequential phosphorylation steps in IRAK-1, which are, in vitro, autophosphorylation. First, IRAK-1 is phosphorylated at Thr209. By fluorescence energy transfer experiments, we demonstrate that Thr209 phosphorylation results in a conformational change of the kinase domain, permitting further phosphorylations to take place. Substitution of Thr209 by alanine results in a kinase-inactive IRAK-1. Second, Thr387 in the activation loop is phosphorylated, leading to full enzymatic activity. Third, IRAK-1 autophosphorylates several times in the proline-, serine-, and threonine-rich ProST region between the N-terminal death domain and kinase domain. Hyperphosphorylation of this region leads to dissociation of IRAK-1 from the upstream adapters MyD88 and Tollip but leaves its interaction with the downstream adapter TRAF6 unaffected. This identifies IRAK-1 as a novel type of adapter protein, which employs its own kinase activity to introduce negative charges adjacent to the protein interaction domain, which anchors IRAK-1 at the active receptor complex. Thus, IRAK-1 regulates its own availability as an adapter molecule by sequential autophosphorylation.

The interleukin-1 receptor-associated kinase (IRAK-1) 1 is the prototype of a small family of serine/threonine protein kinases that are key molecules in the signaling cascade of the Toll/IL-1 receptor (TIR) family (reviewed in Ref. 1). The TIR family comprises the IL-1 receptor subfamily, recognizing the endogenous proinflammatory cytokines IL-1 and IL-18, and the members of the Toll-like receptor (TLR) subfamily, recognizing pathogenassociated molecular patterns (reviewed in Ref. 2). After ligand binding, all members of the TIR family form multimeric receptor complexes. These receptors share the cytoplasmic TIR domain, which is indispensable for signal transduction. The TIR domain serves as a scaffold for the recruitment of adapter proteins. This results in the activation of a core signaling module consisting of MyD88, IRAK-4 and IRAK-1, and TRAF6. Subsequently, several central signaling pathways are activated in parallel, the activation of NF-B being a hallmark of the inflammatory response. The additional use of individual adapter proteins like TIRAP/Mal (3,4) by TLR2 or TLR4 and TICAMI/TRIF by TLR3 (5, 6) accounts for the differential biological response of cells observed after stimulation with distinct pathogens.
IRAK-1 was the first family member identified due to its association with the activated IL-1 receptor complex (16) and its pronounced autophosphorylation (7). During IL-1 stimulation, IRAK-1 is recruited to the IL-1 receptor complex, where it is anchored via protein-protein interactions to the adapter proteins MyD88 (8,17,18), and Tollip, a molecule functioning also as silencer of quiescent IRAK-1 (19,20). Recently, additional adapter molecules have been described that interact with IRAK-1 and participate in the formation of a signaling complex at the receptor. They interact directly with IRAK-1 and include Pellino1 (21) and Pellino2 (22). Others facilitate interaction of IRAK-1 with the downstream adapter TRAF6 such as TIFA (23).
During activation, IRAK-1 becomes heavily phosphorylated, giving rise to a pronounced shift in gels, now accepted as a hallmark of IRAK activation. Hyperphosphorylated IRAK-1 leaves the receptor complex and interacts with TRAF6 (24) and TAK1, TAB1, and TAB2 (25). At least in vitro, IRAK-1 autophosphorylates, demonstrating that it is an active protein kinase. Interestingly, however, no substrate for IRAK-1 has been identified downstream of IRAK-1 yet. In fact, the only substrates known so far are IRAK-1 itself, Tollip (19), and Pellino2, an adapter molecule that is phosphorylated by IRAK-1 and IRAK-4 (26). It was reported that the consequences of phosphorylation of IRAK-1 were dissociation from MyD88 (7), termination of the interaction with Tollip (20), and massive alteration of the half-life of IRAK-1 (27,28). However, none of these reports identified the individual domains or the precise amino acids that actually become phosphorylated in IRAK-1 or explained why IRAK-1 shows such a drastic shift in electrophoretic mobility upon activation. The lack of downstream targets and the observation that IRAK-1 kinase activity is dispensable for IL-1 signaling (9,15,29,30) suggest that the kinase activity of IRAK-1 may serve functions other than relaying and amplifying IL-1 signals. We sought to identify and characterize the sites of IRAK-1 (auto)phosphorylation and to describe the consequences of (auto)phosphorylation for the function of IRAK-1 in IL-1 signaling.
