Cell Fate Decisions Regulated by K63 Ubiquitination of Tumor Necrosis Factor Receptor 1

Abstract

Signaling by tumor necrosis factor (TNF) receptor 1 (TNF-R1), a prototypic member of the death receptor family, mediates pleiotropic biological outcomes ranging from inflammation and cell proliferation to cell death. Although many elements of specific signaling pathways have been identified, the main question of how these selective cell fate decisions are regulated is still unresolved. Here we identified TNF-induced K63 ubiquitination of TNF-R1 mediated by the ubiquitin ligase RNF8 as an early molecular checkpoint in the regulation of the decision between cell death and survival. Downmodulation of RNF8 prevented the ubiquitination of TNF-R1, blocked the internalization of the receptor, prevented the recruitment of the death-inducing signaling complex and the activation of caspase-8 and caspase-3/7, and reduced apoptotic cell death. Conversely, recruitment of the adaptor proteins TRADD, TRAF2, and RIP1 to TNF-R1, as well as activation of NF-κB, was unimpeded and cell growth and proliferation were significantly enhanced in RNF8-deficient cells. Thus, K63 ubiquitination of TNF-R1 can be sensed as a new level of regulation of TNF-R1 signaling at the earliest stage after ligand binding.

INTRODUCTION

The cytokine tumor necrosis factor alpha (TNF-α) is involved in a variety of cellular processes, such as inflammation, differentiation, control of cell proliferation, and initiation of apoptosis. TNF is known to bind to two receptors of the TNF receptor superfamily, TNF receptor 1 (TNF-R1) and TNF-R2. TNF-R1 is a member of the death receptor subgroup of this superfamily (1). The death receptors all have a death domain (DD) in the C-terminal tail that is necessary for activation of apoptosis. Selective recruitment of adaptor proteins to TNF-R1 decides whether nonapoptotic signaling pathways or cell death-inducing pathways will be initiated. Complex I is formed at the TNF-R1 DD by recruitment of TRADD, RIP1, TRAF2, and c-IAP1 (2). In the model of Micheau and Tschopp, induction of apoptosis is initiated by the ubiquitination of most complex I proteins, leading to their dissociation from TNF-R1. Binding of FADD to the DD of cytosolic TRADD facilitates the recruitment of caspase-8 and -10 via their DDs, forming complex II.

Conflicting data exist regarding the complex formation that induces apoptosis after TNF stimulation. In contrast to the model described above (2), we previously reported that after recruitment of TRADD, RIP, and TRAF2 to TNF-R1 at the cell surface, the receptor is internalized and FADD and procaspase-8 are recruited, forming the death-inducing signaling complex (DISC) still associated with the TNF receptor at endosomal vesicles (TNF receptosomes) (3–5). Consecutively, caspase-8 is activated by autocleavage and induces caspase-3 activation either directly or indirectly via a mitochondrial amplification loop involving cytochrome c and APAF-1 release, forming the apoptosome with caspase-9.

Recently, we found that within TNF receptosomes, caspase-8 activates caspase-7, which in turn cleaves A-SMase, initiating the production of ceramide and stimulation of cathepsin D (5), resulting in the cleavage of Bid and the activation of caspase-9 and -3 (6, 7).

The essential role of TNF-R1 internalization in the initiation of proapoptotic signaling was initially demonstrated by us by using pharmacological inhibitors (8), by deletion of a region termed the TNF-R1 internalization domain (TRID) (3), or by transducing cells with adenoviral protein 14.7K (4, 5, 9, 10). Major questions that remain are how the internalization and intracellular trafficking of TNF-R1 are regulated and which molecular events dictate the initial switch between antiapoptotic signaling from the cell surface and proapoptotic signaling from receptosomes.

Protein ubiquitination has been recognized as one crucial regulatory mechanism of target proteins. A multitude of ubiquitination processes have been reported to be important in TNF-R1 signaling, but all of these address events downstream of the ligand-activated receptor. The NF-κB pathway is activated via RIP1, involving its K11 or K63 ubiquitination mediated by TRAF2 and cIAP1/2. These ubiquitin chains serve as a scaffold for the recruitment of the linear ubiquitin chain assembly complex mediating NEMO ubiquitination. This leads to the phosphorylation of IκB and its subsequent K48 ubiquitination and proteasomal degradation, allowing the nuclear translocation of NF-κB (11–13). At the level of internalized TNF-R1, termination of NF-κB activation is regulated by K48 ubiquitination of RIP1 by the E3 ligases CARP-2 and CARP-1 (14, 15).

In this study, we show that TNF-R1 is a novel target of K63 ubiquitination upon stimulation with TNF. This ubiquitination is crucial for the internalization and proapoptotic signaling of TNF-R1.

MATERIALS AND METHODS

Reagents and antibodies.

TNF and biotinylated TNF (biotinTNF) were purchased from R&D Systems, CellMask and streptavidin-Alexa Fluor 488 conjugate were from Invitrogen, tetramethyl rhodamine isocyanate (TRITC)-dextran and Dynasore were from Sigma-Aldrich, Pitstop 2 was from Abcam, Complete protease inhibitor cocktail was from Roche, Benzonase was from Merck, and protein G microbeads were from Miltenyi. The plasmid coding for Fc-tagged TNF was kindly provided by Harald Wajant (University Hospital, Julius Maximilians University, Würzburg, Germany). Fc-tagged TNF was expressed in HEK293T cells and subsequently purified with HiTrap protein G columns (GE Healthcare). The primary antibodies used in this study were specific for FADD (2782), IκB (4814), NF-κB (4764), Ubc13 (4919), TRADD (3684), RIP1 (3493) (New England BioLabs), ubiquitin K48 (05-1307), ubiquitin K63 (05-1308) (Millipore), Lamp-2 (9840-01) (Southern Biotech), TNF-R1 (H5, sc-8436; C20, sc-1068), RNF8 (X21, sc-133971; B2, sc-271462), Rab5 (sc-598), TRAF2 (sc-876), and AMSH (sc-292045) (Santa Cruz). Caspase-8 C15 antibody was a gift from Marcus Peter (Feinberg School of Medicine, Northwestern University, Chicago, IL). The secondary antibodies used were peroxidase-conjugated anti-goat antibody 205-035-108 (Dianova), Alexa Fluor 488-labeled donkey anti-mouse IgG A21202, Alexa Fluor 555-labeled goat anti-mouse IgG A21422, Alexa Fluor 555-labeled donkey anti-rabbit IgG A315572 (Invitrogen), anti-mouse light-chain horseradish peroxidase (HRP)-conjugated antibody AP200P, and anti-rabbit light-chain HRP-conjugated antibody MAB201P (Millipore).

Cell culture.

U937 cells were cultured in RPMI 1640, and HeLa cells were cultured in Dulbecco's modified Eagle's medium (DMEM; Gibco) supplemented with 10% fetal calf serum (FCS; Biochrom) and penicillin-streptomycin (Biochrom). RNF8-deficient and parental mouse embryo fibroblasts (MEFs; both cell lines were kindly provided by Junjie Chen, University of Texas M. D. Anderson Cancer Center, Houston, TX) were cultured in DMEM supplemented with 10% FCS and penicillin-streptomycin.

