HOIL‐1‐catalysed, ester‐linked ubiquitylation restricts IL‐18 signaling in cytotoxic T cells but promotes TLR signalling in macrophages

The atypical E3 ligase HOIL‐1 forms ester bonds between ubiquitin and serine/threonine residues in proteins, but the physiological roles of this unusual modification are unknown. We now report that IL‐18 signalling leading to the production of interferon γ (IFNγ) and granulocyte–macrophage colony‐stimulating factor (GM‐CSF) is enhanced in cytotoxic T cells from knock‐in mice expressing the E3 ligase‐inactive HOIL‐1[C458S] mutant, demonstrating that the formation of HOIL‐1‐catalysed ester‐linked ubiquitin bonds restricts the activation of this pathway. We show that the interaction of IRAK2 with TRAF6 is required for IL‐18‐stimulated IFN‐γ and GM‐CSF production, and that the increased production of these cytokines in cytotoxic T cells from HOIL‐1[C458S] mice correlates with an increase in both the number and size of the Lys63/Met1‐linked hybrid ubiquitin chains attached to IRAK2 in these cells. In contrast, the secretion of IL‐12 and IL‐6 and the formation of il‐12 and il‐6 mRNA induced in bone marrow‐derived macrophages (BMDMs) by prolonged stimulation with TLR‐activating ligands that signal via myddosomes, which also requires the interaction of IRAK2 with TRAF6, were not increased but modestly reduced in HOIL‐1[C458S] BMDM. The decreased production of these cytokines correlated with reduced ubiquitylation of IRAK2. Our results establish that changes in HOIL‐1‐catalysed ester‐linked ubiquitylation can promote or reduce cytokine production depending on the ligand, receptor and immune cell and may be explained by differences in the ubiquitylation of IRAK2.

HOIP is an E3 ubiquitin ligase that catalyses the formation of Met1-linked ubiquitin (M1-Ub) oligomers [1,5], also called linear ubiquitin, which are required for the efficient activation of the canonical IjB kinase (IKK) complex in some cells ( [6], reviewed in Ref. [7]). The IKK complex has multiple functions in vivo, which include activation of the transcription factors nuclear factor j-light-chain-enhancer of activated B cells (NF-jB) and interferon regulatory factor 5 [8,9]. These proteins are required for the transcription of genes encoding pro-inflammatory cytokines, such as interleukin 6 (IL-6), IL-12 and TNF in macrophages [10,11].
Like HOIP, the HOIL-1 component of LUBAC is also a RING-Between-RING E3 ligase family member, but, until recently, the type of ubiquitin linkage that it generated was unknown. We found that HOIL-1 catalyses the attachment of ubiquitin to the hydroxyl side chains of serine and threonine residues in proteins, forming ester bonds [30], instead of the isopeptide bonds formed by linking ubiquitin to the e-amino group of lysine residues, as occurs in nearly all other E3 ubiquitin ligases.
Ubiquitin linked to proteins by ester bonds is cleaved rapidly and specifically by hydroxylamine. By exploiting sensitivity to hydroxylamine and bone marrow-derived macrophages (BMDMs) from knockin mice expressing the E3 ligase-inactive HOIL-1 [C458S] mutant, we identified several physiological substrates of HOIL-1, most notably IRAK1 and IRAK2. Interestingly some, but not all, of the K63/ M1-Ub hybrids attached covalently to IRAK1 and IRAK2 were initiated by HOIL-1-catalysed ester bonds [30]. Thus, the K63/M1 hybrid ubiquitin chains formed when myddosomes are activated are initiated in two different ways.
These unexpected findings raised the question of how ester-linked ubiquitin affects the production of inflammatory mediators. Previous studies aimed at understanding the physiological roles of HOIL-1employed mice in which the expression of HOIL-1 had been largely [6] or completely ablated [31,32]. However, every functional domain is lost in HOIL-1deficient cells, and not just its E3 ligase activity, so that the specific effects attributable to loss of just the E3 ligase activity could not be deduced. For example, the absence of HOIL-1 affects the stability of HOIP and drastically decreases its expression, explaining why HOIL-1 KO mice, like HOIP KO mice, display early embryonic lethality [33]. In contrast, mice expressing the E3 ligase-inactive HOIL-1[C458S] mutant are born at normal Mendelian frequencies, develop normally and are of normal size and weight up to 1 year of age when kept under relatively sterile conditions [30]. Here, we have exploited cytotoxic T cells and BMDMs from the HOIL-1[C458S] mice to identify specific roles of the HOIL-1 E3 ligase in regulating myddosomedependent signalling pathways, which has revealed that HOIL-1 has divergent roles in different immune signalling pathways.