Here we show that phosphorylation of IRAK-1 is due to three sequential phosphorylation steps, which in vitro are autophosphorylation steps. We identify Thr 209 as the critical amino acid for activation of IRAK-1 kinase activity and demonstrate that Thr 387 in the activation loop is required to achieve full enzymatic activity. The result of full kinase activation is hyperphosphorylation in the ProST region. We show that this hyperphosphorylation regulates the interaction of IRAK-1 with its upstream adapters MyD88 and Tollip but leaves its interaction with the downstream adapter TRAF6 unaffected. These data suggest that IRAK-1 uses its kinase activity to autophosphorylate and thereby limit its own availability as an adapter protein in IL-1 signaling.

Expression Vectors and Cloning Procedures
Mammalian expression vectors encoding wild-type human IRAK-1 (pRK5-IRAK) and human IRAK-1 with a defective ATP-binding site (IRAK K239S) were kind gifts of Z. Cao (Tularik) and have been described previously (7). The control plasmid pRK5* was generated from pRK5-IRAK by excision of the insert and religation of the vector. Nterminally FLAG-tagged fragments of IRAK-1 were constructed by inserting PCR-generated cDNA fragments in the mammalian expression vector pcDNA3-FLAG (kind gift of B. Lü scher, Aachen, Germany). Point mutations were introduced using the QuikChange TM site-directed mutagenesis kit (Stratagene, La Jolla, CA) exactly as specified by the manufacturer and as described previously (9,18). IRAK-1 deletion mutants (⌬ProST, ⌬KD) were generated by PCR using the plasmid coding for FLAG-tagged IRAK-1wt as template. The primers were designed to flank the regions to be eliminated with overhangs resulting in a hemagglutinin tag at the site of deletion.

Identification of PEST Sequences
The program PEST-FIND (available on the World Wide Web at www.at.embnet.org/embnet/tools/bio/PESTfind/) (31) was used with window size set to 10. Transient Transfection, Immunoprecipitation, and Immunoblotting 2 ϫ 10 6 HEK 293 cells were seeded into 100-mm Petri dishes (Falcon, BD Biosciences, Heidelberg, Germany) and transfected the following day by calcium phosphate precipitation with a total of 10 g of DNA.
The amounts of coding plasmids used were optimized for each construct and thus differed respectively. By adding empty vector, the total amount of DNA was kept constant in all transfections. After an over-night incubation period, the medium was changed, and 24 h later cells were collected, washed, and lysed for 30 min in buffer containing 50 mM HEPES, pH 7.9, 250 mM NaCl, 20 mM glycerophosphate, 5 mM pnitrophenyl phosphate, 1 mM EDTA, 1 mM sodium orthovanadate, 0.5% (w/v) Nonidet P-40, 10% (w/v) glycerol, 5 mM dithiothreitol, and protease inhibitors (Complete; Roche Applied Science). Nuclei and cellular debris were removed by centrifugation (13,000 ϫ g, 30 min, 4°C). Proteins were precipitated either with M2-agarose beads (Sigma) using their FLAG tag or by the indicated antibodies and Protein G-Sepharose (Amersham Biosciences). After extensive washing with lysis buffer, the beads were either subjected to in vitro kinase assays (see below) or were heated in SDS sample buffer. Proteins were separated by SDS-PAGE, transferred to polyvinylidene difluoride membranes, and immunostained with the indicated antibodies. Reactive bands on Western blots were visualized with horseradish peroxidase-coupled secondary antibodies by an enhanced chemiluminescence detection system (Pierce). Pull-down experiments were performed with the indicated antibodies as described previously (7,18).

In Vitro Kinase Assay
Washed immunoprecipitates were resuspended in a kinase reaction mixture containing 20 mM Hepes, pH 6.5, 150 mM NaCl, 5 mM MgCl 2 , 5 mM MnCl 2 , 1 M ATP, and 1 Ci of [␥-32 P]ATP and incubated for 20 min at 30°C. The kinase reaction was stopped by heating in SDS sample buffer. Phosphorylated proteins were separated by SDS-PAGE. The gels were dried, and autoradiographies were performed.

Mass Spectrometric Detection of Thr 209 Phosphorylation
Enzymatic Digest-The Coomassie-stained IRAK-1 KDL4 and IRAK-1 K239S KDL bands from one-dimensional SDS-PAGE were excised, destained, reduced, alkylated, and digested in-gel with trypsin (Promega Corp., Madison, WI) in 100 mM ammonium bicarbonate by using an enzyme/substrate ratio of 1:50 (w/w). The tryptic digestion was carried out for 12 h at 37°C.