Cell transfection with lentiviral shRNA.

For the generation of stable, RNF8-directed short hairpin RNA (shRNA)-expressing cell lines, lentiviral RNF8 shRNA (sc-61484-V) and control shRNA (sc-108080) particles were purchased from Santa Cruz Biotechnology. The transfection of U937 and HeLa cells was conducted according to the manufacturer's instructions. Stable shRNA-expressing clones were isolated by using puromycin (10 μg/ml; Santa Cruz) as a selection marker.

IP.

Cells were resuspended in 1 ml of immunoprecipitation (IP) lysis buffer (50 mM Tris-HCl [pH 7.4], 150 mM NaCl, 1% NP-40, 0.25% Na-deoxycholate, 1% Triton X-100, 1 mM EDTA, Benzonase [Merck], Complete protease inhibitor cocktail [Roche]) for 45 min on ice and subsequently sheared 10 times with a 21-gauge needle. The lysate was centrifuged for 10 min at 10,000 × g, the respective antibody needed for precipitation (for precipitation via Fc-TNF, no antibody was added) and 50 μl of protein G microbeads were added to the supernatant, and it was incubated overnight. The lysate was applied to μ Columns (Miltenyi), washed with 750 μl lysis buffer, and eluted with preheated SDS loading buffer. For Western blot analysis, 10 μl of the eluate was loaded.

Deubiquitination assay.

For deubiquitination, IPs were performed after 15 min of internalization as described above. After the material was loaded onto μ Columns, it was equilibrated with 500 μl of digestion buffer (for AMSH, 50 mM HEPES [pH 8.0] and 100 mM NaCl; for UCH-L1, 50 mM HEPES [pH 8.0], 100 mM NaCl, 1 mM EDTA, and 10 mM dithiothreitol [DTT]). A 20-μl volume of digestion buffer containing 100 nM AMSH or UCH-L1 was then added, and the mixture was incubated for 16 h at 37°C. Subsequently, the material was eluted and subjected to SDS-PAGE and Western blot analysis.

Immunomagnetic isolation of TNF-R1-containing fractions.

For magnetic isolation of TNF-R1-containing endosomes, 2 × 10⁸ cells were sedimented and cooled on salted ice for 20 min. Cells were subjected to incubation in a total volume of 300 μl containing 100 ng/ml Fc-tagged TNF for 20 min, followed by the addition of 50 μl of protein G microbeads (Miltenyi) and incubation for 1 h. The cells were washed with 25 ml of cold phosphate-buffered saline (PBS), internalization was initiated by adding 5 ml of prewarmed medium, and the cells were incubated for the desired times at 37°C. Internalization was stopped by adding 50 ml of cold PBS, and the cells were washed once with 10 ml of homogenization buffer (15 mM HEPES [pH 7.4], 250 mM sucrose, 0.5 mM MgCl₂, Complete protease inhibitor cocktail [Roche]). The cell sediment was resuspended in a total volume of 800 μl of homogenization buffer, and 12.5 U of Benzonase (Merck) was added. Cells were homogenized at 4°C with a Branson W-450 sonication device equipped with a cup resonator (G. Heinemann, Schwäbisch Gmünd, Germany). The homogenate was centrifuged for 4 min at 1,500 × g, and the postnuclear supernatant (PNS) was used for magnetic separation in a high-gradient magnetic field generated in a free-flow chamber (Hoock GmbH, Kiel, Germany). Magnetic fractions were analyzed by SDS-PAGE and Western blotting of 4 μg of the sample.

SDS-PAGE and Western blotting.

For SDS-PAGE, precast 4 to 20% gradient Mini-PROTEAN TGX gels (Bio-Rad) or self-made 12.5% polyacrylamide gels were used. Proteins were blotted to polyvinylidene difluoride membrane (Carl Roth GmbH). The membranes were blocked with 5% skim milk in Tris-buffered saline–Tween 20 and incubated overnight with the primary antibody diluted 1:500 to 1:5,000. The membranes were then incubated for 1 h with peroxidase-conjugated secondary antibodies diluted 1:10,000. Blots were developed with the ECL kit and films from GE Healthcare. Bands were scanned with a personal densitometer (GE Healthcare) and handled with ImageQuant 5.2 software (GE Healthcare). For densitometric analysis of TNF-R1 ubiquitination, the nonubiquitinated 55-kDa and ubiquitinated bands were analyzed. The ratio of K63-ubiquitinated to nonubiquitinated TNF-R1 is shown.

2D PAGE.

IP material was resuspended in two-dimensional (2D) PAGE sample buffer {8 M urea, 2 M thiourea, 4% 3-[(3-cholamidopropyl)-dimethylammonio]-1-propanesulfonate (CHAPS), 2% ASB-14, 40 mM DTT}. Isoelectric focusing (IEF) was performed at 18°C with 18-cm immobilized pH gradient (IPG) strips (linear gradient of pH 3 to 10; GE Healthcare) on an IPGphor device (GE Healthcare). The strips were rehydrated for 8 h by using a 350-μl sample. Subsequently, IEF was performed at a maximum of 50 mA per IPG strip for 6 h at 30 V, 0.5 h at 500 V, 1 h at 1,000 V, 1 h at 3,000 V, and finally 8,000 V until 80,000 Vh was reached. For the second dimension, the strips were equilibrated for 2 × 15 min with equilibration buffer (50 mM Tris [pH 8.8], 6 M urea, 30% glycerol, 2% SDS, 0.03% bromphenol blue) supplemented with 65 mM DTT or 135 mM iodoacetamide, respectively. The strips were then separated by 11% SDS gel electrophoresis and subjected to Western blot analysis.

Fluorescence microscopy.

Cells were grown on fibronectin-coated coverslips (Becton, Dickinson) and incubated for 60 min at 4°C with biotinTNF (100 ng/ml) together with 5 μg/ml Alexa Fluor 488-conjugated streptavidin. The temperature was shifted to 37°C with prewarmed medium for the times indicated in the figure legends to allow receptor internalization. Subsequently, cells were fixed in 4% paraformaldehyde (PFA)–PBS for 20 min, permeabilized in PBS supplemented with 0.1% saponin and 0.2% bovine serum albumin (BSA), and stained with the respective primary antibodies at a 1:50 dilution for 60 min at room temperature, followed by incubation with the respective secondary antibodies. For measurement of internalization efficiency, cells were treated as described before. During internalization, 0.2 mg/ml TRITC-dextran was added to the prewarmed medium. Subsequently, the cells were fixed in 4% PFA–PBS but not permeabilized. Immunofluorescence analysis was performed with a confocal laser scanning microscope (LSM 510, equipped with an Axiovert 100M; Zeiss), and one picture represents one single plane from a z-stack. All LSM micrographs were made with a 63× objective. Contrast and brightness were adjusted with Adobe Photoshop. Colocalization was analyzed with ImageJ.

ImageStream analysis.

