Axon degeneration induces glial responses through Draper-TRAF4-JNK signalling

Draper/Ced-1/MEGF-10 is an engulfment receptor that promotes clearance of cellular debris in C. elegans, Drosophila and mammals. Draper signals through an evolutionarily conserved Src family kinase cascade to drive cytoskeletal rearrangements and target engulfment through Rac1. Glia also alter gene expression patterns in response to axonal injury but pathways mediating these responses are poorly defined. We show Draper is cell autonomously required for glial activation of transcriptional reporters after axonal injury. We identify TNF receptor associated factor 4 (TRAF4) as a novel Draper binding partner that is required for reporter activation and phagocytosis of axonal debris. TRAF4 and misshapen (MSN) act downstream of Draper to activate c-Jun N-terminal kinase (JNK) signalling in glia, resulting in changes in transcriptional reporters that are dependent on Drosophila AP-1 (dAP-1) and STAT92E. Our data argue injury signals received by Draper at the membrane are important regulators of downstream transcriptional responses in reactive glia.

G lial cells are exquisitely sensitive to perturbations of nervous system physiology [1][2][3][4][5][6][7] . In response to neural injury they coordinate a sophisticated cascade of reactive events, including engagement of phagocytic programmes to clear cellular debris, and secretion growth factors to recruit additional immune cells or promote neuronal survival and repair 8 . Reactive glial responses are driven in important ways by changes in glial gene expression following trauma. Strong upregulation of expression of GFAP and S100b are among the earliest detectable changes in mammalian neural tissues after brain injury, and similar upregulation occurs in chronic neurodegenerative disease 9,10 . In peripheral nerves early events include the activation of JNK signalling, which is required for efficient Schwann cell de-differentiation and initiation of regenerative responses 11 . Initiating events in glial reactive responses are presumably driven by 'injury signals' released by damaged cells that signal to transmembrane receptors on glia, but whether and how cellular debris activates glial transcriptional responses remains mechanistically unclear.
Drosophila glia rapidly respond at the transcriptional level to axotomy, which argues reactive glial responses are an ancient feature of the metazoan nervous system 12 . Transcriptional reporters for the sole Drosophila STAT molecule, STAT92E, are induced in glial cells within hours after local axons are severed 6 . Axotomy also activates the glial JNK signalling cascade and ultimately transcriptional reporters for the Drosophila AP-1 complex (dAP-1) 7 . Activation of STAT92E-dependent transcriptional responses has been shown to require the engulfment receptor Draper 6 , suggesting STAT92E lies genetically downstream of Draper, although how axon degeneration is mechanistically linked to STAT92E-or dAP-1-mediated transcriptional changes is not known.

Results
Axon degeneration activates dAP-1 reporters in glia. To understand how axonal injury leads to glial transcriptional changes we sought to define the mechanisms regulating activation of the dAP-1 complex in glia after axotomy. We used the dAP-1 transcriptional reporters containing multiple tetradecanoylphorbol acetate response elements (TRE, that is the dAP-1 binding sequence) upstream of either eGFP (TRE-eGFP) or RFP (TRE-RFP) 13 . Axon injury was induced by surgical removal of adult Drosophila third antennal segments, which severs B90% of olfactory receptor neurons (ORNs) projecting into the Drosophila antennal lobe and axon fragmentation begins within B12 h by activation of Wallerian degeneration 2 . Prior to axonal injury TRE-RFP was not detectable at significant levels in control animals in any part of the adult brain. However, one day after axonal injury (1 dpant, 1 day post antennal ablation), we observed strong activation of TRE-RFP throughout the brain (Supplementary Fig. 1), as previously observed for TRE-eGFP (ref. 7), along with glial hypertrophy and increased Draper levels in glia adjacent to the injury site. To determine whether activation of TRE-RFP represented a glial response to axon degeneration, we blocked ORN axon degeneration using highwire (hiw) mutants 14 . Glial cells in hiw DN mutant animals exhibited no morphological changes after axonal injury, TRE-RFP activation was not observed in any part of the brain, and Draper levels adjacent to the injury site remained similar to the uninjured controls (Fig. 1a). Hiw is not expressed at detectable levels in glia but is prominent in neurons (Fig. 1b), suggesting the lack of glial activation after axotomy in hiw mutants resulted from blockade of axon degeneration, rather than a cell autonomous role for hiw in glial activation. Consistent with this notion, pan-neuronal expression of Highwire in hiw DN mutant background rescued glial response to axon degeneration ( Supplementary Fig. 2b) while glial expression of a hiw cDNA in the hiw DN mutant background did not lead to activation of the TRE-RFP reporter ( Supplementary Fig. 2a). Activation of the Wallerian degeneration signalling pathway is therefore essential for brain-wide induction of dAP-1 transcriptional reporters in glia after nerve injury. Furthermore, these data indicate glial cells respond specifically to signals generated by degenerating axons, rather than other alterations in axons caused by axotomy (for example loss of neuronal activity).