Liquid Chromatography/Electrospray Ionization-Tandem Mass Spectrometry-Electrospray ionization-tandem mass spectrometry (n ϭ 2, 3) was performed on an LCQ DECA ion trap mass analyzer interfaced with a Surveyor HPLC system (Thermo Finnigan, San Jose, CA), and a 0.18-mm inner diameter ϫ 100-mm length reverse-phase C 18 column BioBasic 18 (ThermoHypersil-Keystone, Bellefonte, PA) was used. The flow rate was 25 l/min. A precolumn split directed ϳ3 l/min of HPLC effluent to the mass spectrometer. A gradient of 10 min of 100% solvent A (2% acetonitrile, 0.1% formic acid) followed by 60 min of 0 -70% solvent B (90% acetonitrile, 0.1% formic acid) was employed. 5 l of in-gel tryptic digest was diluted to 20 l with 14 l of solvent A and 1 l of 10% trifluoroacetic acid before injection into the LC column.

Generation of Fusion Proteins for Fluorescence Resonance Energy Transfer (FRET) Measurements/FRET Measurements
IRAK-1 KDL4 (aa 194 -536) cDNA was amplified via PCR with a 5Ј XhoI recognition sequence, followed by the FLAG coding sequence and a 3Ј EcoRI recognition sequence. The PCR product was inserted into the vector pEGFP-C1 (Clontech, BD Biosciences, Heidelberg, Germany) via XhoI and EcoRI, excised via AgeI and EcoRI together with the enhanced green fluorescent protein (EGFP) cDNA, and inserted into pQBI50-fN1 (Q-BIOgene, Heidelberg, Germany). The plasmids were kindly provided by M. Gaestel (Institute for Biochemistry, Hannover Medical School). HEK 293 cells were transiently transfected with 10 g of plasmid DNA coding for either wild type or T209A mutated EGFP-KDL4-enhanced blue fluorescent protein. 40 h after transfection, cells were washed with phosphate-buffered saline and lysed in 1 ml of lysis buffer (20 mM Tris, 100 mM EDTA, 1 mM EGTA, 270 mM sucrose, 10 mM glycerophosphate, 5 mM sodium pyrophosphate, 1 mM sodium orthovanadate, 1% (v/v) Triton X-100, 0.1% (v/v) ␤-mercaptoethanol, 1ϫ protease inhibitor mixture (Roche Applied Science), pH 7.0). DNA was pelleted by 15,000 ϫ g centrifugation for 20 min and lysates were cleared by an additional 100,000 ϫ g centrifugation step.
Fluorescence emission spectra of lysates were measured from 400 -600 nm with an excitation wavelength of 378 nm (direct excitation of enhanced blue fluorescent protein). After subtraction of the emission spectrum of lysate from empty vector-transfected cells, the relative fluorescence intensity was normalized against the fluorescence intensity at 508 nm when EGFP was directly excited at 473 nm. Additionally, the fusion proteins were affinity-purified on M2-agarose, eluted with FLAG peptide, and then analyzed. Fluorescence measurements were performed at the Biophysics Department of the Hannover Medical School with the kind support of Dr. U. Curth.

Demonstration of Phosphorylation of Endogenous IRAK-1
1 ϫ 10 7 HEK293RI cells (stably overexpressing the type I IL-1 receptor) were seeded into 100-mm Petri dishes and on the next day stimulated with 1 ng/ml IL-1␤ for the indicated periods of time. Cells were washed twice with ice-cold phosphate-buffered saline. Lysis, immunoprecipitation with anti-IRAK-1 antibody 2A9, SDS-PAGE, Western blotting, and immunostaining with 2A9 were performed as described above.
For phosphatase treatment, washed immunoprecipitates were washed twice with phosphate-buffered saline and incubated with 0.125 units/l protein phosphatase 1 (New England Biolabs) for 1 h at 30°C in the provided reaction buffer (50 mM Tris-HCl, 0.1 mM Na 2 EDTA, 5 mM dithiothreitol, 0.01% Brij 35, 1 mM MnCl 2 , pH 7.0). After removing the buffer, beads were heated in SDS sample buffer.

Statistics
All experiments were routinely performed at least three times with comparable results.
Identification of Phosphorylated IRAK-1 Domains-Although it has been known for a long time that IRAK-1 becomes heavily phosphorylated upon IL-1 stimulation of cells or upon overexpression, the domains phosphorylated or the individual amino acids have not yet been defined. At least in vitro, IRAK-1 is capable of autophosphorylation, which results in a shift identical to the one observed in cells stimulated with IL-1. Furthermore, recombinant IRAK-1 produced in insect cells and purified to homogeneity yields the typical IRAK-1 shift after incubation in an in vitro kinase assay in the absence of any further proteins (data not shown).
First we defined the domains and amino acids phosphorylated in IRAK-1.
Since these questions cannot be addressed with endogenous IRAK-1, we chose overexpression systems in which IRAK-1 is known to autophosphorylate (7). This allowed us to generate IRAK-1 mutants lacking individual domains or containing defined point mutations.