IL-18 signalling and cytokine production is enhanced in cytotoxic T cells from E3 ligasedeficient HOIL-1[C458S] mice
IL-18 signals via myddosomes [34] to stimulate the adaptive immune response and the recruitment of neutrophils to sites of infection [35,36]. We investigated whether the HOIL-1 E3 ligase participates in IL-18 signalling in cytotoxic T cells where it triggers the production and secretion of interferon c (IFN-c) and granulocyte-macrophage colony-stimulating factor (GM-CSF). We found that the IL-18-dependent production of the mRNA encoding ifng and csf2 (the gene encoding GM-CSF) (Fig. 1A,B) and the secretion of IFN-c and GM-CSF (Fig. 1C,D) were increased in cytotoxic T cells from knock-in mice expressing the E3 ligase-inactive HOIL-1[C458S] mutant, suggesting that the E3 ligase activity of HOIL-1 restricts IL-18 signalling at or above the level of gene transcription.
The IL-18-stimulated activation of the canonical IKK complex and MAP kinases occurs rapidly and transiently returning to near basal levels after 2 h (Fig. 1G). Interestingly, we found that the IL-18dependent phosphorylation of the activation loops of IKKa and IKKb and the phosphorylation of p105/ NF-jB1, a physiological substrate of IKKb [41,42], were enhanced in cytotoxic T cells from HOIL-1 [C458S] mice. Additionally, the duration of phosphorylation of IKK and p105/NF-jB1, as well as the phosphorylation of the activation loops of p38a and JNKs (which activates these MAP kinase family members), was prolonged in cytotoxic T cells from HOIL-1 [C458S] mice (Fig. 1G). These findings are consistent with the increased production of the mRNA encoding ifng and csf2 observed after 1-2 h (Fig. 1A,B).

IRAK ubiquitylation during IL-18 signalling in cytotoxic T cells
The expression of IL-18 receptor 1 (IL-18R1), MyD88, TRAF6 and the three components of LUBAC was similar in cytotoxic T cells from WT and HOIL-1 [C458S] mice (Fig. 1H), indicating that the enhanced IL-18 signalling seen in HOIL-1[C458S] cells is not a consequence of an increase in the expression of these proteins. The minor, more slowly migrating HOIL-1 band present in extracts from WT cytotoxic not HOIL-1[C458S] T cells, is a monoubiquitylated species arising from HOIL-1-catalysed autoubiquitylation [30].
To investigate why IL-18-dependent signalling pathways might be enhanced and prolonged in cytotoxic T cells from HOIL-1[C458S] mice, we next studied ubiquitylation events that occur 'upstream' of the activation of IKKb and MAP kinases. The ubiquitylation of IRAK1 peaked after 20 min and then declined rapidly, while the IRAK2 ubiquitylation was sustained for up to 6 h ( Fig. 2A).
Strikingly, the size of the ubiquitin chains attached to IRAK1 and IRAK2 that formed during IL-18 signalling was larger and the extent of their ubiquitylation was higher in cytotoxic T cells from HOIL-1 [C458S] mice ( Fig. 2A), indicating that the HOIL-1catalysed formation of ester-linked ubiquitin restricts the size that IRAK-linked ubiquitin chains can attain.
The cleavage of ester-linked ubiquitin with hydroxylamine reduced the size of the ubiquitin chains attached to IRAK1 and IRAK2 in cytotoxic T cells from WT mice and reconverted some of the IRAK1 and IRAK2 to the fully deubiquitylated proteins (Fig. 2B). These observations imply that some but not all of the ubiquitin chains attached covalently to IRAK1 and IRAK2 are initiated by ester-linked monoubiquitylation, similar to observations made in BMDM [30]. In contrast, and as expected, the size of the ubiquitin chains attached to IRAK1 and IRAK2 was unaffected by hydroxylamine in T cells from HOIL-1[C458S] mice and no reconversion to the deubiquitylated proteins took place (Fig. 2B). This result establishes that, similar to our previous findings in BMDM [30], HOIL-1 is the E3 ligase that catalyses the ester-linked monoubiquitylation of IRAK1 and IRAK2 in cytotoxic T cells.
Otulin, a deubiquitylase that cleaves M1-Ub linkages exclusively [43,44], decreased the size of the ubiquitin chains attached to IRAK1 and IRAK2 in both WT and HOIL-1[C458S] T cells, but without any conversion to the monoubiquitylated or deubiquitylated species (Fig. 2C). These results indicate that, as previously found in other cells [5], the first ubiquitin molecule attached to IRAK1 and IRAK2 is initially elongated by another type of ubiquitin linkage, presumably a K63-Ub linkage, prior to the introduction of M1-Ub linkages to form the K63/M1-Ub hybrid The interaction of IRAK2 with TRAF6 is critical for IL-18 signalling in cytotoxic T cells The expression of IRAK1 was decreased drastically during prolonged IL-18 signalling, but the expression of IRAK2 did not and increased slightly (Fig. 2D), consistent with the transient ubiquitylation of IRAK1 and sustained ubiquitylation of IRAK2 ( Fig. 2A). Similar results have been observed previously in BMDM following stimulation with TLR-activating ligands, the rapid disappearance of IRAK1 presumably explaining why IRAK2 becomes rate limiting for cytokine production during prolonged TLR signalling [45,46]. We have reported that the TLR-dependent production of pro-inflammatory cytokines is greatly reduced in BMDM from knock-in mice expressing the IRAK2[E525A] mutant that is unable to interact with TRAF6 [46]. Similar to BMDM, we found that the IL-18-dependent secretion of IFN-c and GM-CSF was reduced by 77% and 72%, respectively, in cytotoxic T cells from IRAK2[E525A] mice (Fig. 2E). These results establish that the interaction of IRAK2 with TRAF6 has a critical role in IL-18-stimulated cytokine production in cytotoxic T cells and raised the question of whether the differences in the number, size and topology of the ubiquitin chains attached to IRAK2 in cytotoxic T cells from HOIL-1[C458S] mice underlies the enhanced production of IFN-c and GM-CSF in these cells.