For the high-throughput microscopic analysis and quantification of TNF-R1 internalization, cells were incubated for 60 min at 4°C with biotinTNF (100 ng/ml) together with 5 μg/ml Alexa Fluor 488-conjugated streptavidin. Subsequently, the temperature was shifted to 37°C with prewarmed medium for the times indicated in the figure legends to allow receptor internalization. The cells were then fixed in 2% PFA–PBS for 20 min without permeabilization. CellMask Deep Red plasma membrane stain (Life Technologies) was added for the final 5 min. The cells were washed twice with 2% PFA–PBS, and the cell pellet was finally resuspended in 50 μl 2% PFA–PBS. To pharmacologically block internalization, cells were pretreated with either 80 μM Dynasore or 20 μM Pitstop 2 for 30 min at 37°C and subsequently cooled on ice for 20 min in the presence of the inhibitors.

The ImageStream Mark II was used for cell analysis. Up to 8,000 cells were counted by detecting the internalization probe biotinTNF-Alexa Fluor 488 (excitation, 488 nm) in channel 2 and the cell surface label CellMask stain (excitation, 642 nm) in channel 5. The bright-field image was acquired in channel 6. For image acquisition, a 60× objective was used. We used the internalization wizard to assay the TNF-R1 endocytosis rates in shControl- and shRNF8-transfected U937 cells. The endocytosis rates after 20 and 40 min were normalized to those of cells kept on ice.

For the p65 NF-κB translocation assay, cells were incubated with TNF for 60 min at 4°C. The temperature was then shifted to 37°C with prewarmed medium for 15 min to allow receptor internalization. Subsequently, cells were fixed in 2% PFA–PBS for 20 min and permeabilized for 20 min with 0.2% saponin–0.1% BSA–PBS and a primary antibody diluted 1:200 in 0.1% BSA–PBS was added for 1 h of incubation; this was followed by washing two times in PBS, the addition of a secondary antibody diluted 1:500 in 0.1% BSA–PBS, and incubation for 1 h. After being washed, the cells were resuspended in 50 μl of PBS containing Hoechst 33258 (1:5,000; Sigma-Aldrich). Up to 10,000 cells were counted by detecting the internalization probe biotinTNF-Alexa Fluor 488 (excitation, 488 nm) in channel 2 and the p65-Alexa Fluor 647 stain (excitation, 642 nm) in channel 5. For image acquisition, a 60× objective was used. We used the colocalization wizard to count the shControl- and shRNF8-transfected U937 cells showing p65 nuclear versus cytosolic localization.

For apoptosis measurement, cells were incubated for the times indicated in the figures with Fc-tagged TNF under standard cell culture conditions. At 30 min before the end, Hoechst stain (Sigma-Aldrich) was added to the culture medium to a final dilution of 1:10,000. Up to 10,000 cells were captured by detecting the nuclear stain (excitation, 405 nm) in channel 1. For image acquisition, a 60× objective was used. The apoptosis wizard was used to count the cells showing nuclear fragmentation or intact nuclei.

Caspase assay.

Caspase activity was assayed with the Caspase-Glo 8 and Caspase-Glo 3/7 kits from Promega by following the manufacturer's instructions. Cells were incubated with Fc-tagged TNF (100 ng/ml) for the indicated times. The respective substrate was added, and luminescence was measured after 45 min of incubation with a TECAN infinite 200. Data analysis was done with Microsoft Excel.

MTS cell proliferation assay.

3-(4,5-Dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium (MTS) assays were performed by following the manufacturer's instructions (Promega). Fc-tagged TNF at 100 ng/ml was added at distinct time points. Subsequently, MTS reagent was added and absorbance was measured after 30 min with a TECAN infinite 200. Data analysis was done with Microsoft Excel.

EdU proliferation assay.

The Click-iT EdU (5-ethynyl-2′-deoxyuridine) cell proliferation assay was used according to the manufacturer's instructions (Invitrogen). After 2 h of preincubation with the EdU reagent, TNF was added and the mixture was incubated for the times shown. Fluorescence was measured with a TECAN infinite 200. Data analysis was done with Microsoft Excel.

TUNEL assay.

The Click-it terminal deoxynucleotidyltransferase-mediated dUTP-biotin nick end labeling (TUNEL) assay was used according to the manufacturer's instructions (Invitrogen). Fluorescence was measured with a TECAN infinite 200. Data analysis was done with Microsoft Excel.

RTCA.

The xCELLigence system from Roche was used for real-time cell analysis (RTCA). The baseline was measured for 30 min with prewarmed medium. Quadruplicates of 5,000 cells per well were then seeded, and attachment was measured for 15 min every 6 h. Subsequently, Fc-tagged TNF was added (100 ng/ml); this was followed by continuous data acquisition with RTCA software 1.2 (Roche). Data analysis was done with Microsoft Excel.

A-SMase activity assay.

A-SMase activity in cell lysates or magnetically isolated material was measured with N-methyl ¹⁴C-sphingomyelin-(0.5 mCi/ml, 0.55 Ci/mmol; Hartmann Analytic) as the substrate in 250 mM Na-acetate (pH 5.0)–1 mM EDTA–0.1% Triton X-100. A 2- to 10-μg protein sample was incubated with the substrate for 2 h at 37°C in a total volume of 150 μl. The reaction mixture was extracted with 750 μl of chloroform-methanol (2:1). The radioactivity of enzymatically liberated radioactive phosphorylcholine in a 250-μl aliquot of the aqueous phase was measured with a β-counter (Beckman Coulter).

RESULTS

TNF-R1 is K63 ubiquitinated upon TNF binding.

In an initial approach to the investigation of the role of TNF-R1 ubiquitination in internalization, we used antibodies recognizing either K48 or K63 polyubiquitin chains (16, 17) for IP of the receptor from TNF-stimulated U937 cells. Figure 1A demonstrates that activated TNF-R1 is coprecipitated only by K63-specific antibodies after stimulation with TNF for 15 and 30 min at 37°C. Two TNF-R1 forms were detected in the Western blot analysis by the anti-TNF-R1 antibody, one at 55 kDa (resembling the nonubiquitinated receptor) and one with a molecular mass shifted to approximately 70 kDa (resembling the K63-ubiquitinated receptor). Vice versa, we analyzed TNF-R1 ubiquitination by incubating cells with Fc-tagged TNF for 0, 5, 15, and 30 min and then immunoprecipitating the Fc-tagged TNF/TNF-R1 complex with protein G microbeads. For the top panel of Fig. 1B, we used an anti-TNF-R1 antibody (H5) for detection of the ubiquitinated receptor and observed a time-dependent increase in ubiquitination. The densitometric analysis depicted as the ratio of K63-ubiquitinated TNF-R1 to 55-kDa-nonubiquitinated TNF-R1 from Fig. 1B (top panel) is shown in Fig. 1C. The middle panel of Fig. 1B shows the signal obtained by probing the same samples with a K63-specific antibody; bands similar to those detected by the anti-TNF-R1 antibody, as shown in the top panel, were detected. In contrast, no signals were obtained when a K48 polyubiquitin antibody was used (bottom panel).

FIG 1.

