TRAIL receptor gene editing unveils TRAIL-R1 as a master player of apoptosis induced by TRAIL and ER stress

TRAIL induces selective tumor cell death through TRAIL-R1 and TRAIL-R2. Despite the fact that these receptors share high structural homologies, induction of apoptosis upon ER stress, cell autonomous motility and invasion have solely been described to occur through TRAIL-R2. Using the TALEN gene-editing approach, we show that TRAIL-R1 can also induce apoptosis during unresolved unfolded protein response (UPR). Likewise, TRAIL-R1 was found to co-immunoprecipitate with FADD and caspase-8 during ER stress. Its deficiency conferred resistance to apoptosis induced by thaspigargin, tunicamycin or brefeldin A. Our data also demonstrate that tumor cell motility and invasion-induced by TRAIL-R2 is not cell autonomous but induced in a TRAIL-dependant manner. TRAIL-R1, on the other hand, is unable to trigger cell migration owing to its inability to induce an increase in calcium flux. Importantly, all the isogenic cell lines generated in this study revealed that apoptosis induced TRAIL is preferentially induced by TRAIL-R1. Taken together, our results provide novel insights into the physiological functions of TRAIL-R1 and TRAIL-R2 and suggest that targeting TRAIL-R1 for anticancer therapy is likely to be more appropriate owing to its lack of pro-motile signaling capability.


INTRODUCTION
TRAIL/APO2L selectively triggers cell death in a large variety of human tumors [1]. This promising antitumor protein belongs to the TNF subfamily of apoptosis-inducing ligands [2] and has attracted much interest in oncology. However, the use of TRAIL or its derivatives in the clinic has, so far, failed to demonstrate its efficacy, despite clear evidence of antitumor capabilities [3]. Apoptosis triggered by TRAIL occurs through its binding www.impactjournals.com/oncotarget/ Oncotarget, 2017, Vol. 8, (No. 6), pp: 9974-9985

Research Paper
to TRAIL-R1 and/or TRAIL-R2, also coined DR4 and DR5, respectively [4]. Activation of these cell surface receptors by TRAIL induces the formation of a socalled death-inducing signaling complex (DISC) on the cytosolic tail of the receptor, where the apoptosis-initiating caspase-8 and caspase-10 proteins are brought in close proximity by the adaptor protein FADD [5]. Recruitment of these two initiator caspases, through homotypic protein/ protein interactions within the TRAIL DISC, allows their activation and the subsequent release of their active cleaved forms to the cytosol where they activate by proteolytic cleavage effector caspases such as caspase-3, driving cell dismantling.
Both TRAIL agonist receptors share high structural homology and the ability to induce apoptosis through the formation of a DISC. Despite similar expression levels of agonist receptors in a given cancer cell type, engagement of apoptosis can occur either through TRAIL-R1 or TRAIL-R2 [6,7]. The reasons for this selective engagement remain unclear, so far. In addition, recent evidence suggests that TRAIL-R2, may also display protumorigenic potential [8,9] and contribute to endoplasmic reticulum stress-induced cell death in a ligand-independent manner [10]. With the aim of assessing selective signaling capabilities of TRAIL-R1 over TRAIL-R2, we generated TRAIL receptor deficient cells using a TALEN approach. The generated isogenic cell lines deficient for TRAIL-R1 and/or TRAIL-R2 allowed us to demonstrate that TRAIL-R1 and TRAIL-R2 display both similar and divergent signaling capacities. Our findings demonstrate that TRAIL-R1 is also able to contribute to cell death induced by unresolved unfolded-protein-response signals and that only TRAIL-R2 is able to induce a calciumdependent cell motility signaling activity. Importantly, although both receptors are able to induce apoptosis upon TRAIL binding, TRAIL-R1 appears to be more critically required than TRAIL-R2 for apoptosis triggering.