The HOIL-1 E3 ligase enhances myddosomedependent cytokine secretion by TLR ligands in BMDM
To investigate whether the HOIL-1 E3 ligase has a similar role in BMDM to that observed in cytotoxic T cells, we stimulated BMDM with Pam3CSK4 (an activator of the TLR1/2 heterodimer) or R848 (an activator of the TLR7/8 heterodimer). Similar to IL-18, these ligands signal exclusively via myddosomes. Unexpectedly, and in contrast to IL-18 signalling in T cells, we found that the Pam3CSK4-or R848-stimulated secretion of several pro-inflammatory cytokines was not increased, but modestly reduced in BMDM from HOIL-1 BMDM only became statistically significant after stimulation for 8 h and was less than that seen in IL-6 and IL-12 secretion. We next investigated whether the TLR-dependent production of these cytokines might be regulated at the transcriptional level by measuring mRNA. Stimulation with Pam3CSK4 or R848 induced an increase in il6, il12 and tnf levels. In BMDM from HOIL-1 [C458S] mice, there was a statistically significant decrease in il6 mRNA 8 h after stimulation with Pam3CSK4. There also appeared to be a small reduction in il6 after stimulation with R848, which did not reach statistical significance (Fig. 3G,J). There was also a statistically significant reduction in il12 mRNA levels, at most time points between 2 and 12 h after stimulation with either Pam3CSK4 or R848 (Fig. 3H, Fig. 1. The HOIL-1 E3 ligase is a negative regulator of IL-18 signalling and cytokine production in CD8 + cytotoxic T cells (CTLs). (A, B) CTLs from WT and HOIL-1[C458S] mice were stimulated for the times indicated with IL-18 (20 ngÁmL À1 ). The increase in ifng mRNA encoding IFN-c and csf2 mRNA encoding GM-CSF is plotted as fold increase in mRNA relative to that determined in unstimulated cells. Error bars represent the mean AE SEM for experiments with 3-4 mice of each genotype. (C, D) As in A, except that plots show the concentrations of IFNc and GM-CSF secreted into the culture medium following stimulation. Error bars represent the mean AE SEM for experiments with three mice of each genotype. Similar results were obtained in three independent experiments. (E, F) As in C, except that CTLs were incubated for 1 h without (À) or with (+) the IKKb inhibitor BI605906 (1 lM), the p38 MAPK inhibitor BIRB0796 (1 lM) or the JNK inhibitor JNK-IN8 (10 lM), prior to stimulation for 8 h with IL-18 (20 ngÁmL À1 ). The IFN-c (E) and GM-CSF (F) secreted into the culture medium were then measured. The graphs show the results for three biological replicates of each genotype AE SEM. * denotes P < 0.05, ** P < 0.01 and *** denotes P < 0.001. Statistical significance was calculated using two-way ANOVA and Sidak's multiple comparisons test (A-D) or oneway ANOVA followed by Dunnett's multiple comparisons test (E,F). (G) As in A, except that after stimulation with IL-18, the CTLs were lysed and cell extracts (20 µg protein) were subjected to SDS/PAGE followed by immunoblotting with the indicated antibodies (p, phosphorylated). Similar results were obtained in three different experiments.  . It is well established that TNF production is regulated at both the translational and posttranslational levels [47,48], and so the reduction in TNF secretion in HOIL-1[C458S] BMDM may arise from a post-transcriptional role(s) of the HOIL-1 E3 ligase activity. To investigate possible reasons for the reduced TLR-dependent production of IL-6 and IL-12 in BMDM and why this differs from the increased cytokine production seen in IL-18 stimulated cytotoxic T cells, we next studied IRAK2 and its ubiquitylation. The expression of IRAK2 in BMDM was a little higher than in cytotoxic T cells (Fig. 4A), while the expression of IRAK1 was slightly lower (Fig. 4A). Stimulation of BMDM from either WT or HOIL-1 [C458S] mice with either Pam3CSK4 or R848 greatly increased the expression of IRAK2 (Fig. 4B,C), and induced the disappearance of IRAK1, as expected. The ubiquitylation of IRAK2 in WT BMDM was increased considerably during prolonged stimulation, presumably as a consequence of its increased expression over this period. IRAK1 ubiquitylation had almost disappeared by 4 h (Fig. 5A,B), consistent with its rate of degradation (Fig. 4B,C). Strikingly, and in contrast to WT BMDM, there was little increase in IRAK2 ubiquitylation in HOIL-1[C458S] BMDM during prolonged TLR ligation (Fig. 5C,D), suggesting that the increased ubiquitylation of IRAK2 during prolonged signalling in WT BMDM might be explained by increased formation of ester-linked ubiquitin chains during prolonged TLR ligation. Consistent with this notion, treatment with hydroxylamine caused a marked decrease in IRAK2 ubiquitylation between 2 and 8 h, especially during stimulation with Pam3CSK4 (Fig. 5E,F). The reduced ubiquitylation of IRAK2 in HOIL-1[C458S] BMDM from 2 to 8 h correlated with the decreased production of il6 and il12 mRNA and secretion of these cytokines.
Similar to IL-18 signalling in cytotoxic T cells, the size of the ubiquitin chains attached to IRAK1 and IRAK2 was increased in BMDM from the HOIL-1 [C458S] mice during the first hour of Pam3CSK4 or R848 stimulation (Fig. 5A,B) [30]. However, in contrast to IL-18 signalling in T cells, the extent of ubiquitylation was not enhanced and nor was the increased size of the ubiquitin chains attached to IRAK2 prolonged. After 4-8 h, the size of the ubiquitin chains attached to IRAK2 became similar in WT and HOIL-1[C458S] BMDM.
Despite the increased size of the ubiquitin chains attached to IRAK2 and IRAK1 in BMDM from HOIL-1[C458S] mice during the first hour of TLR ligation, no consistent increase was seen in the phosphorylation of p105/NF-jB1 or the rate of degradation of IjBa in these cells, which are readouts of IKKb activity. Moreover, the rate of reappearance of IjBa after 45-60 min, a readout of NF-jB-dependent gene transcription, was similar in BMDM from HOIL-1[C458S] and WT mice (Fig. 6A,B). The similar level of tnf mRNA 1 h after stimulating BMDM from WT and HOIL-1 [C458S] mice with Pam3CSK4 (Fig. 3I) or R848 (Fig. 3L) is consistent with this finding. There was also no significant difference in the activation of p38a MAP kinase or JNK1/2 in WT and HOIL-1[C458S] BMDM during the first hour of TLR signalling (Fig. 6A,B), implying that the activities of 'upstream' activators of these pathways, such as TAK1, were also unaffected.