Glia activate dAP-1 reporters after injury through Draper. The engulfment receptor Draper activates glial phagocytic function during Wallerian degeneration 2 . We found draper D5 null mutant animals also exhibited a striking loss of TRE-eGFP induction 1 day after axon injury compared to controls (Fig. 1c,d), indicating that Draper is also required for dAP-1 reporter activation in response to axonal degeneration. ORN axon terminals only project to the antennal lobe. The lack of TRE-eGFP activation in draper mutants observed throughout the adult brain implied that Draper was required for TRE-eGFP activation even in glia at sites distant from degenerating axons in the antennal lobe. To characterize the autonomy of Draper function in activation of dAP-1 transcriptional reporters in brain glia, we generated draper D5 mutant clones using the mosaic analysis with a repressible cell marker (MARCM) approach 15 , severed antennal ORN axons, and assayed TRE-RFP activation 1 day after axotomy. Based on morphological analysis of clones we determined that activation of TRE-RFP occurred in ensheathing (Fig. 1e) and cortex glia (Fig. 1f), but not in astrocytes ( Supplementary Fig. 3), consistent with a previous study demonstrating Draper is expressed solely in these glial subtypes in the Drosophila adult brain 16 . While control clones in each subtype of glial cell showed strong activation of TRE-RFP in response to axonal injury, draper D5 mutant clones did not activate the TRE-RFP reporter (Fig. 1e,f). This was even true when draper D5 mutant clones were made in cortex glia, which are not in direct contact with severed axons (for example Fig. 1f). This was surprising in light of the fact that Draper/Ced-1 is considered a recognition receptor for 'eat me' cues on the surface of engulfment targets such as axonal debris 1 , which we presumed would require cell-cell contact. This observation suggests that Draper may be activated by a soluble ligand released by severed ORNs, or alternatively that Draper-dependent glial-glial signalling may occur in cortex glia upstream of dAP-1 reporter activation, perhaps similar to MEGF10 signalling in starburst amacrine cells during tiling where MEGF10 appears to act genetically as both a receptor and ligand 17 .

TRAF4 signals downstream of Draper after axonal injury.
Based on the potent regulation of TRE-RFP by Draper, we sought to determine how Draper activation might lead to its activation through dAP-1. We screened a large collection of RNAi lines for novel signalling molecules and assayed for roles in glial clearance of degenerating axons. Interestingly, we found that pan-glial expression of UAS-traf4 IR (ref. 18) resulted in strong suppression of glial clearance of axonal debris, compared to controls (Fig. 2a,b, Supplementary Fig. 4c,d). TRAF4 is known to act upstream in activation of the JNK signalling pathway through association with the MAP kinase kinase kinase kinase (MAP4K) misshapen (MSN) (ref. 19), which then phosphorylates the downstream MAP3K Slipper and ultimately drives dAP-1 activation of transcriptional targets 20 . A role for TRAF4 in glial engulfment was intriguing given that we previously implicated a number of downstream components of the JNK signalling pathway in glial engulfment of axonal debris including dAP-1 (ref. 7), but how this pathway might be activated remained unclear.
Pan glial expression of a dominant negative version of MSN (msn DN ) 21 strongly suppressed glial engulfment of axonal debris 5 days after axotomy (Fig. 2c,d). Consistent with a role for TRAF4 and MSN in activation of the JNK signalling pathway we found that glial expression of traf4 IR (Fig. 2e,f) or msn DN (Fig. 2g, and which we previously reported for downstream components of the JNK signalling pathways. We further noted that in traf4 IR animals Draper expression was reduced B50% even before axon injury (Fig. 2i,j). This phenotype is similar to glial stat92e RNAi animals, where Draper levels are reduced in uninjured animals, and engulfment defects can be rescued by overexpression of a draper cDNA 6 . In contrast, we found defects in glial responses to axonal degeneration in the traf4 IR background did not result from insufficient Draper levels: re-expression of a functional Draper-I transcript ( Supplementary Fig. 4a,b) in traf4 IR animals failed to rescue axonal debris clearance defects ( Supplementary Fig. 4c,d).