TRAIL-induced apoptosis proceeds preferentially through TRAIL-R1
TRAIL signal transduction pathways are fairly well understood, but the specific contribution of TRAIL-R1 and TRAIL-R2 in implementing apoptotic or nonapoptotic signaling remains still poorly understood. In order to get a better understanding of TRAIL-R1 and TRAIL-R2 specific signaling capabilities, we generated TRAIL receptor-deficient isogenic cell derivatives of the colorectal carcinoma cell line HCT116 using a Transcription activator-like effector nucleases (TALEN) gene editing approach. TRAIL-R1 and TRAIL-R2 right and left TALE were designed to target exon 1 and exon 3, respectively (Supplementary Figure 1A). TRAIL-R1and/or TRAIL-R2-deficient cells were obtained after transfection and successive sorting of cells lacking TRAIL receptor expression by flow cytometry ( Figure  1A-1D). TRAIL-R2-deficient (DR5 -/-) clones were obtained by limiting dilution and expansion for 14 to 24 days ( Figure 1E). Genomic alterations were confirmed by deep sequencing. As expected, and contrary to parental cells, all HCT116 isogenic TRAIL-R1-deficient (DR4-/-) and TRAIL-R1/TRAIL-R2 double knockout (DKO) cell populations (Pop) as well as the selected TRAIL-R2deficient (DR5-/-) clone displayed extensive insertions, deletions or substitutions events in the corresponding targeted locus (Supplementary Figure 1B). Compared to the parental cells, all DR5 -/-isogenic cells (population or derived clones) displayed complete resistance to cell death induced by the TRAIL-R2 selective peptidomimetic M1d ( Figure 1F), but only moderate protection to His-TRAIL ( Figure 1G). On the other hand, loss of TRAIL-R1 caused a pronounced protection against His-TRAIL-induced cell death, and as expected deletion of both receptors fully protected HCT116 cells (DKO) against both TRAIL and TRAIL derivatives ( Figure 1H). DR4-/-HCT116 cells displayed, a residual sensitivity to high concentrations of the TRAIL-R1 selective ligand 4C9, albeit to a much lower extent than WT His-TRAIL ( Figure 1H). 4C9 is a mutated version of TRAIL, selected for its higher affinity for TRAIL-R1 and low affinity for TRAIL-R2 [11]. Since, HCT116 DKO cells display complete resistance to 4C9, the residual activity of 4C9 in DR4-/-HCT116 cells was attributed to its ability to engage apoptosis through TRAIL-R2 at a high concentration.
The substantial reduction of TRAIL-sensitivity in the DR4-/-HCT116 cells is not due to damaged TRAIL-R2 functionality, as evidenced by their fully retained sensitivity to M1d ( Figure 1H). To confirm that the HCT116 isogenic cell clones or populations generated are strictly interfering with TRAIL signaling, cells were stimulated with increasing concentrations of Fas ligand. Methylene blue staining, demonstrated that Fas ligand-induced cell death remained unaltered in all these isogenic cells, irrespective of the concentration (Supplementary Figure 1C). In order to substantiate these findings, additional TRAIL-receptor-deficient isogenic cell lines were generated generated, from MDA-MB-231 breast cancer cells, SW480 colorectal carcinoma cells and H1703 lung cancer cells (Supplementary Figure  2A). Irrespective of the cell type, deletion of TRAIL-R1 in these cells (DR4-/-) was always more detrimental to TRAIL-induced cell death than deletion of TRAIL-R2 (DR5-/-) (Figures 2A and Supplementary Figure 2B). Remarkably, SW480 and H1703 DR5-/-cells were almost as sensitive as parental WT cells to TRAIL ( Figure 2A). As expected, DR5-/-cells were equally sensitive to the TRAIL-R1 selective TRAIL ligand 4C9 as the parental cells, whereas DR4-/-cells displayed high resistance to 4C9 (Figure 2A). Like HCT116 DR5-/-cells, SW480 cells deficient for TRAIL-R2 were also protected against M1d-induced cell death (Figure 2A). Parental H1703 and MDA-MB-231 and TRAIL receptor-deficient isogenic cells on the other hand displayed complete resistance to M1d (Figure 2A). In agreement with these results, DR4-/-isogenic cells displayed much higher resistance to TRAIL-induced apoptosis than DR5-/-cells, as monitored by annexin V staining ( Figure 2B). As expected, DR4-/and DR5-/-cells were also protected against apoptosis induced by Mapatumumab and Lexatumumab, two agonist antibodies selectively targeting TRAIL-R1 and TRAIL-R2, respectively ( Figure 2B). Although, TRAIL DISC composition was not significantly different in cells expressing either TRAIL-R1 or TRAIL-R2, as compared to parental cells (Figure 3), activation of the caspase-8, as evidenced by its cleaved products, was severely impaired in cells lacking TRAIL-R1 (DR4-/-). These findings altogether indicate that TRAIL-induced apoptosis in these cells is mainly triggered by TRAIL-R1.