LPS-induced cytokine production is similar in BMDM from WT and HOIL-1[C458S] mice
LPS interacts with TLR4, which is unique among TLRs in signalling via TRIF and myddosomes. TRIF-  (Fig. 7A,B). LPS-stimulated TNF secretion was similar in WT and HOIL-1[C458S] BMDM (Fig. 7C). Taken together, these observations indicate that HOIL-1 E3 ligase activity is not critical for the TRIFdependent secretion of these pro-inflammatory cytokines. Nevertheless, it was of interest to study the LPSstimulated ubiquitylation of IRAK1 and IRAK2. We found that, similar to stimulation by Pam3CSK4 and R848, the size of the ubiquitin chains attached to IRAK2 during the first hour of LPS stimulation was increased in BMDM from HOIL-1[C458S] mice, but the extent of ubiquitylation was not increased (Fig. 7D). The LPS-stimulated phosphorylation of p105/NF-jB1, the degradation of IjBa and the activation of MAP kinases over this period were also similar in BMDM from WT and HOIL-1[C458S] mice (Fig. 7E).
In contrast to Pam3CSK4 and R848, the LPSstimulated increase in the size of the ubiquitin chains attached to IRAK2 and the extent of ubiquitylation were both maintained for up to 8 h in BMDM from WT and HOIL-1[C458S] mice (Fig. 7D). However, the significance of these findings is unclear since TRIF, and not IRAK2, is rate limiting for LPS-induced cytokine production during prolonged TLR4 ligation in BMDM from HOIL-1[C458S] mice.