TNF-R1 is K63 ubiquitinated upon TNF stimulation. (A) Ubiquitinated TNF-R1, detected by the anti-TNF-R1 (H5) antibody, is coprecipitated by K63-specific but not by K48-specific antibodies (16, 17). Cells were either incubated at 37°C for 15 or 30 min or kept on ice (0 min). (B) Fc-tagged TNF was bound to TNF-R1 prior to internalization for 5, 15, or 30 min. The top panel shows time-dependent increases in ubiquitination. The middle panel shows the corresponding pattern obtained with K63 polyubiquitin-specific antibodies. No such signals were visible when K48-specific antibodies were used (bottom panel). Lane B, bead control. (C) Densitometric analysis and ratio of ubiquitinated TNF-R1 to nonubiquitinated 55-kDa TNF-R1 from panel B (top panel). (D) Fc-tagged TNF was used to precipitate ligand-bound TNF-R1 from HeLa cells after different internalization times or from cells kept on ice. An increase in TNF-R1 ubiquitination was detected with an anti-TNF-R1 antibody. (E) Densitometric analysis and ratio of K63-TNF-R1 to 55-kDa TNF-R1 from panel D. (F) Precipitation of TNF-R1 after 15 min of stimulation at 37°C, followed by incubation without or with AMSH (a K63-specific protease) or UCH-L1 (a C-terminal ubiquitin hydrolase). The top panel shows the TNF-R1 signals, and the middle and bottom panels show the two proteases. (G) Densitometric analysis of the ratio of K63-ubi-TNF-R1 to 55-kDa TNF-R1. The values to the left are molecular sizes in kilodaltons. WB, Western blotting; n.s., not significant; ubi., ubiquitination.

As shown in the middle panel of Fig. 1B, the K63-specific antibody also recognized a band resembling nonubiquitinated TNF-R1. To rule out the possibility that this antibody cross-reacts with nonubiquitinated TNF-R1, we analyzed material immunoprecipitated after 15 min of incubation by 2D PAGE (see Fig. S1 in the supplemental material). Probing of the material with the TNF-R1 (H5)-, ubiquitin-, and K63-specific ubiquitin antibodies revealed one major spot with the same molecular weight (MW) and isoelectric point (pI) (red squares), indicating that this protein is ubiquitinated TNF-R1. Both ubiquitin antibodies recognized additional spots at lower MWs and different pIs that were not recognized by the TNF-R1 antibody. Thus, these spots are most likely ubiquitinated proteins that were coprecipitated with activated TNF-R1.

IP of activated TNF-R1 from HeLa cells with Fc-TNF resulted in the appearance of a shifted TNF-R1 band similar to that observed in U937 cells (Fig. 1D). Densitometric quantification of the blot is shown in Fig. 1E. The respective lysates used for the IP experiments are shown in Fig. S2 in the supplemental material.

To support our findings on the TNF-R1 ubiquitination state, we precipitated activated TNF-R1 via Fc-tagged TNF after 15 min. The precipitated material was incubated with either AMSH, a K63 linkage type-specific deubiquitinase, or UCH-L1, a ubiquitin C-terminal hydrolase, in vitro. Treatment with both enzymes clearly resulted in a reduction of ubiquitination, as shown by Western blot and densitometric analyses (Fig. 1F to G).

To further characterize TNF-R1 K63 ubiquitination, we treated U937 cells for 20 min with Fc-tagged TNF and then added protein G magnetic microbeads for 1 h of incubation at 4°C. The incubation temperature was then shifted to 37°C to induce simultaneous internalization of the TNF receptors. At various time points, cells were homogenized and TNF receptosomes were isolated as described previously (18, 19). The blot in Fig. 2A shows that the 70-kDa TNF-R1 band appeared within 5 min of TNF treatment and its intensity increased time dependently (top panel). Figure 2B shows a densitometric quantification of the appearance of the ubiquitinated band. Probing of the same samples with the anti-K63 antibody revealed that this 70-kDa protein resembled K63-ubiquitinated TNF-R1 (second panel from top). The respective total cell homogenates used for this experiment are shown in Fig. S2D in the supplemental material.

FIG 2.

TNF-R1 is K63 ubiquitinated in receptosomes and interacts with RNF8 and Ubc13. (A) Kinetics of K63-ubiquitinated TNF-R1 after TNF stimulation in isolated TNF receptosomes. TNF-R1 was detected with the anti-TNF-R1 (H5) antibody. Ubiquitination was detected with a K63-specific antibody. The E2-ubiquitin ligase Ubc13 and the E3-ubiquitin ligase RNF8 [anti-RNF8 (X21)] are concomitantly recruited to immunomagnetically isolated TNF receptosomes. (B) Increase in ubiquitinated-TNF-R1 versus nonubiquitinated TNF-R1 in the top blot in panel A. (C) Coimmunoprecipitation of TNF-R1 and RNF8, demonstrating their presence within the same protein complex. Anti-TNF-R1 (C20) and anti-RNF8 (X21) antibodies precipitate mainly nonubiquitinated TNF-R1, while precipitation of cross-linked TNF-R1 by Fc-tagged TNF coprecipitates K63-ubiquitinated TNF-R1, as detected by the anti-TNF-R1 (H5) antibody. An RNF8 antibody coprecipitates both TNF-R1 variants. RNF8 detected by the B2 antibody can be precipitated with Fc-tagged TNF, TNF-R1 (C20), and RNF8 (X21), respectively. Representative data from two or three experiments are shown. (D) Densitometric analysis of the ratio of K63-ubi-TNF-R1 to 55-kDa TNF-R1. Western blot analyses of the total lysates used for the IP experiments or PNS used for receptosome isolation are shown in Fig. S2 in the supplemental material. The values to the left of panels A and C are molecular sizes in kilodaltons.

Ubiquitination of TNF-R1 was also observed in other cell lines like Jurkat and SKW6.1 (data not shown). Taken together, our data indicate that TNF-R1 is K63 ubiquitinated in a time-dependent manner upon ligand binding.

The E2 ubiquitin-conjugating enzyme Ubc13, as well as the E3 ubiquitin ligase RNF8, interacts with TNF-R1 upon TNF stimulation.

Mass spectrometric analysis of isolated TNF receptosomes revealed several E3 ubiquitin ligases that were copurified (data not shown). We identified the E3 ubiquitin ligase RNF8 as a candidate enzyme responsible for modifying TNF-R1. Figure 2A shows the time-dependent recruitment of E3 ligase to TNF-R1 in isolated TNF receptosomes (third panel). We also detected the recruitment of the E2 ubiquitin-conjugating enzyme Ubc13 in TNF-R1 in these fractions with the same kinetics as RNF8 (fourth panel). In Fig. 2B, the quantification of the bands and the ratio of ubiquitinated TNF-R1 to 55-kDa nonubiquitinated TNF-R1 is shown.

The presence of RNF8 and TNF-R1 within the same protein complex was investigated in the coimmunoprecipitation experiments depicted in Fig. 2C. When a C terminus-specific anti-TNF-R1 (C20) antibody was used to precipitate TNF-R1 from cells not treated with TNF, mainly the nonubiquitinated 55-kDa form was precipitated and only a faint band of the ubiquitinated form appeared. Precipitation of activated TNF-R1 after cell stimulation for 15 min with Fc-tagged TNF resulted in the enrichment of mainly the ubiquitinated higher-MW form. Precipitation with an RNF8 antibody after 15 min of treatment with nontagged TNF revealed that mainly the 55-kDa TNF-R1 variant was coprecipitated (Fig. 2C, top panel). The lower panel of Fig. 2C shows that RNF8 coprecipitated with Fc-tagged, TNF-activated TNF-R1 and was only weakly detected when an anti-TNF-R1 antibody was used for precipitation of the TNF receptor from unstimulated cells. The input material is shown in Fig. S2E in the supplemental material. These results indicate that RNF8 is recruited to activated TNF receptors upon ligand binding.