TRAIL-induced cell motility is solely mediated through TRAIL-R2
Like CD95L/FasL [12], TRAIL has been reported in some circumstances to induce cell motility, invasion and metastasis [8,9,13]. Although non-apoptotic signalling associated with some of these events was assigned to TRAIL-R2 in cancer cells, in a cell autonomous manner, pro-migratory signalling events induced by TRAIL in normal cells remains unknown, except for monocytes for which chemotaxis and migration has been ascribed to TRAIL-R1 [14]. We therefore took advantage of the TRAIL-receptor-deficient isogenic cell lines to determine whether cell motility induced by TRAIL requires TRAIL-R1, TRAIL-R2 or both receptors. In order to address this question and to avoid apoptosis-triggering, we used trimers of TRAIL, which are unable to induce apoptosis (Supplementary Figure 3A). Soluble cleaved TRAIL (cl-TRAIL) was purified from supernatants of  CHO cells transiently transfected with full-length TRAIL ( Figure 4A), or from bacterial extracts (sTRAIL) [15]. After ultracentrifugation of the conditioned media to remove exosomes, the amount of cl-TRAIL recovered was estimated by ELISA, and used in the Boyden-chamber assay with MDA-MB-231 cells to assess its ability to trigger cell motility. Like cleaved CD95L/Fas Ligand, cl-TRAIL was able to increase MDA-MB-231 cell motility ( Figure 4B). Motility induced by cl-TRAIL in these cells was also recapitulated using the purified Flag-tagged TRAIL (sTRAIL) ( Figure 4C), a soluble trimer devoid of pro-apoptotic activity in the absence of cross-linking (Supplementary Figure 3A and [16]). Interestingly, deficiency in TRAIL-R2, but not TRAIL-R1, abolished sTRAIL-induced MDA-MB-231 cell motility ( Figure  4C). Similar results were obtained in TRAIL-R2-deficient HCT116 isogenic cells with both DR5-/-and DKO cells displaying reduced cell motility after sTRAIL stimulation as compared to WT or DR4-/-(Supplementary Figure  3B). Consistent with this observation, sTRAIL-induced MDA-MB-231 cell invasion in vivo involved TRAIL-R2, as assessed using a chick embryo chorioallantoic membrane CAM assay ( Figure 4D-4E). Deficiency in TRAIL-R2 inhibited MDA-MB-231 cell migration ( Figure 4D) and invasion as demonstrated by the absence of Alu sequences in the chick embryo ( Figure 4E). Contrary to DR5-/-cells, WT and DR4-/-cells, which express TRAIL-R2, were able to migrate and invade the host organism ( Figure 4D-4E). Cell motility has been reported to be tightly associated with changes in calcium flux [17]. Accordingly, sTRAIL induced a calcium  Figure 3C). Notwithstanding, all MDA-MB-231 parental or isogenic cells were able to respond to thapsigargin, a non-competitive sarco/endoplasmic reticulum Ca2+ pumps (SERCAs) inhibitor ( Figure 4F-4G), indicating that the calcium response is selectively induced by TRAIL-R2 upon sTRAIL stimulation. These results were confirmed in HCT116 cells (Figures 4H and Supplementary Figure 3D). It should be noted here that migration induced by TRAIL-R2 is non-self autonomous, as spontaneous migration of HCT116 and MDA-MB-231 cells is not altered in the absence of the receptor, as compared to parental or DR4 -/-cells (Supplementary Figure 3E). However, migration induced by FCS was clearly reduced in TRAIL-R2-deficient cells, suggesting that soluble TRAIL may be present in FCS, in addition to other chemoattractants (Supplementary Figure 3E).