Discussion
Although the importance of ubiquitin chains in initiating innate immune signalling has been recognised for 20 years [19,49], only recently has the complexity of the ubiquitin oligomers that are formed begun to be fully appreciated. These chains, which can contain up to three ubiquitin linkage types, namely Lys63-linked, Met1-linked and ester bonds, are formed during signalling by IL-1 family members and ligands that activate TLRs and are attached covalently to proteins that include components of myddosomes, such as IRAK1 and IRAK2. The ubiquitin chains attached to IRAK1 and IRAK2 are not only initiated by HOIL-1catalysed ester bonds but also by initiated by isopeptide bonds formed by the action of 'conventional' E3 ligases, implying that there are at least two polyubiquitin chains attached to each of these proteins [30]. In principle, every protein-linked ubiquitin chain could operate as a distinct signalling node, so that the phosphorylation events widely used as readouts of innate immune signalling pathways may be averaging the outputs of multiple polyubiquitin-triggered events.
The present study has established that the HOIL-1catalysed formation of ester-linked ubiquitin has important roles in controlling Myddosome-dependent signalling processes, and revealed that their effects vary with the ligand, receptor and cell type. We found that the presence of ester-linked ubiquitin restricts the strength and duration of IL-18 signalling in cytotoxic T cells, because failure to produce these linkages enhanced both early signalling events and subsequent events leading to the production and secretion of IFNc and GM-CSF. We found that IRAK2 and its interaction with TRAF6 were critical for the production of IFN-c and GM-CSF during prolonged IL-18 signalling and that stimulation with IL-18 increased both the number and size of the ubiquitin chains attached to IRAK2 in cytotoxic T cells expressing the E3 ligaseinactive HOIL-1[C458S] mutant. The NEMO component of the canonical IKK complex binds with greater avidity to longer M1-Ub and K63-Ub oligomers than to shorter ubiquitin oligomers [25], and so the increased size of the ubiquitin chains attached to IRAK1 and IRAK2 might contribute to the enhanced and more prolonged activation of the IKK complex seen in HOIL-1 E3 ligase-deficient cytotoxic T cells. Similarly, stronger interactions between larger K63/ M1-Ub hybrids and the ubiquitin-binding subunits of TAK1 complexes, TAB2 and TAB3, might enhance TAK1 activation and account for the increased activation of p38 MAP kinase and JNKs. Taken together, our results suggest the following working hypothesis to account for the increased production of cytokines in cytotoxic T cells from HOIL-1 [C458S] mice during prolonged IL-18 signalling. The IRAK2-TRAF6 interaction activates the E3 ligase activity of TRAF6, enabling TRAF6 to produce K63linked ubiquitin oligomers that are attached to IRAK2 by isopeptide linkages, thereby activating TAK1 (in contrast to IRAK1, IRAK2 is a catalytically inactive pseudokinase, which is unable to phosphorylate and activate the Pellino family of E3 ligases-see Introduction). The K63-Ub chains attached to IRAK2 interact specifically with the Npl4 zinc finger domain of HOIP [5] recruiting LUBAC to the signalling complex and enabling the HOIP component of LUBAC to convert the K63-Ub oligomers attached to IRAK2 to K63/M1-Ub hybrids, thereby maintaining a low level of TAK1 and IKK activation needed to sustain cytokine production during prolonged IL-18 signalling. The HOIL-1 component of LUBAC monoubiquitylates a serine or threonine residue(s) on IRAK2 initiating the formation of distinct ester-linked ubiquitin chains. We speculate that the presence of both esterlinked and isopeptide-linked ubiquitin chains attached to IRAK1 and IRAK2 restricts the size that each type of ubiquitin chain can attain, perhaps due to competition between the chains for the limiting amounts of TRAF6 and LUBAC that are available. In cytotoxic T cells expressing the HOIL-1[C458S] mutant, esterlinked ubiquitin chains are not formed permitting the isopeptide-linked ubiquitin chain(s) to become larger, and enhancing and prolonging the activation of TAK1 and IKK to increase the production of cytokines.