A series of constructs was produced including the individual domains, and these were tested either for their autophosphorylative capacity or whether they could serve as a substrate for IRAK-1 (Fig. 1A). All proteins carried an N-terminal FLAG epitope tag. They were transiently expressed in HEK 293 cells, either alone or in combination with full-length IRAK-1. Then co-immunoprecipitations (co-IP) were performed using anti-FLAG monoclonal antibody beads, and the precipitates were subjected to an in vitro kinase assay. We developed this experimental set-up (forced co-IP; schematically depicted in Fig. 1B), because it closely resembles the situation in IRAK-1 dimers or oligomers that form by the homotypic interaction of IRAK-1

FIG. 1. Identification of phosphorylated IRAK-1 domains.
A, schematic representation of human IRAK-1 and the fragments and mutant proteins generated. IRAK-1 is composed of four domains: the N-terminal death domain, followed by the regulatory ProST region, the serine/ threonine-specific kinase domain, and a C-terminal region. IRAK-1 fragments comprise individual domains or combinations thereof. IRAK-1 K239S mutant proteins bear a point mutation in the ATP-binding site that aborts their kinase activity. All recombinant proteins carry an N-terminal FLAG tag. B, depiction of the forced co-IP assay. FLAG-tagged IRAK-1wt and FLAG-tagged IRAK-1 fragments were co-expressed in HEK 293 cells and co-immunoprecipitated using M2-agarose in order to force them into close proximity. Subsequently, an in vitro kinase assay was performed on the washed beads to test whether the fragments serve as substrates for IRAK-1wt (symbolized by the arrow). C, the ProST region and the N-terminally extended kinase domain KDL are phosphorylated after forced co-immunoprecipitation. death domains at the receptor complex (9), possibly facilitated by MyD88 dimers (17). Under these experimental conditions, IRAK-N208, composed of death domain and most of the ProST region and thus not enzymatically active (Fig. 1C, lane 1), served as a substrate for full-length IRAK-1 (lane 2). The kinase domain (KDL) autophosphorylated (lane 3). Neither the death domain (aa 2-109) nor the C terminus (aa 537-712) were phosphorylated by IRAK-1 (data not shown). These results identify the ProST region and the kinase domain as the two domains of IRAK-1 that become phosphorylated.
Kinase Activity of IRAK-1 Depends on the Phosphorylation of Threonine 209 -Interestingly, the protein comprising exactly the minimal homology sequence of the kinase domain (Fig. 1, KD; aa 212-536) did not autophosphorylate (Fig. 1C, lane 5), whereas a protein containing part of the ProST region (KDL; aa 149 -536) was enzymatically active (Fig. 1C, lane 3). We sequentially truncated KDL (aa 149 -536) to KDL4 (aa 194 -536), which was still enzymatically active ( Fig. 2A). This N-terminal extension contained three phosphorylatable amino acids (Ser 196 , Ser 206 , and Thr 209 ). By substituting these three amino acids with alanine residues, Thr 209 was identified as the critical amino acid for maintaining kinase activity. Mutating Thr 209 into alanine completely abolished auto-phosphorylation in the kinase domain of IRAK-1 ( Fig. 2A) or in full-length IRAK-1 (Fig. 2B). Substitution of Thr 209 by the negatively charged amino acids glutamate or aspartate did not result in a constitutively active kinase but rendered these IRAK-1 Thr 209 mutant proteins as inactive as IRAK-1 K239S in which the ATP binding site was mutated (Fig. 2B). These results demonstrate that Thr 209 is essential for enzymatic activity, and they strongly suggest that Thr 209 is a site of phosphorylation.
In order to ascertain whether threonine 209 was indeed phosphorylated, we performed liquid chromatography/electrospray ionization-tandem mass spectrometry (n ϭ 2, 3) experiments on peptides derived from IRAK-1 KDL4 (and IRAK-1 K239S KDL as unphosphorylated control; data not shown). The MS n was operated to isolate and fragment target ions based on the predicted mass of the tryptic peptide containing the phosphorylated threonine 209 (aa 208 -217, GTpHNFSEELK; m/z 1242.2). Fig. 2C shows the MS 2 spectrum of ϩ2 charged peptide ions of aa 208 -217 (m/z 621.6). In agreement with previous findings (35,36), the dominant ion (ϩ2; m/z 572.5) corresponded to the loss of 98 Da (H 3 PO 4 ), but a large series of band y-type ions were also observed. According to Fig. 2C (upper  panel), a partial series of y ions (y5, y7), which did not contain a phosphate group, and b3, which contained a phosphate group, clearly indicated that the phosphorylation site was located at position of threonine 209. To investigate further the phosphorylation of this peptide, a MS 3 experiment was carried out by isolating and fragmenting the dominant product ion (m/z 572.5). Fig. 2C (lower panel) shows the MS 3 spectrum of m/z 572.5. Extensive fragmentation along the peptide backbone was observed because of the facile loss of the phosphate group. In agreement with the MS 2 results, y ion series from y5 to y8 -18 (loss of H 2 O) did not contain dehydroalanine (phosphorylated serine that has lost H 3 PO 4 ), and b3 ion contained dehydroaminobutyric acid (phosphorylated threonine that has lost H 3 PO 4 ). From these results, threonine 209 was unambiguously determined as the phosphorylation site in the peptide aa 208 -217.