Unresolved UPR-mediated apoptosis involves both TRAIL-R2 and TRAIL-R1
TRAIL-R2 is known to be regulated and contribute to ER stress-mediated apoptosis, independent of its cognate ligand [10]. Accordingly, in the three cell lines tested namely HCT116, H1703 and MDA-MB-231, thapsigargin (TG), tunicamycin (TM) and brefeldin A (BfA) induce up-regulation of TRAIL-R2 in parental or DR4 -/-cells, but not in DR5-/-and DKO isogenic cells ( Figure 5A), and in agreement with previous findings [10], deficiency in TRAIL-R2 significantly protected HCT116 cells from apoptosis induced by TG ( Figures 5B and 4A). As expected, HCT116 DKO cells were also significantly less sensitive than parental cells to TG-induced apoptosis. Protection against TG-induced apoptosis mediated by TRAIL-R2 deficiency was also evidenced in H1703 and  . Resistance to ER-stress inducers, as shown here with MDA-MB-231 isogenic cells, was not associated with a downstream disability to transduce apoptosis, since apoptosis-induced by Fas ligand, another ligand of the TNF superfamily, was unaffected whether the cells harbored or not these receptors (Supplementary Figure 4C).
Interestingly, although BIP or CHOP were induced after ER stress stimulation in all cell lines, and despite the fact that they likely contribute to both TRAIL-R1 and TRAIL-R2 up-regulation [10,18] (Figure 5A), the percentage of cells undergoing apoptosis varied both depending on the cell line and the ER stressinducer ( Figure 5A). These finding suggest that UPRmediated apoptosis may not be strictly correlated with transcriptional regulation of TRAIL-R1 and TRAIL-R2. Accordingly, regulation of TRAIL-R1 and TRAIL-R2 expression levels in WT HCT116 and MDA-MB-231 cells did not correlate with apoptosis induced by these compounds. For instance while BfA induced stronger up-regulation of both receptors as compared to TG in both cell lines ( Figure 5A), it induced weaker apoptosis than TG or TM in MDA-MB-231 cells ( Figure 5B). Irrespective of whether apoptosis-induced by TRAIL receptors upon ER stress is directly linked to their transcriptional up-regulation, and in agreement with previous findings [10], apoptosis involved formation of a complex associating caspase-8, FADD, TRAIL-R1 and TRAIL-R2 in a TRAIL-independent manner ( Figure 5D-5F). Likewise, stimulation of parental HCT116 and MDA-MB-231 cells with TG or BfA induced the formation of heteromers of TRAIL-R1 and TRAIL-R2, as evidenced by immunoprecipitation (Figure 5D-5E) and the formation of a DISC-like complex [10] (Figure 5F). Our results therefore support the finding that both TRAIL-R1 and TRAIL-R2 are able to form an active DISC-Like complex upon BfA stimulation and contribute, at least in part, to apoptosis induced by ER stress inducers.

DISCUSSION
TRAIL and its agonist receptors have been considered for anticancer therapies for over two decades [3]. TRAIL-R1 and TRAIL-R2 display 50 to 53% overall homology and share the ability to recruit FADD and caspase-8, allowing apoptosis triggering. Although TRAIL binding affinity for TRAIL-R2 was shown to be slightly stronger than TRAIL-R1 [11,19], apoptosis is preferentially engaged through TRAIL-R1 in chronic leukemic lymphocytic cells, despite high expression levels of TRAIL-R2 [7,20,21]. While both receptors are often co-expressed in tumor cell lines, and albeit loss of TRAIL-R1 can be found in many tumor cell lines, occurrence of a TRAIL-R2-deficiency in TRAIL-R1 expressing cells is rare, and has to our knowledge only been described in the human erythroleukemia cell line Hel [22]. So far, the studies aiming at analyzing the specific contribution of TRAIL-R1 or TRAIL-R2 during TRAIL-induced apoptosis have mostly relied on mutated TRAIL ligands [11], peptidomimetics [23] or agonistic monoclonal antibodies [24]. However, these agonists introduce bias, due either to their residual affinity for nontargeted receptors or to their reduced potency as compared to TRAIL itself.
In this work we have generated TRAIL-receptordeficient isogenic cell lines to analyze the relative contribution of each TRAIL agonist receptor in triggering apoptosis or non-apoptotic signaling using the same ligand. Our findings, that apoptosis induced by TRAIL in a small panel of carcinoma cell lines is preferentially induced by TRAIL-R1 together with the demonstration that TRAIL-induced migration is solely triggered by TRAIL-R2 or that TRAIL-R1 is able to contribute to apoptosis induced by ER stress inducers, uncover TRAIL-R1 as an important target for cancer therapy. Until now, most studies have focused on TRAIL-R2, neglecting TRAIL-R1 [3]. Accordingly, only one anti-TRAIL-R1 antibody (Mapatumumab) has been assessed in clinical trials, whereas 5 distinct anti-TRAIL-R2 antibodies have been tested, so far [3,25]. TRAIL-R2 has probably attracted more attention than TRAIL-R1 owing to the fact that its expression is more prone to transcriptional regulation by genotoxic or non-genotoxic compounds and that its increase in expression is often associated with increased sensitivity to TRAIL-induced cell death [10,26,27]. However, as demonstrated here using TRAILdeficient isogenic cell lines, and in a recently published work using a siRNA approach [8], TRAIL-R2 but not TRAIL-R1, can induce tumor cell motility and contribute to metastasis in vivo. In contrast to the cell autonomous pro-motile signaling capability attributed to TRAIL-R2 [8], our present work indicates that this signaling property requires TRAIL. Along this line, we have shown in the past that even the most potent soluble TRAIL (His-TRAIL) can induce the migration of HCT116 cells exhibiting a deficiency in the intrinsic pathway, due to the absence of Bax expression or to the overexpression of Bcl-2 anti-apoptotic family members [9]. This nonapoptotic signaling is also shared by Fas [12], another member of the TNF superfamily. TRAIL-R2 pro-motile signaling should thus considered as a potential threat for TRAIL-based cancer therapy, as it may potentially lead to increased occurrence of metastasis.
TRAIL-R1, on the other hand is unable to trigger this non-apoptotic signaling pathway, but is extremely potent in triggering TRAIL-induced apoptosis. Moreover, the present work reveals for the first time that, like TRAIL-R2, TRAIL-R1 is also able to contribute to unresolved UPR-mediated apoptosis. In the course of ER stress, TRAIL-R1 is engaged in a macromolecular complex including FADD and caspase-8 and contributes to apoptosis induced by ER stress inducers.
While additional work will be required to fully understand why TRAIL engages apoptosis preferentially through TRAIL-R1, or how TRAIL-R1 engages the caspase-8 machinery in the absence of TRAIL during unresolved UPR, our findings provide novel insights to TRAIL agonist receptors physiological function and delineate TRAIL-R1 as a priority for cancer therapy.