Although HOIL-1 can initiate ubiquitin chain formation de novo by monoubiquitylating IRAK2 and other proteins, which are subsequently elongated and branched by other E3 ligases [5,30], it can also catalyse the formation of ubiquitin dimers in vitro in which the C terminus of one ubiquitin is linked by an ester bond to the hydroxyl side chain of Thr12 of another ubiquitin molecule [30,50]. If such linkages are formed in cells, which has yet to be established, then HOIL-1 may not only initiate ubiquitin chain formation but also incorporate ester linkages within the hybrid ubiquitin chains. The presence of such linkages could have positive or negative effects on the strength and duration of signalling depending on where, when and if these linkages are introduced. For example, Thr12linked ubiquitin dimers present within a hybrid polyubiquitin chain might help (or hinder) the recruitment of specific binding proteins that restrict the strength of signalling, such as ABIN1.
In contrast to cytotoxic T cells, we found that the TLR1/2-and TLR7/8-stimulated secretion of the proinflammatory cytokines IL-6, IL-12 and TNF was not increased but modestly reduced in BMDM from HOIL-1 [C458S] mice, and this correlated with modestly reduced levels of the mRNAs encoding IL-6 and IL-12. Interestingly, the decreased secretion of these pro-inflammatory cytokines also correlated with a marked decrease in the ubiquitylation of IRAK2 during prolonged TLR signalling in HOIL-1[C458S] BMDM. Since IRAK2 and its interaction with TRAF6 are critical for the production of these pro-inflammatory cytokines [45,46], we speculate that it is the decreased ubiquitylation of IRAK2 during prolonged TLR signalling that underlies the reduced production of IL-6 and IL-12 in BMDM from HOIL-1 [C458S] mice. However, the possibility cannot be excluded that the reduced ubiquitylation of other proteins modified by HOIL-1-catalysed ester-linked ubiquitylation, such as MyD88 [30], also contributes to the decreased cytokine production. An interesting possibility that is worth exploring in future work is whether the loss of ester-linked ubiquitylation permits the serine/threonine residues to which they were attached to undergo another post-translational modification, such as phosphorylation or O-GlcNacylation, that affects signalling and cytokine production.
Although we speculate that the increased size of the ubiquitin chains attached to IRAK1/2 may underlie the enhanced and more sustained IL-18 signalling in cytotoxic T cells from HOIL-1[C458S] mice, a significant increase in TLR signalling was not observed in BMDM from HOIL-1[C458S] mice during the first hour of signalling (Fig. 6), even though the size of the ubiquitin chains attached to these IRAKs was enhanced similarly to that seen in cytotoxic T cells (Fig. 5A). However, in contrast to cytotoxic T cells, the extent of ubiquitylation of IRAK1/2 was not increased. It may therefore be that it is the increased extent of IRAK1/2 ubiquitylation, rather than the increased size of the ubiquitin chains, that is responsible for enhancing IL-18 signalling in cytotoxic T cells.
Myddosomes are complexes between the liganded receptor, MyD88 and IRAK family members. Moreover, the relative amounts of different IRAK family members can vary from cell to cell and with the stimuli to which they have been exposed. This plasticity in the composition of myddosomes adds to the difficulty in elucidating the molecular basis for the diverse effects on cytokine production caused by the loss of ester-linked ubiquitylation in different immune cells. The present study has not only highlighted the potential importance of IRAK2 ubiquitylation but also begun to explain why loss of the E3 ligase activity of HOIL-1 can cause immunodeficiency and autoinflammation [51,52].