We next analyzed the compartmentalization of RNF8 and Ubc13 and the interactions between these molecules and activated TNF-R1 by confocal fluorescence microscopy. In cells treated with biotinTNF at 4°C (time zero), no colocalization of TNF-R1 and RNF8 is apparent (Fig. 3A to C). After an incubation temperature shift to 37°C, TNF treatment resulted in the partial colocalization of RNF8 with internalized TNF-R1 (Fig. 3D to G). In Fig. 3G, the colocalization of TNF-R1 and RNF8 is highlighted by arrows. In addition, we also observed colocalization of Ubc13 with internalized TNF-R1 (Fig. 3I and J), while no colocalization was visible in cells prior to the internalization of TNF-R1 at 4°C (Fig. 3H). Earlier studies reported the nuclear localization of RNF8. We also detected RNF8 in the nucleus but also in the cytoplasm of HeLa cells. Intriguingly, after TNF treatment, we observed the translocation of RNF8 from the cell nucleus to the cytoplasm (see Fig. S3 in the supplemental material).

FIG 3.

TNF-induced colocalization of RNF8 and Ubc13 with TNF-R1. (A to C) Clear separation of biotinTNF–TNF-R1 (green) and RNF8 (red) (R_coloc [colocalization coefficent] = 0.05) in unstimulated HeLa cells. (D to G) Internalization of biotinTNF–TNF-R1 (green) upon 15 min of TNF treatment and partial colocalization (R_coloc = 0.1411) (G, indicated by arrows) with RNF8 (red) upon 15 min of TNF stimulation. (H) No colocalization of biotinTNF–TNF-R1 (green) and Ubc13 (red) in untreated HeLa cells (R_coloc = −0.0884). (I) Partial colocalization Ubc13 (R_coloc = 0.0633) with activated TNF-R1 upon TNF-R1 internalization (green) in HeLa cells (J, indicated by arrows). Activated TNF-R1 was labeled with biotinTNF, RNF8 was labeled with the X21 antibody, and Ubc13 was labeled with the 4919 antibody. The dashed blue line marks the nucleus. All confocal images were captured with a 63× oil immersion objective. One single slice of a z-stack is shown.

Downmodulation of RNF8 retards TNF-R1 internalization.

To investigate the role of RNF8 in TNF-R1 signaling, we generated stable RNF8 knockdown clones of both U937 and HeLa cells by lentiviral shRNA transfection (see Fig. S4A and B in the supplemental material). The following experiments were performed with shRNF8-transfected HeLa clones G7 and E1 and shRNF8-transfected U937 clone D5. We first analyzed the role of RNF8 in receptor endocytosis by high-throughput microscopic analysis with the ImageStream device (Amnis/EMD Millipore). The left side of Fig. 4A shows examples of shControl-transfected U937 cells kept on ice, The cell surface is marked in red, and the distribution of TNF-R1 is visible as green dots. After 40 min of internalization (right panel), the surface distribution is clearly reduced and only a few but larger TNF-R1-containing structures appear within the cellular boundary. The fewer TNF-R1-positive compartments can be explained by the accumulation of TNF-R1 in multivesicular compartments (3).

FIG 4.

Downregulation of RNF8 impairs TNF-R1 internalization in U937 and HeLa cells. (A) High-throughput microscopic analysis of shControl-transfected U937 cells with the ImageStream device revealing punctate cell surface labeling with biotinTNF–TNF-R1 (green) at 0 min (left side) compared to fewer but internalized green spots observable after 40 min of TNF treatment (right side). The plasma membrane was stained with CellMask (red), and TNF-R1 was labeled with biotinTNF–streptavidin-Alexa Fluor 488. (B) Analysis of shRNF8-transfected U937 cells in parallel with the ImageStream device. Cell surface-resident TNF-R1 can be observed after both 0 (left side) and 40 (right side) min of TNF treatment. Panels A and B show 5 cells representative of the ∼20,000 analyzed pre- or postinternalization. (C) Quantification of TNF-R1 endocytosis. Control transfected cells display an increase in endocytosis, which is clearly reduced in shRNF8-transfected cells. The data are normalized to those from cells kept on ice. Mean results ± standard deviations of four independent experiments, each counting 7,000 to 9,000 cells, are shown. (D to G) In shControl- and shRNF8-transfected HeLa cells, biotinTNF–TNF-R1 was labeled (green) in addition to the cell surface (red). In control transfected cells, TNF-R1 appears at the cell surface while cells are kept on ice (D) but appears to be internalized after 30 min at 37°C (E). In shRNF8-transfected cells, TNF-R1 remains on the cell surface after 0 (F) and 30 (G) min of TNF treatment. Nuclei are shown in blue. (H, I) HeLa cells labeled with biotinTNF–TNF-R1 (green) and the fluid-phase marker TRITC-dextran (red) that has been cointernalized for 60 min at 37°C. Control transfected cells (H) display mainly intracellular TNF-R1 and many yellow spots, indicating colocalization with TRITC-dextran. In shRNF8-transfected cells (I), TNF-R1 again is located mainly at the cell surface, indicating that TNF-R1 internalization is blocked. In panels A to G, the plasma membrane was stained with CellMask (red), and in panels H and I, TRITC-dextran is red and activated TNF-R1 was indirectly labeled with biotinTNF–streptavidin-Alexa Fluor 488. Cells were not permeabilized for the experiments shown. Panels D to I represent one confocal section. Nuclei were stained with 4′,6-diamidino-2-phenylindole (blue). All confocal images were captured with a 63× oil immersion objective.

Figure 4B shows examples of shRNF8-transfected U937 cells analyzed with the ImageStream device after 0 and 40 min of internalization. The cell surface (red) distribution of TNF-R1 (green) is clearly visible at both time points. The use of the ImageStream device allowed the quantification of cells showing TNF-R1 internalization (Fig. 4C). The number of cells positive for internalized TNF-R1 increased in shControl-transfected U937 cells from 0 to 40 min (black columns), while only a negligible increase in internalization was detected in shRNF8-transfected U937 cells (red columns).

In addition to U937 cells, we used shControl- and shRNF8-transfected HeLa cells to investigate the effect of RNF8 downmodulation on TNF-R1 endocytosis. Figure 4D shows the cell surface distribution of the receptor (green) in cells kept on ice. After 30 min at 37°C, the receptor appears intracellularly in shControl-transfected HeLa cells (Fig. 4E). In contrast, shRNF8-transfected HeLa cells display TNF-R1 surface distribution at both time points (Fig. 4F and G), indicating a block in TNF-R1 internalization due to RNF8 deficiency. To further substantiate these findings, we labeled shControl- and shRNF8-transfected HeLa cells with biotinTNF–streptavidin-Alexa Fluor 488. During internalization for 60 min, we added the fluid-phase marker TRITC-dextran to allow its cointernalization and inclusion in the TNF receptosomes formed. The shControl-transfected HeLa cells (Fig. 4H) display many yellow vesicular structures inside the cell, indicating internalized TNF-R1 containing cointernalized TRITC-dextran, which is hardly observed in shRNF8-transfected HeLa cells (Fig. 4I). Here, TNF-R1 remains mostly on the cell surface.