Cell lines
The human colon carcinoma HCT116 and SW480 (ATCC), breast carcinoma MDA-MB-231 (ATCC) were maintained in DMEM (Lonza, Levallois-Perret, France) with 10% fetal calf serum. The lung carcinoma H1703 (kind gift from Pr Lebecque, Lyon, France) was maintained in RPMI (Lonza) as above. Cells were cultured in 5% CO 2 . Chinese hamster ovary (CHO) cells were maintained in DMEM as above, and in OPTI-MEM (Lonza) after transfection.

TALEN plasmids and locus
Plasmids pTAL.EF1a-Left and pTAL.EF1a-Right encoding TRAIL-R1 and TRAIL-R2 custom TALEN ® were generated by Cellectis (OM1303 and OM1304 for TRAIL-R2 and OM1307 and OM1308 for TRAIL-R1). Equal amounts of each pair of TALENs, ranging from 1 to 6 μg, were introduced by transfection into HCT116 cells grown in 12-well plates using Fugene HD (Promega, Charbonnières, France). Transfected cells were cultured for 6 days at 37°C before sorting out the TRAIL-R1-and TRAIL-R2-deficient cells with cell sorting. The sorted cell populations were allowed to grow for 6-14 days and standard limited dilutions in 96-well plates were performed to expand individual clones. Clonal populations were expanded for 14 to 24 days before screening for loss of TRAIL receptor expression by flow cytometry. To confirm gene disruption, genomic DNA was isolated from individual clones or cell populations or cell populations and the targeted locus was analyzed by deep sequencing.

Deep sequencing
One to 5 million cells were harvested, washed twice in PBS and genomic DNA was extracted using either a genomic DNA lysis buffer (10 mM TRIS, pH8; 0.45% (V/V) NP40; 0.45% (V/V) Tween20; 100 μg/ml proteinase K) or a genomic DNA extraction kit (DNeasy Blood & Tissue Kit from Qiagen), respectively. Lysates were obtained using the genomic DNA lysis buffer after incubation for 10 min at room temperature, followed by 2 h at 55°C and finally for 10 min at 95°C. These lysates were stored at 4°C for up to one week. Genomic extraction using the DNeasy Blood & Tissue Kit was performed according to the manufacturer's instructions. PCR were performed using the primers containing the adaptor sequences needed for deep-sequencing: forward 5'-GGT-ATG-ATG-AAG-ACC-AAG-GTG-GA-3'. PCR products were purified using Agencourt AMPure XP kit (Beckman Coulter, Villepinte, France) and sequenced using one-way reads amplicon sequencing method with the 454 sequencing system from (Roche, Meylan, France) according to manufacturer's instructions. Several thousands of sequences were obtained per PCR product and analyzed for the presence of site-specific insertion or deletion events at the TALEN cleavage site using the Cellectis proprietary algorithm. Analysis took into account one base pair insertion or deletion events.