The Consequence of Threonine 209 Phosphorylation Is a Conformational Change in the Kinase Domain of IRAK-1-Thr 209 is located at the border of the conserved kinase domain. Thus, we hypothesized that Thr 209 could be involved in a conformational change of the IRAK-1 kinase domain. Since the effect of T209A substitution on kinase activity could be observed in KDL4 (aa 194 -536), we used this protein to investigate this question. We generated fusion proteins that allow detection of conformational changes by alterations in the FRET of either KDL4wt or KDL4T209A. These fusion proteins consisted of an N-terminal EGFP, an intermediate FLAG epitope tag, the respective IRAK-1 KDL4 sequence, and a C-terminal enhanced blue fluorescent protein (EBFP) (Fig. 3A). These constructs were transiently expressed in HEK 293 cells, and both fusion proteins were detected with the calculated size (Fig. 3B). The fluorescence properties were as expected, and all transfected cells emitted blue and green fluorescence light after appropri-ate excitation (data not shown). The fusion protein containing the wild type kinase domain autophosphorylated upon overexpression in an in vitro kinase assay, demonstrating that the addition of the two fluorescent moieties at the termini did not distort the kinase domain (Fig. 3B). In the lysates of HEK 293 cells transiently transfected with the fluorescent fusion proteins, the emission spectra of KDL4wt and KDL4T209A were measured (excitation of enhanced blue fluorescent protein at 378 nm) and compared. A strong reduction of the fluorescence energy transfer at 508 nm was observed after substituting threonine 209 by alanine, indicating that the fluorophores were further apart in the mutated form than in the wild type molecule in which stronger fluorescence energy transfer was detected from the enhanced blue fluorescent protein to the EGFP moiety (Fig. 3C). Similar results were obtained with affinitypurified fusion proteins (data not shown).
Full Activation of IRAK-1 Enzymatic Activity Requires Threonine 387 Phosphorylation-Full enzymatic activity of protein kinases is often achieved after phosphorylation of one or more amino acids in the activation loop. Since the kinase domain of IRAK-1 is very conserved, we hypothesized that full enzymatic activity may depend on phosphorylation(s) in the activation loop. The activation loop of IRAK-1 contains nine serine and threonine residues that could serve as targets for phosphorylation. By mutational analysis, we identified Thr 387 as the critical amino acid required for full kinase activity. The T387A mutation resulted in alteration of the autophosphorylation pattern. First, overall phosphorylation intensity was strongly reduced compared with IRAK-1wt, and second, only a partial shift was observed ( Fig. 4; also see Fig. 2B). Any other single mutation of serines or threonines in the activation loop had no effect; these mutant proteins autophosphorylated and shifted like IRAK-1wt (Fig. 4A). Since in some kinases introduction of negative charges at critical phosphorylatable amino acids mimics phosphorylation and is sufficient to activate kinases constitutively, we exchanged Thr 387 for aspartate or glutamate. How- ever, these mutant proteins behaved like the T387A protein and were not constitutively active (data not shown).
Since many kinases require double phosphorylations in two neighboring amino acids to become fully activated, we generated double mutations on the basis of T387A. Although subtle differences in autophosphorylation were observed, the overall pattern was comparable with that of T387A, and no double mutation in the activation loop yielded an enzymatically inactive IRAK-1 (Fig. 4B). These data show that Thr 387 is required to achieve full autophosphorylative capacity of IRAK-1, which results in the typical shift in SDS-PAGE. Thus, Thr 387 regulates hyperphosphorylation of IRAK-1.