Together these data support a model whereby TRAF4 signals downstream of Draper to promote target engulfment and glial transcriptional reporter activation mediated by STAT92E and dAP-1.
Based on our observation that TRAF4 and MSN regulated injury-induced changes in Draper levels and transcriptional reporters in glia, we sought to define signalling interactions between TRAF4 and MSN, the JNK signalling pathway, and STAT92E. STAT92E activity after axotomy was assayed using the degradable GFP reporter 10xSTAT92E-dGFP (ref. 22), which is transiently activated in glia 1 day after axotomy 6 . In control uninjured animals, GFP expression was absent in brains, but strongly activated in glia surrounding the antennal lobe 1 day after antennal ORN axotomy (Fig. 2l). Glial expression of traf4 IR suppressed this increase of GFP expression after axon injury (Fig. 2l,m), indicating that TRAF4 is upstream of the injuryinduced activation of 10xSTAT92E-dGFP. Similarly, blockade of the JNK signalling pathway with basket RNAi (Drosophila JNK), or overexpression of the JNK inhibitor Puckered eliminated the injury-induced upregulation of 10xSTAT92E-dGFP (Fig. 2l,m). Finally, knockdown of Kayak, a component of the dAP-1 transcriptional complex, also led to a significant decrease in 10xSTAT92E-dGFP activation (Fig. 2l,m). Thus STAT92E is genetically downstream of dAP-1 in the context of 10xSTAT92E-dGFP reporter activation. Taken together, these genetic data argue that glia activate these transcriptional reporters after axonal injury through a TRAF4-JNK pathway-dAP-1-STAT92E signalling cascade that lies downstream of Draper.
TRAF4 binds the Draper intracellular domain. The mechanism by which Draper activates JNK signalling and dAP-1-mediated transcriptional reporter activation has not been defined. The only molecule known to physically associate with Draper to promote downstream signalling events is the Src family kinase Shark, which binds an ITAM in the Draper intracellular domain after phosphorylation by Src42a (ref. 4). We found that depletion of Shark from all glia by pan-glial expression of UAS-shark RNAi (using the repo-Gal4 driver) blocked activation of TRE-eGFP 1 day after axotomy (Supplementary Fig. 5a,b), and similar RNAi-mediated knockdown of rac1 or src42a also significantly decreased injury-induced activation of TRE-eGFP ( Supplementary Fig. 5c,d). We therefore conclude that dAP-1 transcriptional reporters are genetically downstream of Draper/Shark signalling interactions.
Given that TRAF4 is an adaptor protein, we speculated that TRAF4 might interact with Draper and thereby directly couple Draper to activation of JNK kinase signalling. TRAF4/ MSN have well-defined roles in the activation of Slipper (ref. 20), and in turn the Drosophila JNK signalling pathway. We therefore performed co-immunoprecipitation (co-IP) experiments using Drosophila S2 cell lysates with a Myc-tagged TRAF4 (TRAF4-myc) and HA-tagged Draper I (Draper-I-HA). We found that pull down of Draper-I-HA led to the coprecipitation of TRAF4-myc, indicating that TRAF4 can physically associate with the Draper signalling complex (Fig. 3a). To identify the region through which TRAF4 associates with Draper we assayed the ability of epitope-tagged versions of the N-terminal TRAF4, which is enriched with zinc-finger domains (ZF, aa 1-927) or the C-terminal TRAF domain (TD, aa 928-1458) 19 to associate with Draper. We found that the N-terminal TRAF4 domain co-immunoprecipitated with Draper I-HA, while the C-terminal TRAF domain, which is known to bind MSN (ref. 19), was not detected in the same assay (Fig. 3b,c). We therefore conclude that Draper is bound by the N-terminal domain of TRAF.