Chick chorioallantoic membrane (CAM) assay
Chick embryos from 3-days-old eggs were opened and placed in double Petri dishes with water for humidity. After incubation for 7 days at 37.5°C, 5×10 6 in Matrigel/ serum (B&D) MDA-MB-231 cells expressing both receptors (WT), TRAIL-R2 only (DR4 -/-) or TRAIL-R1 onle (DR5-/-cells) were applied on the CAM on a filter paper disk (diameter of 6 mm). Tumour cell migration was analyzed after 7 days of incubation at 37°C. Cell migration was observed using a digital camera at 10 × magnification and by PCR on whole chick embryo using human Alu sequences. Detection of human Alu sequences from CAM studies was performed on whole chick embryo tissue samples. Genomic DNA was isolated from CAM's tissues using DNA Isolation Kit (DNeasy Blood & Tissue Kit, Qiagen) and Alu sequences were amplified using the following primers : Alu-Forward 5' ACG-CCT-GTA-ATC-CCA-GCA-CTT 3'and Alu-Reverse 5' TCG-CCC-AGG-CTG-GAG-TGC-A 3'. PCR analysis was performed in a 50 μl PCR reaction containing 1 μg of cDNA, 2 μl dNTPs (2.5 mmol/l each), 1 μl each of the specific primers (20 μM), 5 μl hot start Taq buffer 10X and 0.5 μl hot start www.impactjournals.com/oncotarget Taq DNA polymerase (5 U/μl; Promega) and amplified as follows: 2 min at 95°C followed by 30 cycles of 30 sec at 95°C, 45 sec at 58°C, 45 sec at 72°C; and a final step of 10 min at 72°C. Products were visualized as described above.

Cell viability and annexin V assays
TRAIL agonists (His-hTRAIL, 4C9 and M1d), or FasL agonist (Fc-FasL) were titrated in 96-well plates using 5x10 4 cells per well, incubated at 37°C for 16 hours. Plates were incubated at 37°C for 16 hours with 5% CO 2 , and cell viability was assessed using methylene blue staining as described previously [29]. Apoptosis induced by TRAIL and the above-mentioned TRAIL agonists, as well as Mapatumumab and Lexatumumab was analyzed by flow cytometry using the Annexin V-FITC/ PI staining kit purchased from Miltenyi Biotec (Miltenyi Biotec, Paris, France) according to the manufacturer's instructions. Stained cells were analyzed with a BD LSR2 flow cytometer (BD Biosciences, Le Pont de Claix, France). The percentage of Annexin V-positive cells was calculated as the number of cells demonstrating Annexin V staining (PI negative or positive) divided by the total number of cells examined. Experiments were repeated at least three times.

In vitro cell motility assays
For in vitro motility assays, cells were serum-starved overnight, detached using trypsin and 10 5 cells were added onto 8 μm pore-sized membrane top Boyden chambers placed in 24-well plates (Millipore, Molsheim, France). Top chambers were filled with low serum (1%)-containing medium, while the bottom chambers were filled as above in the presence or absence of 100 ng/ml of cleaved CD95L (cl-CD95L), TRAIL (cl-TRAIL) or Flag-TRAIL and the plates were cultured for 24 h at 37°C in a 5% CO2, humidified incubator. Cell invasion was quantified as follows: the non-motile cells from the top side of the membrane were removed mechanically using cottontipped swabs, and the invading cells from the reverse side of the membrane were fixed with methanol and stained with Giemsa. For each experiment, representative pictures were taken for each insert, then cells were lyzed and absorbance at 560 nm correlating to the amount of Giemsa stain was measured.