The ProST Region Is the Site of IRAK-1 Hyperphosphorylation-Our initial fragmentation studies identified the ProST region of IRAK-1 as a target for autophosphorylation. We observed that phosphorylation of proteins including the complete ProST region always resulted in several phosphorylated bands, suggesting multiple phosphorylations (Fig. 1C). The ProST region (aa 110 -211) contains 23 proline, 20 serine, and 6 threonine residues in a total of 102 amino acids. Two stretches of this sequence yield relatively high scores of 1.7 (aa 117-133) and 7.3 (aa 153-180) in a program designed to identify potential PEST sequences (31,37) (Fig. 5A) and thus represent potential regulatory areas. In HEK 293 cells, fragments of IRAK-1 containing the death domain and stretches of different lengths of the adjacent ProST region were co-expressed with IRAK-1wt, and the phosphorylation pattern was analyzed in forced co-immu-noprecipitation assays. Consecutive deletion of phosphorylatable serines and threonines identified several critical amino acids in the ProST region and resulted in altered phosphorylation patterns or reduced phosphorylation intensities (Fig. 5B). Clear alteration in the phosphorylation pattern began after reducing IRAK-1 N143 to IRAK-1 N139, with a dramatic change again observed after truncation to IRAK-1 N127, suggesting that the SSAST motif is target for multiple phosphorylations. Phosphorylation was practically lost after truncation to IRAK-1 N109, which is the IRAK-1 death domain. The death domain was not phosphorylated by IRAK-1 in forced co-IP experiments (Fig. 5B, IRAK-N109). The shifts in either IRAK-1 N195 (data not shown) or IRAK-1wt (see Fig.  7) were completely reversible by phosphatase treatment, demonstrating that the shift of IRAK-1 is due to phosphorylation. These results identify the ProST region as the site of hyperphosphorylation.

Hyperphosphorylation in the ProST Region Terminates Interaction of IRAK-1 with the Upstream Adapters MyD88 and
Tollip-We defined the ProST region and the kinase domain as the structural elements that become phosphorylated in IRAK-1. We also established the sequence of phosphorylation events and identified the consequences of the first and second phosphorylation event. We then investigated the consequence of the hyperphosphorylation in the ProST region, which is the third phosphorylation event.
MyD88 and Tollip play critical roles in the recruitment of Two potential PEST sequences can be identified in the ProST region (boldface type). B, identification of multiple phosphorylation sites in the ProST region. IRAK-1 N211 (the death domain plus the complete ProST region) was sequentially truncated. These mutant proteins were transiently co-expressed together with IRAK-1wt, and forced co-immunoprecipitation with subsequent in vitro kinase assays was carried out. Phosphorylation of the ProST region (IRAK-1 N211) by IRAK-1wt gave rise to a characteristic pattern of several bands in the autoradiogram, which was reduced in IRAK-1 N127 and disappeared completely in IRAK-1 N109 (the death domain).
IRAK-1 to the active receptor complex. In order to clarify which of the three individual phosphorylation events in IRAK-1 controls the interaction with MyD88 or Tollip, we transiently co-expressed these molecules in HEK 293 cells with different forms of IRAK-1 that lacked either enzymatic activity (IRAK K239S, IRAK-1 T209A), the ProST region (IRAK-1 ⌬ProST), or the kinase domain (IRAK-1 ⌬KD). Then we performed co-immunoprecipitation experiments. We confirmed that interaction of IRAK-1 with either MyD88 or Tollip was sensitive to phosphorylation, whereas interaction with the downstream adapter TRAF6 was not (Fig. 6). IRAK-1wt could not be pulled down with MyD88 or Tollip, whereas TRAF6 was interacting with hyperphosphorylated IRAK-1. If the kinase domain was deleted and the ProST region left intact, this deletion mutant behaved like IRAK-1wt and could only be co-precipitated with TRAF6 and not with MyD88 or Tollip. However, deletion of the ProST region, as the site of hyperphosphorylation, allowed the interaction with up-and downstream adapters similar to IRAK-1 K239S or IRAK-1 T209A.
This shows that the protein interaction of the death domains of MyD88 and IRAK-1 is sensitive to the introduction of phosphates in the neighboring ProST region, whereas the interaction of IRAK-1 with TRAF6, which is facilitated by the C terminus, is not. With respect to the interaction of Tollip with IRAK-1, we showed that removal of the kinase domain completely abrogated interaction with Tollip, whereas removal of the ProST region still allowed it. Thus, the ProST region is not required for the interaction of Tollip with IRAK-1, but hyperphosphorylation of the ProST region affects the Tollip/IRAK-1 interaction negatively.
These results suggest that introduction of phosphates in the ProST region serves the regulatory function of rapidly releasing activated and hyperphosphorylated IRAK-1 from the adapter molecules anchoring it at the receptor complex, whereas it does not influence the protein interaction with the downstream adapter TRAF6.
Phosphorylation Pattern of Endogenous IRAK-1 upon IL-1 Stimulation-The experiments described above were carried out in overexpression systems or in vitro. In order to show the validity of our findings in the natural situation, we stimulated HEK293RI cells with IL-1␤ in a time course, immunoprecipitated endogenous IRAK-1, and analyzed its phosphorylation status by Western blotting. Without stimulation, only the unphosphorylated form of IRAK-1 is visible (Fig. 7). After 5 min, incomplete shifted intermediate forms of IRAK-1 appear, and the hyperphosphorylated high molecular forms of IRAK-1 are visible after 10 min of stimulation (Fig. 7A, upper panel). The lower panel of Fig. 7A shows that unphosphorylated IRAK-1 is depleted over time. When we extended the time course, we could no longer detect hyperphosphorylated IRAK-1 after 4 h, concomitant with the reduction of nonactivated IRAK-1 (Fig.  7B). Finally, the shift of IRAK-1 could be reversed by treatment with protein phosphatase 1 (Fig. 7B, right lane), demonstrating that the shift is due to phosphorylation or modification via phosphoester bond.