To address the question of whether ubiquitination precedes internalization, we precipitated activated TNF-R1 via Fc-TNF from wild-type (WT) U937 cells with or without pretreatment with the endocytosis inhibitor Dynasore (inhibitor of dynamin) or Pitstop 2 (inhibitor of clathrin). In both cases, TNF-R1 ubiquitination could be observed (see Fig. S5 in the supplemental material), showing that ubiquitination occurs at TNF-R1 on the cell surface prior to internalization.

In summary, our findings obtained by conventional, as well as high-throughput, microscopic analysis of TNF-R1 internalization in U937 and HeLa cells suggest that RNF8 deficiency impedes proper receptor endocytosis.

The ubiquitination of ligand-activated TNF-R1 is blocked in RNF8-downmodulated cells, which does not affect the recruitment of TRAF2 and RIP1 but prevents the binding of FADD and caspase-8 to TNF-R1.

To investigate the effect of RNF8 downmodulation on the assembly of TNF-R1 adaptor proteins, we next analyzed TNF-R1 magnetic fractions isolated from shControl-transfected U937 and shRNF8-transfected cells (Fig. 5A). The top three panels of Fig. 5A demonstrate the lack of K63 ubiquitination of TNF-R1 in RNF8-downmodulated U937 cells compared to control-transfected cells; TNF-R1 is K63 but not K48 ubiquitinated in shControl-transfected U937 cells but not in shRNF8-transfected U937 cells. The densitometric analysis and quantification of the ubiquitinated band from the first panel are shown in Fig. 5B. Although TNF-R1 itself appears not to be K48 ubiquitinated, some other K48-ubiquitinated protein migrating at a higher MW is recruited in shControl-transfected U937 cells but not in shRNF8-transfected cells. This could be due to the absence of RNF8 enzymatic E3 ligase activity, which may also mediate K48 ubiquitination. Another explanation could be that K48 ubiquitination of the protein detected only in shControl-transfected U937 cells requires receptosome formation. Since this K48 ubiquitination appears only in isolated receptosomes and not in TNF-R1 immunoprecipitates (Fig. 1B), the K48-ubiquitinated protein is not associated with the TNF receptor complex.

FIG 5.

Downregulation of RNF8 prevents TNF-R1 trafficking and DISC recruitment. (A) Western blot analysis of TNF receptosomes. TNF-R1 K63 ubiquitination occurs only in shControl-transfected cells and not in shRNF8-transfected cells (top three panels). Further analysis of TNF-R1-containing magnetically isolated TNF receptosomes shows that RNF8 downmodulation induces a change in complex II protein recruitment. FADD and the receptor-bound caspase-8 p41/43 fragment are recruited only in shControl-transfected U937 cells, and the active caspase-8 p18 fragment is also formed only in shControl-transfected U937 cells and not in shRNF8-transfected U937 cells. The complex I proteins TRADD, TRAF2, and RIP1 are recruited to TNF-R1 in both cell lines independently of RNF8. The lack of recruitment of Rab5 and Lamp-2 in shRNF8-transfected U937 cells, in contrast to shControl-transfected U937 cells, supports the finding that TNF-R1 internalization is impaired in RNF8 knockdown HeLa cells. Representative data from three experiments are shown. Western blot analyses of the PNS used for the isolation of TNF-R1-containing membrane fractions are shown in Fig. S6 in the supplemental material. MW is shown to the left in thousands. (B) Quantification of TNF-R1 forms from the first part of panel A, showing the increase in TNF-R1 ubiquitination in shControl-transfected U937 cells (left panel) and its absence from shRNF8-transfected U937 cells (right panel). The ratio of K63-ubiquitinated to nonubiquitinated TNF-R1 is presented. ubi., ubiquitination; Casp., caspase.

Additionally, in shControl-transfected U937 cells, the adaptor proteins TRAF2, RIP1, TRADD, FADD, and caspase-8 were rapidly recruited to TNF receptosomes, and caspase-8 is activated, as indicated by the receptor-bound 41/43-kDa form and the p18 active caspase fragment. In TNF-R1 magnetic fractions from shRNF8-transfected U937 cells, low levels of TRADD but significant amounts of TRAF2 and RIP1 could be detected, but the DISC proteins FADD and caspase-8 were completely lacking (Fig. 5A, middle panels). These findings mirror our observations on cells transfected with recombinant TNF-R1 carrying mutations in the internalization domain or when internalization is blocked by adenovirus infection or transfection with the adenovirus 14.7-kDa protein (3, 4). As shown in the lower two panels of Fig. 5A, the endosomal protein Rab5 and the lysosomal protein Lamp-2 were transiently enriched only in control transfected U937 magnetic fractions and not in the magnetic fractions from shRNF8-transfected U937 cells, again supporting our observations of impaired TNF-R1 trafficking in U937 and HeLa cells lacking RNF8. Fig. S6 in the supplemental material shows a Western blot analysis of the cell homogenates from which the corresponding magnetic fractions were isolated.

Downmodulation of RNF8 prevents TNF-induced activation of caspase-8, caspase-3/7, A-SMase, and apoptosis.

We next investigated the role of RNF8 in the TNF-induced activation of caspase-8 and the downstream effectors caspase-3/7 in whole cells. As shown in Fig. 6A, an increase in caspase-8 activity in response to TNF was observed only in WT or shControl-transfected U937 cells and not in shRNF8-downmodulated cells. TNF induction of caspase-3/7 was also completely abrogated in shRNF8-transfected cells, in contrast to WT U937 cells (Fig. 6B). To confirm our observations on the role of RNF8 ubiquitination in TNF-R1 signaling in another cellular system, we next used RNF8^−/− MEFs. The results shown in Fig. 6C and D confirmed our data obtained with shRNF8-downmodulated U937 cells, Caspase-8 and caspases-3/7 were activated only in WT MEFs and not in RNF8^−/− MEFs. The active caspase-8 fragment p18 was detected by Western blot analysis only in U937 cells and WT MEFs and not in the respective cells lacking RNF8 (see Fig. S7A and B in the supplemental material).

FIG 6.

Proapoptotic signaling is abrogated in cells lacking RNF8. TNF-induced activation of caspase-8, as well as caspase-3/7, is inhibited in RNF8-downmodulated U937 cells (A and B) and RNF8^−/− MEFs (C and D), while profound activation occurs in the respective WT cells. Data from three experiments are shown with standard deviations. See also Fig. S7 in the supplemental material. Additionally, A-SMase activation is blocked in shRNF8-transfected U937 cells (E) and in magnetically isolated TNF-R1 membrane fractions from shRNF8-transfected U937 cells (F). Data from two experiments performed with triplicate assays are shown with standard deviations. (G) Analysis of apoptosis by ImageStream, measuring the increase in nuclear fragmentation in dead cells. RNF8 deficiency blocks TNF-induced apoptosis in U937 clone D5 cells, in contrast to WT U937 cells. Mean values from three experiments are shown. Etoposide (Etop) was used as a positive control for apoptosis induction. (H) In the same line, analysis of HeLa cells and MEFs by TUNEL assays also revealed that WT cells had an apoptosis rate higher than that of the respective RNF8-lacking cells after 12 h of TNF treatment. Representative results from two experiments are shown. Etoposide (Etop) treatment was used as a positive control for apoptosis induction.