Calcium video imaging in living cells
Single-cell [Ca2+]cyt imaging was performed ratiometrically, using either Fura-2 AM or FuraPE3-AM calcium indicators. Serum-starved cells were loaded with 1 μM Fura-2 AM or 1 μM FuraPE3-AM at room temperature in Hank's Balanced Salt Solution (HBSS) for 30 min. FuraPE3-AM exhibits limited compartmentalization in intracellular stores and is leakage resistant [30]. Cells were rinsed with HBSS and incubated in the absence of the Ca2+ probe for 15 min to complete dye de-esterification. For Fura-2 AM staining, cells were placed under an inverted fluorescence microscope (Axiovert 40 CFL) and stimulated with Flag-TRAIL (100 ng/ml) or thapsigargin (for the indicated period of time). Images were captured at 510 nm by a fast-scan camera. Fura2-AM was alternately excited at 340 and 380 nm, and ratios of the resulting images (excitations at 340 and 380 nm and emission filter at 520 nm) were produced at constant intervals (2s or 5s according to the stimulus). Fura-2 ratio images (Fratio 340/380) were displayed and the Fratio values from the regions of interest (ROIs) drawn on individual cells were monitored over time and analyzed with the Axiovision software. Each experiment was independently repeated 3 times, and for each experimental condition, an average of more than 10 single-cell traces was displayed. For FuraPE3-AM analysis, cells were stimulated as above and fluorescence micrograph images were captured at 510 nm, using an inverted epifluorescence microscope (Olympus IX70) equipped with a 40x UApo/340-1.15W objective (Olympus), and fluorescence micrograph images were captured at 510 nm and 12-bit resolution by a fast-scan camera (CoolSNAP fx Monochrome, Photometrics). To minimize UV light exposure, a 4×4 binning function was used. Fura-PE3 was alternately excited at 340 and 380 nm, and the ratios of the resulting images (emission filter at 520 nm) were produced at constant 10s intervals. ROIs were drawn on selected recorded cells to restrict data collection to specific regions. Fratio reflected the changes in intracellular Ca2+ concentration. Fluorescence of regions of interest was normalized to baseline fluorescence (F/F0). Each experiment was repeated 3 times, and the average of more than 20 single-cell traces was analyzed. Data was processed using OriginPro 7.5 software (Origin Lab).

Immunoprecipitations
For native DISC analysis, 8x10 7 cells were stimulated in 1 ml of complete medium with 1 μg of His-hTRAIL ligand for the specified times at 37°C. Cells were then washed with PBS before lysis in NP40 buffer (1 % NP40, 20 mM Tris-HCl pH 7.5, 150 mM NaCl, and 10 % glycerol) containing a protease inhibitor cocktail. Alternatively, adherent cells were stimulated with thapsigargin or brefeldin A for the indicated periods of time in the absence of His-TRAIL. Cells were then detached and collected using trypsin, and lysed in Triton-X100 (1 % Triton-X100, 30 mM Tris-HCl pH 7.5, 150 mM NaCl). Cell lysates were pre-cleared with Sepharose 6B (Sigma-Aldrich) for 1 hour at 4°C. After centrifugation, the TRAIL receptor DISC was immunoprecipitated overnight at 4°C from the cell lysates in the presence of protein G-coated beads (Amersham Biosciences, Les Ullis, France) and antibodies targeting caspase-8 (C20), 6xHis (clone AD1.1.10), TRAIL-R1 (wB-T29) or TRAIL-R2 (B-D37). Beads were washed four times with the corresponding NP40 or Triton-X100 lysis buffer, and the immunoprecipitated complexes were eluted in loading buffer (63 mM Tris-HCl pH 6.8, 2 % SDS, 0.03 % phenol red, 10% glycerol, 100 mM DTT), after a 5 min incubation at 95°C. Samples were then processed for immunoblot analysis.

Western blotting
Immunoprecipitates or cell lysates (in NP40 lysis buffer) were separated by SDS-PAGE then transferred to nitrocellulose membranes. Nonspecific binding sites were blocked by incubation in PBS containing 0.05% Tween 20 and 5% powdered milk. Immunoblots were incubated with a specific primary antibody, then an HRP-conjugated secondary antibody. Blots were developed using the Advansta WersternBright chemiluminescence substrate according to the manufacturer's protocol (Diagomics, Blagnac, France).

Statistical analysis
Statistical analyses were performed by one-way or two-way ANOVA using the GraphPad Prism program. Significant differences in cytotoxicity were determined by the Bonferroni's multiple comparison test. P-values *< 0.05, **< 0.01, ***< 0.001 were considered significant. All experiments were performed at least 3 times.