DISCUSSION
The IRAK family members IRAK-1 and IRAK-4 are receptorproximal protein kinases in the signaling cascade of the TIR family. They are enzymatically active and therefore represent attractive targets for the development of anti-inflammatory drugs, since TIR family members are involved in the initiation and perpetuation of inflammatory disorders.
Presently, however, some major questions remain open concerning the potential of inhibiting the kinase activity of IRAK-1. Normally, cytokine receptor-associated protein kinases use their enzymatic activity to phosphorylate downstream substrates to relay and amplify the signal initiated at the receptor complex. IRAK-4 follows this rule, since its kinase activity is absolutely necessary for IL-1 signaling (10). For IRAK-1, this is not that clear. Although IRAK-1 becomes heavily phosphorylated after IL-1 stimulation, at least in vitro due to autophosphorylation, IRAK-1 kinase activity was reported to be dispensable for IL-1 signaling (9,15,29,30). In addition, no downstream substrate has yet been identified that is involved in signal amplification. This suggests that the kinase activity of IRAK-1 may serve functions other than relaying and amplifying signals. In order to better understand the function of IRAK-1 enzymatic activity, we aimed to identify the domains and amino acids autophosphorylated in IRAK-1 and to characterize the molecular mechanism of activation of the enzymatic activity of IRAK-1. Finally, we wanted to understand the consequences of IRAK-1 phosphorylation for its function as a signaling molecule. As a prototype for the TIR family, we chose the IL-1 receptor system for our studies.
We identified the ProST region and the kinase domain as phosphorylation sites in IRAK-1. Recently, it was suggested that Thr 66 in the death domain was phosphorylated (20,38). We did not observe phosphorylation in either death domain or C terminus. However, since we only ascertained auto-or crossphosphorylation of IRAK-1, we cannot exclude that IRAK-1 is a substrate for other protein kinases. Thus, it has been shown that Thr 100 is phosphorylated by a calcium/calmodulin-dependent protein kinase with implications for NF-B signaling (39).
Phosphorylation of the kinase domain resulted in one band, whereas phosphorylation in the ProST region gave rise to several bands, suggesting that several phosphorylations take place in the ProST region that are responsible for the IRAK-1 typical shift.
We observed that the core kinase domain, as defined by homology searches in protein data bases, was enzymatically inactive and did not serve as a substrate for IRAK-1wt. A short N-terminal extension including the critical Thr 209 sufficed to allow autophosphorylation. Like mutating Lys 239 in the ATP binding site, mutating Thr 209 completely destroyed the enzymatic activity of the kinase domain or full-length IRAK-1. Thus, Thr 209 was identified as a second amino acid that controls the kinase activity of IRAK-1. By mass spectrometry, we showed that Thr 209 is phosphorylated in cells. By FRET analysis, we demonstrated that Thr 209 controls a conformational change of the kinase domain. In the absence of Thr 209 (IRAK-1 KDs), the kinase domain did not serve as a substrate for wild type IRAK-1, whereas in the presence of Thr 209 (IRAK-1 KDL4), this molecule could autophosphorylate. In combination, these results suggest that Thr 209 phosphorylation regulates the opening of IRAK-1 kinase domain, allowing weak enzymatic activity and giving access to the activation loop.
Full enzymatic activity is achieved after phosphorylation of Thr 387 in the activation loop of IRAK-1. Mutation of Thr 387 resulted in a severe reduction of autophosphorylation and in a reduced shift. This is in accordance with the model that Thr 209 phosphorylation allows a weak enzymatic activity, which gives rise to a phosphorylated band showing an intermediate shift.
This result also demonstrates that Thr 387 regulates full enzymatic activity necessary for the pronounced shift of IRAK-1. Mutation of Thr 387 did not abolish kinase activity completely, suggesting an additional phosphorylation site in the activation loop. We could not identify a distinct second phosphorylation site in the activation loop. However, it is conceivable that IRAK-1 shows redundancy in its substrate specificity, since the activation loop contains nine serine and threonine residues in close proximity.