To identify the region of Draper-I to which TRAF4 bound, we next made serial deletions in the intracellular domain (Fig. 3d), and performed co-IP with full-length TRAF4-myc. We found that aa 826-894 were essential for Draper-I-TRAF4 interactions (Fig. 3e, D826-894), while all other regions were dispensable. We note this includes the evolutionarily conserved NPXY motif that is likely bound by the PTB domain binding protein dCED-6 (ref. 23), which is consistent with the observation that dCED-6 is not required for activation of 10xSTAT92E-dGFP (ref. 6) or JNK signalling ( Supplementary  Fig. 6) in glia after axotomy. Further analysis of this subdomain revealed that aa 826-854 and 855-594 were both able to bind TRAF4 (Fig. 3d,f), suggesting there may be multiple TRAF4 binding sites in the Draper intracellular domain. The NPXY motif is not absolutely required for TRAF4 binding, as TRAF4 associated strongly with the D855-894 construct lacking this domain (Fig. 3f). Finally, given that both Src42A and Shark are required for activation of TRE-eGFP in response to injury, we assayed whether TRAF4 and Shark binding were reciprocally interdependent. Phosphorylation of Y949 in ITAM, which is Control ensheathing glial clones ( þ / þ ) became hypertrophic and exhibited strong TRE reporter activation (red) 1 day after antennal ablation, while draper null (drpr D5 ) clones did not respond to axotomy (that is lack of membrane hypertrophy, and activation of TRE reporter) while neighbouring control cells were activated. þ / þ : TRE-RFP-16, UAS-mCD8::GFP/repo-FLP 6-2 ; FRT2A, FRT82B/tub-Gal80, FRT2A, repo-Gal4. drpr D5 : TRE-RFP-16, UAS-mCD8:: GFP/repo-FLP 6-2 ; drpr D5 , FRT2A/tub-Gal80, FRT2A, repo-Gal4. Representative images were shown from four animals. Scale bar ¼ 10 mm. (f) TRE-RFP reporter activity in cortex glial MARCM clones (dotted line, green) 1 day after antennal ablation. Similar to ensheathing glia, cortex glial clones that lacked Draper (drpr D5 ) failed to activate TRE reporter. Genotypes as in (e). Representative images were shown from four animals. Scale bar ¼ 10 mm. mediated by Src42A and is bound by Shark (ref. 4), was also dispensable for TRAF4 to associate with Draper (Fig. 3g). Reciprocally, elimination of the TRAF4 interaction domain in Draper I (D826-894) did not alter Shark association with Draper ( Supplementary Fig. 7). Taken together these data indicate that TRAF4 physically associates with the Draper intracellular domain in the region of the NPXY motif, and this interaction occurs in a Shark-independent fashion. These data argue Draper activates the JNK signalling pathway and dAP-1-and STAT92E-dependent transcriptional reporters in

Discussion
Activation of JNK/AP-1 and STAT transcriptional activity during reactive glial responses is an evolutionarily conserved feature of Drosophila and mammalian glia 24,25 . However, transmembrane receptors on glia that receive extracellular injury signals and in turn activate AP-1 or STAT-dependent transcriptional changes have not been identified. Our work supports a model whereby the engulfment receptor Draper is activated in response to axonal injury, which in turn promotes JNK signalling through TRAF4 and MSN to alter glial transcription via dAP-1 and STAT92E. Our reliance on well-characterized transcriptional reporters for dAP-1 and STAT92E supports this conclusion, but we note that we have not formally shown transcriptional activation of a downstream target gene, primarily because none have been identified. Axon degeneration is required to activate glial cells after injury. Blockade of axon degeneration using hiw mutations fully suppresses glial responses to axon injury, including morphological changes and activation of transcriptional reporters, indicating that axonal self-destruction leads to the production of cues that activate glial responses through Draper. Interestingly, based on the activation of glial transcriptional reporters (that is TRE-RFP) even at a distance from the injury site (Fig. 1f), it appears that glia can be activated by degenerating axons even when not in direct contact with severed axons. This observation suggests that axonal 'injury signals' that stimulate Draper signalling may be soluble, or, alternatively that glial-glial signalling may propagate through the glial network in a Draper-dependent manner. Our identification of Draper/TRAF4 interactions as a mechanism for JNK activation raises the intriguing possibility that MEGF10 might play a conserved role in mammalian reactive astrocytes to activate AP-1 or STAT signalling, or other contexts. Indeed, MEGF10 and several other engulfment genes are expressed in astrocytes and MEGF10 is required for developmental synapse pruning 26 .