Recently, we showed that activation of caspase-8 and subsequently caspase-7 is needed for TNF-induced A-SMase activation (6, 7). Figure 6E shows that A-SMase activation in WT U937 cells starts at about 15 min, with peak activity at 30 min. In contrast, U937 clone D5 cells displayed delayed and significantly reduced activity at 60 min after stimulation. A similar effect of decreased A-SMase induction in RNF8 knockdown versus WT cells was observed in isolated TNF-R1 magnetic fractions. TNF receptosomes isolated from WT U937 cells contained enhanced enzymatic A-SMase activity after 30 and 60 min of internalization, while TNF-R1-containing membrane fractions isolated from U937 clone D5 cells displayed no increase in A-SMase activity after 30 and 60 min, respectively (Fig. 6F).

Since our results so far suggest an important role for RNF8-mediated TNF-R1 K63 ubiquitination in the initiation and propagation of proapoptotic signaling, we next analyzed the impact of RNF8 deficiency on apoptosis by high-throughput microscopic analysis with the ImageStream device (Fig. 6G). After the cells were treated with either TNF or etoposide for the times indicated in Fig. 6G, Hoechst stain was added. Subsequently, the number of cells showing fragmented nuclei (apoptotic) and the number of cells showing intact nuclei (nonapoptotic) were determined. We found a massive decrease in TNF-induced apoptosis in shRNF8-transfected U937 cells (5 h TNF, no increase; 20 h TNF, ∼10%) compared to that in WT U937 cells (5 h TNF, ∼24%; 20 h TNF, ∼50%). Treatment with etoposide as a positive control, in contrast, resulted in similar apoptosis induction in both cell lines. A similar decrease in the apoptotic response was revealed by TUNEL assays, in which shRNF8-transfected HeLa cells and RNF8^−/− MEFs showed significantly weaker responses to TNF than the respective WT cells (Fig. 6H).

Downmodulation of RNF8 does not interfere with TNF-induced NF-κB activation but enhances TNF-mediated cell growth.

We next investigated effects of RNF8 depletion on NF-κB induction and cell growth. Degradation of I-κB occurred in response to TNF in WT U937 cells, as well as in shControl-transfected and shRNF8-downmodulated U937 cells (Fig. 7A). shRNF8-transfected U937 cells displayed even more rapid I-κB degradation after 5 min than WT and shControl-transfected cells (10 to 15 min). In addition, TNF-induced translocation of NF-κB could be observed in shControl-transfected U937 cells, as well as in shRNF8-transfected U937 cells, by ImageStream microscopic analysis (Fig. 7B). We observed an ∼23% increase in NF-κB p65 nuclear localization in both cell lines upon 15 min of TNF stimulation. These data fit the capability of RNF8-deficient U937 cells to recruit TRAF2 and RIP1 and indicate that proinflammatory and growth-stimulating signaling is transmitted independently from cell surface-arrested TNF-R1.

FIG 7.

RNF8 deficiency does not interfere with TNF-mediated NF-κB activation and potentiates TNF-induced proliferation. (A) I-κB is comparably degraded upon TNF stimulation in transfected WT and shControl- and shRNF8-transfected U937 cells. The top half of each panel shows the protein IκB level revealed by Western blot analysis, and the bottom half of each panel shows Rab5, which was used as a loading control. (B) High-throughput microscopic analysis of NF-κB p65 translocation. The top of the panel shows representative examples of untreated (0 min) and TNF-treated (15 min) shControl- and shRNF8-transfected U937 cells. The similarity of nuclear and p65 staining was analyzed with the ImageStream device. The results are depicted as histograms at the bottom. (C) TNF treatment stimulated proliferation in both U937 clone D5 cells treated with RNF8 shRNA and MEFs deficient in RNF8 compared with the respective WT cells. Data were obtained by triplicate measurements. (D) Similar results were obtained by measuring proliferation as EdU incorporation after 12 h in HeLa cells and MEFs (D). RNF8-deficient cells display a significantly smaller increase in proliferation than WT cells. Etoposide (Etop) treatment was used as a positive control for cell death induction. (E) RTCA with HeLa cells showed a similar effect. WT HeLa, HeLa clone E1, and HeLa clone G7 cells were left untreated (− TNF) or incubated with TNF for the times shown. Both shRNF8-downmodulated HeLa clones grew better in response to TNF than did untreated cells, while TNF-treated WT HeLa cells clearly grew worse than untreated cells. Data are from quadruplicate assays.

To investigate the effect of RNF8 depletion on cell growth, we analyzed WT MEFs, RNF8-deficient MEFs, WT U937 cells, and shRNF8-downmodulated U937 cells in MTS proliferation assays. Compared to the respective WT cells, both RNF8^−/− MEFs and shRNF8-transfected U937 cells showed significantly enhanced proliferative responses (Fig. 7C). A similar effect was observed in an EdU incorporation assay, where TNF stimulation induced greater proliferation in RNF8-lacking cells than in the respective WT cells (Fig. 7D).

These findings were further corroborated by a label-free RTCA system (Roche). In this system, information about cell number and viability is provided by measuring changes in the impedance of adherent cells over a given time. Monitoring of the response of adherent HeLa cells to 9 h of TNF stimulation revealed that TNF treatment of WT HeLa cells led to growth slower than that of untreated cells (Fig. 7E, top). In striking contrast, TNF treatment of HeLa clones E1 (Fig. 7E, middle) and G7 (Fig. 7E, bottom) led to greater vitality, and thus a higher cell number, than that of untreated cells.

Together, these data suggest that nonubiquitinated TNF-R1 preferentially transmits proliferative signals from the cell surface. The RNF8-mediated K63 ubiquitination of TNF-R1 is involved in the negative regulation of cell proliferation by inducing the internalization of the receptor. This eventually leads to the recruitment of DISC proteins to internalized TNF-R1, activation of caspase-8 and caspase-3/7, and subsequent activation of the proapoptotic signaling cascade.

DISCUSSION

Our present findings are summarized in the model depicted in Fig. 8, where activation of TNF-R1 results in the rapid recruitment of adaptor proteins TRADD, RIP1, and TRAF2/5 to the TNF-R1 DD, forming complex I (3, 4), which transmits inflammatory and cell survival signals via NF-κB induction. On the basis of our data presented in this study, activated TNF-R1 on the cell surface rapidly recruits the E3-ubiquitin ligase RNF8. This enzyme binds to TNF-R1 together with the ubiquitin-conjugating enzyme Ubc13. RNF8 modifies TNF-R1 by adding at least two K63-linked ubiquitin molecules to the receptor. This K63 ubiquitination initiates the internalization of activated TNF-R1, which is required for DISC recruitment, caspase-8 and caspase-3/7 activation, and propagation of apoptosis (3, 4).