Does IRAK-4 activate IRAK-1? In peptide phosphorylation assays, we identified Thr 387 in the activation loop of IRAK-1 as a potential target for IRAK-4 (10), suggesting that IRAK-4 may phosphorylate the activation loop of IRAK-1 in cells. Since recombinant IRAK-1 derived from insect cells and purified to homogeneity shows full kinase activity, hyperphosphorylation, and the IRAK-1 typical shift (data not shown here), we must assume that at least in overexpression systems, IRAK-1 can autophosphorylate at the two critical positions Thr 209 and Thr 387 . Peptide phosphorylation experiments showed that IRAK-4 can phosphorylate peptides including Thr 209 of IRAK-1 (data not shown here). Interestingly, IRAK-1 co-precipitated with IRAK-4 does not show a significant shift (10), suggesting that IRAK-4 phosphorylates IRAK-1 at Thr 209 , thereby initiating the autophosphorylation sequence which finally results in the hyperphosphorylation of IRAK-1.
We mapped hyperphosphorylation of IRAK-1 to the ProST region, a hydrophilic sequence of 102 amino acids rich in serine, threonine, and proline residues, which contains two potential PEST sequences. PEST sequences have been implicated in the regulation of protein half-life by serving as targets for phosphorylation and often also for subsequent polyubiquitination (31,37). The question arises whether the ProST region is phosphorylated by IRAK-4. In forced co-IP experiments with ProST fragments and IRAK-4, we found only very marginal phosphorylation, which was by no means comparable with that observed with IRAK-1 (data not shown). This suggests that hyperphosphorylation of the ProST region is predominantly, if not exclusively, catalyzed by IRAK-1.
Although phosphorylation of IRAK-1 is dispensable for its function as an adapter molecule in transfection experiments, it does have consequences for IRAK-1 in intact cells. It has been shown that interaction with the receptor complex or the upstream adapters MyD88 and Tollip was not detectable after IRAK-1 phosphorylation (7,8,17,18). Here we demonstrate that hyperphosphorylation in the ProST region terminates the interaction with both adapter molecules, resulting in the release of hyperphosphorylated IRAK-1 from the active receptor complex. Hyperphosphorylation did not affect interaction with the downstream adapter TRAF6 as described by Cao et al. (7).
Several studies correlate phosphorylation with the reduction of IRAK-1 protein half-life by proteolytical degradation (27,28,40). We suggest that hyperphosphorylation in or adjacent to the two PEST sequences in the ProST region regulates proteolytical cleavage. Frequently, phosphorylation is the first step toward polyubiquitination, which targets proteins to the proteasome. Although we were unable to detect IRAK-1 ubiquitination employing several different techniques in our cell types, others have found IRAK-1 ubiquitination in their cellular systems (15,33). We do not know the reason for this discrepancy, but our inhibitor studies (data not shown) support the observation that the hyperphosphorylated form of IRAK-1 is proteolytically degraded, resulting in a time-dependent depletion of IRAK-1 within the stimulated cells.
In conclusion, our study defines IRAK-1 as a novel type of adapter molecule that uses its kinase activity to limit its own availability in IL-1 signaling by autophosphorylation. A working model is depicted in Fig. 8. The results leading to the depicted three-step model of IRAK-1 activation were obtained in overexpression experiments. The sequential phosphorylation steps of first Thr 209 and then Thr 387 and finally the hyperphosphorylation in the ProST region resulted in experimentally detectable shifts of IRAK-1 in gel electrophoresis. We stimulated HEK293RI cells with IL-1 and ascertained by Western blot analysis the behavior of endogenous human IRAK-1. We observed exactly the same shifting pattern as seen in overexpression systems: a 76-kDa form in unstimulated cells (nonphosphorylated IRAK-1), the transient appearance of intermediate forms (corresponding to IRAK-1 phosphorylated at Thr 209 and Thr 387 ), and the maximally shifted form of 110 kDa (hyperphosphorylated IRAK-1). We were able to reverse the shift in samples immunoprecipitated from IL-1 stimulated cells by phosphatase treatment, demonstrating that the modifications causing the shift were due to phosphorylation (or modifications via a phosphate-ester bond). In summary, these results allow us to draw the conclusion that the sequential autophosphorylation mechanism seen in overexpression experiments also takes place with endogenous IRAK-1 in intact cells after IL-1 stimulation.
Our results have major implications for the development of IRAK-1 inhibitors. In contrast to IRAK-4, where kinase activity is required for signal transduction, inhibiting the kinase activity of IRAK-1 may turn out to result in an effect opposite to the one desired. One must assume that inhibiting autophosphorylation will inhibit dissociation of IRAK-1 from the receptor complex and at the same time will increase the half-life of IRAK-1. This would lead to an increase and prolongation of the IL-1 signal rather than an inhibition. The availability of IRAK-1-specific kinase inhibitors will enable us to verify the working model derived from the results reported here.