We propose a simple model where axon self-destruction leads to injury signals that activate Draper, which in turn drives two primary downstream signalling events (Fig. 3h). In the first, Draper promotes cytoskeletal rearrangements essential for phagocytic activity through Src42a, Shark and Traf4 in direct association with Draper, and downstream GEF complexes such as Crk-II/Mbc/dCed-12 and DRK/DOS/SOS (refs 5,27). These signalling pathways mediate membrane extension to, and internalization of engulfment targets. At the same time we propose that Draper activation leads to glial transcriptional responses to injury, including upregulation of dAP-1 and STAT92E transcriptional targets (ref. 12), although these need to be identified and validated. Consistent with the draper gene being a primary transcriptional target of this cascade, re-expression of Draper in glia is sufficient to rescue engulfment defects after knockdown of JNK, dAP-1 and STAT92E (refs 6,7). However, re-expression of Draper is not sufficient to rescue engulfment defects in animals lacking glial Src42a or TRAF4 (Supplementary Fig. 4). The simplest interpretation of these data is that Draper, TRAF4 and Shark function upstream in the signalling pathway and are required for both internalization of targets and activation of transcriptional responses. This model is supported by our finding that TRAF4 and Shark directly bind Draper. Depletion of TRAF4 or STAT92E from glia alters basal levels of Draper, while, surprisingly, RNAi mediates suppression of dJNK or dAP-1 does not. This may indicate additional signalling pathways promote cross-talk between TRAF4 and STAT92e during transcriptional regulation of basal levels of draper, which is independent of dJNK.
Draper/MEGF10-mediated signalling through TRAF4 may also be engaged during developmental synapse elimination or reactive gliosis in mammals, but this awaits further exploration. Draper/MEGF10 has recently been shown to have potentially non-phagocytic signalling roles including activation of autophagy in Drosophila salivary glands 28  a-tubulin was used as loading control. Representative images were shown from three repeats. Control: repo-Gal4/ þ . Repo4traf4 IR : UAS-traf4 IR / þ ; repo-Gal4/ þ . (j) Quantification of (i). n ¼ 3. Unpaired t-test, 2-tailed. (k) Higher magnification images of dashed squares in (g) showed the lack of Draper increase 1 day after antennal ablation in msn DN animals. (l) STAT92E-mediated dGFP (degradable GFP) expression was induced in glia 1 day after antennal ablation in control animals (10xSTAT92E-dGFP/ þ ; repo-Gal4/ þ ), but blocked in animals where JNK signalling cascade was interrupted by glial expression of traf4 IR (10xSTAT92E-dGFP/UAS-traf4 IR ; repo-Gal4/ þ ), basket RNAi (10xSTAT92E-dGFP/UAS-basket RNAi#34138 ; repo-Gal4/ þ ) and puckered WT (10xSTAT92E-dGFP/ þ ; repo-Gal4/UAS-puckered), or kayak RNAi (repo4kayak RNAi :10xSTAT92E-dGFP/ þ ; repo-Gal4/UAS-kayak RNAi#6212 ). (m) Quantification of the increase of dGFP immunoreactivity in (l). Repo4traf4 IR : n ¼ 15 each. Repo4basket RNAi , repo4puckered WT and repo4kayak RNAi : n ¼ 25 each. Unpaired t-test for each group, 2-tailed. Injury protocol and adult brain dissection. Drosophila cultures were kept at 25°C except for the experiments involving UAS-msn DN and UAS-src42A RNAi#26019 , due to the lethality they caused when overexpressed using repo-Gal4. Temperaturesensitive GAL80 (Gal80 ts ) 38 was used to inhibit the activity of GAL4 at 18°C before fly eclosion and then deactivated at 29°C for at least 5 days before injury. Flies continued to grow at 29°C until the day of dissection. Maxillary palp and the third antennal segment ablation were performed using a pair of fine forceps under dissection microscope. For immunocytochemistry, heads were first fixed and permeabilized in 4% formaldehyde/0.1% Triton X-100/PBS for 20 min at room temperature (R.T.). After washed with PBST (0.1% Triton X-100/PBS) for five times, brains were dissected in PBST and post-fixed with 4% formaldehyde/0.1% Triton X-100/PBS for another 15 min. Fixed brains were then washed for five times with PBST, blocked in 0.1% Triton X-100/5% BSA/PBS for 30 min at R.T., and incubated with primary antibodies in PBST at 4°C overnight or two nights for anti-Draper antibody. The next day, the brains were washed five times with PBST and then incubated with secondary antibodies in PBST for 2 h at R.T. or 4 h for Draper staining. After washed with PBST for five times, the brains were then mounted with VECTASHIELD Mounting Medium (Vector Labs) and stored at 4°C before imaging within 2 weeks. For western blot analysis, heads from adult flies aged at 25°C for at least 5 days after eclosion were isolated and immediately dissected in chilled PBS. Only the central brains (without the optic lobes) were collected and prepared for western blot analysis. Equal number of males and females were used for each experiment. Dissected brains were spinned down at the highest speed for 1 min and then lysed in 1X Laemmli lysis buffer (2 ml per brain) on ice for 20 min. The mixture was ground for 30 s on ice and centrifuged at the highest speed for 10 min at 4°C. Only the supernatants were collected for SDS-PAGE analysis and western blotting. Images have been cropped for presentation in Fig. 2i. Full-size images are presented in Supplementary Fig. 8.