FIG 8.

K63 ubiquitination of TNF-R1 mediated by RNF8 and Ubc13 regulates receptor endocytosis and thereby proliferative versus apoptotic signaling. Binding of TNF to its receptor initiates the formation of complex I, consisting of TRADD, RIP1, and TRAF2, mediating cell survival signaling via NF-κB. Simultaneously, the E3 ubiquitin ligase RNF8 is recruited to TNF-R1. In cooperation with the E2 enzyme Ubc13, RNF8 K63 ubiquitinates TNF-R1. This ubiquitination allows clathrin-dependent endocytosis of TNF-R1, which is required for recruitment of the DISC proteins FADD and caspase-8 to TNF-R1 and activation of caspase-8, leading to the induction of apoptosis.

Induction of apoptosis via death receptors has been well studied for several years. It is widely accepted that the most prominent death receptors (TNF-R1, CD95, and TRAIL-R1/2) can respond to ligand binding with seemingly contradictory biological outcomes, i.e., induction of inflammation, proliferation, or cell death. How this is regulated at the level of receptor modifications is still unknown. Our group recently showed that internalization and endosomal trafficking of TNF-R1 are absolutely mandatory for apoptosis induction. TNF-R1 has to enter the endolysosomal compartment to sequentially activate caspase-8 and -7, which induce A-SMase activity. This activates cathepsin D, which in turn cleaves BID to tBID, triggering the mitochondrial apoptosis amplification loop (3, 5, 8, 9, 19). TNF-R1 internalization is blocked when its TRID is deleted or mutated or when cells ectopically express the adenoviral protein E3-14.7K. The TRID contains a canonical Tyr-based internalization motif that is required for recruitment of the AP2 clathrin adaptor (3, 4, 20). Of note, all three of the experimental approaches described in this report, deletion of the TNF-R1 internalization domain, adenovirus 14.7K transfection, and experimentally induced RNF8 deficiency, mediated either by shRNA downmodulation or by genetic deletion in RNF8 knockout MEFs, lead to inhibition of TNF-R1 internalization, inhibition of DISC recruitment, caspase activation, and inhibition of cell death, leaving NF-κB activation and proliferative signaling intact. Interestingly, in RNF8-deficient cells, RIP-1 and TRAF2 could still be recruited to TNF-R1 for activation of NF-κB even at low levels of TRADD as an assembly platform, as also observed in our previous reports (3, 4). Although it is generally accepted that TRADD is required for RIP-1 recruitment to TNF-R1, this observation is in line with previous reports on a direct interaction of RIP-1 with TNF-R1 and CD95 (21, 22). In addition, Zheng et al. (23) and Jin and El Deiry (24) showed that TRADD and RIP-1 can independently and competitively associate with TNF-R1.

Supporting our findings on the important role of TNF-R1 internalization in apoptosis signaling, a very recent report revealed that CerS2-null mice that cannot synthesize very long acyl chain ceramides were resistant to lipopolysaccharide/galactosamine-mediated fulminant hepatic failure and cultured hepatocytes from these mice were also insensitive to TNF-α-mediated apoptosis (25). Both in the liver and in hepatocytes, caspase activity was not elevated, consistent with inhibition of TNF-R1 proapoptotic signaling. Caspase activation was blocked because of the inability of CerS2-null mice to internalize TNF-R1. These results indicate that alteration of the acyl chain composition of sphingolipids inhibits TNF-R1 internalization and inhibits selective proapoptotic downstream signaling for apoptosis both in vitro and in vivo.

Ubiquitination of receptors can lead to either their proteasomal degradation or trafficking of the receptor into the endolysosomal compartment and toward multivesicular bodies (MVB). Sorting of receptors into MVB provides intracellular signaling platforms and/or subsequent recycling of the receptor to the plasma membrane and/or leads to their lysosomal degradation (26). The importance of compartmentalized signaling has already been described for a number of other receptor systems, where ubiquitination and subsequent endocytosis of the receptors are needed for proper signal transduction (27–29).

Ubiquitination of TNF-R1 upon stimulation with TNF has been described before by Legler and colleagues (30), but the type and role of TNF-R1 ubiquitination remained unresolved. Using ubiquitin linkage-specific antibodies (16, 17), we showed here that TNF-R1 is K63 ubiquitinated upon stimulation and thereby regulates receptor internalization and proapoptotic signaling.

Computational prediction of putative ubiquitination sites (31) within TNF-R1 reveals several consensus sites for the ligation of ubiquitin molecules. We did not succeed in identifying the ubiquitination site by mutating the highest-scoring lysines within TNF-R1 (data not shown). TNF-R1 might have a preferential ubiquitination site, but mutation of this site may be compensated for by alternative lysines, which has been described in several cases (32–37).

Current data suggest that K63 ubiquitin chains may serve as a scaffold for the recruitment of other proteins via special K63-polyubiquitin interaction motifs (38–40). Such motifs are part of the proteins of the ESCRT (endosomal sorting complex required for transport) involved in the selection of ubiquitinated cargo and its sequential delivery into multivesicular compartments (41–43). Of note, recent reports by Mahul-Mellier and colleagues showed that an Alix/ALG-2 protein complex interacts with procaspase-8 and is needed for its translocation to TNF receptosomes. Alix, in turn, can interact with the ESCRT, which mediates the sorting of protein targets to MVB. Defects in Alix abrogate TNF-mediated apoptosis, probably by inhibiting the translocation of procaspase-8 to activated TNF-R1 (44, 45).

Of >700 E3 enzymes, only a few had been previously reported to be potentially involved in K63 ubiquitination. Among those, TRAF2, -6, and Itch/AIP-4 have been implicated in TNF-R1 signaling (13, 46). We found RNF8 to be recruited to TNF-R1 with the same kinetics as the occurrence of TNF-R1 ubiquitination.

RNF8 localization and activity have previously been detected in the cell nucleus (47–49). We detected both nuclear and cytoplasmic staining in our cell systems. Of note, we observed rapid nuclear-to-cytoplasmic translocation of RNF8 after TNF treatment. Differences in cellular localization could be due to the use of tagged RNF8 versions and the application of different antibodies and cell lines. The anti-RNF8 (X21) antibody used in the present study recognizes two RNF8 bands, presumably variants with and without a RING finger domain. Figure 2A demonstrates that mainly the larger, RING finger domain-containing protein, and thus the enzyme possessing ubiquitin ligase activity, is recruited to TNF receptosomes. The presence of RNF8 and TNF-R1 in one protein complex, and thus the possibility of direct interaction, was revealed in our coimmunoprecipitation and colocalization experiments.

In conclusion, our findings add a new level of regulation to TNF-R1 signaling at the earliest stage after ligand binding, K63 ubiquitination of the receptor as a first checkpoint for a cell to decide whether to survive after TNF treatment or to die by apoptosis. This molecular mechanism may serve as a novel target for pharmaceutical intervention strategies to modulate the biological responses to TNF by either stimulating RNF8 binding to TNF-R1 to facilitate ubiquitin ligation to the receptor and induce cell death or to inhibit RNF8-mediated K63 ubiquitination to enhance cell proliferation and rescue the cells from apoptosis.