Stimulation of bone marrow-derived macrophages (BMDMs)
BMDMs were obtained by differentiating bone marrow from the femur and tibia using L929 preconditioned medium [46]. Adherent BMDMs were replated in fresh culture medium in 24-well (0.25 9 10 6 cells) or 6-well (1 9 10 6 cells) tissue culture plates or 15-cm tissue culture plates (10 9 10 6 cells), and stimulated with TLR ligands for up to 12 h at the concentrations specified in figure legends.

Dephosphorylation and deubiquitylation of IRAK1 and IRAK2 in cell extracts
Cells were stimulated, washed and lysed as described above except that EDTA and the protein phosphatase inhibitors sodium fluoride, sodium pyrophosphate and sodium orthovanadate were omitted, and the deubiquitylase inhibitor iodoacetamide was also excluded. Cell extract (50 µg protein) was incubated in a total volume of 50 µL containing 50 mM HEPES pH 7.5, 100 mM NaCl, 2 mM DTT, 0.01% (wt/vol) Brij-35, 1 mM MnCl 2 , kPPase (100 units per reaction) and 1 µM USP2. After incubation for 60 min at 37°C, the samples were denatured in SDS plus 2.5% (vol/ vol) 2-mercaptoethanol, subjected to SDS/PAGE, transferred to PVDF membrane and immunoblotted as described [30,54].

Capture of ubiquitylated proteins, treatment with hydroxylamine and deubiquitylases, and immunoblotting
Ubiquitylated proteins containing M1-Ub oligomers were captured from cell extracts using Halo-NEMO beads and treated with hydroxylamine or deubiquitylases [30], then

Cytokine measurements
After stimulation with ligands, the cell culture medium was removed, clarified by centrifugation (10 min at 14 000 g ) frozen in liquid nitrogen and stored in aliquots at À80°C. The levels of IL-6, IL-10 and IL-12 in the cell culture medium were measured using the Bio-Plex Pro Assay multiplex system from Bio-Rad Laboratories (Hercules, CA, USA) and the level of TNF by ELISA (R&D Systems) according to the supplier's instructions. When TNF was measured using the Bio-Plex assay, an apparent decrease in the level of TNF was measured in the culture medium between 8 and 12 h of stimulation. The reason is unclear, but it is possible that the antibody used in the Bio-Plex assay does not recognise a modified species of TNF that is formed in the culture medium during prolonged stimulation with TLR ligands. Such a situation has been described for IL-33, where oxidation of cysteine residues prevents its detection by some antibodies, but not by others [55]. The cell culture medium was diluted to ensure that the cytokine measurement was made within the linear range of the standard curve.

Isolation of mRNA and quantitative real-time PCR
RNA was isolated from cytotoxic T cells using the MicroElute Total RNA Kit (Omega Bio-Tek, Norcross, GA, USA) following the manufacturers' instructions and reverse transcription of RNA to cDNA was performed using the iScript cDNA Synthesis Kit (Bio-Rad Laboratories) using 300 ng total RNA. Quantitative RT-PCR was performed using Sso-Fast EvaGreen Supermix (Bio-Rad Laboratories). The primers for measuring mouse ifng mRNA were as follows: forward 5 0 -AGCCAAGACTGTGATTGC-3 0 , reverse 5 0 -TTATTGGTCAGTGAAGTAAAGG-3 0 , and for csf2 were as follows: forward 5 0 -CTCACCCATCACTGTCACCC-3 0 , reverse: 5 0 -TGAAATTGCCCCGTAGACCC-3 0 . The primers for measuring il6, il12 and tnf mRNA were described previously [46]. Normalisation and quantitation were performed using gapdh RNA and the DDCt method.

Statistical analysis of the data
Graphs were generated using GRAPHPAD PRISM 9 (San Diego, CA, USA), Microsoft PowerPoint (WA, USA) and Adobe Illustrator 2021 (CA, USA). Statistical analysis was done in GRAPHPAD PRISM 9 using one-way ANOVA (analysis of variance) followed by Dunnett's multiple comparisons test or two-way ANOVA followed by Sidak's multiple comparisons test.