Confocal microcopy and image analysis. Confocal microscopy settings were always kept constant throughout the same set of experiments. Zeiss LSM5 Pascal confocal microscope was used for all axonal debris clearance, TRE-eGFP reporter (except for UAS-msn DN ), and STAT92E reporter assays with Zeiss Plan-Apochromat Â 63/1.4NA oil objective lens. Glial MARCM clones, glial activation in hiw DN animals and TRE-eGFP reporter activity in msn DN animals were analysed using Zeiss Axio Imager.M2 spinning-disk confocal microscope with Zeiss C-Apochromat Â 40/1.2NA water objective lens. For quantification of the fluorescence intensity, the centre single z-section was identified from each image stack of the relevance glomerulus or central adult brain, and pixel intensity was measured using ImageJ (National Institute of Health Schneider 2 cell culture and co-immunoprecipitation. Drosophila S2 cells were cultured and maintained in 10% HyClone Fetal Bovine Serine/SFX-Insect Media (Fisher Scientific) with 1% penicillin-streptomycin solution (Sigma-Aldrich) at 25°C. For immunoprecipitation of Draper I-HA, 2 Â 10 6 S2 cells were seeded per well in a six-well plate the day before DNA transfection, where 100 ng of each appropriate DNA vector were mixed and incubated with Effectene Transfection Reagents (QIAGEN), as described in manufacturer's instruction. Two days after transfection, cells were lysed in our modified NP-40 lysis buffer (1% NP-40, 10 mM Tris-HCl pH 7.5, 50 mM NaCl, 30 mM Na 4 P 2 O 7 , 50 mM NaF, 100 mM Na 3 VO 4 , 5 mM ZnCl 2 and protease inhibitor cocktail purchased from Roche) 12 . Rat anti-HA monoclonal antibody (0.3 mg per sample) was incubated with Protein G beads (Sigma-Aldrich) at 4°C for 3 h before incubated with pre-cleared cell lysates at 4°C overnight. After washed at least eight times to reduce non-specific binding, proteins on beads were eluted in SDS loading buffer (60 mM Tris pH 6.8, 10% glycerol, 2% SDS, 1% b-mercaptoethanol and 0.01% bromophenol blue) and boiled at 100°C for 5 min, following a 2-min centrifugation at 16,000 g. Supernatants were loaded into 4-15% gradient Tris-HCl ReadyGel (Bio-Rad). SeeBlue Pre-stained Protein Standard (Life Technologies) was used (10 ml per well) to determine protein molecular weight. Samples were transferred to nitrocellulose membrane (Bio-Rad) after electrophoresis. Before incubated with corresponding primary antibodies at 4°C overnight, membrane was blocked in 5% milk/0.01% Tween-20/PBS at R.T. for 1 h. Membrane was then washed with 0.01% Tween-20/ PBS two times in 30 min and then probed with corresponding secondary antibodies at R.T. for 1 h, followed by another two washes in 30 min. Immunoreactivity was detected using Clarity Western ECL Substrate (Bio-Rad) or ECL Prime Western Blotting Detection Reagent (Amersham) for TRAF4-myc after co-IP. Image acquisition and protein quantification were conducted using ChemiDoc TM MP system (Bio-Rad) and signal saturation was avoided during acquisition. Images have been cropped for presentation in Fig. 3a,c,e-g. Full-size images are presented in Supplementary Fig. 8. All results showed were representative blots after at least three repeats.
Statistical analysis. All error bars represent s.e.m. Unpaired Student's t-test (2-tailed), one-way ANOVA, and post-hoc analyses were carried out using GraphPad Prism 6 (GraphPad Software). Data distribution was assumed to be normal, although this was not formally tested. No statistical methods were used to predetermine sample sizes, but our sample sizes are similar to those reported in previous publications. Also, data collection and analysis were not performed blind owing to the conditions of the experiments. Data were not collected and processed randomly. Animals were assigned to the various experimental groups on the basis of genotype.
Data availability. The authors declare that the experimental results supporting the findings are included in the article and the Supplementary Information or are available upon request.